	ArXiv	Old scans math	Tables	Old scans	Headers & footers	Multi column	Long tiny text	Base	Overall
Mistral OCR API	77.2	67.5	60.6	29.3	93.6	71.3	77.1	99.4	72.0±1.1
Marker 1.10.1	83.8	66.8	72.9	33.5	86.6	80.0	85.7	99.3	76.1±1.1
MinerU 2.5.4*	76.6	54.6	84.9	33.7	96.6	78.2	83.5	93.7	75.2±1.1
DeepSeek-OCR	77.2	73.6	80.2	33.3	96.1	66.4	79.4	99.8	75.7±1.0
Nanonets-OCR2-3B	75.4	46.1	86.8	40.9	32.1	81.9	93.0	99.6	69.5±1.1
PaddleOCR-VL*	85.7	71.0	84.1	37.8	97.0	79.9	85.7	98.5	80.0±1.0
Infinity-Parser 7B*	84.4	83.8	85.0	47.9	88.7	84.2	86.4	99.8	82.5±?
Chandra OCR 0.1.0*	82.2	80.3	88.0	50.4	90.8	81.2	92.3	99.9	83.1±0.9

olmOCR v0.4.0	83.0	82.3	84.9	47.7	96.1	83.7	81.9	99.7	82.4±1.1

ArXiv

Old
scans
math

Tables

Old
scans

Headers
&
footers

Multi
column

Long
tiny
text

Base

Overall

Mistral OCR API

77.2

67.5

60.6

29.3

93.6

71.3

77.1

99.4

72.0±1.1

Marker 1.10.1

83.8

66.8

72.9

33.5

86.6

80.0

85.7

99.3

76.1±1.1

MinerU 2.5.4*

76.6

54.6

84.9

33.7

96.6

78.2

83.5

93.7

75.2±1.1

DeepSeek-OCR

77.2

73.6

80.2

33.3

96.1

66.4

79.4

99.8

75.7±1.0

Nanonets-OCR2-3B

75.4

46.1

86.8

40.9

32.1

81.9

93.0

99.6

69.5±1.1

PaddleOCR-VL*

85.7

71.0

84.1

37.8

97.0

79.9

85.7

98.5

80.0±1.0

Infinity-Parser 7B*

84.4

83.8

85.0

47.9

88.7

84.2

86.4

99.8

82.5±?

Chandra OCR 0.1.0*

82.2

80.3

88.0

50.4

90.8

81.2

92.3

99.9

83.1±0.9

olmOCR v0.4.0

83.0

82.3

84.9

47.7

96.1

83.7

81.9

99.7

82.4±1.1

	ArXiv	Old scans math	Tables	Old scans	Headers & footers	Multi column	Long tiny text	Base	Overall
Mistral OCR API	77.2	67.5	60.6	29.3	93.6	71.3	77.1	99.4	72.0±1.1
Marker 1.10.1	83.8	66.8	72.9	33.5	86.6	80.0	85.7	99.3	76.1±1.1
MinerU 2.5.4*	76.6	54.6	84.9	33.7	96.6	78.2	83.5	93.7	75.2±1.1
DeepSeek-OCR	77.2	73.6	80.2	33.3	96.1	66.4	79.4	99.8	75.7±1.0
Nanonets-OCR2-3B	75.4	46.1	86.8	40.9	32.1	81.9	93.0	99.6	69.5±1.1
PaddleOCR-VL*	85.7	71.0	84.1	37.8	97.0	79.9	85.7	98.5	80.0±1.0
Infinity-Parser 7B*	84.4	83.8	85.0	47.9	88.7	84.2	86.4	99.8	82.5±?
Chandra OCR 0.1.0*	82.2	80.3	88.0	50.4	90.8	81.2	92.3	99.9	83.1±0.9

olmOCR (first release)	63.3	67.5	62.3	38.6	93.4	67.6	54.8	97.9	68.2±1.1
v0.1.60 + Dynamic temp scaling	71.4	73.1	65.6	40.5	93.2	76.6	64.9	96.7	72.8±1.2
v0.1.68 + Better prompting	76.3	76.0	70.2	43.2	94.1	77.5	71.9	96.8	75.8±1.0
v0.2.0 + New trainer, YAML, img resize, Qwen 2.5 VL	78.8	77.5	71.9	45.4	94.2	78.6	81.4	99.8	78.5±1.1
v0.3.0 + Handle blank pages	78.6	79.9	72.9	43.9	95.1	77.3	81.2	98.9	78.5±1.1
v0.4.0 + Synth data, RLVR, souping	83.0	82.3	84.9	47.7	96.1	83.7	81.9	99.7	82.4±1.1

ArXiv

Old
scans
math

Tables

Old
scans

Headers
&
footers

Multi
column

Long
tiny
text

Base

Overall

Mistral OCR API

77.2

67.5

60.6

29.3

93.6

71.3

77.1

99.4

72.0±1.1

Marker 1.10.1

83.8

66.8

72.9

33.5

86.6

80.0

85.7

99.3

76.1±1.1

MinerU 2.5.4*

76.6

54.6

84.9

33.7

96.6

78.2

83.5

93.7

75.2±1.1

DeepSeek-OCR

77.2

73.6

80.2

33.3

96.1

66.4

79.4

99.8

75.7±1.0

Nanonets-OCR2-3B

75.4

46.1

86.8

40.9

32.1

81.9

93.0

99.6

69.5±1.1

PaddleOCR-VL*

85.7

71.0

84.1

37.8

97.0

79.9

85.7

98.5

80.0±1.0

Infinity-Parser 7B*

84.4

83.8

85.0

47.9

88.7

84.2

86.4

99.8

82.5±?

Chandra OCR 0.1.0*

82.2

80.3

88.0

50.4

90.8

81.2

92.3

99.9

83.1±0.9

olmOCR (first release)

63.3

67.5

62.3

38.6

93.4

67.6

54.8

97.9

68.2±1.1

v0.1.60 + Dynamic temp scaling

71.4

73.1

65.6

40.5

93.2

76.6

64.9

96.7

72.8±1.2

v0.1.68 + Better prompting

76.3

76.0

70.2

43.2

94.1

77.5

71.9

96.8

75.8±1.0

v0.2.0 + New trainer, YAML, img resize, Qwen 2.5 VL

78.8

77.5

71.9

45.4

94.2

78.6

81.4

99.8

78.5±1.1

v0.3.0 + Handle blank pages

78.6

79.9

72.9

43.9

95.1

77.3

81.2

98.9

78.5±1.1

v0.4.0 + Synth data, RLVR, souping

83.0

82.3

84.9

47.7

96.1

83.7

81.9

99.7

82.4±1.1

with class 'image'.\n" + "4. Render math equations inline using \$ \$ or \\[ \\].\n" + "5. Preserve any multi-column layout using CSS flexbox or grid.\n" + f"6. The viewport is fixed at {png_width // 2}x{png_height // 2} pixels.\n\n" + "Enclose your HTML in a ```html code block." + ), + }, + ], + } + ], + ) + + initial_html = "" + for content in initial_response.content: + if content.type == "text": + initial_html += content.text + + return extract_code_block(initial_html) + except Exception as e: + print(f"Error calling Claude API: {e}") + return None + + +def generate_tests_from_html(html_content: str, pdf_id: str, page_num: int) -> List[Dict]: + """ + Generate order tests from HTML content by splitting the main text into sentences and pairing them. + Only order tests are generated. + """ + tests = [] + pdf_filename = pdf_id + soup = BeautifulSoup(html_content, "html.parser") + full_text = soup.get_text(separator=" ").strip() + + sentences = [] + for paragraph in process(full_text): + for sentence in paragraph: + sentence_str = "".join(token.spacing + token.value for token in sentence).strip() + if sentence_str: + sentences.append(sentence_str) + + if len(sentences) < 2: + return tests + + all_indexes = list(range(len(sentences))) + random.shuffle(all_indexes) + random_pairs = [(all_indexes[i * 2], all_indexes[i * 2 + 1]) for i in range(len(all_indexes) // 2)] + random_pairs = [(min(i, j), max(i, j)) for (i, j) in random_pairs] + + num_order_tests = 0 + for i, j in random_pairs: + first_sentence = sentences[i] + second_sentence = sentences[j] + if len(first_sentence) < 10 or len(second_sentence) < 10: + continue + first_sentence = first_sentence.split("\n")[0].strip() + second_sentence = second_sentence.split("\n")[0].strip() + max_diffs = round(max(len(first_sentence), len(second_sentence)) * 0.05) + if max_diffs > len(first_sentence) // 2 or max_diffs > len(second_sentence) // 2: + continue + tests.append( + { + "pdf": pdf_filename, + "page": page_num, + "id": f"{pdf_id}_order_{uuid.uuid4().hex[:8]}", + "type": "order", + "before": first_sentence, + "after": second_sentence, + "max_diffs": max_diffs, + } + ) + num_order_tests += 1 + if num_order_tests >= 5: + break + + return tests + + +async def render_pdf_with_playwright(html_content, output_pdf_path, png_width, png_height): + """ + Render HTML content using Playwright and save it as a PDF. + Tries different scale factors until the output PDF has exactly one page. + """ + scale_factors = [1.0, 0.9, 0.8, 0.7, 0.6, 0.5] + for scale in scale_factors: + try: + async with async_playwright() as p: + browser = await p.chromium.launch() + page = await browser.new_page(viewport={"width": int(png_width // 2 * scale), "height": int(png_height // 2 * scale)}) + await page.set_content(html_content) + + # Add KaTeX assets for math rendering + katex_dir = os.path.join(os.path.dirname(os.path.abspath(__file__)), "..", "katex") + katex_css_path = os.path.join(katex_dir, "katex.min.css") + katex_js_path = os.path.join(katex_dir, "katex.min.js") + katex_autorender_js_path = os.path.join(katex_dir, "auto-render.min.js") + + await page.add_style_tag(path=katex_css_path) + await page.add_script_tag(path=katex_js_path) + await page.add_script_tag(path=katex_autorender_js_path) + + await page.evaluate(""" + renderMathInElement(document.body, { + delimiters: [ + {left: '\\\$begin:math:text$', right: '\\\\\\$end:math:text$', display: false}, + {left: '\\\$begin:math:display$', right: '\\\\\\$end:math:display$', display: true} + ], + throwOnError: false + }); + """) + + await page.pdf(path=output_pdf_path, scale=scale, print_background=True) + await browser.close() + + try: + reader = pypdf.PdfReader(output_pdf_path) + if len(reader.pages) == 1: + print(f"Successfully rendered as a single page PDF with scale factor {scale}") + return True + else: + print(f"PDF has {len(reader.pages)} pages with scale factor {scale}, trying a smaller scale...") + except Exception as pdf_check_error: + print(f"Error checking PDF page count: {pdf_check_error}") + return False + + except Exception as e: + print(f"Error rendering PDF with Playwright at scale {scale}: {e}") + print("Failed to render PDF as a single page with any scale factor") + return False + + +def process_pdf(pdf_info, args, client): + """ + Process a single PDF from a local folder: + - Select a random page from the PDF. + - Render that page as a base64 PNG. + - Generate HTML from the image using Claude. + - Optionally render the HTML to a PDF using Playwright. + - Generate order tests from the HTML. + """ + local_pdf_path, index = pdf_info + original_pdf_name = os.path.basename(local_pdf_path) + pdf_id = original_pdf_name + temp_pdf_dir = os.path.join(args.temp_dir, f"{os.path.splitext(original_pdf_name)[0]}") + os.makedirs(temp_pdf_dir, exist_ok=True) + + try: + reader = pypdf.PdfReader(local_pdf_path) + num_pages = len(reader.pages) + if num_pages == 0: + print(f"PDF has no pages: {local_pdf_path}") + return None + page_num = random.randint(1, num_pages) + except Exception as e: + print(f"Error reading {local_pdf_path}: {e}") + return None + + try: + image_base64 = render_pdf_to_base64png(local_pdf_path, page_num, target_longest_image_dim=2048) + except Exception as e: + print(f"Error rendering page {page_num} from {local_pdf_path}: {e}") + return None + + try: + html_content = generate_html_from_image(client, image_base64) + if not html_content: + print(f"Failed to generate HTML for {local_pdf_path}, page {page_num}") + return None + except Exception as e: + print(f"Error generating HTML for {local_pdf_path}: {e}") + return None + + html_dir = os.path.join(args.output_dir, "html") + pdfs_dir = os.path.join(args.output_dir, "pdfs") + os.makedirs(html_dir, exist_ok=True) + os.makedirs(pdfs_dir, exist_ok=True) + + html_path = os.path.join(html_dir, f"{os.path.splitext(original_pdf_name)[0]}_page{page_num}.html") + try: + with open(html_path, "w") as f: + f.write(html_content) + except Exception as e: + print(f"Error saving HTML for {local_pdf_path}: {e}") + return None + + playwright_pdf_path = None + render_success = False + playwright_pdf_filename = f"{os.path.splitext(original_pdf_name)[0]}_page{page_num}.pdf" + if not args.skip_playwright: + playwright_pdf_path = os.path.join(pdfs_dir, playwright_pdf_filename) + try: + png_width, png_height = get_png_dimensions_from_base64(image_base64) + render_success = asyncio.run(render_pdf_with_playwright(html_content, playwright_pdf_path, png_width, png_height)) + if render_success: + print(f"Successfully rendered with Playwright: {playwright_pdf_path}") + else: + print(f"Failed to render as a single page PDF: {playwright_pdf_path}") + playwright_pdf_path = None + except Exception as e: + print(f"Failed to render with Playwright: {e}") + playwright_pdf_path = None + render_success = False + if not args.skip_playwright and not render_success: + return None + + tests = generate_tests_from_html(html_content, pdf_id, page_num) + # IMPORTANT: Preserve the original PDF filename in the tests. + for test in tests: + test["pdf"] = original_pdf_name + + try: + if os.path.exists(temp_pdf_dir): + subprocess.run(["rm", "-rf", temp_pdf_dir]) + except Exception as e: + print(f"Error cleaning up temp directory {temp_pdf_dir}: {e}") + + return { + "pdf_id": pdf_id, + "pdf_path": local_pdf_path, + "page_number": page_num, + "html_path": html_path, + "playwright_pdf_path": playwright_pdf_path, + "tests": tests, + "num_tests": len(tests), + } + + +def main(): + parser = argparse.ArgumentParser(description="Convert PDFs in a folder to HTML templates and render with Playwright (order tests only)") + parser.add_argument("--input_dir", required=True, help="Folder containing PDF files") + parser.add_argument("--output_dir", required=True, help="Directory to store HTML and tests") + parser.add_argument("--temp_dir", default="/tmp/mine_tables", help="Directory for temporary files") + parser.add_argument("--max_tests", type=int, default=100, help="Maximum number of PDFs to process (randomly selected)") + parser.add_argument("--parallel", type=int, default=1, help="Number of parallel threads to use") + parser.add_argument("--api_key", help="Claude API key (or set ANTHROPIC_API_KEY environment variable)") + parser.add_argument("--skip_playwright", action="store_true", help="Skip Playwright PDF rendering") + parser.add_argument("--verbose", action="store_true", help="Enable verbose output") + args = parser.parse_args() + + os.makedirs(args.output_dir, exist_ok=True) + os.makedirs(args.temp_dir, exist_ok=True) + + api_key = args.api_key or os.environ.get("ANTHROPIC_API_KEY") + if not api_key: + print("Error: API key not provided. Use --api_key or set ANTHROPIC_API_KEY environment variable.") + return + + client = Anthropic(api_key=api_key) + + pdf_paths = [] + for root, dirs, files in os.walk(args.input_dir): + for file in files: + if file.lower().endswith(".pdf"): + pdf_paths.append(os.path.join(root, file)) + print(f"Found {len(pdf_paths)} PDF files in {args.input_dir}") + + random.shuffle(pdf_paths) + pdf_paths = pdf_paths[: args.max_tests] + + synthetic_json_path = os.path.join(args.output_dir, "synthetic.jsonl") + open(synthetic_json_path, "w").close() + + test_counter = 0 + test_types = {"order": 0} + results = [] + + import threading + + file_lock = threading.Lock() + + with ThreadPoolExecutor(max_workers=args.parallel) as executor: + futures = {executor.submit(process_pdf, (pdf_path, i), args, client): pdf_path for i, pdf_path in enumerate(pdf_paths)} + for future in tqdm(concurrent.futures.as_completed(futures), total=len(futures), desc="Processing PDFs"): + pdf_path = futures[future] + try: + result = future.result() + if result and result.get("tests"): + results.append(result) + with file_lock: + with open(synthetic_json_path, "a") as f: + for test in result["tests"]: + f.write(json.dumps(test) + "\n") + test_counter += len(result["tests"]) + for test in result["tests"]: + if test.get("type") == "order": + test_types["order"] += 1 + print(f"Added {len(result['tests'])} tests from {result['pdf_id']}, total: {test_counter}") + except Exception as e: + print(f"Error processing {pdf_path}: {e}") + + print(f"Generated {len(results)} HTML templates") + if not args.skip_playwright: + playwright_success = sum(1 for r in results if r and r.get("playwright_pdf_path")) + print(f"Playwright PDF rendering: {playwright_success}/{len(results)} successful") + print(f"Saved {test_counter} tests to {synthetic_json_path}") + print(f"Generated a total of {test_counter} tests across {len(results)} templates") + print("Test type distribution:") + for test_type, count in test_types.items(): + print(f" - {test_type}: {count} tests") + + +if __name__ == "__main__": + main() diff --git a/olmocr/bench/miners/mine_multilingual_gpt.py b/olmocr/bench/miners/mine_multilingual_gpt.py new file mode 100644 index 0000000..1857f70 --- /dev/null +++ b/olmocr/bench/miners/mine_multilingual_gpt.py @@ -0,0 +1,295 @@ +#!/usr/bin/env python3 +""" +mine_multilingual_gpt.py - Identify PDF documents with tables and copy them. + +This script: +1. Takes a file containing S3 paths to PDF documents as input +2. For each PDF, renders a random page and uses GPT-4o to check the primary language +3. Identifies PDFs where the page is not in english +4. Copies those PDF files to a new output folder + +Usage: + python mine_multilingual_gpt.py --input_list path/to/s3_paths.txt --output_dir path/to/output --api_key your_openai_api_key +""" + +import argparse +import os +import random +from concurrent.futures import ThreadPoolExecutor, as_completed +from typing import Optional + +import boto3 +import pypdf +from lingua import Language +from openai import OpenAI +from pydantic import BaseModel +from tqdm import tqdm + +from olmocr.data.renderpdf import render_pdf_to_base64png +from olmocr.filter import PdfFilter + +TARGET_IMAGE_DIM = 1024 + + +class LanguageDetectionResponse(BaseModel): + """Structured output for table detection.""" + + two_letter_language_code: str + + +def download_pdf_from_s3(s3_path: str, local_path: str) -> bool: + """ + Download a PDF file from S3. + + Args: + s3_path: The S3 path (s3://bucket/path/to/file.pdf) + local_path: The local path to save the file + + Returns: + bool: True if download was successful, False otherwise + """ + try: + # Parse S3 path + parts = s3_path.replace("s3://", "").split("/", 1) + bucket = parts[0] + key = parts[1] + + # Create S3 client + s3 = boto3.client("s3") + + # Create directory if it doesn't exist + os.makedirs(os.path.dirname(local_path), exist_ok=True) + + # Download file + s3.download_file(bucket, key, local_path) + return True + except Exception as e: + print(f"Error downloading {s3_path}: {str(e)}") + return False + + +def check_for_table(pdf_path: str, page_num: int, api_key: str) -> Optional[bool]: + """ + Use GPT-4o to check if a page contains a table. + + Args: + pdf_path: Path to the PDF file + page_num: The page number to analyze (0-indexed) + api_key: OpenAI API key + + Returns: + Optional[bool]: True if page contains a table, False otherwise, None if detection fails + """ + # Initialize OpenAI client + client = OpenAI(api_key=api_key) + + try: + # Render the PDF page as an image (render_pdf_to_base64png is 1-indexed) + image_base64 = render_pdf_to_base64png(pdf_path, page_num=page_num + 1, target_longest_image_dim=TARGET_IMAGE_DIM) + + # Simple prompt asking about tables + prompt = "Respond with the primary language that this document is written in. Reply in JSON, and use a two letter ISO 639 language code, such as 'en', 'pt', 'fr', etc." + + response = client.beta.chat.completions.parse( + model="gpt-4o-2024-08-06", + messages=[ + { + "role": "user", + "content": [{"type": "text", "text": prompt}, {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{image_base64}"}}], + } + ], + temperature=0.1, + max_tokens=100, + response_format=LanguageDetectionResponse, + ) + + if not response.choices or len(response.choices) == 0: + print(f"No response generated for {pdf_path} page {page_num}") + return None + + # Parse the structured response + parsed_response = response.choices[0].message.parsed + + if parsed_response is None: + print(f"Failed to parse response for {pdf_path} page {page_num}") + return None + + is_non_english = parsed_response.two_letter_language_code.lower() != "en" + + if is_non_english: + print(f"Found {parsed_response.two_letter_language_code} language in {pdf_path} page {page_num + 1}") + + return is_non_english + + except Exception as e: + print(f"Error checking {pdf_path} page {page_num}: {str(e)}") + return None + + +def process_pdf(s3_path: str, temp_dir: str, output_dir: str, api_key: str) -> bool: + """ + Process a single PDF from S3. + + Args: + s3_path: S3 path to the PDF + temp_dir: Directory for temporary files + output_dir: Directory for output files + api_key: OpenAI API key + + Returns: + bool: True if the PDF has a table and was copied, False otherwise + """ + # Extract filename from S3 path + pdf_filename = os.path.basename(s3_path) + local_pdf_path = os.path.join(temp_dir, pdf_filename) + + # Download PDF from S3 + if not download_pdf_from_s3(s3_path, local_pdf_path): + return False + + pdf_filter = PdfFilter(languages_to_keep=Language.all()) + + if pdf_filter.filter_out_pdf(local_pdf_path): + print(f"Filtering out {pdf_filename}") + return False + + try: + # Read the PDF to get the number of pages + reader = pypdf.PdfReader(local_pdf_path) + num_pages = len(reader.pages) + + if num_pages == 0: + print(f"PDF {pdf_filename} has no pages") + return False + + # Select a random page to check + page_num = random.randint(0, num_pages - 1) + page_num = random.choice([page_num, 0]) # Bias 50% of the time to do the first page + + # Check if the page contains a table + has_table = check_for_table(local_pdf_path, page_num, api_key) + + if has_table: + # Extract just the page with the table and save it as a new PDF + os.makedirs(output_dir, exist_ok=True) + + # Create output filename with basename_pgnum.pdf format + pdf_basename = os.path.splitext(pdf_filename)[0] + output_pdf_path = os.path.join(output_dir, f"{pdf_basename}_pg{page_num+1}.pdf") + + # Extract the single page + writer = pypdf.PdfWriter() + writer.add_page(reader.pages[page_num]) + + # Write the output PDF + with open(output_pdf_path, "wb") as output_file: + writer.write(output_file) + + print(f"Extracted page {page_num+1} with table from {pdf_filename} to {os.path.basename(output_pdf_path)}") + return True + + return False + + except Exception as e: + print(f"Error processing {pdf_filename}: {str(e)}") + return False + finally: + if os.path.exists(local_pdf_path): + os.remove(local_pdf_path) + + +def main(): + parser = argparse.ArgumentParser(description="Identify and copy PDFs with tables") + parser.add_argument("--input_list", required=True, help="Path to a file containing S3 paths to PDFs") + parser.add_argument("--output_dir", required=True, help="Directory to copy PDFs with tables") + parser.add_argument("--api_key", help="OpenAI API key (if not provided, will use OPENAI_API_KEY environment variable)") + parser.add_argument("--temp_dir", default="/tmp/mine_tables", help="Directory for temporary files") + parser.add_argument("--max_pdfs", type=int, default=100, help="Maximum number of PDFs with tables to find") + parser.add_argument("--parallel", type=int, default=1, help="Number of parallel workers (default: 1 for sequential)") + parser.add_argument("--reservoir_multiplier", type=int, default=100, help="Multiplier for reservoir sampling (default: 100x max_pdfs)") + args = parser.parse_args() + + # Get API key + api_key = args.api_key or os.environ.get("OPENAI_API_KEY") + if not api_key: + print("Error: OpenAI API key not provided. Use --api_key or set OPENAI_API_KEY environment variable.") + return + + os.makedirs(args.temp_dir, exist_ok=True) + os.makedirs(args.output_dir, exist_ok=True) + + # Reservoir sampling to get random subset of PDFs + reservoir_size = args.max_pdfs * args.reservoir_multiplier + pdf_paths = [] + n = 0 # Total number of items seen + + print(f"Using reservoir sampling with size {reservoir_size}") + + with open(args.input_list, "r") as f: + for line in f: + n += 1 + path = line.strip() + if not path: + continue + + if len(pdf_paths) < reservoir_size: + pdf_paths.append(path) + else: + # Randomly decide whether to include this item + s = random.randint(1, n) + if s <= reservoir_size: + pdf_paths[s - 1] = path + + # Shuffle the reservoir + random.shuffle(pdf_paths) + + print(f"Sampled {len(pdf_paths)} PDF paths from {n} total paths") + + table_pdfs_found = 0 + + if args.parallel > 1: + # Parallel processing + print(f"Processing PDFs with {args.parallel} parallel workers") + + with ThreadPoolExecutor(max_workers=args.parallel) as executor: + futures = [] + + # Submit all tasks + for s3_path in pdf_paths: + if table_pdfs_found >= args.max_pdfs: + break + future = executor.submit(process_pdf, s3_path, args.temp_dir, args.output_dir, api_key) + futures.append(future) + + # Process results as they complete + with tqdm(total=min(len(pdf_paths), args.max_pdfs), desc="Processing PDFs") as pbar: + for future in as_completed(futures): + try: + result = future.result() + if result: + table_pdfs_found += 1 + pbar.update(1) + + if table_pdfs_found >= args.max_pdfs: + print(f"Reached maximum number of PDFs with tables ({args.max_pdfs}), stopping") + # Cancel remaining futures + for f in futures: + f.cancel() + break + except Exception as e: + print(f"Error in parallel processing: {str(e)}") + else: + # Sequential processing + for s3_path in tqdm(pdf_paths, desc="Processing PDFs"): + if process_pdf(s3_path, args.temp_dir, args.output_dir, api_key): + table_pdfs_found += 1 + + if table_pdfs_found >= args.max_pdfs: + print(f"Reached maximum number of PDFs with tables ({args.max_pdfs}), stopping") + break + + print(f"Found and copied {table_pdfs_found} PDFs to {args.output_dir}") + + +if __name__ == "__main__": + main() diff --git a/olmocr/bench/miners/mine_old_scan_pdf.py b/olmocr/bench/miners/mine_old_scan_pdf.py new file mode 100644 index 0000000..bff7065 --- /dev/null +++ b/olmocr/bench/miners/mine_old_scan_pdf.py @@ -0,0 +1,120 @@ +import json +import os +import time + +import requests +from bs4 import BeautifulSoup + + +def process_jsonl(jsonl_file, output_dir="downloaded_images"): + """ + Process each line in the JSONL file and download the corresponding images. + + Args: + jsonl_file (str): Path to the JSONL file + output_dir (str): Directory to save downloaded images + """ + + os.makedirs(output_dir, exist_ok=True) + with open(jsonl_file, "r") as f: + for line_num, line in enumerate(f, 1): + try: + data = json.loads(line.strip()) + pdf_path = data.get("pdf", "") + pdf_filename = os.path.basename(pdf_path) + url = data.get("url", "") + if not url: + print(f"Line {line_num}: Missing URL, skipping...") + continue + + print(f"Processing line {line_num}: {pdf_filename} - {url}") + image_path = download_image(url, pdf_filename, output_dir) + + if image_path: + print(f"Successfully downloaded: {image_path}") + else: + print(f"Failed to download image for {pdf_filename}") + time.sleep(1) + + except json.JSONDecodeError: + print(f"Line {line_num}: Invalid JSON, skipping...") + except Exception as e: + print(f"Line {line_num}: Error processing - {str(e)}") + + +def download_image(url, output_filename, output_dir): + """ + Download the highest resolution JPEG (before JPEG2000) from the Library of Congress URL. + + Args: + url (str): The initial URL from the JSONL file + output_filename (str): Filename for the downloaded image + output_dir (str): Directory to save the image + + Returns: + str or None: Path to the downloaded image if successful, None otherwise + """ + try: + response = requests.get(url) + response.raise_for_status() + soup = BeautifulSoup(response.text, "html.parser") + loc_link = soup.find("a", class_="btn btn-outline-primary text-nowrap", title="View the original source for this item in a new tab") + if not loc_link: + print(f"Could not find 'View on www.loc.gov' link on {url}") + return None + + loc_url = loc_link["href"] + print(f"Found LOC URL: {loc_url}") + response = requests.get(loc_url) + response.raise_for_status() + soup = BeautifulSoup(response.text, "html.parser") + download_select = soup.find("select", id="download") + if not download_select: + print(f"Could not find download options on {loc_url}") + return None + + jpeg_options = [option for option in download_select.find_all("option") if "JPEG" in option.text and "JPEG2000" not in option.text] + + if not jpeg_options: + print(f"No JPEG options found on {loc_url}") + return None + + highest_jpeg = jpeg_options[-1] + image_url = highest_jpeg["value"] + + print(f"Found highest resolution JPEG: {image_url}") + response = requests.get(image_url, stream=True) + response.raise_for_status() + + if not output_filename.lower().endswith((".jpg", ".jpeg")): + output_filename = f"{os.path.splitext(output_filename)[0]}.jpg" + + output_path = os.path.join(output_dir, output_filename) + + with open(output_path, "wb") as f: + for chunk in response.iter_content(chunk_size=8192): + f.write(chunk) + + return output_path + + except Exception as e: + print(f"Error downloading image: {str(e)}") + return None + + +def main(): + import argparse + + parser = argparse.ArgumentParser(description="Download images from Library of Congress based on JSONL file") + parser.add_argument("jsonl_file", help="Path to the JSONL file containing URLs") + parser.add_argument("--output", "-o", default="downloaded_images", help="Directory to save downloaded images (default: downloaded_images)") + + args = parser.parse_args() + + print(f"Processing JSONL file: {args.jsonl_file}") + process_jsonl(args.jsonl_file, args.output) + print("Processing complete!") + + +if __name__ == "__main__": + main() diff --git a/olmocr/bench/miners/mine_old_scans.py b/olmocr/bench/miners/mine_old_scans.py new file mode 100644 index 0000000..b0d99fa --- /dev/null +++ b/olmocr/bench/miners/mine_old_scans.py @@ -0,0 +1,122 @@ +import json +import random +import re + + +def extract_random_segment(text, min_words=7, max_words=15): + """Extract a random segment of 7-15 words from the text.""" + words = text.split() + if len(words) <= max_words: + return text # Return full text if it's shorter than max_words + + max_start = len(words) - min_words + start = random.randint(0, max_start) + remaining_words = len(words) - start + segment_length = random.randint(min_words, min(max_words, remaining_words)) + segment = words[start : start + segment_length] + return " ".join(segment) + + +def process_jsonl_file_present(input_file, output_file): + """Process a JSONL file and create multiple random cases for each PDF.""" + with open(input_file, "r") as infile, open(output_file, "w") as outfile: + for line in infile: + if line.strip(): # Skip empty lines + data = json.loads(line) + image = data["image"] + original_text = data["text"] + num_cases = random.randint(1, 3) + + for _ in range(num_cases): + processed_num = random.randint(5, 10) + processed_id = f"{image}_processed{processed_num:02d}" + max_diffs = random.randint(1, 2) + text_segment = extract_random_segment(original_text) + + new_case = { + "pdf": f"{image}.pdf", + "page": 1, + "id": processed_id, + "type": "present", + "max_diffs": max_diffs, + "text": text_segment, + "case_sensitive": True, + "first_n": None, + "last_n": None, + } + outfile.write(json.dumps(new_case) + "\n") + + +def extract_ordered_segments(text, min_words=7, max_words=15): + """Extract two ordered segments from the text.""" + sentences = re.split(r"(?<=[.!?])\s+", text) + + if len(sentences) < 2: + return None, None + valid_indices = list(range(len(sentences))) + if len(valid_indices) <= 2: + before_idx, after_idx = 0, 1 + else: + before_idx = random.randint(0, len(valid_indices) - 2) + after_idx = random.randint(before_idx + 1, len(valid_indices) - 1) + before_sentence = sentences[before_idx] + after_sentence = sentences[after_idx] + + before_words = before_sentence.split() + after_words = after_sentence.split() + + if len(before_words) > max_words: + start = random.randint(0, len(before_words) - min_words) + length = random.randint(min_words, min(max_words, len(before_words) - start)) + before_segment = " ".join(before_words[start : start + length]) + else: + before_segment = before_sentence + + if len(after_words) > max_words: + start = random.randint(0, len(after_words) - min_words) + length = random.randint(min_words, min(max_words, len(after_words) - start)) + after_segment = " ".join(after_words[start : start + length]) + else: + after_segment = after_sentence + + return before_segment, after_segment + + +def process_jsonl_file_order(input_file, output_file): + """Process a JSONL file and create order-type cases.""" + with open(input_file, "r") as infile, open(output_file, "w") as outfile: + for line in infile: + if line.strip(): # Skip empty lines + data = json.loads(line) + image = data["image"] + original_text = data["text"] + num_cases = random.randint(1, 3) + + for _ in range(num_cases): + before_text, after_text = extract_ordered_segments(original_text) + if not before_text or not after_text: + continue + processed_num = random.randint(11, 16) + processed_id = f"{image}_processed{processed_num:02d}" + max_diffs = random.randint(1, 3) + + new_case = { + "pdf": f"{image}.pdf", + "page": 1, + "id": processed_id, + "type": "order", + "before": before_text, + "after": after_text, + "max_diffs": max_diffs, + "checked": "verified", + "url": f"https://example.com/document/{image}", + } + + outfile.write(json.dumps(new_case) + "\n") + + +if __name__ == "__main__": + input_file = "olmoce/bench/sample_data/old_scans.jsonl" + output_file = "order_cases.jsonl" + process_jsonl_file_present(input_file, output_file) + process_jsonl_file_order(input_file, output_file) diff --git a/olmocr/bench/miners/mine_old_scans_math.py b/olmocr/bench/miners/mine_old_scans_math.py new file mode 100644 index 0000000..a18e8c3 --- /dev/null +++ b/olmocr/bench/miners/mine_old_scans_math.py @@ -0,0 +1,277 @@ +#!/usr/bin/env python3 +""" +mine_old_scans_math.py - Extract mathematical equations from PDF documents. + +This script: +1. Takes a folder containing PDF documents as input +2. For each PDF, extracts a random page and renders it to an image +3. Uses Gemini to identify mathematical equations in the rendered image +4. Creates a test file asserting that the equation text should be present +5. Extracts the page from the PDF and saves it to an output folder + +Usage: + python mine_old_scans_math.py --input_dir path/to/pdf_folder --output_dir path/to/output --api_key your_gemini_api_key +""" + +import argparse +import base64 +import json +import os +import random +from typing import List, Optional + +import pypdf +from google import genai +from google.genai import types +from tqdm import tqdm + +from olmocr.bench.tests import TextPresenceTest, save_tests +from olmocr.data.renderpdf import render_pdf_to_base64png +from olmocr.filter import PdfFilter + + +def extract_page_from_pdf(input_path: str, output_path: str, page_num: int) -> bool: + """ + Extract a specific page from a PDF and save it as a new PDF. + + Args: + input_path: Path to the input PDF + output_path: Path to save the extracted page + page_num: The page number to extract (0-indexed) + + Returns: + bool: True if extraction was successful, False otherwise + """ + try: + # Ensure output directory exists + os.makedirs(os.path.dirname(output_path), exist_ok=True) + + # Read the input PDF + reader = pypdf.PdfReader(input_path) + + # Check if page number is valid + if page_num >= len(reader.pages): + print(f"Page number {page_num} out of range for {input_path} with {len(reader.pages)} pages") + return False + + # Create a new PDF with just the selected page + writer = pypdf.PdfWriter() + writer.add_page(reader.pages[page_num]) + + # Write the output PDF + with open(output_path, "wb") as output_file: + writer.write(output_file) + + return True + except Exception as e: + print(f"Error extracting page {page_num} from {input_path}: {str(e)}") + raise + + +def detect_equations(pdf_path: str, page_num: int, api_key: str) -> Optional[List[str]]: + """ + Use Gemini to detect mathematical equations in a rendered PDF page. + + Args: + pdf_path: Path to the PDF file + page_num: The page number to analyze (0-indexed) + api_key: Gemini API key + + Returns: + Optional[List[str]]: List of detected equations, or None if detection failed + """ + client = genai.Client( + api_key=os.environ.get("GEMINI_API_KEY"), + ) + model = "gemini-2.0-flash" + + # Render the PDF page as an image + try: + image_base64 = render_pdf_to_base64png(pdf_path, page_num=page_num + 1, target_longest_image_dim=2048) # render_pdf_to_base64png is 1-indexed + except Exception as e: + print(f"Error rendering PDF page: {str(e)}") + return None + + image_part = types.Part(inline_data=types.Blob(mime_type="image/png", data=base64.b64decode(image_base64))) + + contents = [ + types.Content( + role="user", + parts=[ + image_part, + types.Part.from_text( + text="""Please extract the mathematical equations from the document without omission. Always output the mathematical equations as Latex escaped with $$. Do not hallucinate""" + ), + ], + ), + ] + + generate_content_config = types.GenerateContentConfig( + temperature=1, + top_p=0.95, + top_k=40, + max_output_tokens=8192, + response_mime_type="application/json", + response_schema=genai.types.Schema( + type=genai.types.Type.OBJECT, + properties={ + "equations": genai.types.Schema( + type=genai.types.Type.ARRAY, + items=genai.types.Schema( + type=genai.types.Type.STRING, + ), + ), + }, + ), + ) + + response = client.models.generate_content(model=model, contents=contents, config=generate_content_config) + + assert len(response.candidates) > 0, "No candidates found" + assert response.candidates[0].finish_reason == types.FinishReason.STOP, "Finish reason was not STOP, likely a processing error or repetition failure" + + data = json.loads(response.candidates[0].content.parts[0].text) + + return data.get("equations", []) + + +def process_pdf(pdf_path: str, output_dir: str, api_key: str, tests: List[TextPresenceTest], max_pages_per_pdf: int = 10) -> None: + """ + Process a single PDF, extracting equations from multiple pages. + + Args: + pdf_path: Path to the PDF + output_dir: Directory for output files + api_key: Gemini API key + tests: List to append tests to + max_pages_per_pdf: Maximum number of pages to process per PDF + """ + # Extract filename from path + pdf_filename = os.path.basename(pdf_path) + + pdf_filter = PdfFilter() + + if pdf_filter.filter_out_pdf(pdf_path): + print("Filtering out", pdf_filename) + return + + try: + # Read the PDF to get the number of pages + reader = pypdf.PdfReader(pdf_path) + num_pages = len(reader.pages) + + if num_pages == 0: + print(f"PDF {pdf_filename} has no pages") + return + + # Get all pages and shuffle them to select a random subset + all_pages = list(range(num_pages)) + random.shuffle(all_pages) + + # Take only the specified maximum number of pages + pages_to_process = all_pages[: min(max_pages_per_pdf, num_pages)] + + processed_pages = 0 + + for page_num in pages_to_process: + # Detect equations + equations = detect_equations(pdf_path, page_num, api_key) + + # Only keep equations that are non-empty + equations = [eq for eq in equations if len(eq.strip()) > 3] + + if not equations: + print(f"No equations detected in {pdf_filename} page {page_num+1}") + continue + + # Extract the page and save to output dir + pdf_basename = os.path.splitext(pdf_filename)[0] + output_pdf_path = os.path.join(output_dir, "pdfs", f"{pdf_basename}_pg{page_num+1}.pdf") + + extract_page_from_pdf(pdf_path, output_pdf_path, page_num) + + # Create tests for each equation + for i, equation in enumerate(equations): + test_id = f"{pdf_basename}_pg{page_num+1}_equation_{i:02d}" + test = TextPresenceTest( + id=test_id, + pdf=f"{pdf_basename}_pg{page_num+1}.pdf", + page=1, # The extracted PDF has only one page + type="present", + text=equation, + max_diffs=0, + ) + tests.append(test) + + print(f"Processed {pdf_filename} page {page_num+1}, found {len(equations)} equations") + processed_pages += 1 + + print(f"Completed processing {processed_pages} pages from {pdf_filename}") + + except Exception as e: + print(f"Error processing {pdf_filename}: {str(e)}") + + +def get_pdf_files_from_directory(directory: str) -> List[str]: + """ + Get a list of all PDF files in a directory. + + Args: + directory: Path to the directory containing PDFs + + Returns: + List[str]: List of full paths to PDF files + """ + pdf_files = [] + + for filename in os.listdir(directory): + if filename.lower().endswith(".pdf"): + full_path = os.path.join(directory, filename) + if os.path.isfile(full_path): + pdf_files.append(full_path) + + return pdf_files + + +def main(): + parser = argparse.ArgumentParser(description="Extract mathematical equations from PDF documents") + parser.add_argument("--input_dir", required=True, help="Directory containing PDF files") + parser.add_argument("--output_dir", required=True, help="Directory to store extracted pages and tests") + parser.add_argument("--api_key", help="Gemini API key (if not provided, will use GEMINI_API_KEY environment variable)") + args = parser.parse_args() + + # Get API key + api_key = args.api_key or os.environ.get("GEMINI_API_KEY") + if not api_key: + print("Error: Gemini API key not provided. Use --api_key or set GEMINI_API_KEY environment variable.") + return + + # Create directories + os.makedirs(os.path.join(args.output_dir, "pdfs"), exist_ok=True) + + # Get PDF files from input directory + pdf_files = get_pdf_files_from_directory(args.input_dir) + + if not pdf_files: + print(f"No PDF files found in {args.input_dir}") + return + + print(f"Found {len(pdf_files)} PDF files in input directory") + + # Process each PDF + tests = [] + for pdf_path in tqdm(pdf_files, desc="Processing PDFs"): + process_pdf(pdf_path, args.output_dir, api_key, tests) + + # Save tests after each PDF to avoid losing data in case of crashes + if tests: + save_tests(tests, os.path.join(args.output_dir, "equation_tests.jsonl")) + + # if len(tests) > 100: + # break + + print(f"Saved {len(tests)} tests to {os.path.join(args.output_dir, 'equation_tests.jsonl')}") + + +if __name__ == "__main__": + main() diff --git a/olmocr/bench/miners/mine_reading_order.py b/olmocr/bench/miners/mine_reading_order.py new file mode 100644 index 0000000..e1608f3 --- /dev/null +++ b/olmocr/bench/miners/mine_reading_order.py @@ -0,0 +1,530 @@ +#!/usr/bin/env python3 +""" +analyze_documents.py - Analyze document layout and extract content from PDF documents. + +This script: +1. Takes a file containing S3 paths to PDF documents as input +2. For each PDF, extracts a random page and renders it to an image +3. Uses Gemini to analyze document layout features (columns, articles, text inserts, etc.) +4. If specific layout features are detected, proceeds with full document content extraction +5. Extracts the page from the PDF and saves it to an output folder along with analysis results + +Usage: + python analyze_documents.py --input_list path/to/s3_paths.txt --output_dir path/to/output --api_key your_gemini_api_key [--parallel 4] +""" + +import argparse +import base64 +import concurrent.futures +import json +import os +import random +import threading +from typing import Any, Dict, List, Optional, Tuple + +import boto3 +import pypdf +from google import genai +from google.genai import types +from tqdm import tqdm + +from olmocr.data.renderpdf import render_pdf_to_base64png +from olmocr.filter import PdfFilter + +# Create a thread-safe lock for writing to output files +file_lock = threading.Lock() + + +def download_pdf_from_s3(s3_path: str, local_path: str) -> bool: + """ + Download a PDF file from S3. + + Args: + s3_path: The S3 path (s3://bucket/path/to/file.pdf) + local_path: The local path to save the file + + Returns: + bool: True if download was successful, False otherwise + """ + try: + # Parse S3 path + parts = s3_path.replace("s3://", "").split("/", 1) + bucket = parts[0] + key = parts[1] + + # Create S3 client + s3 = boto3.client("s3") + + # Create directory if it doesn't exist + os.makedirs(os.path.dirname(local_path), exist_ok=True) + + # Download file + s3.download_file(bucket, key, local_path) + return True + except Exception as e: + print(f"Error downloading {s3_path}: {str(e)}") + return False + + +def extract_page_from_pdf(input_path: str, output_path: str, page_num: int) -> bool: + """ + Extract a specific page from a PDF and save it as a new PDF. + + Args: + input_path: Path to the input PDF + output_path: Path to save the extracted page + page_num: The page number to extract (0-indexed) + + Returns: + bool: True if extraction was successful, False otherwise + """ + try: + # Ensure output directory exists + os.makedirs(os.path.dirname(output_path), exist_ok=True) + + # Read the input PDF + reader = pypdf.PdfReader(input_path) + + # Check if page number is valid + if page_num >= len(reader.pages): + print(f"Page number {page_num} out of range for {input_path} with {len(reader.pages)} pages") + return False + + # Create a new PDF with just the selected page + writer = pypdf.PdfWriter() + writer.add_page(reader.pages[page_num]) + + # Write the output PDF + with open(output_path, "wb") as output_file: + writer.write(output_file) + + return True + except Exception as e: + print(f"Error extracting page {page_num} from {input_path}: {str(e)}") + raise + + +def analyze_document_layout(pdf_path: str, page_num: int, api_key: str) -> Optional[Tuple[Dict[str, Any], str]]: + """ + Use Gemini to analyze document layout features in a rendered PDF page. + + Args: + pdf_path: Path to the PDF file + page_num: The page number to analyze (0-indexed) + api_key: Gemini API key + + Returns: + Optional[Tuple[Dict[str, Any], str]]: + A tuple with the layout analysis results as a dictionary and the base64 string of the rendered page image. + Returns None if analysis fails. + """ + # Initialize Gemini client + client = genai.Client( + api_key=api_key, + ) + model = "gemini-2.0-flash" + + # Render the PDF page as an image (render_pdf_to_base64png is 1-indexed) + try: + image_base64 = render_pdf_to_base64png(pdf_path, page_num=page_num + 1, target_longest_image_dim=2048) + except Exception as e: + print(f"Error rendering PDF page: {str(e)}") + return None + + image_part = types.Part(inline_data=types.Blob(mime_type="image/png", data=base64.b64decode(image_base64))) + + # Prepare prompt for Gemini to analyze document layout + contents = [ + types.Content( + role="user", + parts=[ + image_part, + types.Part.from_text( + text=( + "Please answer the following questions about the document in JSON format:\n" + "-How many columns are used in the main text document layout?\n" + "-How many unique articles are captured in main text on this page?\n" + "-Are there any text inserts in the main article content?\n" + "-Do any of the main content articles start with a dropcap?\n" + "-Are there any boxed out regions of text that need to be read separately from the main article content?\n" + "-Are there any regions of text with a different orientation/rotation?" + ) + ), + ], + ), + ] + + generate_content_config = types.GenerateContentConfig( + temperature=0.2, + top_p=0.95, + top_k=40, + max_output_tokens=2048, + response_mime_type="application/json", + response_schema=types.Schema( + type=types.Type.OBJECT, + required=[ + "num_columns", + "num_unique_articles", + "contains_text_inserts", + "contains_dropcaps", + "contains_boxed_regions", + "contains_text_different_orientation", + ], + properties={ + "num_columns": types.Schema( + type=types.Type.INTEGER, + ), + "num_unique_articles": types.Schema( + type=types.Type.INTEGER, + ), + "contains_text_inserts": types.Schema( + type=types.Type.BOOLEAN, + ), + "contains_dropcaps": types.Schema( + type=types.Type.BOOLEAN, + ), + "contains_boxed_regions": types.Schema( + type=types.Type.BOOLEAN, + ), + "contains_text_different_orientation": types.Schema( + type=types.Type.BOOLEAN, + ), + }, + ), + ) + + try: + # Call Gemini API + response = client.models.generate_content(model=model, contents=contents, config=generate_content_config) + + print(response) + + if not response.candidates or len(response.candidates) == 0: + print(f"No response generated for {pdf_path} page {page_num}") + return None + + if response.candidates[0].finish_reason != types.FinishReason.STOP: + print(f"Response generation incomplete for {pdf_path} page {page_num}") + return None + + # Parse the response + response_text = response.candidates[0].content.parts[0].text + + layout_analysis = json.loads(response_text) + + print(f"Layout analysis for {pdf_path} page {page_num}:") + print(json.dumps(layout_analysis, indent=2)) + + # Return both the layout analysis and the rendered image (base64 string) + return (layout_analysis, image_base64) + + except Exception as e: + print(f"Error analyzing document layout in {pdf_path} page {page_num}: {str(e)}") + return None + + +def extract_document_content(pdf_path: str, page_num: int, image_base64: str, api_key: str) -> Optional[str]: + """ + Use Gemini to extract full document content from a rendered PDF page. + + Args: + pdf_path: Path to the PDF file + page_num: The page number to analyze (0-indexed) + image_base64: The base64 string of the rendered page image + api_key: Gemini API key + + Returns: + Optional[str]: The extracted document content in markdown format, or None if extraction fails. + """ + # Initialize Gemini client + client = genai.Client( + api_key=api_key, + ) + model = "gemini-2.0-flash" + + image_part = types.Part(inline_data=types.Blob(mime_type="image/png", data=base64.b64decode(image_base64))) + + # Prepare prompt for Gemini to extract document content + contents = [ + types.Content( + role="user", + parts=[ + image_part, + types.Part.from_text( + text=( + "Analyze the document attached and output it in markdown format. " + "Output equations as Latex escaped with $$. " + "Output tables in HTML format that preserves the structure and content exactly, do not use
tags. " + "Instead of the markdown table format, be sure to output tables in HTML, even though the rest of the document is styled in markdown. " + "Output figures with just a simple markdown image placeholder." + ) + ), + ], + ), + ] + + generate_content_config = types.GenerateContentConfig(temperature=0.2, top_p=0.95, top_k=40, max_output_tokens=8192) + + try: + # Call Gemini API + response = client.models.generate_content(model=model, contents=contents, config=generate_content_config) + + if not response.candidates or len(response.candidates) == 0: + print(f"No response generated for content extraction in {pdf_path} page {page_num}") + return None + + if response.candidates[0].finish_reason != types.FinishReason.STOP: + print(f"Content extraction incomplete for {pdf_path} page {page_num}") + return None + + # Get the extracted content + content = response.candidates[0].content.parts[0].text + return content + + except Exception as e: + print(f"Error extracting document content from {pdf_path} page {page_num}: {str(e)}") + return None + + +def should_extract_full_content(layout_analysis: Dict[str, Any]) -> bool: + """ + Determine if full content extraction is needed based on layout analysis results. + + Args: + layout_analysis: Dictionary containing layout analysis results + + Returns: + bool: True if any of the special layout features are detected, False otherwise + """ + # Check for special layout features that warrant full content extraction + features_to_check = ["text_inserts", "dropcaps", "boxed_regions", "rotated_text"] + + # Also check if there are multiple columns or articles + try: + columns = layout_analysis.get("columns", 0) + if isinstance(columns, str): + columns = int(columns) if columns.isdigit() else 0 + + articles = layout_analysis.get("articles", 0) + if isinstance(articles, str): + articles = int(articles) if articles.isdigit() else 0 + + if columns > 1 or articles > 1: + return True + except (ValueError, TypeError): + # If we can't parse the values, assume we need to extract + pass + + # Check for any True values in the features + for feature in features_to_check: + value = layout_analysis.get(feature, False) + if isinstance(value, str): + if value.lower() in ["yes", "true", "1"]: + return True + elif value: + return True + + return False + + +def process_pdf(s3_path: str, temp_dir: str, output_dir: str, api_key: str) -> Dict: + """ + Process a single PDF from S3. + + Args: + s3_path: S3 path to the PDF + temp_dir: Directory for temporary files + output_dir: Directory for output files + api_key: Gemini API key + + Returns: + Dict: Results of processing the PDF + """ + # Create a thread-specific temp directory to avoid conflicts + thread_id = threading.get_ident() + thread_temp_dir = os.path.join(temp_dir, f"thread_{thread_id}") + os.makedirs(thread_temp_dir, exist_ok=True) + + # Extract filename from S3 path + pdf_filename = os.path.basename(s3_path) + local_pdf_path = os.path.join(thread_temp_dir, pdf_filename) + + # Download PDF from S3 + if not download_pdf_from_s3(s3_path, local_pdf_path): + return {"error": f"Failed to download {s3_path}"} + + pdf_filter = PdfFilter() + + if pdf_filter.filter_out_pdf(local_pdf_path): + print(f"Filtering out {pdf_filename}") + if os.path.exists(local_pdf_path): + os.remove(local_pdf_path) + return {"error": f"PDF {pdf_filename} filtered out"} + + try: + # Read the PDF to get the number of pages + reader = pypdf.PdfReader(local_pdf_path) + num_pages = len(reader.pages) + + if num_pages == 0: + print(f"PDF {pdf_filename} has no pages") + return {"error": f"PDF {pdf_filename} has no pages"} + + all_pages = list(range(len(reader.pages))) + random.shuffle(all_pages) + + results = {"filename": pdf_filename, "s3_path": s3_path} + + for page_num in all_pages: + # Analyze document layout + layout_result = analyze_document_layout(local_pdf_path, page_num, api_key) + if not layout_result: + print(f"Failed to analyze layout in {pdf_filename} page {page_num+1}") + continue + + layout_analysis, image_base64 = layout_result + results["layout_analysis"] = layout_analysis + + # Determine if we need to extract full content + full_extraction_needed = should_extract_full_content(layout_analysis) + results["full_extraction_needed"] = full_extraction_needed + + # Extract full content if needed + if full_extraction_needed: + content = extract_document_content(local_pdf_path, page_num, image_base64, api_key) + results["content"] = content if content else "Content extraction failed" + + # Extract the page and save to output dir + pdf_basename = os.path.splitext(pdf_filename)[0] + output_pdf_path = os.path.join(output_dir, "pdfs", f"{pdf_basename}_pg{page_num+1}.pdf") + with file_lock: # Use lock when writing to shared output directory + extract_page_from_pdf(local_pdf_path, output_pdf_path, page_num) + + # Save analysis results + output_json_path = os.path.join(output_dir, "results", f"{pdf_basename}_pg{page_num+1}.json") + with file_lock: + os.makedirs(os.path.join(output_dir, "results"), exist_ok=True) + with open(output_json_path, "w") as f: + json.dump(results, f, indent=2) + + print(f"Processed {pdf_filename} page {page_num+1}, analysis saved to {output_json_path}") + + # Process only one page per PDF + break + + return results + + except Exception as e: + print(f"Error processing {pdf_filename}: {str(e)}") + return {"error": f"Error processing {pdf_filename}: {str(e)}"} + finally: + # Cleanup + if os.path.exists(local_pdf_path): + os.remove(local_pdf_path) + + +def process_pdfs_parallel(s3_paths: List[str], temp_dir: str, output_dir: str, api_key: str, max_docs: int, num_workers: int): + """ + Process PDFs in parallel using a thread pool. + + Args: + s3_paths: List of S3 paths to PDFs + temp_dir: Directory for temporary files + output_dir: Directory for output files + api_key: Gemini API key + max_docs: Maximum number of documents to process + num_workers: Number of parallel workers to use + """ + # Create output directory structure + os.makedirs(os.path.join(output_dir, "pdfs"), exist_ok=True) + os.makedirs(os.path.join(output_dir, "results"), exist_ok=True) + + # Create a summary file + summary_file = os.path.join(output_dir, "summary.jsonl") + + # Track processed documents + processed_count = 0 + + # Create a ThreadPoolExecutor + with concurrent.futures.ThreadPoolExecutor(max_workers=num_workers) as executor: + # Submit tasks and track futures + futures = {executor.submit(process_pdf, s3_path, temp_dir, output_dir, api_key): s3_path for s3_path in s3_paths} + + # Process results as they complete + for future in concurrent.futures.as_completed(futures): + s3_path = futures[future] + try: + # Get the result from this worker + result = future.result() + + # Add to summary file + with file_lock: + with open(summary_file, "a") as f: + f.write(json.dumps(result) + "\n") + + # Increment counter if no error + if "error" not in result: + processed_count += 1 + print(f"Successfully processed {os.path.basename(s3_path)}, total: {processed_count}") + + # Check if we've reached the maximum number of documents + if processed_count >= max_docs: + print(f"Reached maximum number of documents ({max_docs}), stopping") + # Cancel any pending futures + for f in futures: + if not f.done(): + f.cancel() + break + + except Exception as e: + print(f"Task for {os.path.basename(s3_path)} generated an exception: {e}") + + +def main(): + parser = argparse.ArgumentParser(description="Analyze document layout and extract content from PDF documents") + parser.add_argument("--input_list", required=True, help="Path to a file containing S3 paths to PDFs") + parser.add_argument("--output_dir", required=True, help="Directory to store extracted pages and analysis results") + parser.add_argument("--api_key", help="Gemini API key (if not provided, will use GEMINI_API_KEY environment variable)") + parser.add_argument("--temp_dir", default="/tmp/analyze_documents", help="Directory for temporary files") + parser.add_argument("--max_docs", type=int, default=100, help="Maximum number of documents to process") + parser.add_argument("--parallel", type=int, default=1, help="Number of parallel threads to use") + args = parser.parse_args() + + # Get API key + api_key = args.api_key or os.environ.get("GEMINI_API_KEY") + if not api_key: + print("Error: Gemini API key not provided. Use --api_key or set GEMINI_API_KEY environment variable.") + return + + os.makedirs(args.temp_dir, exist_ok=True) + os.makedirs(os.path.join(args.output_dir, "pdfs"), exist_ok=True) + os.makedirs(os.path.join(args.output_dir, "results"), exist_ok=True) + + # Reservoir sampling implementation + s3_paths = [] + with open(args.input_list, "r") as f: + for i, line in enumerate(tqdm(f)): + line = line.strip() + if not line: + continue + + if i < 100000: + s3_paths.append(line) + else: + # Randomly replace elements with decreasing probability + j = random.randint(0, i) + if j < 100000: + s3_paths[j] = line + + print(f"Found {len(s3_paths)} PDF paths in input list") + + # Determine number of workers to use + num_workers = max(1, min(args.parallel, len(s3_paths))) + print(f"Processing PDFs using {num_workers} parallel workers") + + # Process PDFs in parallel + process_pdfs_parallel(s3_paths, args.temp_dir, args.output_dir, api_key, args.max_docs, num_workers) + + +if __name__ == "__main__": + main() diff --git a/olmocr/bench/miners/mine_tables_gemini.py b/olmocr/bench/miners/mine_tables_gemini.py new file mode 100644 index 0000000..f829124 --- /dev/null +++ b/olmocr/bench/miners/mine_tables_gemini.py @@ -0,0 +1,491 @@ +#!/usr/bin/env python3 +""" +mine_tables.py - Extract tables from PDF documents and create table tests. + +This script: +1. Takes a file containing S3 paths to PDF documents as input +2. For each PDF, extracts a random page and renders it to an image +3. Uses Gemini to identify tables in the rendered image +4. Extracts table content and creates table relationship tests by making a second Gemini request + that now includes the page image alongside the prompt (e.g., "Given cell with {cell_value}, which cell is directly to the left of it?") +5. Extracts the page from the PDF and saves it to an output folder + +Usage: + python mine_tables.py --input_list path/to/s3_paths.txt --output_dir path/to/output --api_key your_gemini_api_key [--parallel 4] +""" + +import argparse +import base64 +import concurrent.futures +import os +import random +import threading +from typing import Dict, List, Optional, Tuple + +import boto3 +import numpy as np +import pypdf +from bs4 import BeautifulSoup +from google import genai +from google.genai import types +from tqdm import tqdm + +from olmocr.bench.tests import TableTest, save_tests +from olmocr.data.renderpdf import render_pdf_to_base64png +from olmocr.filter import PdfFilter + +# Create a thread-safe lock for writing to the output file +file_lock = threading.Lock() +tests_lock = threading.Lock() + + +def download_pdf_from_s3(s3_path: str, local_path: str) -> bool: + """ + Download a PDF file from S3. + + Args: + s3_path: The S3 path (s3://bucket/path/to/file.pdf) + local_path: The local path to save the file + + Returns: + bool: True if download was successful, False otherwise + """ + try: + # Parse S3 path + parts = s3_path.replace("s3://", "").split("/", 1) + bucket = parts[0] + key = parts[1] + + # Create S3 client + s3 = boto3.client("s3") + + # Create directory if it doesn't exist + os.makedirs(os.path.dirname(local_path), exist_ok=True) + + # Download file + s3.download_file(bucket, key, local_path) + return True + except Exception as e: + print(f"Error downloading {s3_path}: {str(e)}") + return False + + +def extract_page_from_pdf(input_path: str, output_path: str, page_num: int) -> bool: + """ + Extract a specific page from a PDF and save it as a new PDF. + + Args: + input_path: Path to the input PDF + output_path: Path to save the extracted page + page_num: The page number to extract (0-indexed) + + Returns: + bool: True if extraction was successful, False otherwise + """ + try: + # Ensure output directory exists + os.makedirs(os.path.dirname(output_path), exist_ok=True) + + # Read the input PDF + reader = pypdf.PdfReader(input_path) + + # Check if page number is valid + if page_num >= len(reader.pages): + print(f"Page number {page_num} out of range for {input_path} with {len(reader.pages)} pages") + return False + + # Create a new PDF with just the selected page + writer = pypdf.PdfWriter() + writer.add_page(reader.pages[page_num]) + + # Write the output PDF + with open(output_path, "wb") as output_file: + writer.write(output_file) + + return True + except Exception as e: + print(f"Error extracting page {page_num} from {input_path}: {str(e)}") + raise + + +def detect_tables(pdf_path: str, page_num: int, api_key: str) -> Optional[Tuple[List[np.ndarray], str]]: + """ + Use Gemini to detect tables in a rendered PDF page. + + Args: + pdf_path: Path to the PDF file + page_num: The page number to analyze (0-indexed) + api_key: Gemini API key + + Returns: + Optional[Tuple[List[np.ndarray], str]]: + A tuple with a list of detected tables (as numpy arrays) and the base64 string of the rendered page image. + Returns None if detection fails. + """ + # Initialize Gemini client + client = genai.Client( + api_key=api_key, + ) + model = "gemini-2.0-flash" + + # Render the PDF page as an image (render_pdf_to_base64png is 1-indexed) + try: + image_base64 = render_pdf_to_base64png(pdf_path, page_num=page_num + 1, target_longest_image_dim=2048) + except Exception as e: + print(f"Error rendering PDF page: {str(e)}") + return None + + image_part = types.Part(inline_data=types.Blob(mime_type="image/png", data=base64.b64decode(image_base64))) + + # Prepare prompt for Gemini to extract tables + contents = [ + types.Content( + role="user", + parts=[ + image_part, + types.Part.from_text( + text=( + "Analyze the document attached and output it in markdown format. " + "Output equations as Latex escaped with $$. " + "Output tables in HTML format that preserves the structure and content exactly, do not use
tags. " + "Instead of the markdown table format, be sure to output tables in HTML, even though the rest of the document is styled in markdown. " + "Output figures with just a simple markdown image placeholder." + ) + ), + ], + ), + ] + + generate_content_config = types.GenerateContentConfig(temperature=0.2, top_p=0.95, top_k=40, max_output_tokens=8192) + + try: + # Call Gemini API + response = client.models.generate_content(model=model, contents=contents, config=generate_content_config) + + if not response.candidates or len(response.candidates) == 0: + print(f"No response generated for {pdf_path} page {page_num}") + return None + + if response.candidates[0].finish_reason != types.FinishReason.STOP: + print(f"Response generation incomplete for {pdf_path} page {page_num}") + return None + + # Parse the response + response_text = response.candidates[0].content.parts[0].text + + print(response_text) + + # Parse tables from HTML + parsed_tables = [] + soup = BeautifulSoup(response_text, "html.parser") + tables = soup.find_all("table") + + for table in tables: + rows = table.find_all("tr") + table_data = [] + for row in rows: + cells = row.find_all(["th", "td"]) + row_data = [cell.get_text().strip() for cell in cells] + table_data.append(row_data) + # Ensure all rows have the same number of columns + if table_data: + max_cols = max(len(row) for row in table_data) + padded_data = [row + [""] * (max_cols - len(row)) for row in table_data] + table_array = np.array(padded_data) + parsed_tables.append(table_array) + + # Return both the parsed tables and the rendered image (base64 string) + return (parsed_tables, image_base64) if parsed_tables else None + + except Exception as e: + print(f"Error detecting tables in {pdf_path} page {page_num}: {str(e)}") + return None + + +def generate_table_tests(tables: List[np.ndarray], pdf_image: str, api_key: str, max_tests_per_table: int = 3) -> List[Dict]: + """ + Generate table tests from the detected tables by making a second Gemini request for each candidate cell. + + For each candidate cell in a table, the function selects one valid relationship (e.g., "left", "up", "top_heading", etc.) + and sends a prompt to Gemini including the page image. For example: + "Given a cell in a table with value 'XYZ', please answer: which cell is directly to the left of it? Provide only the cell's text." + + Args: + tables: List of tables as numpy arrays + pdf_image: Base64 string of the rendered page image + api_key: Gemini API key to use for generating relationship tests + max_tests_per_table: Maximum number of tests to generate per table + + Returns: + List of table test dictionaries + """ + tests = [] + # Initialize Gemini client for test queries + client = genai.Client(api_key=api_key) + model = "gemini-2.0-flash" + config = types.GenerateContentConfig(temperature=0.2, top_p=0.95, top_k=40, max_output_tokens=100) + + # Mapping for relationship prompts + prompt_map = { + "up": "which cell is directly above it?", + "down": "which cell is directly below it?", + "left": "which cell is directly to the left of it?", + "right": "which cell is directly to the right of it?", + "top_heading": "what is the top heading (the heading for the column at the top of the table) for this cell?", + "left_heading": "what is the left heading (the heading for this row on the left part of the table) for this cell?", + } + + # Create an image part from the rendered pdf image + image_part = types.Part(inline_data=types.Blob(mime_type="image/png", data=base64.b64decode(pdf_image))) + + for table in tables: + rows, cols = table.shape + if table.size == 0 or rows < 2 or cols < 2: + continue # Skip tables that are too small + + # Try up to 3x max_tests_per_table candidate cells + candidate_positions = [] + for row in range(rows): + for col in range(cols): + if not table[row, col].strip(): + continue + if row > 0: + candidate_positions.append((row, col, "up")) + if row < rows - 1: + candidate_positions.append((row, col, "down")) + if col > 0: + candidate_positions.append((row, col, "left")) + if col < cols - 1: + candidate_positions.append((row, col, "right")) + if row > 0: + candidate_positions.append((row, col, "top_heading")) + if col > 0: + candidate_positions.append((row, col, "left_heading")) + + random.shuffle(candidate_positions) + tests_for_this_table = 0 + + for row, col, relationship in candidate_positions: + if tests_for_this_table >= max_tests_per_table: + break + + cell_value = table[row, col].strip() + + prompt = ( + f"Given a cell in a table with value '{cell_value}', please answer: " + f"{prompt_map[relationship]} Provide only the cell's text or output 'null' if there is not a matching cell." + ) + + try: + contents = [types.Content(role="user", parts=[image_part, types.Part.from_text(text=prompt)])] + response = client.models.generate_content(model=model, contents=contents, config=config) + if not response.candidates or len(response.candidates) == 0 or response.candidates[0].finish_reason != types.FinishReason.STOP: + continue + answer_text = response.candidates[0].content.parts[0].text.strip() + if answer_text and "null" not in answer_text: + test_data = {"cell": cell_value, relationship: answer_text} + tests.append(test_data) + tests_for_this_table += 1 + except Exception as e: + print(f"Error querying Gemini for cell '{cell_value}' and relationship '{relationship}': {str(e)}") + + return tests + + +def process_pdf(s3_path: str, temp_dir: str, output_dir: str, api_key: str) -> List[TableTest]: + """ + Process a single PDF from S3. + + Args: + s3_path: S3 path to the PDF + temp_dir: Directory for temporary files + output_dir: Directory for output files + api_key: Gemini API key + + Returns: + List[TableTest]: List of generated table tests + """ + # Create a thread-specific temp directory to avoid conflicts + thread_id = threading.get_ident() + thread_temp_dir = os.path.join(temp_dir, f"thread_{thread_id}") + os.makedirs(thread_temp_dir, exist_ok=True) + + # Extract filename from S3 path + pdf_filename = os.path.basename(s3_path) + local_pdf_path = os.path.join(thread_temp_dir, pdf_filename) + + # Download PDF from S3 + if not download_pdf_from_s3(s3_path, local_pdf_path): + return [] + + pdf_filter = PdfFilter() + + if pdf_filter.filter_out_pdf(local_pdf_path): + print(f"Filtering out {pdf_filename}") + if os.path.exists(local_pdf_path): + os.remove(local_pdf_path) + return [] + + try: + # Read the PDF to get the number of pages + reader = pypdf.PdfReader(local_pdf_path) + num_pages = len(reader.pages) + + if num_pages == 0: + print(f"PDF {pdf_filename} has no pages") + return [] + + all_pages = list(range(len(reader.pages))) + random.shuffle(all_pages) + + local_tests = [] + + for page_num in all_pages: + # Detect tables and obtain the rendered image for this page + result = detect_tables(local_pdf_path, page_num, api_key) + if not result: + print(f"No tables detected in {pdf_filename} page {page_num+1}") + continue + + tables, image_base64 = result + + # Generate table tests using the new Gemini query approach with the page image + table_tests_data = generate_table_tests(tables, image_base64, api_key, max_tests_per_table=5) + + if not table_tests_data: + print(f"Could not generate valid tests for tables in {pdf_filename} page {page_num+1}") + continue + + # Extract the page and save to output dir + pdf_basename = os.path.splitext(pdf_filename)[0] + output_pdf_path = os.path.join(output_dir, "pdfs", f"{pdf_basename}_pg{page_num+1}.pdf") + with file_lock: # Use lock when writing to shared output directory + extract_page_from_pdf(local_pdf_path, output_pdf_path, page_num) + + # Create table tests + for i, test_data in enumerate(table_tests_data): + test_id = f"{pdf_basename}_pg{page_num+1}_table_{i:02d}" + test = TableTest( + id=test_id, + pdf=f"{pdf_basename}_pg{page_num+1}.pdf", + page=1, # The extracted PDF has only one page + type="table", + cell=test_data["cell"], + url=s3_path, # Added the S3 path as the url field + up=test_data.get("up", None), + down=test_data.get("down", None), + left=test_data.get("left", None), + right=test_data.get("right", None), + top_heading=test_data.get("top_heading", None), + left_heading=test_data.get("left_heading", None), + ) + local_tests.append(test) + + print(f"Processed {pdf_filename} page {page_num+1}, found {len(tables)} tables, created {len(table_tests_data)} tests") + break # Process only one page per PDF + + return local_tests + + except Exception as e: + print(f"Error processing {pdf_filename}: {str(e)}") + return [] + finally: + # Cleanup + if os.path.exists(local_pdf_path): + os.remove(local_pdf_path) + + +def process_pdfs_parallel(s3_paths: List[str], temp_dir: str, output_dir: str, api_key: str, max_tests: int, num_workers: int): + """ + Process PDFs in parallel using a thread pool. + + Args: + s3_paths: List of S3 paths to PDFs + temp_dir: Directory for temporary files + output_dir: Directory for output files + api_key: Gemini API key + max_tests: Maximum number of tests to generate + num_workers: Number of parallel workers to use + """ + # Create shared resources + all_tests = [] + output_file = os.path.join(output_dir, "table_tests.jsonl") + + # Create a ThreadPoolExecutor + with concurrent.futures.ThreadPoolExecutor(max_workers=num_workers) as executor: + # Submit tasks and track futures + futures = {executor.submit(process_pdf, s3_path, temp_dir, output_dir, api_key): s3_path for s3_path in s3_paths} + + # Process results as they complete + for future in concurrent.futures.as_completed(futures): + s3_path = futures[future] + try: + # Get the tests produced by this worker + new_tests = future.result() + + # If we got new tests, add them to our collection + if new_tests: + all_tests.extend(new_tests) + save_tests(all_tests, output_file) + print(f"Added {len(new_tests)} tests from {os.path.basename(s3_path)}, total: {len(all_tests)}") + + # Check if we've reached the maximum number of tests + if len(all_tests) >= max_tests: + print(f"Reached maximum number of tests ({max_tests}), stopping") + # Cancel any pending futures + for f in futures: + if not f.done(): + f.cancel() + break + + except Exception as e: + print(f"Task for {os.path.basename(s3_path)} generated an exception: {e}") + + +def main(): + parser = argparse.ArgumentParser(description="Extract tables from PDF documents and create table tests") + parser.add_argument("--input_list", required=True, help="Path to a file containing S3 paths to PDFs") + parser.add_argument("--output_dir", required=True, help="Directory to store extracted pages and tests") + parser.add_argument("--api_key", help="Gemini API key (if not provided, will use GEMINI_API_KEY environment variable)") + parser.add_argument("--temp_dir", default="/tmp/mine_tables", help="Directory for temporary files") + parser.add_argument("--max_tests", type=int, default=100, help="Maximum number of tests to generate") + parser.add_argument("--parallel", type=int, default=1, help="Number of parallel threads to use") + args = parser.parse_args() + + # Get API key + api_key = args.api_key or os.environ.get("GEMINI_API_KEY") + if not api_key: + print("Error: Gemini API key not provided. Use --api_key or set GEMINI_API_KEY environment variable.") + return + + os.makedirs(args.temp_dir, exist_ok=True) + os.makedirs(os.path.join(args.output_dir, "pdfs"), exist_ok=True) + + # Reservoir sampling implementation + s3_paths = [] + with open(args.input_list, "r") as f: + for i, line in enumerate(tqdm(f)): + line = line.strip() + if not line: + continue + + if i < 100000: + s3_paths.append(line) + else: + # Randomly replace elements with decreasing probability + j = random.randint(0, i) + if j < 100000: + s3_paths[j] = line + + print(f"Found {len(s3_paths)} PDF paths in input list") + + # Determine number of workers to use + num_workers = max(1, min(args.parallel, len(s3_paths))) + print(f"Processing PDFs using {num_workers} parallel workers") + + # Process PDFs in parallel + process_pdfs_parallel(s3_paths, args.temp_dir, args.output_dir, api_key, args.max_tests, num_workers) + + +if __name__ == "__main__": + main() diff --git a/olmocr/bench/miners/mine_tables_gpt.py b/olmocr/bench/miners/mine_tables_gpt.py new file mode 100644 index 0000000..53010bb --- /dev/null +++ b/olmocr/bench/miners/mine_tables_gpt.py @@ -0,0 +1,418 @@ +#!/usr/bin/env python3 +""" +mine_tables.py - Extract tables from PDF documents and create table tests. + +This script: +1. Takes a file containing S3 paths to PDF documents as input +2. For each PDF, extracts a random page and renders it to an image +3. Uses GPT-4o to identify tables in the rendered image +4. Extracts table content and creates table relationship tests by making a second GPT-4o request + that now includes the page image alongside the prompt (e.g., "Given cell with {cell_value}, which cell is directly to the left of it?") +5. Extracts the page from the PDF and saves it to an output folder + +Usage: + python mine_tables.py --input_list path/to/s3_paths.txt --output_dir path/to/output --api_key your_openai_api_key +""" + +import argparse +import os +import random +from typing import Dict, List, Optional, Tuple + +import boto3 +import numpy as np +import pypdf +from bs4 import BeautifulSoup +from openai import OpenAI +from tqdm import tqdm + +from olmocr.bench.tests import TableTest, save_tests +from olmocr.data.renderpdf import render_pdf_to_base64png +from olmocr.filter import PdfFilter + + +def download_pdf_from_s3(s3_path: str, local_path: str) -> bool: + """ + Download a PDF file from S3. + + Args: + s3_path: The S3 path (s3://bucket/path/to/file.pdf) + local_path: The local path to save the file + + Returns: + bool: True if download was successful, False otherwise + """ + try: + # Parse S3 path + parts = s3_path.replace("s3://", "").split("/", 1) + bucket = parts[0] + key = parts[1] + + # Create S3 client + s3 = boto3.client("s3") + + # Create directory if it doesn't exist + os.makedirs(os.path.dirname(local_path), exist_ok=True) + + # Download file + s3.download_file(bucket, key, local_path) + return True + except Exception as e: + print(f"Error downloading {s3_path}: {str(e)}") + return False + + +def extract_page_from_pdf(input_path: str, output_path: str, page_num: int) -> bool: + """ + Extract a specific page from a PDF and save it as a new PDF. + + Args: + input_path: Path to the input PDF + output_path: Path to save the extracted page + page_num: The page number to extract (0-indexed) + + Returns: + bool: True if extraction was successful, False otherwise + """ + try: + # Ensure output directory exists + os.makedirs(os.path.dirname(output_path), exist_ok=True) + + # Read the input PDF + reader = pypdf.PdfReader(input_path) + + # Check if page number is valid + if page_num >= len(reader.pages): + print(f"Page number {page_num} out of range for {input_path} with {len(reader.pages)} pages") + return False + + # Create a new PDF with just the selected page + writer = pypdf.PdfWriter() + writer.add_page(reader.pages[page_num]) + + # Write the output PDF + with open(output_path, "wb") as output_file: + writer.write(output_file) + + return True + except Exception as e: + print(f"Error extracting page {page_num} from {input_path}: {str(e)}") + raise + + +def detect_tables(pdf_path: str, page_num: int, api_key: str) -> Optional[Tuple[List[np.ndarray], str]]: + """ + Use GPT-4o to detect tables in a rendered PDF page. + + Args: + pdf_path: Path to the PDF file + page_num: The page number to analyze (0-indexed) + api_key: OpenAI API key + + Returns: + Optional[Tuple[List[np.ndarray], str]]: + A tuple with a list of detected tables (as numpy arrays) and the base64 string of the rendered page image. + Returns None if detection fails. + """ + # Initialize OpenAI client + client = OpenAI(api_key=api_key) + model = "gpt-4o" + + # Render the PDF page as an image (render_pdf_to_base64png is 1-indexed) + try: + image_base64 = render_pdf_to_base64png(pdf_path, page_num=page_num + 1, target_longest_image_dim=2048) + except Exception as e: + print(f"Error rendering PDF page: {str(e)}") + return None + + # Prepare prompt for GPT-4o to extract tables + try: + # Call OpenAI API + response = client.chat.completions.create( + model=model, + messages=[ + { + "role": "user", + "content": [ + {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{image_base64}", "detail": "high"}}, + { + "type": "text", + "text": ( + "Analyze the document attached and output it in markdown format. " + "Output equations as Latex escaped with $$. " + "Output tables in valid HTML format that preserves the structure and content exactly. " + "Output figures with just a simple markdown image placeholder." + ), + }, + ], + } + ], + temperature=0.2, + ) + + if not response.choices or len(response.choices) == 0: + print(f"No response generated for {pdf_path} page {page_num}") + return None + + # Parse the response + response_text = response.choices[0].message.content + + print(response_text) + + # Parse tables from HTML + parsed_tables = [] + soup = BeautifulSoup(response_text, "html.parser") + tables = soup.find_all("table") + + for table in tables: + rows = table.find_all("tr") + table_data = [] + for row in rows: + cells = row.find_all(["th", "td"]) + row_data = [cell.get_text().strip() for cell in cells] + table_data.append(row_data) + # Ensure all rows have the same number of columns + if table_data: + max_cols = max(len(row) for row in table_data) + padded_data = [row + [""] * (max_cols - len(row)) for row in table_data] + table_array = np.array(padded_data) + parsed_tables.append(table_array) + + # Return both the parsed tables and the rendered image (base64 string) + return (parsed_tables, image_base64) if parsed_tables else None + + except Exception as e: + print(f"Error detecting tables in {pdf_path} page {page_num}: {str(e)}") + return None + + +def generate_table_tests(tables: List[np.ndarray], pdf_image: str, api_key: str, max_tests_per_table: int = 3) -> List[Dict]: + """ + Generate table tests from the detected tables by making a second GPT-4o request for each candidate cell. + + For each candidate cell in a table, the function selects one valid relationship (e.g., "left", "up", "top_heading", etc.) + and sends a prompt to GPT-4o including the page image. For example: + "Given a cell in a table with value 'XYZ', please answer: which cell is directly to the left of it? Provide only the cell's text." + + Args: + tables: List of tables as numpy arrays + pdf_image: Base64 string of the rendered page image + api_key: OpenAI API key to use for generating relationship tests + max_tests_per_table: Maximum number of tests to generate per table + + Returns: + List of table test dictionaries + """ + tests = [] + # Initialize OpenAI client for test queries + client = OpenAI(api_key=api_key) + model = "gpt-4o" + + # Mapping for relationship prompts + prompt_map = { + "up": "which cell is directly above it?", + "down": "which cell is directly below it?", + "left": "which cell is directly to the left of it?", + "right": "which cell is directly to the right of it?", + "top_heading": "what is the top heading for this cell?", + "left_heading": "what is the left heading for this cell?", + } + + for table in tables: + rows, cols = table.shape + if table.size == 0 or rows < 2 or cols < 2: + continue # Skip tables that are too small + + # Try up to 3x max_tests_per_table candidate cells + candidate_positions = [] + for _ in range(max_tests_per_table * 3): + row = random.randint(0, rows - 1) + col = random.randint(0, cols - 1) + if not table[row, col].strip(): + continue + candidate_positions.append((row, col)) + + random.shuffle(candidate_positions) + tests_for_this_table = 0 + + for row, col in candidate_positions: + if tests_for_this_table >= max_tests_per_table: + break + + cell_value = table[row, col].strip() + # Determine valid relationship types based on candidate's position + valid_relationships = [] + if row > 0: + valid_relationships.append("up") + if row < rows - 1: + valid_relationships.append("down") + if col > 0: + valid_relationships.append("left") + if col < cols - 1: + valid_relationships.append("right") + if row > 0: + valid_relationships.append("top_heading") + if col > 0: + valid_relationships.append("left_heading") + if not valid_relationships: + continue + + relationship = random.choice(valid_relationships) + prompt = ( + f"Given a cell in a table with value '{cell_value}', please answer: " + f"{prompt_map[relationship]} Provide only the cell's text or output 'null' if there is not a matching cell." + ) + + try: + response = client.chat.completions.create( + model=model, + messages=[ + { + "role": "user", + "content": [ + {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{pdf_image}", "detail": "high"}}, + {"type": "text", "text": prompt}, + ], + } + ], + temperature=0.2, + ) + + if not response.choices or len(response.choices) == 0: + continue + + answer_text = response.choices[0].message.content.strip() + if answer_text and "null" not in answer_text: + test_data = {"cell": cell_value, relationship: answer_text} + tests.append(test_data) + tests_for_this_table += 1 + except Exception as e: + print(f"Error querying GPT-4o for cell '{cell_value}' and relationship '{relationship}': {str(e)}") + + return tests + + +def process_pdf(s3_path: str, temp_dir: str, output_dir: str, api_key: str, tests: List[TableTest]) -> None: + """ + Process a single PDF from S3. + + Args: + s3_path: S3 path to the PDF + temp_dir: Directory for temporary files + output_dir: Directory for output files + api_key: OpenAI API key + tests: List to append tests to + """ + # Extract filename from S3 path + pdf_filename = os.path.basename(s3_path) + local_pdf_path = os.path.join(temp_dir, pdf_filename) + + # Download PDF from S3 + if not download_pdf_from_s3(s3_path, local_pdf_path): + return + + pdf_filter = PdfFilter() + + if pdf_filter.filter_out_pdf(local_pdf_path): + print(f"Filtering out {pdf_filename}") + return + + try: + # Read the PDF to get the number of pages + reader = pypdf.PdfReader(local_pdf_path) + num_pages = len(reader.pages) + + if num_pages == 0: + print(f"PDF {pdf_filename} has no pages") + return + + all_pages = list(range(len(reader.pages))) + random.shuffle(all_pages) + + for page_num in all_pages: + # Detect tables and obtain the rendered image for this page + result = detect_tables(local_pdf_path, page_num, api_key) + if not result: + print(f"No tables detected in {pdf_filename} page {page_num+1}") + continue + + tables, image_base64 = result + + # Generate table tests using the new GPT-4o query approach with the page image + table_tests_data = generate_table_tests(tables, image_base64, api_key, max_tests_per_table=5) + + if not table_tests_data: + print(f"Could not generate valid tests for tables in {pdf_filename} page {page_num+1}") + continue + + # Extract the page and save to output dir + pdf_basename = os.path.splitext(pdf_filename)[0] + output_pdf_path = os.path.join(output_dir, "pdfs", f"{pdf_basename}_pg{page_num+1}.pdf") + extract_page_from_pdf(local_pdf_path, output_pdf_path, page_num) + + # Create table tests + for i, test_data in enumerate(table_tests_data): + test_id = f"{pdf_basename}_pg{page_num+1}_table_{i:02d}" + test = TableTest( + id=test_id, + pdf=f"{pdf_basename}_pg{page_num+1}.pdf", + page=1, # The extracted PDF has only one page + type="table", + cell=test_data["cell"], + up=test_data.get("up", None), + down=test_data.get("down", None), + left=test_data.get("left", None), + right=test_data.get("right", None), + top_heading=test_data.get("top_heading", None), + left_heading=test_data.get("left_heading", None), + ) + tests.append(test) + + print(f"Processed {pdf_filename} page {page_num+1}, found {len(tables)} tables, created {len(table_tests_data)} tests") + return # Process only one page per PDF + + except Exception as e: + print(f"Error processing {pdf_filename}: {str(e)}") + finally: + if os.path.exists(local_pdf_path): + os.remove(local_pdf_path) + + +def main(): + parser = argparse.ArgumentParser(description="Extract tables from PDF documents and create table tests") + parser.add_argument("--input_list", required=True, help="Path to a file containing S3 paths to PDFs") + parser.add_argument("--output_dir", required=True, help="Directory to store extracted pages and tests") + parser.add_argument("--api_key", help="OpenAI API key (if not provided, will use OPENAI_API_KEY environment variable)") + parser.add_argument("--temp_dir", default="/tmp/mine_tables", help="Directory for temporary files") + parser.add_argument("--max_tests", type=int, default=100, help="Maximum number of tests to generate") + args = parser.parse_args() + + # Get API key + api_key = args.api_key or os.environ.get("OPENAI_API_KEY") + if not api_key: + print("Error: OpenAI API key not provided. Use --api_key or set OPENAI_API_KEY environment variable.") + return + + os.makedirs(args.temp_dir, exist_ok=True) + os.makedirs(os.path.join(args.output_dir, "pdfs"), exist_ok=True) + + with open(args.input_list, "r") as f: + s3_paths = [line.strip() for line in f if line.strip()] + + print(f"Found {len(s3_paths)} PDF paths in input list") + tests = [] + for s3_path in tqdm(s3_paths, desc="Processing PDFs"): + process_pdf(s3_path, args.temp_dir, args.output_dir, api_key, tests) + + if tests: + save_tests(tests, os.path.join(args.output_dir, "table_tests.jsonl")) + + if len(tests) >= args.max_tests: + print(f"Reached maximum number of tests ({args.max_tests}), stopping") + break + + print(f"Saved {len(tests)} table tests to {os.path.join(args.output_dir, 'table_tests.jsonl')}") + + +if __name__ == "__main__": + main() diff --git a/olmocr/bench/miners/mine_tables_gpt_simple.py b/olmocr/bench/miners/mine_tables_gpt_simple.py new file mode 100644 index 0000000..32f3199 --- /dev/null +++ b/olmocr/bench/miners/mine_tables_gpt_simple.py @@ -0,0 +1,333 @@ +#!/usr/bin/env python3 +""" +mine_tables_gpt_simple.py - Identify PDF documents with tables and copy them. + +This script: +1. Takes a file containing S3 paths to PDF documents as input +2. For each PDF, renders a random page and uses GPT-4o to check for tables +3. Identifies PDFs where the page contains a table +4. Copies those PDF files to a new output folder + +Usage: + python mine_tables_gpt_simple.py --input_list path/to/s3_paths.txt --output_dir path/to/output --api_key your_openai_api_key +""" + +import argparse +import os +import random +from concurrent.futures import ThreadPoolExecutor, as_completed +from typing import Optional + +import boto3 +import pypdf +from openai import OpenAI +from pydantic import BaseModel +from tqdm import tqdm + +from olmocr.data.renderpdf import render_pdf_to_base64png +from olmocr.filter import PdfFilter + +TARGET_IMAGE_DIM = 1024 + + +class TableInfo(BaseModel): + """Information about a single table.""" + + num_rows: int + num_cols: int + + +class TableDetectionResponse(BaseModel): + """Structured output for table detection.""" + + tables: list[TableInfo] + + +def download_pdf_from_s3(s3_path: str, local_path: str) -> bool: + """ + Download a PDF file from S3. + + Args: + s3_path: The S3 path (s3://bucket/path/to/file.pdf) + local_path: The local path to save the file + + Returns: + bool: True if download was successful, False otherwise + """ + try: + # Parse S3 path + parts = s3_path.replace("s3://", "").split("/", 1) + bucket = parts[0] + key = parts[1] + + # Create S3 client + s3 = boto3.client("s3") + + # Create directory if it doesn't exist + os.makedirs(os.path.dirname(local_path), exist_ok=True) + + # Download file + s3.download_file(bucket, key, local_path) + return True + except Exception as e: + print(f"Error downloading {s3_path}: {str(e)}") + return False + + +def get_cell_count_bucket(total_cells: int) -> str: + """ + Get the folder name for a given cell count, bucketed by powers of 2. + + Args: + total_cells: Total number of cells across all tables + + Returns: + str: Folder name like "0_cells", "1_cell", "2_cells", "4_cells", etc. + """ + if total_cells == 0: + return "0_cells" + elif total_cells == 1: + return "1_cell" + else: + # Find the next power of 2 >= total_cells + power = 1 + while power < total_cells: + power *= 2 + return f"{power}_cells" + + +def check_for_table(pdf_path: str, page_num: int, api_key: str) -> Optional[tuple[bool, int]]: + """ + Use GPT-4o to check if a page contains a table. + + Args: + pdf_path: Path to the PDF file + page_num: The page number to analyze (0-indexed) + api_key: OpenAI API key + + Returns: + Optional[tuple[bool, int]]: Tuple of (has_table, total_cells) or None if detection fails + """ + # Initialize OpenAI client + client = OpenAI(api_key=api_key) + + try: + # Render the PDF page as an image (render_pdf_to_base64png is 1-indexed) + image_base64 = render_pdf_to_base64png(pdf_path, page_num=page_num + 1, target_longest_image_dim=TARGET_IMAGE_DIM) + + # Prompt asking for detailed table information + prompt = "Identify all tables on this page. For each table, count the number of rows and columns. Return an empty list if there are no tables." + + response = client.beta.chat.completions.parse( + model="gpt-5.1", + messages=[ + { + "role": "user", + "content": [{"type": "text", "text": prompt}, {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{image_base64}"}}], + } + ], + max_completion_tokens=1000, + response_format=TableDetectionResponse, + ) + + if not response.choices or len(response.choices) == 0: + print(f"No response generated for {pdf_path} page {page_num}") + return None + + # Parse the structured response + parsed_response = response.choices[0].message.parsed + + if parsed_response is None: + print(f"Failed to parse response for {pdf_path} page {page_num}") + return None + + tables = parsed_response.tables + has_table = len(tables) > 0 + total_cells = sum(table.num_rows * table.num_cols for table in tables) + + if has_table: + print(f"Found {len(tables)} table(s) in {pdf_path} page {page_num + 1}, total cells: {total_cells}") + for i, table in enumerate(tables, 1): + print(f" Table {i}: {table.num_rows} rows × {table.num_cols} cols = {table.num_rows * table.num_cols} cells") + + return (has_table, total_cells) + + except Exception as e: + print(f"Error checking {pdf_path} page {page_num}: {str(e)}") + return None + + +def process_pdf(s3_path: str, temp_dir: str, output_dir: str, api_key: str) -> bool: + """ + Process a single PDF from S3. + + Args: + s3_path: S3 path to the PDF + temp_dir: Directory for temporary files + output_dir: Directory for output files + api_key: OpenAI API key + + Returns: + bool: True if the PDF has a table and was copied, False otherwise + """ + # Extract filename from S3 path + pdf_filename = os.path.basename(s3_path) + local_pdf_path = os.path.join(temp_dir, pdf_filename) + + # Download PDF from S3 + if not download_pdf_from_s3(s3_path, local_pdf_path): + return False + + pdf_filter = PdfFilter() + + if pdf_filter.filter_out_pdf(local_pdf_path): + print(f"Filtering out {pdf_filename}") + return False + + try: + # Read the PDF to get the number of pages + reader = pypdf.PdfReader(local_pdf_path) + num_pages = len(reader.pages) + + if num_pages == 0: + print(f"PDF {pdf_filename} has no pages") + return False + + # Select a random page to check + page_num = random.randint(0, num_pages - 1) + page_num = random.choice([page_num, 0]) # Bias 50% of the time to do the first page + + # Check if the page contains a table + result = check_for_table(local_pdf_path, page_num, api_key) + + if result is None: + return False + + has_table, total_cells = result + + if has_table: + # Get the cell count bucket for organizing output + bucket_name = get_cell_count_bucket(total_cells) + bucket_dir = os.path.join(output_dir, bucket_name) + os.makedirs(bucket_dir, exist_ok=True) + + # Create output filename with basename_pgnum.pdf format + pdf_basename = os.path.splitext(pdf_filename)[0] + output_pdf_path = os.path.join(bucket_dir, f"{pdf_basename}_pg{page_num+1}.pdf") + + # Extract the single page + writer = pypdf.PdfWriter() + writer.add_page(reader.pages[page_num]) + + # Write the output PDF + with open(output_pdf_path, "wb") as output_file: + writer.write(output_file) + + print(f"Extracted page {page_num+1} with table from {pdf_filename} to {bucket_name}/{os.path.basename(output_pdf_path)}") + return True + + return False + + except Exception as e: + print(f"Error processing {pdf_filename}: {str(e)}") + return False + finally: + if os.path.exists(local_pdf_path): + os.remove(local_pdf_path) + + +def main(): + parser = argparse.ArgumentParser(description="Identify and copy PDFs with tables") + parser.add_argument("--input_list", required=True, help="Path to a file containing S3 paths to PDFs") + parser.add_argument("--output_dir", required=True, help="Directory to copy PDFs with tables") + parser.add_argument("--api_key", help="OpenAI API key (if not provided, will use OPENAI_API_KEY environment variable)") + parser.add_argument("--temp_dir", default="/tmp/mine_tables", help="Directory for temporary files") + parser.add_argument("--max_pdfs", type=int, default=100, help="Maximum number of PDFs with tables to find") + parser.add_argument("--parallel", type=int, default=1, help="Number of parallel workers (default: 1 for sequential)") + parser.add_argument("--reservoir_multiplier", type=int, default=100, help="Multiplier for reservoir sampling (default: 100x max_pdfs)") + args = parser.parse_args() + + # Get API key + api_key = args.api_key or os.environ.get("OPENAI_API_KEY") + if not api_key: + print("Error: OpenAI API key not provided. Use --api_key or set OPENAI_API_KEY environment variable.") + return + + os.makedirs(args.temp_dir, exist_ok=True) + os.makedirs(args.output_dir, exist_ok=True) + + # Reservoir sampling to get random subset of PDFs + reservoir_size = args.max_pdfs * args.reservoir_multiplier + pdf_paths = [] + n = 0 # Total number of items seen + + print(f"Using reservoir sampling with size {reservoir_size}") + + with open(args.input_list, "r") as f: + for line in tqdm(f): + n += 1 + path = line.strip() + if not path: + continue + + if len(pdf_paths) < reservoir_size: + pdf_paths.append(path) + else: + # Randomly decide whether to include this item + s = random.randint(1, n) + if s <= reservoir_size: + pdf_paths[s - 1] = path + + # Shuffle the reservoir + random.shuffle(pdf_paths) + + print(f"Sampled {len(pdf_paths)} PDF paths from {n} total paths") + + table_pdfs_found = 0 + + if args.parallel > 1: + # Parallel processing + print(f"Processing PDFs with {args.parallel} parallel workers") + + with ThreadPoolExecutor(max_workers=args.parallel) as executor: + futures = [] + + # Submit all tasks + for s3_path in pdf_paths: + if table_pdfs_found >= args.max_pdfs: + break + future = executor.submit(process_pdf, s3_path, args.temp_dir, args.output_dir, api_key) + futures.append(future) + + # Process results as they complete + with tqdm(total=min(len(pdf_paths), args.max_pdfs), desc="Processing PDFs") as pbar: + for future in as_completed(futures): + try: + result = future.result() + if result: + table_pdfs_found += 1 + pbar.update(1) + + if table_pdfs_found >= args.max_pdfs: + print(f"Reached maximum number of PDFs with tables ({args.max_pdfs}), stopping") + # Cancel remaining futures + for f in futures: + f.cancel() + break + except Exception as e: + print(f"Error in parallel processing: {str(e)}") + else: + # Sequential processing + for s3_path in tqdm(pdf_paths, desc="Processing PDFs"): + if process_pdf(s3_path, args.temp_dir, args.output_dir, api_key): + table_pdfs_found += 1 + + if table_pdfs_found >= args.max_pdfs: + print(f"Reached maximum number of PDFs with tables ({args.max_pdfs}), stopping") + break + + print(f"Found and copied {table_pdfs_found} PDFs with tables to {args.output_dir}") + + +if __name__ == "__main__": + main() diff --git a/olmocr/bench/miners/pick_mediod.py b/olmocr/bench/miners/pick_mediod.py new file mode 100755 index 0000000..f9564c3 --- /dev/null +++ b/olmocr/bench/miners/pick_mediod.py @@ -0,0 +1,162 @@ +#!/usr/bin/env python3 +""" +pick_mediod.py - Identify representative examples from repeated OCR outputs + +This code will take as arguments two directories: +--input and --output +Each of those is going to be a directory that was generated by convert.py and is a candidate to be evaluated as part of benchmark.py +What it will do is find and group all of the .md files into their repeats +ex. input_dir/tables/buildingnotes_pg1_repeat1.md, input_dir/tables/buildingnotes_pg1_repeat2.md, etc. +Then, for each repeat, it will use string similarity metrics to calculate the edit distance to every other repeat +The repeat with the lowest mean edit distance will then get output as ..._repeat1.md in the output folder +""" + +import argparse +import glob +import os +import re +import shutil +from typing import Dict, List + +from rapidfuzz import distance as fuzz_distance +from tqdm import tqdm + + +def compute_distance(text1: str, text2: str) -> float: + """ + Compute the edit distance between two text strings using rapidfuzz. + Returns a normalized distance between 0.0 (identical) and 1.0 (completely different). + """ + # Use Levenshtein distance for string comparison + return fuzz_distance.Levenshtein.normalized_distance(text1, text2) + + +def find_mediod(texts: List[str]) -> int: + """ + Find the index of the mediod from a list of texts. + The mediod is the text with the minimum average distance to all other texts. + """ + if not texts: + return -1 + + if len(texts) == 1: + return 0 + + # Calculate pairwise distances between all texts + n = len(texts) + distances = [[0.0 for _ in range(n)] for _ in range(n)] + + for i in range(n): + for j in range(i + 1, n): + dist = compute_distance(texts[i], texts[j]) + distances[i][j] = dist + distances[j][i] = dist + + # Calculate average distance of each text to all others + avg_distances = [] + for i in range(n): + avg_dist = sum(distances[i]) / (n - 1) # Don't include distance to self + avg_distances.append(avg_dist) + + # Return the index of the text with the minimum average distance + min_avg_dist = min(avg_distances) + return avg_distances.index(min_avg_dist) + + +def group_repeats(md_files: List[str]) -> Dict[str, List[str]]: + """ + Group MD files by their base name (without the repeat number). + Returns a dictionary mapping base names to lists of file paths. + """ + grouped = {} + + for md_path in md_files: + base_name = re.sub(r"_repeat\d+\.md$", "", os.path.basename(md_path)) + if base_name not in grouped: + grouped[base_name] = [] + grouped[base_name].append(md_path) + + return grouped + + +def main(): + parser = argparse.ArgumentParser(description="Find mediod (most representative) examples from repeated OCR outputs.") + parser.add_argument( + "--input", type=str, required=True, help="Path to the directory containing repeated OCR outputs (e.g., *_repeat1.md, *_repeat2.md, etc.)" + ) + parser.add_argument("--output", type=str, required=True, help="Path to the directory where mediod examples will be copied") + parser.add_argument("--min_repeats", type=int, default=3, help="Minimum number of repeats required to compute a mediod (default: 3)") + args = parser.parse_args() + + input_dir = args.input + output_dir = args.output + min_repeats = args.min_repeats + + # Create output directory if it doesn't exist + os.makedirs(output_dir, exist_ok=True) + + # Find all markdown files in the input directory (recursive) + md_files = glob.glob(os.path.join(input_dir, "**/*.md"), recursive=True) + + if not md_files: + print(f"No markdown files found in {input_dir}") + return + + # Group files by their base name + grouped_files = group_repeats(md_files) + + # Process each group + successful = 0 + skipped = 0 + + print(f"Found {len(grouped_files)} unique test cases with repeats") + + for base_name, file_paths in tqdm(grouped_files.items(), desc="Processing test cases"): + # Skip if there aren't enough repeats + if len(file_paths) < min_repeats: + print(f"Skipping {base_name}: only {len(file_paths)} repeats (minimum {min_repeats} required)") + skipped += 1 + continue + + # Read all text content + texts = [] + for path in file_paths: + try: + with open(path, "r", encoding="utf-8") as f: + texts.append(f.read()) + except Exception as e: + print(f"Error reading {path}: {e}") + continue + + # Find the mediod + mediod_idx = find_mediod(texts) + if mediod_idx == -1: + print(f"Failed to find mediod for {base_name}") + skipped += 1 + continue + + # Get the path of the mediod file + mediod_path = file_paths[mediod_idx] + + # Create the output path, preserving the directory structure relative to input_dir + rel_path = os.path.relpath(mediod_path, input_dir) + # Change the repeat number to 1 in the output filename + output_filename = re.sub(r"_repeat\d+\.md$", "_repeat1.md", os.path.basename(rel_path)) + output_subdir = os.path.dirname(rel_path) + output_path = os.path.join(output_dir, output_subdir, output_filename) + + # Create directories if needed + os.makedirs(os.path.dirname(output_path), exist_ok=True) + + # Copy the mediod file + try: + shutil.copy2(mediod_path, output_path) + successful += 1 + except Exception as e: + print(f"Error copying {mediod_path} to {output_path}: {e}") + + print(f"Processing complete: {successful} mediods copied, {skipped} cases skipped") + + +if __name__ == "__main__": + main() diff --git a/olmocr/bench/prompts.py b/olmocr/bench/prompts.py new file mode 100644 index 0000000..b8c355a --- /dev/null +++ b/olmocr/bench/prompts.py @@ -0,0 +1,62 @@ +def build_basic_prompt() -> str: + return "Please provide a natural, plain text representation of the document, formatted in Markdown. Skip any headers and footers. For ALL mathematical expressions, use LaTeX notation with $ and $ for inline equations and \[ and \] for display equations. Convert any tables into Markdown format." + + +def build_openai_silver_data_prompt_no_document_anchoring(_base_text: str) -> str: + return ( + "Below is the image of one page of a PDF document. " + "Just return the plain text representation of this document as if you were reading it naturally.\n" + "Turn equations into a LaTeX representation, and tables into markdown format. Remove the headers and footers, but keep references and footnotes.\n" + "Read any natural handwriting.\n" + "This is likely one page out of several in the document, so be sure to preserve any sentences that come from the previous page, or continue onto the next page, exactly as they are.\n" + "If there is no text at all that you think you should read, you can output null.\n" + "Do not hallucinate." + ) + + +def claude_response_format_schema() -> dict: + return ( + { + "name": "page_response", + "description": "Extracts text from pdf's.", + "input_schema": { + "type": "object", + "properties": { + "primary_language": { + "type": ["string", "null"], + "description": "The primary language of the text using two-letter codes or null if there is no text at all that you think you should read.", + }, + "is_rotation_valid": { + "type": "boolean", + "description": "Is this page oriented correctly for reading? Answer only considering the textual content, do not factor in the rotation of any charts, tables, drawings, or figures.", + }, + "rotation_correction": { + "type": "integer", + "description": "Indicates the degree of clockwise rotation needed if the page is not oriented correctly.", + "enum": [0, 90, 180, 270], + "default": 0, + }, + "is_table": { + "type": "boolean", + "description": "Indicates if the majority of the page content is in tabular format.", + }, + "is_diagram": { + "type": "boolean", + "description": "Indicates if the majority of the page content is a visual diagram.", + }, + "natural_text": { + "type": ["string", "null"], + "description": "The natural text content extracted from the page.", + }, + }, + "required": [ + "primary_language", + "is_rotation_valid", + "rotation_correction", + "is_table", + "is_diagram", + "natural_text", + ], + }, + }, + ) diff --git a/olmocr/bench/report.py b/olmocr/bench/report.py new file mode 100644 index 0000000..4f61b3b --- /dev/null +++ b/olmocr/bench/report.py @@ -0,0 +1,300 @@ +import glob +import os +from typing import Dict, List, Optional, Tuple + +from tqdm import tqdm + +from olmocr.data.renderpdf import render_pdf_to_base64webp + +from .tests import BasePDFTest + + +def _filter_by_max_reports( + test_results_by_candidate: Dict[str, Dict[str, Dict[int, List[Tuple[BasePDFTest, bool, str]]]]], + test_to_jsonl: Dict[str, str], + max_reports: int, +) -> Dict[str, Dict[str, Dict[int, List[Tuple[BasePDFTest, bool, str]]]]]: + """Filter test results to include at most max_reports unique PDFs per .jsonl file.""" + filtered = {} + for candidate, pdf_results in test_results_by_candidate.items(): + # Track which unique PDFs we've included per jsonl file + jsonl_pdfs: Dict[str, set] = {} + filtered_pdfs: Dict[str, Dict[int, List[Tuple[BasePDFTest, bool, str]]]] = {} + + for pdf_name in sorted(pdf_results.keys()): + pages = pdf_results[pdf_name] + for page in sorted(pages.keys()): + for test_tuple in pages[page]: + test, passed, explanation = test_tuple + jsonl_file = test_to_jsonl.get(test.id, "unknown") + jsonl_pdfs.setdefault(jsonl_file, set()) + # If this PDF is new for this jsonl file, check the limit + if pdf_name not in jsonl_pdfs[jsonl_file]: + if len(jsonl_pdfs[jsonl_file]) >= max_reports: + continue + jsonl_pdfs[jsonl_file].add(pdf_name) + filtered_pdfs.setdefault(pdf_name, {}).setdefault(page, []).append(test_tuple) + + filtered[candidate] = filtered_pdfs + return filtered + + +def generate_html_report( + test_results_by_candidate: Dict[str, Dict[str, Dict[int, List[Tuple[BasePDFTest, bool, str]]]]], + pdf_folder: str, + output_file: str, + max_reports: Optional[int] = None, + test_to_jsonl: Optional[Dict[str, str]] = None, +) -> None: + """ + Generate a simple static HTML report of test results. + + Args: + test_results_by_candidate: Dictionary mapping candidate name to dictionary mapping PDF name to dictionary + mapping page number to list of (test, passed, explanation) tuples. + pdf_folder: Path to the folder containing PDF files. + output_file: Path to the output HTML file. + max_reports: If set, limit to at most N tests per .jsonl file in the report. + test_to_jsonl: Dictionary mapping test IDs to their source jsonl filenames. + """ + # If max_reports is set, filter test_results_by_candidate to limit per jsonl file + if max_reports is not None and test_to_jsonl is not None: + test_results_by_candidate = _filter_by_max_reports(test_results_by_candidate, test_to_jsonl, max_reports) + candidates = list(test_results_by_candidate.keys()) + + # Create HTML report + html = """ + + + + + OLMOCR Bench Test Report + + + +

OLMOCR Bench Test Report

+""" + + # Process all candidates + print("Generating test report...") + for candidate in candidates: + html += f"

Candidate: {candidate}

\n" + + # Get all PDFs for this candidate + all_pdfs = sorted(test_results_by_candidate[candidate].keys()) + + for pdf_name in tqdm(all_pdfs, desc=f"Processing {candidate}"): + pages = sorted(test_results_by_candidate[candidate][pdf_name].keys()) + + for page in pages: + # Get tests for this PDF page + tests = test_results_by_candidate[candidate][pdf_name][page] + + for test, passed, explanation in tests: + result_class = "pass" if passed else "fail" + status_text = "PASSED" if passed else "FAILED" + status_class = "pass-status" if passed else "fail-status" + + # Begin test block + html += f""" +

Test ID: {test.id} {status_text}

PDF: {pdf_name} | Page: {page} | Type: {test.type}

+ +

+""" + + # Add test details based on type + test_type = getattr(test, "type", "").lower() + if test_type == "present" and hasattr(test, "text"): + text = getattr(test, "text", "") + html += f"""

Text to find: "{text}"

\n""" + elif test_type == "absent" and hasattr(test, "text"): + text = getattr(test, "text", "") + html += f"""

Text should not appear: "{text}"

\n""" + elif test_type == "order" and hasattr(test, "before") and hasattr(test, "after"): + before = getattr(test, "before", "") + after = getattr(test, "after", "") + html += f"""

Text order: "{before}" should appear before "{after}"

\n""" + elif test_type == "table": + if hasattr(test, "cell"): + cell = getattr(test, "cell", "") + html += f"""

Table cell: "{cell}"

\n""" + if hasattr(test, "up") and getattr(test, "up", None): + up = getattr(test, "up") + html += f"""

Above: "{up}"

\n""" + if hasattr(test, "down") and getattr(test, "down", None): + down = getattr(test, "down") + html += f"""

Below: "{down}"

\n""" + if hasattr(test, "left") and getattr(test, "left", None): + left = getattr(test, "left") + html += f"""

Left: "{left}"

\n""" + if hasattr(test, "right") and getattr(test, "right", None): + right = getattr(test, "right") + html += f"""

Right: "{right}"

\n""" + elif test_type == "math" and hasattr(test, "math"): + math = getattr(test, "math", "") + html += f"""

Math equation: {math}

\n""" + elif test_type == "format" and hasattr(test, "text"): + text = getattr(test, "text", "") + fmt = getattr(test, "format", "") + html += f"""

Text: "{text}" should be formatted as {fmt}

\n""" + elif test_type == "footnote" and hasattr(test, "marker"): + marker = getattr(test, "marker", "") + html += f"""

Footnote marker: {marker}

\n""" + before = getattr(test, "appears_before_marker", None) + after = getattr(test, "appears_after_marker", None) + if before: + html += f"""

Text before marker: "{before}"

\n""" + if after: + html += f"""

Text after marker: "{after}"

\n""" + elif test_type == "baseline": + max_length = getattr(test, "max_length", None) + if max_length is not None: + html += f"""

Baseline check: max length {max_length} (blank page check)

\n""" + else: + html += f"""

Baseline check: non-blank, no repeats, valid characters

\n""" + + html += """

\n""" + + # Add explanation for failed tests + if not passed: + html += f"""

+ Explanation: {explanation} +

\n""" + + # Render PDF page + pdf_path = os.path.join(pdf_folder, pdf_name) + try: + html += """

PDF Render:

\n""" + image_data = render_pdf_to_base64webp(pdf_path, page, 1024) + html += f""" PDF Page {page}

\n""" + except Exception as e: + html += f"""

Error rendering PDF: {str(e)}

\n""" + + # Get the Markdown content for this page + md_content = None + try: + md_base = os.path.splitext(pdf_name)[0] + md_files = list(glob.glob(os.path.join(os.path.dirname(pdf_folder), candidate, f"{md_base}_pg{page}_repeat*.md"))) + if md_files: + md_file_path = md_files[0] # Use the first repeat as an example + with open(md_file_path, "r", encoding="utf-8") as f: + md_content = f.read() + except Exception as e: + md_content = f"Error loading Markdown content: {str(e)}" + + if md_content: + html += """

Markdown Content:

\n""" + html += f"""

{md_content}

\n""" + + # End test block + html += """

\n""" + + # Add separator between pages + html += """

\n""" + + # Close HTML + html += """ + +""" + + with open(output_file, "w", encoding="utf-8") as f: + f.write(html) + + print(f"Simple HTML report generated: {output_file}") diff --git a/olmocr/bench/review_app.py b/olmocr/bench/review_app.py new file mode 100644 index 0000000..cf1495e --- /dev/null +++ b/olmocr/bench/review_app.py @@ -0,0 +1,300 @@ +#!/usr/bin/env python3 +import argparse +import json +import os +import shutil +import sys +import tempfile +from collections import defaultdict +from typing import Dict, List, Optional, Tuple + +from flask import Flask, jsonify, redirect, render_template, request, send_file, url_for + +app = Flask(__name__) +# Add static folder for KaTeX files +app.static_folder = os.path.join(os.path.dirname(os.path.abspath(__file__)), "katex") + +# Global state +DATASET_DIR = "" +DATASET_FILE = None +CURRENT_PDF = None +PDF_TESTS = {} +ALL_PDFS = [] +FORCE = False # New global flag + + +def find_next_unchecked_pdf() -> Optional[str]: + """Find the next PDF with at least one unchecked test.""" + global PDF_TESTS, ALL_PDFS + + for pdf_name in ALL_PDFS: + pdf_tests = PDF_TESTS[pdf_name] + for test in pdf_tests: + if test.get("checked") is None: + return pdf_name + return None + + +def calculate_stats() -> dict: + """Calculate statistics for all tests in the dataset.""" + global PDF_TESTS + + total_tests = 0 + null_status = 0 + verified_status = 0 + rejected_status = 0 + + for pdf_tests in PDF_TESTS.values(): + total_tests += len(pdf_tests) + + for test in pdf_tests: + status = test.get("checked") + if status is None: + null_status += 1 + elif status == "verified": + verified_status += 1 + elif status == "rejected": + rejected_status += 1 + + completion = 0 + if total_tests > 0: + completion = (verified_status + rejected_status) / total_tests * 100 + + return {"total": total_tests, "null": null_status, "verified": verified_status, "rejected": rejected_status, "completion": completion} + + +def save_dataset(jsonl_file: str) -> None: + """Save the tests to a JSONL file, using temp file for atomic write.""" + global PDF_TESTS + + # Flatten all tests + all_tests = [] + for pdf_tests in PDF_TESTS.values(): + all_tests.extend(pdf_tests) + + # Create temp file and write updated content + with tempfile.NamedTemporaryFile(mode="w", delete=False) as temp_file: + for test in all_tests: + temp_file.write(json.dumps(test) + "\n") + + # Atomic replace + shutil.move(temp_file.name, jsonl_file) + + +@app.route("/pdf/") +def serve_pdf(pdf_name): + """Serve the PDF file directly.""" + pdf_path = os.path.join(DATASET_DIR, "pdfs", pdf_name) + return send_file(pdf_path, mimetype="application/pdf") + + +@app.route("/") +def index(): + """Main page displaying the current PDF and its tests.""" + global CURRENT_PDF, PDF_TESTS, DATASET_DIR, ALL_PDFS, FORCE + + # If no current PDF is set, find the next one with unchecked tests + if CURRENT_PDF is None: + CURRENT_PDF = find_next_unchecked_pdf() + + # If still no PDF, either show the "All done" page or force display the first PDF + if CURRENT_PDF is None: + if FORCE and ALL_PDFS: + CURRENT_PDF = ALL_PDFS[0] + else: + return render_template("all_done.html") + + # Get the tests for the current PDF + current_tests = PDF_TESTS.get(CURRENT_PDF, []) + + # Create PDF URL for pdf.js to load + pdf_url = url_for("serve_pdf", pdf_name=CURRENT_PDF) + + # Calculate statistics + stats = calculate_stats() + + return render_template( + "review.html", + pdf_name=CURRENT_PDF, + tests=current_tests, + pdf_path=pdf_url, + pdf_index=ALL_PDFS.index(CURRENT_PDF) if CURRENT_PDF in ALL_PDFS else 0, + total_pdfs=len(ALL_PDFS), + stats=stats, + ) + + +@app.route("/update_test", methods=["POST"]) +def update_test(): + """API endpoint to update a test.""" + global PDF_TESTS, DATASET_DIR, DATASET_FILE + + data = request.json + pdf_name = data.get("pdf") + test_id = data.get("id") + field = data.get("field") + value = data.get("value") + + # Find and update the test + for test in PDF_TESTS.get(pdf_name, []): + if test.get("id") == test_id: + test[field] = value + break + + # Save the updated tests + save_dataset(DATASET_FILE) + + return jsonify({"status": "success"}) + + +@app.route("/reject_all", methods=["POST"]) +def reject_all(): + """API endpoint to reject all tests for a PDF.""" + global PDF_TESTS, DATASET_DIR, DATASET_FILE + + data = request.json + pdf_name = data.get("pdf") + + if pdf_name and pdf_name in PDF_TESTS: + # Update all tests for this PDF to rejected + for test in PDF_TESTS[pdf_name]: + test["checked"] = "rejected" + + # Save the updated tests + save_dataset(DATASET_FILE) + + return jsonify({"status": "success", "count": len(PDF_TESTS[pdf_name])}) + + return jsonify({"status": "error", "message": "PDF not found"}) + + +@app.route("/next_pdf", methods=["POST"]) +def next_pdf(): + """Move to the next PDF in the list.""" + global CURRENT_PDF, ALL_PDFS, FORCE + + if CURRENT_PDF in ALL_PDFS: + current_index = ALL_PDFS.index(CURRENT_PDF) + if current_index < len(ALL_PDFS) - 1: + CURRENT_PDF = ALL_PDFS[current_index + 1] + else: + # If in force mode, cycle back to the beginning instead of checking for an unchecked PDF + if FORCE and ALL_PDFS: + CURRENT_PDF = ALL_PDFS[0] + else: + CURRENT_PDF = find_next_unchecked_pdf() + else: + if FORCE and ALL_PDFS: + CURRENT_PDF = ALL_PDFS[0] + else: + CURRENT_PDF = find_next_unchecked_pdf() + + return redirect(url_for("index")) + + +@app.route("/prev_pdf", methods=["POST"]) +def prev_pdf(): + """Move to the previous PDF in the list.""" + global CURRENT_PDF, ALL_PDFS + + if CURRENT_PDF in ALL_PDFS: + current_index = ALL_PDFS.index(CURRENT_PDF) + if current_index > 0: + CURRENT_PDF = ALL_PDFS[current_index - 1] + + return redirect(url_for("index")) + + +@app.route("/goto_pdf/", methods=["POST"]) +def goto_pdf(index): + """Go to a specific PDF by index.""" + global CURRENT_PDF, ALL_PDFS + + if 0 <= index < len(ALL_PDFS): + CURRENT_PDF = ALL_PDFS[index] + + return redirect(url_for("index")) + + +def load_dataset(dataset_file: str) -> Tuple[Dict[str, List[Dict]], List[str]]: + """Load tests from the dataset file and organize them by PDF.""" + if not os.path.exists(dataset_file): + raise FileNotFoundError(f"Dataset file not found: {dataset_file}") + + pdf_tests = defaultdict(list) + + with open(dataset_file, "r") as f: + for line in f: + line = line.strip() + if not line: + continue + + try: + test = json.loads(line) + pdf_name = test.get("pdf") + if pdf_name: + pdf_tests[pdf_name].append(test) + except json.JSONDecodeError: + print(f"Warning: Could not parse line as JSON: {line}") + + all_pdfs = list(pdf_tests.keys()) + + return pdf_tests, all_pdfs + + +def create_templates_directory(): + """Create templates directory for Flask if it doesn't exist.""" + templates_dir = os.path.join(os.path.dirname(__file__), "templates") + os.makedirs(templates_dir, exist_ok=True) + + +def main(): + """Main entry point with command-line arguments.""" + global DATASET_DIR, DATASET_FILE, PDF_TESTS, ALL_PDFS, CURRENT_PDF, FORCE + + parser = argparse.ArgumentParser(description="Interactive Test Review App") + parser.add_argument("dataset_file", help="Path to the dataset jsonl file") + parser.add_argument("--port", type=int, default=5000, help="Port for the Flask app") + parser.add_argument("--host", default="127.0.0.1", help="Host for the Flask app") + parser.add_argument("--debug", action="store_true", help="Run Flask in debug mode") + parser.add_argument("--force", action="store_true", help="Force show each file one by one and never do the 'All done' page") + + args = parser.parse_args() + FORCE = args.force # Set the global FORCE flag + + # Validate dataset directory + if not os.path.exists(args.dataset_file): + print(f"Error: Dataset not found: {args.dataset_file}") + return 1 + + # Store dataset directory globally + DATASET_DIR = os.path.dirname(os.path.abspath(args.dataset_file)) + DATASET_FILE = args.dataset_file + + pdf_dir = os.path.join(DATASET_DIR, "pdfs") + if not os.path.isdir(pdf_dir): + print(f"Error: PDF directory not found: {pdf_dir}") + return 1 + + # Load dataset + try: + PDF_TESTS, ALL_PDFS = load_dataset(args.dataset_file) + except Exception as e: + print(f"Error loading dataset: {str(e)}") + return 1 + + # Create templates directory + create_templates_directory() + + # Find first PDF with unchecked tests + CURRENT_PDF = find_next_unchecked_pdf() + + # Start Flask app + print(f"Starting server at http://{args.host}:{args.port}") + app.run(host=args.host, port=args.port, debug=args.debug) + + return 0 + + +if __name__ == "__main__": + sys.exit(main()) diff --git a/olmocr/bench/review_app_latex.py b/olmocr/bench/review_app_latex.py new file mode 100644 index 0000000..04fd57e --- /dev/null +++ b/olmocr/bench/review_app_latex.py @@ -0,0 +1,716 @@ +#!/usr/bin/env python3 +import argparse +import json +import os +import shutil +import sys +import tempfile +from collections import defaultdict +from typing import Dict, List, Optional, Tuple + +from flask import Flask, jsonify, redirect, render_template, request, send_file, url_for + +app = Flask(__name__) + +# Global state +DATASET_DIR = "" +DATASET_FILE = None +CURRENT_PDF = None +PDF_TESTS = {} +ALL_PDFS = [] +FORCE = False + + +def find_next_unchecked_pdf() -> Optional[str]: + """Find the next PDF with at least one unchecked test.""" + global PDF_TESTS, ALL_PDFS + + for pdf_name in ALL_PDFS: + pdf_tests = PDF_TESTS[pdf_name] + for test in pdf_tests: + if test.get("checked") is None: + return pdf_name + return None + + +def calculate_stats() -> dict: + """Calculate statistics for all tests in the dataset.""" + global PDF_TESTS + + total_tests = 0 + null_status = 0 + verified_status = 0 + rejected_status = 0 + + for pdf_tests in PDF_TESTS.values(): + total_tests += len(pdf_tests) + + for test in pdf_tests: + status = test.get("checked") + if status is None: + null_status += 1 + elif status == "verified": + verified_status += 1 + elif status == "rejected": + rejected_status += 1 + + completion = 0 + if total_tests > 0: + completion = (verified_status + rejected_status) / total_tests * 100 + + return {"total": total_tests, "null": null_status, "verified": verified_status, "rejected": rejected_status, "completion": completion} + + +def save_dataset(jsonl_file: str) -> None: + """Save the tests to a JSONL file, using temp file for atomic write.""" + global PDF_TESTS + + # Flatten all tests + all_tests = [] + for pdf_tests in PDF_TESTS.values(): + all_tests.extend(pdf_tests) + + # Create temp file and write updated content + with tempfile.NamedTemporaryFile(mode="w", delete=False) as temp_file: + for test in all_tests: + temp_file.write(json.dumps(test) + "\n") + + # Atomic replace + shutil.move(temp_file.name, jsonl_file) + + +@app.route("/pdf/") +def serve_pdf(pdf_name): + """Serve the PDF file directly.""" + pdf_path = os.path.join(DATASET_DIR, "pdfs", pdf_name) + return send_file(pdf_path, mimetype="application/pdf") + + +@app.route("/") +def index(): + """Main page displaying the current PDF and its tests.""" + global CURRENT_PDF, PDF_TESTS, DATASET_DIR, ALL_PDFS, FORCE + + # If no current PDF is set, find the next one with unchecked tests + if CURRENT_PDF is None: + CURRENT_PDF = find_next_unchecked_pdf() + + # If still no PDF, either show the "All done" page or force display the first PDF + if CURRENT_PDF is None: + if FORCE and ALL_PDFS: + CURRENT_PDF = ALL_PDFS[0] + else: + return render_template("all_done_latex.html") + + # Get the tests for the current PDF + current_tests = PDF_TESTS.get(CURRENT_PDF, []) + + # Create PDF URL for pdf.js to load + pdf_url = url_for("serve_pdf", pdf_name=CURRENT_PDF) + + # Calculate statistics + stats = calculate_stats() + + return render_template( + "review_latex.html", + pdf_name=CURRENT_PDF, + tests=current_tests, + pdf_path=pdf_url, + pdf_index=ALL_PDFS.index(CURRENT_PDF) if CURRENT_PDF in ALL_PDFS else 0, + total_pdfs=len(ALL_PDFS), + stats=stats, + ) + + +@app.route("/update_test", methods=["POST"]) +def update_test(): + """API endpoint to update a test.""" + global PDF_TESTS, DATASET_DIR, DATASET_FILE + + data = request.json + pdf_name = data.get("pdf") + test_id = data.get("id") + field = data.get("field") + value = data.get("value") + + # Find and update the test + for test in PDF_TESTS.get(pdf_name, []): + if test.get("id") == test_id: + test[field] = value + break + + # Save the updated tests + save_dataset(DATASET_FILE) + + return jsonify({"status": "success"}) + + +@app.route("/reject_all", methods=["POST"]) +def reject_all(): + """API endpoint to reject all tests for a PDF.""" + global PDF_TESTS, DATASET_DIR, DATASET_FILE + + data = request.json + pdf_name = data.get("pdf") + + if pdf_name and pdf_name in PDF_TESTS: + # Update all tests for this PDF to rejected + for test in PDF_TESTS[pdf_name]: + test["checked"] = "rejected" + + # Save the updated tests + save_dataset(DATASET_FILE) + + return jsonify({"status": "success", "count": len(PDF_TESTS[pdf_name])}) + + return jsonify({"status": "error", "message": "PDF not found"}) + + +@app.route("/next_pdf", methods=["POST"]) +def next_pdf(): + """Move to the next PDF in the list.""" + global CURRENT_PDF, ALL_PDFS, FORCE + + if CURRENT_PDF in ALL_PDFS: + current_index = ALL_PDFS.index(CURRENT_PDF) + if current_index < len(ALL_PDFS) - 1: + CURRENT_PDF = ALL_PDFS[current_index + 1] + else: + # If in force mode, cycle back to the beginning instead of checking for an unchecked PDF + if FORCE and ALL_PDFS: + CURRENT_PDF = ALL_PDFS[0] + else: + CURRENT_PDF = find_next_unchecked_pdf() + else: + if FORCE and ALL_PDFS: + CURRENT_PDF = ALL_PDFS[0] + else: + CURRENT_PDF = find_next_unchecked_pdf() + + return redirect(url_for("index")) + + +@app.route("/prev_pdf", methods=["POST"]) +def prev_pdf(): + """Move to the previous PDF in the list.""" + global CURRENT_PDF, ALL_PDFS + + if CURRENT_PDF in ALL_PDFS: + current_index = ALL_PDFS.index(CURRENT_PDF) + if current_index > 0: + CURRENT_PDF = ALL_PDFS[current_index - 1] + + return redirect(url_for("index")) + + +@app.route("/goto_pdf/", methods=["POST"]) +def goto_pdf(index): + """Go to a specific PDF by index.""" + global CURRENT_PDF, ALL_PDFS + + if 0 <= index < len(ALL_PDFS): + CURRENT_PDF = ALL_PDFS[index] + + return redirect(url_for("index")) + + +def load_dataset(dataset_file: str) -> Tuple[Dict[str, List[Dict]], List[str]]: + """Load tests from the dataset file and organize them by PDF.""" + if not os.path.exists(dataset_file): + raise FileNotFoundError(f"Dataset file not found: {dataset_file}") + + pdf_tests = defaultdict(list) + + with open(dataset_file, "r") as f: + for line in f: + line = line.strip() + if not line: + continue + + try: + test = json.loads(line) + pdf_name = test.get("pdf") + if pdf_name: + pdf_tests[pdf_name].append(test) + except json.JSONDecodeError: + print(f"Warning: Could not parse line as JSON: {line}") + + all_pdfs = list(pdf_tests.keys()) + + return pdf_tests, all_pdfs + + +def create_templates_directory(): + """Create templates directory for Flask if it doesn't exist.""" + templates_dir = os.path.join(os.path.dirname(__file__), "templates") + os.makedirs(templates_dir, exist_ok=True) + + # Create the review_latex.html template with MathJax support + review_html = """ + + + + + + + Equation Verification + + + + + + +

+ + + +

Equations ({{ tests|length }})

+ {% for test in tests %} + +

+ {{ test.text|safe }} +

+ + + + +

+ {% endfor %} +

+ + + + + """ + + # Create the all_done_latex.html template + all_done_html = """ + + + + + + All Done! + + + +

All Done! 🎉

You have reviewed all equations in the dataset.

+ + + +

+ + + """ + + with open(os.path.join(templates_dir, "review_latex.html"), "w") as f: + f.write(review_html) + + with open(os.path.join(templates_dir, "all_done_latex.html"), "w") as f: + f.write(all_done_html) + + +def main(): + """Main entry point with command-line arguments.""" + global DATASET_DIR, DATASET_FILE, PDF_TESTS, ALL_PDFS, CURRENT_PDF, FORCE + + parser = argparse.ArgumentParser(description="Interactive Test Review App") + parser.add_argument("dataset_file", help="Path to the dataset jsonl file") + parser.add_argument("--port", type=int, default=5000, help="Port for the Flask app") + parser.add_argument("--host", default="127.0.0.1", help="Host for the Flask app") + parser.add_argument("--debug", action="store_true", help="Run Flask in debug mode") + parser.add_argument("--force", action="store_true", help="Force show each file one by one and never do the 'All done' page") + + args = parser.parse_args() + FORCE = args.force + + if not os.path.exists(args.dataset_file): + print(f"Error: Dataset not found: {args.dataset_file}") + return 1 + + DATASET_DIR = os.path.dirname(os.path.abspath(args.dataset_file)) + DATASET_FILE = args.dataset_file + + pdf_dir = os.path.join(DATASET_DIR, "pdfs") + if not os.path.isdir(pdf_dir): + print(f"Error: PDF directory not found: {pdf_dir}") + return 1 + + try: + PDF_TESTS, ALL_PDFS = load_dataset(args.dataset_file) + except Exception as e: + print(f"Error loading dataset: {str(e)}") + return 1 + + create_templates_directory() + CURRENT_PDF = find_next_unchecked_pdf() + + print(f"Starting server at http://{args.host}:{args.port}") + app.run(host=args.host, port=args.port, debug=args.debug) + + return 0 + + +if __name__ == "__main__": + sys.exit(main()) diff --git a/olmocr/bench/runners/__init__.py b/olmocr/bench/runners/__init__.py new file mode 100644 index 0000000..e69de29 diff --git a/olmocr/bench/runners/run_chatgpt.py b/olmocr/bench/runners/run_chatgpt.py new file mode 100644 index 0000000..d66d1d7 --- /dev/null +++ b/olmocr/bench/runners/run_chatgpt.py @@ -0,0 +1,128 @@ +import json +import logging +import os +from typing import Literal + +from openai import OpenAI + +from olmocr.bench.prompts import ( + build_basic_prompt, + build_openai_silver_data_prompt_no_document_anchoring, +) +from olmocr.data.renderpdf import ( + get_png_dimensions_from_base64, + render_pdf_to_base64png, +) +from olmocr.prompts.anchor import get_anchor_text +from olmocr.prompts.prompts import ( + PageResponse, + build_finetuning_prompt, + build_openai_silver_data_prompt, + build_openai_silver_data_prompt_v2, + build_openai_silver_data_prompt_v2_simple, + build_openai_silver_data_prompt_v3_simple, + openai_response_format_schema, +) + +# Set up logger +logger = logging.getLogger(__name__) + +# Global variables to track token usage and document count +TOTAL_INPUT_TOKENS = 0 +TOTAL_OUTPUT_TOKENS = 0 +TOTAL_DOCUMENTS = 0 + + +def run_chatgpt( + pdf_path: str, + page_num: int = 1, + model: str = "gpt-4o-2024-08-06", + temperature: float = 0.1, + target_longest_image_dim: int = 2048, + max_completion_tokens: int = 10000, + prompt_template: Literal["full", "full_no_document_anchoring", "basic", "finetune", "fullv2", "fullv2simple", "fullv3simple"] = "finetune", + response_template: Literal["plain", "json"] = "json", +) -> str: + """ + Convert page of a PDF file to markdown using the commercial openAI APIs. + + See run_server.py for running against an openai compatible server + + Args: + pdf_path (str): The local path to the PDF file. + + Returns: + str: The OCR result in markdown format. + """ + global TOTAL_INPUT_TOKENS, TOTAL_OUTPUT_TOKENS, TOTAL_DOCUMENTS + # Convert the first page of the PDF to a base64-encoded PNG image. + image_base64 = render_pdf_to_base64png(pdf_path, page_num=page_num, target_longest_image_dim=target_longest_image_dim) + anchor_text = get_anchor_text(pdf_path, page_num, pdf_engine="pdfreport") + + if not os.getenv("OPENAI_API_KEY"): + raise SystemExit("You must specify an OPENAI_API_KEY") + + client = OpenAI(api_key=os.getenv("OPENAI_API_KEY")) + + if prompt_template == "full": + prompt = build_openai_silver_data_prompt(anchor_text) + elif prompt_template == "full_no_document_anchoring": + prompt = build_openai_silver_data_prompt_no_document_anchoring(anchor_text) + elif prompt_template == "finetune": + prompt = build_finetuning_prompt(anchor_text) + elif prompt_template == "basic": + prompt = build_basic_prompt() + elif prompt_template == "fullv2": + prompt = build_openai_silver_data_prompt_v2(anchor_text) + elif prompt_template == "fullv2simple": + width, height = get_png_dimensions_from_base64(image_base64) + prompt = build_openai_silver_data_prompt_v2_simple(width, height) + elif prompt_template == "fullv3simple": + width, height = get_png_dimensions_from_base64(image_base64) + prompt = build_openai_silver_data_prompt_v3_simple(width, height) + else: + raise ValueError("Unknown prompt template") + + response = client.chat.completions.create( + model=model, + messages=[ + { + "role": "user", + "content": [ + {"type": "text", "text": prompt}, + {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{image_base64}"}}, + ], + } + ], + temperature=temperature, + max_completion_tokens=max_completion_tokens, + # reasoning_effort="high", + response_format=openai_response_format_schema() if response_template == "json" else None, + safety_identifier="olmocr-bench-runner", + ) + + # Accumulate token counts from the response + if response.usage: + TOTAL_INPUT_TOKENS += response.usage.prompt_tokens + TOTAL_OUTPUT_TOKENS += response.usage.completion_tokens + + # Increment document counter + TOTAL_DOCUMENTS += 1 + + raw_response = response.choices[0].message.content + + assert len(response.choices) > 0 + assert response.choices[0].message.refusal is None + assert response.choices[0].finish_reason == "stop" + + if response_template == "json": + data = json.loads(raw_response) + data = PageResponse(**data) + + # Log token counts before returning + logger.warning(f"Token Usage - Documents: {TOTAL_DOCUMENTS}, Input: {TOTAL_INPUT_TOKENS}, Output: {TOTAL_OUTPUT_TOKENS}") + return data.natural_text + else: + # Log token counts before returning + logger.warning(f"Token Usage - Documents: {TOTAL_DOCUMENTS}, Input: {TOTAL_INPUT_TOKENS}, Output: {TOTAL_OUTPUT_TOKENS}") + return raw_response diff --git a/olmocr/bench/runners/run_claude.py b/olmocr/bench/runners/run_claude.py new file mode 100644 index 0000000..5ce3f1d --- /dev/null +++ b/olmocr/bench/runners/run_claude.py @@ -0,0 +1,61 @@ +import json +import os + +from anthropic import Anthropic +from prompts import build_openai_silver_data_prompt, claude_response_format_schema + +from olmocr.data.renderpdf import render_pdf_to_base64png +from olmocr.prompts.anchor import get_anchor_text + + +def run_claude(pdf_path: str, page_num: int = 1, model: str = "claude-3-7-sonnet-20250219", temperature: float = 0.1) -> str: + """ + Convert page of a PDF file to markdown using Claude OCR. + This function renders the specified page of the PDF to an image, runs OCR on that image, + and returns the OCR result as a markdown-formatted string. + + Args: + pdf_path (str): The local path to the PDF file. + page_num (int): The page number to process (starting from 1). + model (str): The Claude model to use. + temperature (float): The temperature parameter for generation. + + Returns: + str: The OCR result in markdown format. + """ + + if not os.getenv("ANTHROPIC_API_KEY"): + raise SystemExit("You must specify an ANTHROPIC_API_KEY") + + image_base64 = render_pdf_to_base64png(pdf_path, page_num=page_num, target_longest_image_dim=2048) + anchor_text = get_anchor_text(pdf_path, page_num, pdf_engine="pdfreport") + client = Anthropic(api_key=os.getenv("ANTHROPIC_API_KEY")) + response = client.messages.create( + model=model, + max_tokens=3000, + temperature=temperature, + # system=system_prompt, + tools=claude_response_format_schema(), + messages=[ + { + "role": "user", + "content": [ + {"type": "image", "source": {"type": "base64", "media_type": "image/png", "data": image_base64}}, + { + "type": "text", + "text": f"{build_openai_silver_data_prompt(anchor_text)}. Use the page_response tool to respond. If the propeties are true, then extract the text from them and respond in natural_text.", + }, + ], + } + ], + ) + + json_sentiment = None + for content in response.content: + if content.type == "tool_use" and content.name == "page_response": + json_sentiment = content.input + break + + if json_sentiment: + response = json.dumps(json_sentiment, indent=2) + return response diff --git a/olmocr/bench/runners/run_docling.py b/olmocr/bench/runners/run_docling.py new file mode 100644 index 0000000..b69d249 --- /dev/null +++ b/olmocr/bench/runners/run_docling.py @@ -0,0 +1,76 @@ +import asyncio +import os +import tempfile +from typing import Literal + +from pypdf import PdfReader, PdfWriter + + +async def run_docling( + pdf_path: str, + page_num: int = 1, + output_format: Literal["markdown"] = "markdown", + use_smoldocling: bool = False, +) -> str: + """Run docling CLI on a PDF file and return the results. + + Args: + pdf_path: Path to the PDF file + page_num: Page number to process (1-indexed) + output_format: Output format (only markdown is supported for CLI version) + + Returns: + String containing the markdown output + """ + if output_format != "markdown": + raise ValueError("Only markdown output format is supported for CLI version") + + # Extract the specific page using pypdf + pdf_reader = PdfReader(pdf_path) + pdf_writer = PdfWriter() + + # Convert from 1-indexed to 0-indexed + zero_based_page_num = page_num - 1 + + if zero_based_page_num >= len(pdf_reader.pages) or zero_based_page_num < 0: + raise ValueError(f"Page number {page_num} is out of bounds for PDF with {len(pdf_reader.pages)} pages") + + # Add the selected page to the writer + pdf_writer.add_page(pdf_reader.pages[zero_based_page_num]) + + # Create temporary files for the single-page PDF and output markdown + with tempfile.NamedTemporaryFile(suffix=".pdf", delete=False) as tmp_pdf_file, tempfile.NamedTemporaryFile(suffix=".md", delete=False) as tmp_md_file: + tmp_pdf_path = tmp_pdf_file.name + tmp_md_path = tmp_md_file.name + + try: + # Write the single-page PDF to the temporary file + with open(tmp_pdf_path, "wb") as f: + pdf_writer.write(f) + + # Build the command to run docling on the single-page PDF + if use_smoldocling: + cmd = ["docling", tmp_pdf_path, "-o", tmp_md_path] # Output file + else: + cmd = ["docling", "--pipeline", "vlm", "--vlm-model", "smoldocling", tmp_pdf_path, "-o", tmp_md_path] # Output file + + # Run the command asynchronously + proc = await asyncio.create_subprocess_exec(*cmd, stdout=asyncio.subprocess.PIPE, stderr=asyncio.subprocess.PIPE) + + stdout, stderr = await proc.communicate() + + if proc.returncode != 0: + error_msg = stderr.decode() if stderr else "Unknown error" + raise RuntimeError(f"docling command failed with return code {proc.returncode}: {error_msg}") + + # Read the results from the temporary markdown file + with open(tmp_md_path, "r", encoding="utf-8") as f: + result = f.read() + + return result + + finally: + # Clean up the temporary files + for path in [tmp_pdf_path, tmp_md_path]: + if os.path.exists(path): + os.unlink(path) diff --git a/olmocr/bench/runners/run_dotsocr.py b/olmocr/bench/runners/run_dotsocr.py new file mode 100644 index 0000000..aa3ebef --- /dev/null +++ b/olmocr/bench/runners/run_dotsocr.py @@ -0,0 +1,190 @@ +import asyncio +import os +import sys +import tempfile +from pathlib import Path +from typing import Dict, Optional, Tuple +from urllib.parse import urlparse + +from pypdf import PdfReader, PdfWriter + +REPO_ROOT = Path(__file__).resolve().parents[3] +DOTS_OCR_PATH = REPO_ROOT / "dots.ocr" + +if not DOTS_OCR_PATH.exists(): + raise ImportError(f"Could not find dots.ocr checkout at {DOTS_OCR_PATH}") + +if str(DOTS_OCR_PATH) not in sys.path: + sys.path.insert(0, str(DOTS_OCR_PATH)) + +from dots_ocr.parser import DotsOCRParser # noqa: E402 + +_PARSER_CACHE: Dict[ + Tuple[str, str, int, str, float, float, int, int, Optional[int], Optional[int], bool], + DotsOCRParser, +] = {} + + +def _parse_server(server: Optional[str]) -> Tuple[str, str, int]: + normalized = (server or "").strip() or "http://localhost:8000" + if "://" not in normalized: + normalized = f"http://{normalized}" + + parsed = urlparse(normalized) + if not parsed.scheme or not parsed.hostname: + raise ValueError(f"Invalid server URL: {server}") + + port = parsed.port + if port is None: + port = 443 if parsed.scheme == "https" else 80 + + return parsed.scheme, parsed.hostname, port + + +def _get_parser( + *, + protocol: str, + ip: str, + port: int, + model_name: str, + temperature: float, + top_p: float, + max_completion_tokens: int, + num_thread: int, + dpi: int, + min_pixels: Optional[int], + max_pixels: Optional[int], + use_hf: bool, +) -> DotsOCRParser: + cache_key = ( + protocol, + ip, + port, + model_name, + temperature, + top_p, + max_completion_tokens, + num_thread, + dpi, + min_pixels, + max_pixels, + use_hf, + ) + + parser = _PARSER_CACHE.get(cache_key) + if parser is None: + parser = DotsOCRParser( + protocol=protocol, + ip=ip, + port=port, + model_name=model_name, + temperature=temperature, + top_p=top_p, + max_completion_tokens=max_completion_tokens, + num_thread=num_thread, + dpi=dpi, + output_dir=tempfile.mkdtemp(prefix="dotsocr-cache-"), + min_pixels=min_pixels, + max_pixels=max_pixels, + use_hf=use_hf, + ) + _PARSER_CACHE[cache_key] = parser + + return parser + + +def _extract_markdown_from_pdf( + parser: DotsOCRParser, + pdf_path: str, + page_num: int, + prompt_mode: str, + fitz_preprocess: bool, +) -> str: + if page_num < 1: + raise ValueError("page_num must be >= 1") + + with open(pdf_path, "rb") as pdf_file: + reader = PdfReader(pdf_file) + zero_index = page_num - 1 + if zero_index >= len(reader.pages): + raise ValueError(f"Page {page_num} does not exist in {pdf_path}") + + writer = PdfWriter() + writer.add_page(reader.pages[zero_index]) + + with tempfile.TemporaryDirectory(prefix="dotsocr-") as tmp_dir: + single_page_pdf = os.path.join(tmp_dir, "page.pdf") + with open(single_page_pdf, "wb") as tmp_pdf: + writer.write(tmp_pdf) + + output_dir = os.path.join(tmp_dir, "output") + os.makedirs(output_dir, exist_ok=True) + + results = parser.parse_file( + single_page_pdf, + output_dir=output_dir, + prompt_mode=prompt_mode, + fitz_preprocess=fitz_preprocess, + ) + + if not results: + raise RuntimeError("DotsOCR did not return any results") + + # Single-page PDF means the first result corresponds to our request. + page_result = results[0] + md_path = page_result.get("md_content_path") or page_result.get("md_content_nohf_path") + + if not md_path: + raise RuntimeError(f"DotsOCR (prompt_mode={prompt_mode}) did not produce markdown output for page {page_num}") + + if not os.path.exists(md_path): + raise RuntimeError(f"DotsOCR reported markdown at '{md_path}', but the file does not exist") + + with open(md_path, "r", encoding="utf-8") as md_file: + return md_file.read() + + +async def run_dotsocr( + pdf_path: str, + page_num: int = 1, + server: Optional[str] = "http://localhost:8000", + prompt_mode: str = "prompt_layout_all_en", + model_name: str = "rednote-hilab/dots.ocr", + temperature: float = 0.1, + top_p: float = 1.0, + max_completion_tokens: int = 16384, + num_thread: int = 16, + dpi: int = 200, + min_pixels: Optional[int] = None, + max_pixels: Optional[int] = None, + use_hf: bool = False, + fitz_preprocess: bool = False, +) -> str: + """ + Run DotsOCR on a single PDF page and return the markdown output. + """ + + protocol, ip, port = _parse_server(server) + parser = _get_parser( + protocol=protocol, + ip=ip, + port=port, + model_name=model_name, + temperature=temperature, + top_p=top_p, + max_completion_tokens=max_completion_tokens, + num_thread=num_thread, + dpi=dpi, + min_pixels=min_pixels, + max_pixels=max_pixels, + use_hf=use_hf, + ) + + return await asyncio.to_thread( + _extract_markdown_from_pdf, + parser, + pdf_path, + page_num, + prompt_mode, + fitz_preprocess, + ) diff --git a/olmocr/bench/runners/run_gemini.py b/olmocr/bench/runners/run_gemini.py new file mode 100644 index 0000000..7857fc7 --- /dev/null +++ b/olmocr/bench/runners/run_gemini.py @@ -0,0 +1,163 @@ +import base64 +import json +import logging +import os +from typing import Literal + +from google import genai +from google.genai import types + +from olmocr.bench.prompts import ( + build_openai_silver_data_prompt_no_document_anchoring, +) +from olmocr.data.renderpdf import ( + get_png_dimensions_from_base64, + render_pdf_to_base64png, +) +from olmocr.prompts.anchor import get_anchor_text +from olmocr.prompts.prompts import ( + build_openai_silver_data_prompt, + build_openai_silver_data_prompt_v3_simple, +) + +# Set up logger +logger = logging.getLogger(__name__) + +# Global variables to track token usage and documents +TOTAL_INPUT_TOKENS = 0 +TOTAL_OUTPUT_TOKENS = 0 +TOTAL_DOCUMENTS = 0 + + +def run_gemini( + pdf_path: str, + page_num: int = 1, + model: str = "gemini-2.0-flash", + temperature: float = 0.1, + target_longest_image_dim: int = 2048, + prompt_template: Literal["full", "full_no_document_anchoring", "basic", "finetune", "fullv3simple"] = "finetune", + response_template: Literal["plain", "json"] = "json", +) -> str: + """ + Convert page of a PDF file to markdown using Gemini's vision capabilities. + This function renders the specified page of the PDF to an image, runs OCR on that image, + and returns the OCR result as a markdown-formatted string. + + Args: + pdf_path (str): The local path to the PDF file. + page_num (int): The page number to process (starting from 1). + model (str): The Gemini model to use. + temperature (float): The temperature parameter for generation. + + Returns: + str: The OCR result in markdown format. + """ + global TOTAL_INPUT_TOKENS, TOTAL_OUTPUT_TOKENS, TOTAL_DOCUMENTS + TOTAL_DOCUMENTS += 1 + if not os.getenv("GEMINI_API_KEY"): + raise SystemExit("You must specify an GEMINI_API_KEY") + + image_base64 = render_pdf_to_base64png(pdf_path, page_num=page_num, target_longest_image_dim=2048) + anchor_text = get_anchor_text(pdf_path, page_num, pdf_engine="pdfreport") + api_key = os.getenv("GEMINI_API_KEY") + client = genai.Client(api_key=api_key) + image_part = types.Part(inline_data=types.Blob(mime_type="image/png", data=base64.b64decode(image_base64))) + + if prompt_template == "full": + text_part = types.Part(text=f"""{build_openai_silver_data_prompt(anchor_text)}""") + elif prompt_template == "full_no_document_anchoring": + text_part = types.Part(text=f"""{build_openai_silver_data_prompt_no_document_anchoring(anchor_text)}""") + elif prompt_template == "fullv3simple": + width, height = get_png_dimensions_from_base64(image_base64) + prompt = build_openai_silver_data_prompt_v3_simple(width, height) + text_part = types.Part(text=prompt) + else: + raise NotImplementedError() + + if response_template == "json": + generation_config = types.GenerateContentConfig( + temperature=temperature, + top_p=1.0, + top_k=32, + max_output_tokens=10000, + response_mime_type="application/json", + response_schema=genai.types.Schema( + type=genai.types.Type.OBJECT, + required=["primary_language", "is_rotation_valid", "rotation_correction", "is_table", "is_diagram", "natural_text"], + properties={ + "primary_language": genai.types.Schema( + type=genai.types.Type.STRING, + ), + "is_rotation_valid": genai.types.Schema( + type=genai.types.Type.BOOLEAN, + ), + "rotation_correction": genai.types.Schema( + type=genai.types.Type.STRING, + enum=["0", "90", "180", "270"], + ), + "is_table": genai.types.Schema( + type=genai.types.Type.BOOLEAN, + ), + "is_diagram": genai.types.Schema( + type=genai.types.Type.BOOLEAN, + ), + "natural_text": genai.types.Schema( + type=genai.types.Type.STRING, + ), + }, + ), + ) + + response = client.models.generate_content( + model=f"models/{model}", + contents=[types.Content(parts=[image_part, text_part])], + config=generation_config, + ) + + assert len(response.candidates) > 0, "No candidates found" + assert response.candidates[0].finish_reason == types.FinishReason.STOP, "Finish reason was not STOP, likely a processing error or repetition failure" + + # Extract token counts from usage metadata + if hasattr(response, "usage_metadata"): + input_tokens = getattr(response.usage_metadata, "prompt_token_count", 0) + output_tokens = getattr(response.usage_metadata, "candidates_token_count", 0) + TOTAL_INPUT_TOKENS += input_tokens + TOTAL_OUTPUT_TOKENS += output_tokens + + result = response.candidates[0].content.parts[0].text + parsed = json.loads(result) + + # The json schema is slightly off with gemini vs chatgpt, so we don't verify it + logger.warning( + f"[Before Return - JSON] Total Documents: {TOTAL_DOCUMENTS}, Total Input Tokens: {TOTAL_INPUT_TOKENS}, Total Output Tokens: {TOTAL_OUTPUT_TOKENS}" + ) + return parsed["natural_text"] + else: + generation_config = types.GenerateContentConfig( + temperature=temperature, + top_p=1.0, + top_k=32, + max_output_tokens=4096, + ) + + response = client.models.generate_content( + model=f"models/{model}", + contents=[types.Content(parts=[image_part, text_part])], + config=generation_config, + ) + + assert len(response.candidates) > 0, "No candidates found" + assert response.candidates[0].finish_reason == types.FinishReason.STOP, "Finish reason was not STOP, likely a processing error or repetition failure" + + # Extract token counts from usage metadata + if hasattr(response, "usage_metadata"): + input_tokens = getattr(response.usage_metadata, "prompt_token_count", 0) + output_tokens = getattr(response.usage_metadata, "candidates_token_count", 0) + TOTAL_INPUT_TOKENS += input_tokens + TOTAL_OUTPUT_TOKENS += output_tokens + + result = response.candidates[0].content.parts[0].text + logger.warning( + f"[Before Return - Plain] Total Documents: {TOTAL_DOCUMENTS}, Total Input Tokens: {TOTAL_INPUT_TOKENS}, Total Output Tokens: {TOTAL_OUTPUT_TOKENS}" + ) + return result diff --git a/olmocr/bench/runners/run_gotocr.py b/olmocr/bench/runners/run_gotocr.py new file mode 100644 index 0000000..8cae98e --- /dev/null +++ b/olmocr/bench/runners/run_gotocr.py @@ -0,0 +1,67 @@ +import base64 +import os +import tempfile + +import torch +from transformers import AutoModel, AutoTokenizer + +from olmocr.data.renderpdf import render_pdf_to_base64png + +# Global cache for the model and tokenizer. +_device = "cuda" if torch.cuda.is_available() else "cpu" +_model = None +_tokenizer = None + + +def load_model(): + """ + Load the GOT-OCR model and tokenizer if they haven't been loaded already. + Returns: + model: The GOT-OCR model loaded on the appropriate device. + tokenizer: The corresponding tokenizer. + """ + global _model, _tokenizer + if _model is None or _tokenizer is None: + _tokenizer = AutoTokenizer.from_pretrained("ucaslcl/GOT-OCR2_0", trust_remote_code=True) + _model = AutoModel.from_pretrained( + "ucaslcl/GOT-OCR2_0", + trust_remote_code=True, + use_safetensors=True, + revision="979938bf89ccdc949c0131ddd3841e24578a4742", + pad_token_id=_tokenizer.eos_token_id, + ) + _model = _model.eval().to(_device) + return _model, _tokenizer + + +def run_gotocr(pdf_path: str, page_num: int = 1, ocr_type: str = "ocr") -> str: + """ + Convert page of a PDF file to markdown using GOT-OCR. + + This function renders the first page of the PDF to an image, runs OCR on that image, + and returns the OCR result as a markdown-formatted string. + + Args: + pdf_path (str): The local path to the PDF file. + + Returns: + str: The OCR result in markdown format. + """ + # Ensure the model is loaded (cached across calls) + model, tokenizer = load_model() + + # Convert the first page of the PDF to a base64-encoded PNG image. + base64image = render_pdf_to_base64png(pdf_path, page_num=page_num, target_longest_image_dim=1024) + + # Write the image to a temporary file. + with tempfile.NamedTemporaryFile("wb", suffix=".png", delete=False) as tmp: + tmp.write(base64.b64decode(base64image)) + tmp_filename = tmp.name + + # Run GOT-OCR on the saved image. + result = model.chat(tokenizer, tmp_filename, ocr_type=ocr_type) + + # Clean up the temporary file. + os.remove(tmp_filename) + + return result diff --git a/olmocr/bench/runners/run_marker.py b/olmocr/bench/runners/run_marker.py new file mode 100644 index 0000000..86bd75f --- /dev/null +++ b/olmocr/bench/runners/run_marker.py @@ -0,0 +1,69 @@ +import os +import tempfile + +from marker.config.parser import ConfigParser +from marker.converters.pdf import PdfConverter +from marker.models import create_model_dict +from marker.output import text_from_rendered +from pypdf import PdfReader, PdfWriter + +_marker_converter = None + + +def run_marker(pdf_path: str, page_num: int = 1) -> str: + global _marker_converter + + if _marker_converter is None: + # Create a configuration dictionary with the necessary settings + config = { + "force_ocr": True, # This enables conversion of inline math to LaTeX + "use_llm": False, # We would prefer to run just plain marker for reporting bench results, not hybrid mode + "disable_tqdm": True, # Disable tqdm for cleaner output + "recognition_batch_size": 256, + "layout_batch_size": 48, + "detection_batch_size": 48, + "equation_batch_size": 64, + "table_rec_batch_size": 48, + "ocr_error_batch_size": 64, + } + config_parser = ConfigParser(config) + + _marker_converter = PdfConverter( + artifact_dict=create_model_dict(), + config=config_parser.generate_config_dict(), + ) + + # Extract the specific page from the PDF + pdf_to_process = pdf_path + temp_file = None + + if page_num > 0: # If a specific page is requested + reader = PdfReader(pdf_path) + + # Check if the requested page exists + if page_num > len(reader.pages): + raise ValueError(f"Page {page_num} does not exist in the PDF. PDF has {len(reader.pages)} pages.") + + # Create a new PDF with just the requested page + writer = PdfWriter() + # pypdf uses 0-based indexing, so subtract 1 from page_num + writer.add_page(reader.pages[page_num - 1]) + + # Save the extracted page to a temporary file + temp_file = tempfile.NamedTemporaryFile(suffix=".pdf", delete=False) + temp_file.close() # Close the file but keep the name + + with open(temp_file.name, "wb") as output_pdf: + writer.write(output_pdf) + + pdf_to_process = temp_file.name + + try: + # Process the PDF (either original or single-page extract) + rendered = _marker_converter(pdf_to_process) + text, _, images = text_from_rendered(rendered) + return text + finally: + # Clean up the temporary file if it was created + if temp_file and os.path.exists(temp_file.name): + os.unlink(temp_file.name) diff --git a/olmocr/bench/runners/run_mineru.py b/olmocr/bench/runners/run_mineru.py new file mode 100644 index 0000000..583bffb --- /dev/null +++ b/olmocr/bench/runners/run_mineru.py @@ -0,0 +1,74 @@ +import os +import tempfile + +from magic_pdf.config.enums import SupportedPdfParseMethod +from magic_pdf.data.data_reader_writer import FileBasedDataReader, FileBasedDataWriter +from magic_pdf.data.dataset import PymuDocDataset +from magic_pdf.model.doc_analyze_by_custom_model import doc_analyze +from pypdf import PdfReader, PdfWriter + + +def run_mineru(pdf_path: str, page_num: int = 1) -> str: + output_folder = tempfile.TemporaryDirectory() + image_output_folder = tempfile.TemporaryDirectory() + + # Initialize writers (same for all PDFs) + image_writer = FileBasedDataWriter(image_output_folder.name) + md_writer = FileBasedDataWriter(output_folder.name) + + if page_num > 0: # If a specific page is requested + reader = PdfReader(pdf_path) + + # Check if the requested page exists + if page_num > len(reader.pages): + raise ValueError(f"Page {page_num} does not exist in the PDF. PDF has {len(reader.pages)} pages.") + + # Create a new PDF with just the requested page + writer = PdfWriter() + # pypdf uses 0-based indexing, so subtract 1 from page_num + writer.add_page(reader.pages[page_num - 1]) + + # Save the extracted page to a temporary file + temp_file = tempfile.NamedTemporaryFile(suffix=".pdf", delete=False) + temp_file.close() # Close the file but keep the name + + with open(temp_file.name, "wb") as output_pdf: + writer.write(output_pdf) + + pdf_to_process = temp_file.name + else: + pdf_to_process = pdf_path + + try: + # Read the PDF file bytes + reader = FileBasedDataReader("") + pdf_bytes = reader.read(pdf_to_process) + + # Create dataset instance + ds = PymuDocDataset(pdf_bytes) + + # Inference: decide whether to run OCR mode based on dataset classification + if ds.classify() == SupportedPdfParseMethod.OCR: + infer_result = ds.apply(doc_analyze, ocr=True) + pipe_result = infer_result.pipe_ocr_mode(image_writer) + else: + infer_result = ds.apply(doc_analyze, ocr=False) + pipe_result = infer_result.pipe_txt_mode(image_writer) + + # Generate markdown content; the image directory is the basename of the images output folder + image_dir_basename = os.path.basename(image_output_folder.name) + # md_content = pipe_result.get_markdown(image_dir_basename) + + # Dump markdown file + with tempfile.NamedTemporaryFile("w+", suffix="md") as tf: + pipe_result.dump_md(md_writer, tf.name, image_dir_basename) + tf.flush() + + tf.seek(0) + md_data = tf.read() + + return md_data + finally: + # Clean up the temporary file if it was created + if temp_file and os.path.exists(temp_file.name): + os.unlink(temp_file.name) diff --git a/olmocr/bench/runners/run_mistral.py b/olmocr/bench/runners/run_mistral.py new file mode 100644 index 0000000..20f20ac --- /dev/null +++ b/olmocr/bench/runners/run_mistral.py @@ -0,0 +1,74 @@ +import os +import tempfile + +from mistralai import Mistral +from pypdf import PdfReader, PdfWriter + + +def run_mistral(pdf_path: str, page_num: int = 1) -> str: + """ + Convert page of a PDF file to markdown using the mistral OCR api + https://docs.mistral.ai/capabilities/document/ + + Args: + pdf_path (str): The local path to the PDF file. + + Returns: + str: The OCR result in markdown format. + """ + if not os.getenv("MISTRAL_API_KEY"): + raise SystemExit("You must specify an MISTRAL_API_KEY") + + api_key = os.environ["MISTRAL_API_KEY"] + client = Mistral(api_key=api_key) + + if page_num > 0: # If a specific page is requested + reader = PdfReader(pdf_path) + + # Check if the requested page exists + if page_num > len(reader.pages): + raise ValueError(f"Page {page_num} does not exist in the PDF. PDF has {len(reader.pages)} pages.") + + # Create a new PDF with just the requested page + writer = PdfWriter() + # pypdf uses 0-based indexing, so subtract 1 from page_num + writer.add_page(reader.pages[page_num - 1]) + + # Save the extracted page to a temporary file + temp_file = tempfile.NamedTemporaryFile(suffix=".pdf", delete=False) + temp_file.close() # Close the file but keep the name + + with open(temp_file.name, "wb") as output_pdf: + writer.write(output_pdf) + + pdf_to_process = temp_file.name + else: + pdf_to_process = pdf_path + + try: + with open(pdf_to_process, "rb") as pf: + uploaded_pdf = client.files.upload( + file={ + "file_name": os.path.basename(pdf_path), + "content": pf, + }, + purpose="ocr", + ) + + signed_url = client.files.get_signed_url(file_id=uploaded_pdf.id) + + ocr_response = client.ocr.process( + model="mistral-ocr-2503", + document={ + "type": "document_url", + "document_url": signed_url.url, + }, + ) + + client.files.delete(file_id=uploaded_pdf.id) + + return ocr_response.pages[0].markdown + finally: + # Clean up the temporary file if it was created + if temp_file and os.path.exists(temp_file.name): + os.unlink(temp_file.name) diff --git a/olmocr/bench/runners/run_nanonetsocr.py b/olmocr/bench/runners/run_nanonetsocr.py new file mode 100644 index 0000000..22d8579 --- /dev/null +++ b/olmocr/bench/runners/run_nanonetsocr.py @@ -0,0 +1,89 @@ +import base64 +import os +import re +import tempfile + +import torch +from PIL import Image +from transformers import AutoModelForImageTextToText, AutoProcessor, AutoTokenizer + +from olmocr.data.renderpdf import render_pdf_to_base64png + +_model = None +_tokenizer = None +_processor = None +_device = None + + +def load_model(model_path: str = "nanonets/Nanonets-OCR-s"): + global _model, _tokenizer, _processor, _device + + if _model is None: + _device = "cuda" if torch.cuda.is_available() else "cpu" + _model = AutoModelForImageTextToText.from_pretrained( + model_path, + torch_dtype="auto", + device_map="auto", + # attn_implementation="flash_attention_2" + ) + _model.eval() + _tokenizer = AutoTokenizer.from_pretrained(model_path) + _processor = AutoProcessor.from_pretrained(model_path) + + return _model, _tokenizer, _processor + + +async def run_nanonetsocr(pdf_path: str, page_num: int = 1, model_path: str = "nanonets/Nanonets-OCR-s", max_new_tokens: int = 4096, **kwargs) -> str: + """ + Convert page of a PDF file to markdown using NANONETS-OCR. + + This function renders the first page of the PDF to an image, runs OCR on that image, + and returns the OCR result as a markdown-formatted string. + + Args: + pdf_path (str): The local path to the PDF file. + + Returns: + str: The OCR result in markdown format. + """ + + model, tokenizer, processor = load_model(model_path) + + image_base64 = render_pdf_to_base64png(pdf_path, page_num=page_num, target_longest_image_dim=1024) + + with tempfile.NamedTemporaryFile(suffix=".png", delete=False) as temp_file: + image_data = base64.b64decode(image_base64) + temp_file.write(image_data) + temp_image_path = temp_file.name + + try: + image = Image.open(temp_image_path) + prompt = """Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document and image caption is not present, add a small description of the image inside the tag; otherwise, add the image caption inside . Watermarks should be wrapped in brackets. Ex: OFFICIAL COPY. Page numbers should be wrapped in brackets. Ex: 14 or 9/22. Prefer using ☐ and ☑ for check boxes.""" + messages = [ + {"role": "system", "content": "You are a helpful assistant."}, + { + "role": "user", + "content": [ + {"type": "image", "image": f"file://{temp_image_path}"}, + {"type": "text", "text": prompt}, + ], + }, + ] + text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True) + + inputs = processor(text=[text], images=[image], padding=True, return_tensors="pt", use_fast=True) + inputs = inputs.to(model.device) + with torch.no_grad(): + output_ids = model.generate(**inputs, max_new_tokens=max_new_tokens, do_sample=False) + + generated_ids = [output_ids[len(input_ids) :] for input_ids, output_ids in zip(inputs.input_ids, output_ids)] + output_text = processor.batch_decode(generated_ids, skip_special_tokens=True, clean_up_tokenization_spaces=True) + cleaned_text = re.sub(r"\d+", "", output_text[0]) + + return cleaned_text + + finally: + try: + os.unlink(temp_image_path) + except Exception as e: + print(f"Warning: Failed to remove temporary file {temp_image_path}: {e}") diff --git a/olmocr/bench/runners/run_nanonetsocr_2.py b/olmocr/bench/runners/run_nanonetsocr_2.py new file mode 100644 index 0000000..f3bdc6f --- /dev/null +++ b/olmocr/bench/runners/run_nanonetsocr_2.py @@ -0,0 +1,86 @@ +import base64 +import subprocess + +import httpx + +from olmocr.data.renderpdf import get_pdf_media_box_width_height + + +# Logic to set min size from here: https://github.com/NanoNets/Nanonets-OCR2/blob/main/Nanonets-OCR2-Cookbook/image2md.ipynb +def render_pdf_to_base64png_min_short_size(local_pdf_path: str, page_num: int, target_shortest_dim: int = 2048) -> str: + shortest_dim = min(get_pdf_media_box_width_height(local_pdf_path, page_num)) + + # Convert PDF page to PNG using pdftoppm + pdftoppm_result = subprocess.run( + [ + "pdftoppm", + "-png", + "-f", + str(page_num), + "-l", + str(page_num), + "-r", + str(target_shortest_dim * 72 / shortest_dim), # 72 pixels per point is the conversion factor + local_pdf_path, + ], + timeout=120, + stdout=subprocess.PIPE, + stderr=subprocess.PIPE, + ) + assert pdftoppm_result.returncode == 0, pdftoppm_result.stderr + return base64.b64encode(pdftoppm_result.stdout).decode("utf-8") + + +async def run_server( + pdf_path: str, + page_num: int = 1, + server: str = "localhost:30000", + model: str = "nanonets/Nanonets-OCR2-3B", + temperature: float = 0.0, + page_dimensions: int = 1280, +) -> str: + """ + Convert page of a PDF file to markdown by calling a request + running against an openai compatible server. + + You can use this for running against vllm, sglang, servers + as well as mixing and matching different model's. + + It will only make one direct request, with no retries or error checking. + + Returns: + str: The OCR result in markdown format. + """ + # Convert the first page of the PDF to a base64-encoded PNG image. + image_base64 = render_pdf_to_base64png_min_short_size(pdf_path, page_num=page_num, target_shortest_dim=page_dimensions) + + # Now use th + prompt = """Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document and image caption is not present, add a small description of the image inside the tag; otherwise, add the image caption inside . Watermarks should be wrapped in brackets. Ex: OFFICIAL COPY. Page numbers should be wrapped in brackets. Ex: 14 or 9/22. Prefer using ☐ and ☑ for check boxes.""" + + request = { + "model": model, + "messages": [ + {"role": "system", "content": "You are a helpful assistant."}, + { + "role": "user", + "content": [ + {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{image_base64}"}}, + {"type": "text", "text": prompt}, + ], + }, + ], + "temperature": temperature, + "max_tokens": 4096, + } + + # Make request and get response using httpx + url = f"http://{server}/v1/chat/completions" + + async with httpx.AsyncClient(timeout=300) as client: + response = await client.post(url, json=request) + + response.raise_for_status() + data = response.json() + + choice = data["choices"][0] + return choice["message"]["content"] diff --git a/olmocr/bench/runners/run_olmocr_pipeline.py b/olmocr/bench/runners/run_olmocr_pipeline.py new file mode 100644 index 0000000..d85c324 --- /dev/null +++ b/olmocr/bench/runners/run_olmocr_pipeline.py @@ -0,0 +1,112 @@ +import asyncio +import logging +from dataclasses import dataclass +from typing import Optional + +# Import necessary components from olmocr +from olmocr.pipeline import ( + MetricsKeeper, + PageResult, + WorkerTracker, + process_page, + vllm_server_host, + vllm_server_ready, +) + +# Setup basic logging +logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(name)s - %(levelname)s - %(message)s") +logger = logging.getLogger("olmocr_runner") + + +# Basic configuration +@dataclass +class Args: + model: str = "allenai/olmOCR-2-7B-1025-FP8" + server: str = "http://localhost:30044/v1" + port: int = 30044 + model_chat_template: str = "qwen2-vl" + max_model_len: int = 16384 + guided_decoding: bool = False + gpu_memory_utilization: float = 0.8 + target_longest_image_dim: int = 1288 + target_anchor_text_len: int = -1 + max_page_retries: int = 8 + max_page_error_rate: float = 0.004 + tensor_parallel_size: int = 1 + data_parallel_size: int = 1 + + +server_check_lock = asyncio.Lock() + + +async def run_olmocr_pipeline(pdf_path: str, page_num: int = 1, model: str = "allenai/olmOCR-2-7B-1025-FP8") -> Optional[str]: + """ + Process a single page of a PDF using the official olmocr pipeline's process_page function + + Args: + pdf_path: Path to the PDF file + page_num: Page number to process (1-indexed) + + Returns: + The extracted text from the page or None if processing failed + """ + # Ensure global variables are initialized + global metrics, tracker + if "metrics" not in globals() or metrics is None: + metrics = MetricsKeeper(window=60 * 5) + if "tracker" not in globals() or tracker is None: + tracker = WorkerTracker() + + args = Args() + args.model = model + semaphore = asyncio.Semaphore(1) + worker_id = 0 # Using 0 as default worker ID + + # Ensure server is running + async with server_check_lock: + _server_task = None + try: + await asyncio.wait_for(vllm_server_ready(args), timeout=5) + logger.info("Using existing vllm server") + except Exception: + logger.info("Starting new vllm server") + _server_task = asyncio.create_task(vllm_server_host(args.model, args, semaphore)) + await vllm_server_ready(args) + + # Sets the model name used in the pipeline code, it's a hack sadly + args.model = "olmocr" + + try: + # Process the page using the pipeline's process_page function + # Note: process_page expects both original path and local path + # In our case, we're using the same path for both + page_result: PageResult = await process_page(args=args, worker_id=worker_id, pdf_orig_path=pdf_path, pdf_local_path=pdf_path, page_num=page_num) + + # Return the natural text from the response + if page_result and page_result.response and not page_result.is_fallback: + return page_result.response.natural_text + return None + + except Exception as e: + logger.error(f"Error processing page: {type(e).__name__} - {str(e)}") + return None + + finally: + # We leave the server running for potential reuse + pass + + +async def main(): + # Example usage + pdf_path = "your_pdf_path.pdf" + page_num = 1 + + result = await run_olmocr_pipeline(pdf_path, page_num) + if result: + print(f"Extracted text: {result[:200]}...") # Print first 200 chars + else: + print("Failed to extract text from the page") + + +if __name__ == "__main__": + asyncio.run(main()) diff --git a/olmocr/bench/runners/run_paddlepaddle.py b/olmocr/bench/runners/run_paddlepaddle.py new file mode 100644 index 0000000..8968b6b --- /dev/null +++ b/olmocr/bench/runners/run_paddlepaddle.py @@ -0,0 +1,31 @@ +from threading import Lock + +from paddleocr import PPStructureV3 + +# Run's paddle paddle as in the docs here: https://huggingface.co/PaddlePaddle/PP-OCRv5_server_det +# text_detection_model_name="PP-OCRv5_server_det", +# and using the PP-StructureV3 pipeline to create markdown +paddle_pipeline = None +paddle_pipeline_lock = Lock() + + +def run_paddlepaddle(pdf_path: str, page_num: int = 1, **kwargs) -> str: + global paddle_pipeline + + with paddle_pipeline_lock: + if paddle_pipeline is None: + paddle_pipeline = PPStructureV3( + text_detection_model_name="PP-OCRv5_server_det", + use_doc_orientation_classify=False, # Use use_doc_orientation_classify to enable/disable document orientation classification model + use_doc_unwarping=False, # Use use_doc_unwarping to enable/disable document unwarping module + use_textline_orientation=False, # Use use_textline_orientation to enable/disable textline orientation classification model + device="gpu:0", # Use device to specify GPU for model inference + ) + + output = paddle_pipeline.predict(pdf_path) + result = "" + for cur_page_0_indexed, res in enumerate(output): + if cur_page_0_indexed == page_num - 1: + result = res.markdown["markdown_texts"] + + return result diff --git a/olmocr/bench/runners/run_paddlevl.py b/olmocr/bench/runners/run_paddlevl.py new file mode 100644 index 0000000..e2c6635 --- /dev/null +++ b/olmocr/bench/runners/run_paddlevl.py @@ -0,0 +1,23 @@ +from threading import Lock + +from paddleocr import PaddleOCRVL + +# Using docs from here: https://huggingface.co/PaddlePaddle/PaddleOCR-VL +paddle_pipeline = None +paddle_pipeline_lock = Lock() + + +def run_paddlevl(pdf_path: str, page_num: int = 1, **kwargs) -> str: + global paddle_pipeline + + with paddle_pipeline_lock: + if paddle_pipeline is None: + paddle_pipeline = PaddleOCRVL() + + output = paddle_pipeline.predict(pdf_path) + result = "" + for cur_page_0_indexed, res in enumerate(output): + if cur_page_0_indexed == page_num - 1: + result = res.markdown["markdown_texts"] + + return result diff --git a/olmocr/bench/runners/run_rolmocr.py b/olmocr/bench/runners/run_rolmocr.py new file mode 100644 index 0000000..71b7352 --- /dev/null +++ b/olmocr/bench/runners/run_rolmocr.py @@ -0,0 +1,58 @@ +import httpx + +from olmocr.data.renderpdf import render_pdf_to_base64png + + +async def run_rolmocr( + pdf_path: str, + page_num: int = 1, + server: str = "localhost:30000", + model: str = "reducto/RolmOCR", + temperature: float = 0.2, + target_longest_image_dim: int = 1024, +) -> str: + """ + + + Returns: + str: The OCR result in markdown format. + """ + # Convert the first page of the PDF to a base64-encoded PNG image. + image_base64 = render_pdf_to_base64png(pdf_path, page_num=page_num, target_longest_image_dim=target_longest_image_dim) + + request = { + "model": model, + "messages": [ + { + "role": "user", + "content": [ + { + "type": "image_url", + "image_url": {"url": f"data:image/png;base64,{image_base64}"}, + }, + { + "type": "text", + "text": "Return the plain text representation of this document as if you were reading it naturally.\n", + }, + ], + } + ], + "temperature": temperature, + "max_tokens": 4096, + } + + # Make request and get response using httpx + url = f"http://{server}/v1/chat/completions" + + async with httpx.AsyncClient(timeout=300) as client: + response = await client.post(url, json=request) + + response.raise_for_status() + data = response.json() + + choice = data["choices"][0] + assert ( + choice["finish_reason"] == "stop" + ), "Response from server did not finish with finish_reason stop as expected, this is probably going to lead to bad data" + + return choice["message"]["content"] diff --git a/olmocr/bench/runners/run_server.py b/olmocr/bench/runners/run_server.py new file mode 100644 index 0000000..5e7f86a --- /dev/null +++ b/olmocr/bench/runners/run_server.py @@ -0,0 +1,100 @@ +import json +from typing import Literal + +import httpx + +from olmocr.bench.prompts import ( + build_basic_prompt, + build_openai_silver_data_prompt_no_document_anchoring, +) +from olmocr.data.renderpdf import ( + get_png_dimensions_from_base64, + render_pdf_to_base64png, +) +from olmocr.prompts.anchor import get_anchor_text +from olmocr.prompts.prompts import ( + PageResponse, + build_finetuning_prompt, + build_no_anchoring_v4_yaml_prompt, + build_openai_silver_data_prompt, + build_openai_silver_data_prompt_v3_simple, +) + + +async def run_server( + pdf_path: str, + page_num: int = 1, + endpoint: str = "http://localhost:8000/v1", + model: str = "allenai/olmOCR-7B-0225-preview", + temperature: float = 0.1, + target_longest_image_dim: int = 1024, + prompt_template: Literal["full", "full_no_document_anchoring", "fullv3simple", "finetune_v4_yaml", "basic", "finetune"] = "fullv3simple", + response_template: Literal["plain", "json"] = "plain", +) -> str: + """ + Convert page of a PDF file to markdown by calling a request + running against an openai compatible server. + + You can use this for running against vllm, sglang, servers + as well as mixing and matching different model's. + + It will only make one direct request, with no retries or error checking. + + Returns: + str: The OCR result in markdown format. + """ + # Convert the first page of the PDF to a base64-encoded PNG image. + image_base64 = render_pdf_to_base64png(pdf_path, page_num=page_num, target_longest_image_dim=target_longest_image_dim) + anchor_text = get_anchor_text(pdf_path, page_num, pdf_engine="pdfreport") + + if prompt_template == "full": + prompt = build_openai_silver_data_prompt(anchor_text) + elif prompt_template == "full_no_document_anchoring": + prompt = build_openai_silver_data_prompt_no_document_anchoring(anchor_text) + elif prompt_template == "finetune": + prompt = build_finetuning_prompt(anchor_text) + elif prompt_template == "basic": + prompt = build_basic_prompt() + elif prompt_template == "finetune_v4_yaml": + prompt = build_no_anchoring_v4_yaml_prompt() + elif prompt_template == "fullv3simple": + width, height = get_png_dimensions_from_base64(image_base64) + prompt = build_openai_silver_data_prompt_v3_simple(width, height) + else: + raise ValueError("Unknown prompt template") + + request = { + "model": model, + "messages": [ + { + "role": "user", + "content": [ + {"type": "text", "text": prompt}, + {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{image_base64}"}}, + ], + } + ], + "temperature": temperature, + "max_tokens": 8000, + } + + # Make request and get response using httpx + url = f"{endpoint.rstrip('/')}/chat/completions" + + async with httpx.AsyncClient(timeout=300) as client: + response = await client.post(url, json=request) + + response.raise_for_status() + data = response.json() + + choice = data["choices"][0] + assert ( + choice["finish_reason"] == "stop" + ), "Response from server did not finish with finish_reason stop as expected, this is probably going to lead to bad data" + + if response_template == "json": + page_data = json.loads(choice["message"]["content"]) + page_response = PageResponse(**page_data) + return page_response.natural_text + elif response_template == "plain": + return choice["message"]["content"] diff --git a/olmocr/bench/runners/run_transformers.py b/olmocr/bench/runners/run_transformers.py new file mode 100644 index 0000000..70bd58c --- /dev/null +++ b/olmocr/bench/runners/run_transformers.py @@ -0,0 +1,129 @@ +import base64 +import json +from io import BytesIO +from typing import Literal + +import torch +from PIL import Image +from transformers import ( + AutoProcessor, + Qwen2_5_VLForConditionalGeneration, +) + +from olmocr.data.renderpdf import render_pdf_to_base64png +from olmocr.prompts.anchor import get_anchor_text +from olmocr.prompts.prompts import ( + PageResponse, + build_finetuning_prompt, + build_no_anchoring_yaml_prompt, + build_openai_silver_data_prompt, +) +from olmocr.train.front_matter import FrontMatterParser + +_cached_model = None +_cached_processor = None + + +def run_transformers( + pdf_path: str, + page_num: int = 1, + model_name: str = "allenai/olmOCR-7B-0725-FP8", + temperature: float = 0.1, + target_longest_image_dim: int = 1024, + prompt_template: Literal["full", "finetune", "yaml"] = "yaml", + response_template: Literal["plain", "json", "yaml"] = "yaml", +) -> str: + """ + Convert page of a PDF file to markdown by calling a request + running against an openai compatible server. + + You can use this for running against vllm, sglang, servers + as well as mixing and matching different model's. + + It will only make one direct request, with no retries or error checking. + + Returns: + str: The OCR result in markdown format. + """ + # Initialize the model + global _cached_model, _cached_processor + device = torch.device("cuda" if torch.cuda.is_available() else "cpu") + + if _cached_model is None: + model = Qwen2_5_VLForConditionalGeneration.from_pretrained( + model_name, torch_dtype=torch.bfloat16, device_map="auto", attn_implementation="flash_attention_2" + ).eval() + processor = AutoProcessor.from_pretrained(model_name) + + model = model.to(device) + + _cached_model = model + _cached_processor = processor + else: + model = _cached_model + processor = _cached_processor + + # Convert the first page of the PDF to a base64-encoded PNG image. + image_base64 = render_pdf_to_base64png(pdf_path, page_num=page_num, target_longest_image_dim=target_longest_image_dim) + + if prompt_template == "yaml": + prompt = build_no_anchoring_yaml_prompt() + else: + anchor_text = get_anchor_text(pdf_path, page_num, pdf_engine="pdfreport") + if prompt_template == "full": + prompt = build_openai_silver_data_prompt(anchor_text) + else: + prompt = build_finetuning_prompt(anchor_text) + + messages = [ + { + "role": "user", + "content": [ + {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{image_base64}"}}, + {"type": "text", "text": prompt}, + ], + } + ] + + # Apply the chat template and processor + text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True) + main_image = Image.open(BytesIO(base64.b64decode(image_base64))) + + inputs = processor( + text=[text], + images=[main_image], + padding=True, + return_tensors="pt", + ) + inputs = {key: value.to(device) for (key, value) in inputs.items()} + + # Generate the output + MAX_NEW_TOKENS = 3000 + with torch.no_grad(): + output = model.generate( + **inputs, + temperature=temperature, + max_new_tokens=MAX_NEW_TOKENS, + num_return_sequences=1, + do_sample=True, + ) + + # Decode the output + prompt_length = inputs["input_ids"].shape[1] + new_tokens = output[:, prompt_length:] + text_output = processor.tokenizer.batch_decode(new_tokens, skip_special_tokens=True)[0] + + assert new_tokens.shape[1] < MAX_NEW_TOKENS, "Output exceed max new tokens" + + if response_template == "json": + page_data = json.loads(text_output) + page_response = PageResponse(**page_data) + return page_response.natural_text if page_response.natural_text else "" + elif response_template == "yaml": + # Parse YAML front matter and extract natural text + parser = FrontMatterParser(front_matter_class=PageResponse) + front_matter, text = parser._extract_front_matter_and_text(text_output) + page_response = parser._parse_front_matter(front_matter, text) + return page_response.natural_text if page_response.natural_text else "" + elif response_template == "plain": + return text_output diff --git a/olmocr/bench/sample_data/blanks.jsonl b/olmocr/bench/sample_data/blanks.jsonl new file mode 100644 index 0000000..7d4641c --- /dev/null +++ b/olmocr/bench/sample_data/blanks.jsonl @@ -0,0 +1 @@ +{"pdf": "blank_book_pg1.pdf", "page": 1, "id": "test1_blank", "type": "baseline", "checked": "verified", "max_length": 10} \ No newline at end of file diff --git a/olmocr/bench/sample_data/dataset.jsonl b/olmocr/bench/sample_data/dataset.jsonl new file mode 100644 index 0000000..17799cc --- /dev/null +++ b/olmocr/bench/sample_data/dataset.jsonl @@ -0,0 +1,117 @@ +{"pdf": "multi_column_miss.pdf", "page": 1, "id": "multi_column_miss_00", "type": "present", "text": "Corporate social responsibility and the tobacco industry: hope or hype?"} +{"pdf": "multi_column_miss.pdf", "page": 1, "id": "multi_column_miss_01", "type": "present", "text": "this leaves BAT to argue why it should not be held to be largely accountable for the annual deaths of some 754 600 smokers, and Philip Morris some 803 600 smokers."} +{"pdf": "multi_column_miss.pdf", "page": 1, "id": "multi_column_miss_02", "type": "present", "text": "The term \"corporate social responsibility\" is in vogue at the moment but as a concept it is vague and means different things to different people.", "max_diffs": 2} +{"pdf": "multi_column_miss.pdf", "page": 1, "id": "multi_column_miss_03", "type": "present", "text": "Over the past three decades increasing pressure from non-governmental"} +{"pdf": "multi_column_miss.pdf", "page": 1, "id": "multi_column_miss_04", "type": "absent", "text": "Downloaded from http://tobaccocontrol.bmj.com/"} + +{"pdf": "multi_column_miss.pdf", "page": 1, "id": "multi_column_miss_10", "type": "order", "before": "Corporate social responsibility and the tobacco industry: hope or hype?", "after": "The unprecedented expansion of power and influence of TNCs over the past three decades has accelerated global trade and development, but also environmental damage and abuses of", "max_diffs": 2} +{"pdf": "multi_column_miss.pdf", "page": 1, "id": "multi_column_miss_11", "type": "order", "before": "It now looks like that with vigilance", "after": "this leaves BAT to argue why it should not be held to be largely accountable for the annual deaths", "max_diffs": 2} +{"pdf": "multi_column_miss.pdf", "page": 1, "id": "multi_column_miss_12", "type": "order", "before": "Corporate social responsibility (CSR) emerged from a realisation among transnational corporations", "after": " perspective on its own behaviour; and reflects on whether marketing tobacco is antithetical to social responsibility.", "max_diffs": 2} + +{"pdf": "discoverworld_crazy_table4.pdf", "page": 1, "id": "discoverworld_crazy_table4_00", "type": "present", "text": "Table 4: Baseline model performance on each of the three scoring metrics"} +{"pdf": "discoverworld_crazy_table4.pdf", "page": 1, "id": "discoverworld_crazy_table4_01", "type": "present", "text": "Table 5: Baseline model performance on each of the three scoring metrics"} +{"pdf": "discoverworld_crazy_table4.pdf", "page": 1, "id": "discoverworld_crazy_table4_02", "type": "present", "text": "We use the GPT-4O model for all our agents due to its higher performance and lower cost compared to other models. For space we provide"} + +{"pdf": "mattsnotes.pdf", "page": 1, "id": "mattsnotes_minediff_00", "type": "present", "checked": "verified", "text": "The-Stack-V2"} +{"pdf": "mattsnotes.pdf", "page": 1, "id": "mattsnotes_minediff_01", "type": "present", "checked": "verified", "text": "SE, whatever we've scraped"} +{"pdf": "mattsnotes.pdf", "page": 1, "id": "mattsnotes_minediff_02", "type": "present", "checked": "verified", "text": "HQ DCLM"} +{"pdf": "mattsnotes.pdf", "page": 2, "id": "mattsnotes_minediff_03", "type": "present", "checked": "verified", "text": "Order by repo"} +{"pdf": "mattsnotes.pdf", "page": 3, "id": "mattsnotes_minediff_04", "type": "present", "checked": "verified", "text": "ARCH + TRAINING"} + +{"pdf": "buildingnotes.pdf", "page": 1, "id": "building_notes_00", "type": "present", "checked": "verified", "text": "Master Bath", "case_sensitive": false} +{"pdf": "buildingnotes.pdf", "page": 1, "id": "building_notes_01", "type": "present", "checked": "verified", "text": "Laundry", "case_sensitive": false} +{"pdf": "buildingnotes.pdf", "page": 1, "id": "building_notes_02", "type": "present", "checked": "verified", "text": "Guest Bath", "case_sensitive": false} + +{"pdf": "lincoln_letter.pdf", "page": 1, "id": "lincoln_letter_minediff_00", "type": "present", "checked": "verified", "text": "January 10th 1864."} +{"pdf": "lincoln_letter.pdf", "page": 1, "id": "lincoln_letter_minediff_01", "type": "present", "checked": "verified", "text": "Major General Hitchcock, Commissioner of Exchanges, is authorized and directed to offer Brigadier General Trimble, now a prisoner of war in Fort McHenry, in exchange for Major White, who is held as a prisoner at Richmond."} +{"pdf": "lincoln_letter.pdf", "page": 1, "id": "lincoln_letter_minediff_03", "type": "present", "checked": "verified", "text": "He is also directed to send forward the offer of exchange by Henry M. Warfield, Esq. of Baltimore, under a flag of truce, and give him a pass to City Point."} + +{"pdf": "openstax_caculus_pg_273.pdf", "page": 1, "id": "openstax_caculus_pg_273_minediff_02", "type": "present", "checked": "verified", "text": "Use the graph of the position function to determine the time intervals when the velocity is positive, negative, or zero."} +{"pdf": "openstax_caculus_pg_273.pdf", "page": 1, "id": "openstax_caculus_pg_273_minediff_03", "type": "present", "checked": "verified", "text": "Use the graph of the velocity function to determine the time intervals when the acceleration is positive, negative, or zero."} +{"pdf": "openstax_caculus_pg_273.pdf", "page": 1, "id": "openstax_caculus_pg_273_minediff_04", "type": "order", "before": "150.", "after": "157."} +{"pdf": "openstax_caculus_pg_273.pdf", "page": 1, "id": "openstax_caculus_pg_273_minediff_05", "type": "order", "before": "150.", "after": "158."} +{"pdf": "openstax_caculus_pg_273.pdf", "page": 1, "id": "openstax_caculus_pg_273_minediff_06", "type": "order", "before": "150.", "after": "159."} + +{"pdf": "multi_column_miss.pdf", "page": 1, "id": "multi_column_miss_minediff_01", "type": "present", "checked": "verified", "text": "This report first provides the context and development of CSR; then, from internal company documents, examines how PM came to its own version."} +{"pdf": "multi_column_miss.pdf", "page": 1, "id": "multi_column_miss_minediff_02", "type": "present", "checked": "verified", "text": "This paper examines whether a tobacco company espousing CSR should be judged simply as a corporate entity along standards of business ethics, or as an irretrievably negative force in the realm of public health, thereby rendering CSR an oxymoron."} + +{"pdf": "olmo2-pg4.pdf", "page": 1, "id": "olmo2-pg4_minediff_00", "type": "present", "checked": "verified", "text": "Table 1 Composition of the pretraining data for OLMo 2."} + +{"pdf": "olmo2-pg4.pdf", "page": 1, "id": "olmo2-pg4_table00", "type": "table", "cell": "Type"} +{"pdf": "olmo2-pg4.pdf", "page": 1, "id": "olmo2-pg4_table01", "type": "table", "cell": "3.32T", "left": "3.71T"} +{"pdf": "olmo2-pg4.pdf", "page": 1, "id": "olmo2-pg4_table02", "type": "table", "cell": "3.32T", "right": "21.32T"} +{"pdf": "olmo2-pg4.pdf", "page": 1, "id": "olmo2-pg4_table03", "type": "table", "cell": "11.8B", "up": "12.2B"} +{"pdf": "olmo2-pg4.pdf", "page": 1, "id": "olmo2-pg4_table04", "type": "table", "cell": "11.8B", "down": "3.7B"} +{"pdf": "olmo2-pg4.pdf", "page": 1, "id": "olmo2-pg4_table05", "type": "table", "cell": "3.32T", "top_heading": "Words"} +{"pdf": "olmo2-pg4.pdf", "page": 1, "id": "olmo2-pg4_table06", "type": "table", "cell": "arXiv", "top_heading": "Source"} +{"pdf": "olmo2-pg4.pdf", "page": 1, "id": "olmo2-pg4_table07", "type": "table", "cell": "47.2B", "top_heading": "Bytes"} +{"pdf": "olmo2-pg4.pdf", "page": 1, "id": "olmo2-pg4_table08", "type": "table", "cell": "Math proofs code", "left_heading": "Algebraic Stack"} + +{"pdf": "discoverworld_crazy_table4.pdf", "page": 1, "id": "olmo2-discoverworld_crazy_table4_t00", "type": "table", "cell": "Quadratic regression", "left": "Challenge"} +{"pdf": "discoverworld_crazy_table4.pdf", "page": 1, "id": "olmo2-discoverworld_crazy_table4_t01", "type": "table", "cell": "Instrument Use", "left": "Normal"} +{"pdf": "discoverworld_crazy_table4.pdf", "page": 1, "id": "olmo2-discoverworld_crazy_table4_t02", "type": "table", "cell": "0.87", "top_heading": "Procedure"} +{"pdf": "discoverworld_crazy_table4.pdf", "page": 1, "id": "olmo2-discoverworld_crazy_table4_t03", "type": "table", "cell": "0.87", "top_heading": "ReACT"} +{"pdf": "discoverworld_crazy_table4.pdf", "page": 1, "id": "olmo2-discoverworld_crazy_table4_t04", "type": "table", "cell": "Pick-and-place object", "left_heading": "27"} +{"pdf": "discoverworld_crazy_table4.pdf", "page": 1, "id": "olmo2-discoverworld_crazy_table4_t05", "type": "table", "cell": "0.66", "right": "0.44"} +{"pdf": "discoverworld_crazy_table4.pdf", "page": 1, "id": "olmo2-discoverworld_crazy_table4_t06", "type": "table", "cell": "Interact with a moving agent", "top_heading": "Unit Test Topic"} + +{"pdf": "earnings.pdf", "page": 1, "id": "earnings_table00", "type": "table", "cell": "1,136", "top_heading": "Year Ended"} +{"pdf": "earnings.pdf", "page": 1, "id": "earnings_table01", "type": "table", "cell": "Year Ended"} +{"pdf": "earnings.pdf", "page": 1, "id": "earnings_table02", "type": "table", "cell": "680", "up": "1,892"} +{"pdf": "earnings.pdf", "page": 1, "id": "earnings_table03", "type": "table", "cell": "2,532", "left_heading": "Research and development"} +{"pdf": "earnings.pdf", "page": 1, "id": "earnings_table04", "type": "absent", "text": "62"} + + +{"pdf": "mathfuncs.pdf", "page": 1, "id": "mathfuncs_00", "type": "order", "before": "Euler's Identity", "after": "Pythagorean Theorem"} +{"pdf": "mathfuncs.pdf", "page": 1, "id": "mathfuncs_01", "type": "order", "before": "Pythagorean Theorem", "after": "The Fundamental Theorem of Calculus"} +{"pdf": "mathfuncs.pdf", "page": 1, "id": "mathfuncs_02", "type": "order", "before": "The Fundamental Theorem of Calculus", "after": "Maxwell's Equations"} +{"pdf": "mathfuncs.pdf", "page": 1, "id": "mathfuncs_03", "type": "math", "math": "e^{i \\pi}+1=0"} +{"pdf": "mathfuncs.pdf", "page": 1, "id": "mathfuncs_04", "type": "math", "math": "\\int_{a}^{b} f(x) d x=F(b)-F(a)"} +{"pdf": "mathfuncs.pdf", "page": 1, "id": "mathfuncs_05", "type": "math", "math": "a^{2}+b^{2}=c^{2}"} +{"pdf": "mathfuncs.pdf", "page": 1, "id": "mathfuncs_06", "type": "math", "math": "\\nabla \\times \\mathbf{E}=-\\frac{\\partial \\mathbf{B}}{\\partial t}"} + +{"pdf": "mathfuncs_colswitch.pdf", "page": 1, "id": "mathfuncscol_00", "type": "order", "before": "Euler's Identity", "after": "Pythagorean Theorem"} +{"pdf": "mathfuncs_colswitch.pdf", "page": 1, "id": "mathfuncscol_01", "type": "order", "before": "Pythagorean Theorem", "after": "The Fundamental Theorem of Calculus"} +{"pdf": "mathfuncs_colswitch.pdf", "page": 1, "id": "mathfuncscol_02", "type": "order", "before": "The Fundamental Theorem of Calculus", "after": "Maxwell's Equations"} +{"pdf": "mathfuncs_colswitch.pdf", "page": 1, "id": "mathfuncscol_03", "type": "math", "math": "e^{i \\pi}+1=0"} +{"pdf": "mathfuncs_colswitch.pdf", "page": 1, "id": "mathfuncscol_04", "type": "math", "math": "\\int_{a}^{b} f(x) d x=F(b)-F(a)"} +{"pdf": "mathfuncs_colswitch.pdf", "page": 1, "id": "mathfuncscol_05", "type": "math", "math": "a^{2}+b^{2}=c^{2}"} +{"pdf": "mathfuncs_colswitch.pdf", "page": 1, "id": "mathfuncscol_06", "type": "math", "math": "\\nabla \\times \\mathbf{E}=-\\frac{\\partial \\mathbf{B}}{\\partial t}"} + +{"pdf": "math_2503_04086.pdf", "page": 1, "id": "math_2503_04086_00", "type": "math", "math": "\\lambda_{g}=\\sum_{s \\in S} \\zeta_{n}^{\\psi(g s)}=\\sum_{i=1}^{k}\\left[\\sum_{s, R s=\\mathcal{I}_{i}} \\zeta_{n}^{\\psi(g s)}\\right]"} +{"pdf": "math_2503_04086.pdf", "page": 1, "id": "math_2503_04086_01", "type": "math", "math": "\\lambda_{g}=\\lambda_{g^{\\prime}}"} +{"pdf": "math_2503_04086.pdf", "page": 1, "id": "math_2503_04086_02", "type": "math", "math": "u \\in\\left(R / \\operatorname{Ann}_{R}\\left(x_{i}\\right)\\right)^{\\times}"} +{"pdf": "math_2503_04086.pdf", "page": 1, "id": "math_2503_04086_03", "type": "math", "math": "\\lambda_{g}=\\sum_{i=1}^{k} c\\left(g, R / \\operatorname{Ann}_{R}\\left(x_{i}\\right)\\right)"} +{"pdf": "math_2503_04086.pdf", "page": 1, "id": "math_2503_04086_04", "type": "present", "text": "We also thank Ján Mináč for his constant encouragement and support."} +{"pdf": "math_2503_04086.pdf", "page": 1, "id": "math_2503_04086_05", "type": "present", "text": "Allgemeine theorie der Gaußschen Summen in endlichen kommutativen Ringe"} + +{"pdf": "test-graphical-text.pdf", "page": 1, "id": "test_graphical_text_00", "type": "present", "text": "THE VISION"} +{"pdf": "test-graphical-text.pdf", "page": 1, "id": "test_graphical_text_01", "type": "present", "text": "by FutureSkill"} +{"pdf": "test-graphical-text.pdf", "page": 1, "id": "test_graphical_text_02", "type": "present", "text": "THE POWER OF STORYTELLING"} + +{"pdf": "small_page_size.pdf", "page": 1, "id": "small_page_size_00", "type": "present", "text": "Since the use of bones has, however, become general, the turnip crop has been, in many instances, ten-fold, and in few less than four or five-fold its former bulk"} +{"pdf": "small_page_size.pdf", "page": 1, "id": "small_page_size_01", "type": "order", "before": "Earthy and saline matter", "after": "Cartilage and jelly"} +{"pdf": "small_page_size.pdf", "page": 1, "id": "small_page_size_02", "type": "present", "text": "The cartilage, indeed, when the bones have been buried in a dry situation, is very indestructible"} +{"pdf": "small_page_size.pdf", "page": 1, "id": "small_page_size_03", "type": "absent", "text": "BRITISH HUSBANDRY"} +{"pdf": "small_page_size.pdf", "page": 1, "id": "small_page_size_04", "type": "absent", "text": "400"} +{"pdf": "small_page_size.pdf", "page": 1, "id": "small_page_size_05", "type": "absent", "text": "XIX"} + +{"pdf": "headers_footers/ff0f0b22c55d8b90dd77d153f48e144fc9db_pg2.pdf", "page": 1, "id": "ff0f0b22c55d8b90dd77d153f48e144fc9db_02a", "type": "absent", "text": "PLOS Neglected Tropical Diseases | www.plosntds.org"} +{"pdf": "headers_footers/ff0f0b22c55d8b90dd77d153f48e144fc9db_pg2.pdf", "page": 1, "id": "ff0f0b22c55d8b90dd77d153f48e144fc9db_02b", "type": "absent", "text": "March 2014 | Volume 8 | Issue 3 | e2748"} +{"pdf": "headers_footers/ff0f0b22c55d8b90dd77d153f48e144fc9db_pg2.pdf", "page": 1, "id": "ff0f0b22c55d8b90dd77d153f48e144fc9db_02c", "type": "absent", "text": "1", "last_n": 20} +{"pdf": "headers_footers/ff0f0b22c55d8b90dd77d153f48e144fc9db_pg2.pdf", "page": 1, "id": "ff0f0b22c55d8b90dd77d153f48e144fc9db_02d", "type": "absent", "text": "OPEN ACCESS Freely available online"} +{"pdf": "headers_footers/ff0f0b22c55d8b90dd77d153f48e144fc9db_pg2.pdf", "page": 1, "id": "ff0f0b22c55d8b90dd77d153f48e144fc9db_02e", "type": "absent", "text": "PLOS NEGLECTED TROPICAL DISEASES"} +{"pdf": "headers_footers/ff1fc6a205ad039139ce566851b6b260c929_pg1.pdf", "page": 1, "id": "ff1fc6a205ad039139ce566851b6b260c929_01a", "type": "absent", "text": "TELEDYNE ENERGY SYSTEMS, INC. A Teledyne Technologies Company"} +{"pdf": "headers_footers/ff1fc6a205ad039139ce566851b6b260c929_pg1.pdf", "page": 1, "id": "ff1fc6a205ad039139ce566851b6b260c929_01b", "type": "absent", "text": "AEROJET ROCKETDYNE"} +{"pdf": "headers_footers/ff1fc6a205ad039139ce566851b6b260c929_pg1.pdf", "page": 1, "id": "ff1fc6a205ad039139ce566851b6b260c929_01c", "type": "absent", "text": "1", "last_n": 20} +{"pdf": "headers_footers/ff3d6e051903fe5ca9bc172ece14964c5632_pg1.pdf", "page": 1, "id": "ff3d6e051903fe5ca9bc172ece14964c5632_01a", "type": "absent", "text": "farbod4ever@gmail.com :\u0646\u0648\u064a\u0633\u0646\u062f\u0647 \u0631\u0627\u0628\u0637", "max_diffs": 4} +{"pdf": "headers_footers/ff3d6e051903fe5ca9bc172ece14964c5632_pg1.pdf", "page": 1, "id": "ff3d6e051903fe5ca9bc172ece14964c5632_01b", "type": "absent", "text": "Downloaded from jipm.irandoc.ac.ir at 6:51 IRST on Monday November 11th 2019"} +{"pdf": "headers_footers/ff4f7dad78081cff727d19ab51c181d4a661_pg1.pdf", "page": 1, "id": "ff4f7dad78081cff727d19ab51c181d4a661_01a", "type": "absent", "text": "Download date: 28 Dec 2018"} +{"pdf": "headers_footers/ff518b1240a66978f22035528ccb029450b5_pg2.pdf", "page": 1, "id": "ff518b1240a66978f22035528ccb029450b5_02a", "type": "absent", "text": "Woodworth et al.: Brief Notices"} +{"pdf": "headers_footers/ff518b1240a66978f22035528ccb029450b5_pg2.pdf", "page": 1, "id": "ff518b1240a66978f22035528ccb029450b5_02b", "type": "absent", "text": "Published by BYU ScholarsArchive, 1997"} +{"pdf": "headers_footers/ff518b1240a66978f22035528ccb029450b5_pg2.pdf", "page": 1, "id": "ff518b1240a66978f22035528ccb029450b5_02c", "type": "absent", "text": "199", "last_n": 20} +{"pdf": "headers_footers/ffaac214730d2b8c2ec842e3618ccb9c4259_pg1.pdf", "page": 1, "id": "ffaac214730d2b8c2ec842e3618ccb9c4259_01a", "type": "absent", "text": "TILBURG UNIVERSITY"} +{"pdf": "headers_footers/ffaac214730d2b8c2ec842e3618ccb9c4259_pg1.pdf", "page": 1, "id": "ffaac214730d2b8c2ec842e3618ccb9c4259_01b", "type": "absent", "text": "Download date: 15. Sep. 2023"} +{"pdf": "headers_footers/fff590bed29a2854ac1f874dad5752ede1aa_pg1.pdf", "page": 1, "id": "fff590bed29a2854ac1f874dad5752ede1aa_01a", "type": "absent", "text": "Revision: 2.4"} +{"pdf": "headers_footers/fff590bed29a2854ac1f874dad5752ede1aa_pg1.pdf", "page": 1, "id": "fff590bed29a2854ac1f874dad5752ede1aa_01b", "type": "absent", "text": "P/N 119-036"} +{"pdf": "headers_footers/fff590bed29a2854ac1f874dad5752ede1aa_pg1.pdf", "page": 1, "id": "fff590bed29a2854ac1f874dad5752ede1aa_01c", "type": "absent", "text": "10 June 2019"} \ No newline at end of file diff --git a/olmocr/bench/sample_data/olmocr_pipeline/blank_book_pg1_pg1_repeat1.md b/olmocr/bench/sample_data/olmocr_pipeline/blank_book_pg1_pg1_repeat1.md new file mode 100644 index 0000000..e69de29 diff --git a/olmocr/bench/sample_data/olmocr_pipeline/buildingnotes_pg1_repeat1.md b/olmocr/bench/sample_data/olmocr_pipeline/buildingnotes_pg1_repeat1.md new file mode 100644 index 0000000..8878ea7 --- /dev/null +++ b/olmocr/bench/sample_data/olmocr_pipeline/buildingnotes_pg1_repeat1.md @@ -0,0 +1,11 @@ +Master - 7 1/4 - 36" +Master Bath - 7 1/4 - 30" +Laundry - 4 3/4 - 36" +Bath - 7 1/4 - 24" +MUD - 7 - 36" +UTIL - 8 1/4 - 36" +DOWN BATH - 7 1/4 - 32" +BUT KIT - 6 3/4 - 30 +PANTRY - 4 3/4 - 24 +6 WEST - 32 9/8 - 32 +6 WEST BATH 5" - 24" \ No newline at end of file diff --git a/olmocr/bench/sample_data/olmocr_pipeline/discoverworld_crazy_table4_pg1_repeat1.md b/olmocr/bench/sample_data/olmocr_pipeline/discoverworld_crazy_table4_pg1_repeat1.md new file mode 100644 index 0000000..401ea71 --- /dev/null +++ b/olmocr/bench/sample_data/olmocr_pipeline/discoverworld_crazy_table4_pg1_repeat1.md @@ -0,0 +1,51 @@ +Table 4: Baseline model performance on each of the three scoring metrics (task completion, task process, explanatory knowledge discovery) across all 24 DISCOVERY WORLD tasks. Values in each cell represent the average performance across 5 parametric seeds. Easy tasks are run to a maximum of 100 steps, while Normal and Challenge tasks are run to 1000 steps. + +| # | Topic | Task | ReACT | Plan+Execute | Hypothesizer | +|---|---|---|---|---|---| +| 1 | Proteomics | Clustering | 0.87 | 0.20 | 0.20 | 0.80 | 0.00 | 0.00 | 0.90 | 0.40 | 1.00 | +| 2 | Chemistry | Exploring Combinations and Hill Climbing | 0.88 | 0.40 | 0.40 | 0.68 | 0.20 | 0.00 | 0.93 | 0.40 | 0.40 | +| 3 | Archaeology | Correlations | 0.88 | 0.40 | 0.60 | 0.55 | 0.20 | 0.00 | 0.93 | 0.40 | 0.60 | +| 4 | Easy | Single substances | 0.87 | 1.00 | 1.00 | 0.70 | 0.60 | 0.40 | 0.90 | 0.00 | 0.40 | +| 5 | Normal | Mix of 3 substances | 0.82 | 0.00 | 0.00 | 0.87 | 0.40 | 0.00 | 0.93 | 0.60 | 0.40 | +| 6 | Challenge | Mix of 4 substances | 0.90 | 0.40 | 0.00 | 0.90 | 0.40 | 0.00 | 0.97 | 0.00 | 0.00 | +| 7 | Easy | Simple instrument | 0.27 | 0.60 | 0.00 | 0.33 | 0.20 | 0.00 | 0.60 | 0.20 | 0.50 | +| 8 | Normal | Instrument Use | 0.72 | 0.40 | 0.30 | 0.74 | 0.00 | 0.00 | 0.64 | 0.40 | 0.40 | +| 9 | Challenge | Correlation | 0.46 | 0.20 | 0.00 | 0.46 | 0.00 | 0.05 | 0.55 | 0.20 | 0.05 | +| 10 | Reactor Lab | Regression | 0.42 | 0.00 | 0.40 | 0.44 | 0.00 | 0.10 | 0.38 | 0.00 | 0.20 | +| 11 | Normal | Linear regression | 0.44 | 0.00 | 0.20 | 0.49 | 0.00 | 0.00 | 0.51 | 0.00 | 0.00 | +| 12 | Challenge | Quadratic regression | 0.43 | 0.00 | 0.20 | 0.39 | 0.00 | 0.00 | 0.39 | 0.00 | 0.00 | +| 13 | Easy | Simplified rules | 0.80 | 0.20 | 0.20 | 0.70 | 0.20 | 0.20 | 0.60 | 0.00 | 0.00 | +| 14 | Normal | Presence rules | 0.91 | 0.60 | 0.00 | 0.84 | 0.40 | 0.00 | 0.56 | 0.00 | 0.00 | +| 15 | Challenge | Logical Rules | 0.89 | 0.40 | 0.00 | 0.73 | 0.40 | 0.00 | 0.62 | 0.00 | 0.00 | +| 16 | Easy | Open-ended discovery | 0.78 | 0.60 | 0.00 | 0.68 | 0.40 | 0.10 | 0.80 | 1.00 | 0.60 | +| 17 | Normal | Multiple instruments | 0.58 | 0.00 | 0.13 | 0.45 | 0.00 | 0.13 | 0.16 | 0.00 | 0.33 | +| 18 | Challenge | Novel instruments | 0.55 | 0.00 | 0.00 | 0.26 | 0.00 | 0.00 | 0.20 | 0.00 | 0.00 | +| 19 | Easy | Look-up variables | 0.33 | 0.00 | 0.00 | 0.53 | 0.00 | 0.07 | 0.13 | 0.40 | 0.00 | +| 20 | Normal | Measure 2 variables | 0.51 | 0.00 | 0.05 | 0.34 | 0.00 | 0.00 | 0.11 | 0.00 | 0.00 | +| 21 | Challenge | Measure 5 variables | 0.43 | 0.00 | 0.00 | 0.15 | 0.00 | 0.00 | 0.22 | 0.00 | 0.03 | +| 22 | Easy | Rosetta-stone style linguistic discovery of alien language | 0.40 | 0.40 | 0.20 | 0.30 | 0.00 | 0.00 | 0.20 | 0.20 | 0.00 | +| 23 | Normal | Noun and verb | 0.20 | 0.00 | 0.00 | 0.68 | 0.40 | 0.00 | 0.84 | 0.40 | 0.00 | +| 24 | Challenge | Noun, adj., and verb | 0.49 | 0.00 | 0.00 | 0.55 | 0.20 | 0.05 | 0.15 | 0.00 | 0.00 | +| Average (Easy) | 0.59 | 0.38 | 0.25 | 0.56 | 0.18 | 0.11 | 0.56 | 0.28 | 0.34 | +| Average (Normal) | 0.63 | 0.18 | 0.14 | 0.64 | 0.18 | 0.02 | 0.58 | 0.23 | 0.19 | +| Average (Challenge) | 0.63 | 0.18 | 0.10 | 0.50 | 0.15 | 0.01 | 0.49 | 0.08 | 0.08 | + +Table 5: Baseline model performance on each of the three scoring metrics (task completion, task process, explanatory knowledge discovery) across all 10 unit test tasks. Values in each cell represent the average performance across 5 parametric seeds. Unit tests tasks are run to a maximum of 100 steps. + +| # | Unit Test Topic | ReACT | Plan+Execute | Hypothesizer | +|---|---|---|---|---| +| 25 | Multi-turn dialog with an agent | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | +| 26 | Measure an object with an instrument | 0.87 | 0.60 | 0.73 | 0.40 | 1.00 | 1.00 | +| 27 | Pick-and-place object | 0.90 | 0.80 | 0.80 | 0.60 | 1.00 | 1.00 | +| 28 | Pick-and-give object | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | +| 29 | Read DiscoveryFeed posts | 1.00 | 1.00 | 0.90 | 0.80 | 1.00 | 1.00 | +| 30 | Move through doors | 0.58 | 0.20 | 0.25 | 0.00 | 0.30 | 0.00 | +| 31 | Using keys with doors | 0.69 | 0.20 | 0.54 | 0.00 | 0.69 | 0.00 | +| 32 | Navigate to a specific room in a house | 0.20 | 0.20 | 0.20 | 0.00 | 0.20 | 0.20 | +| 33 | Search an environment for an object | 0.80 | 0.80 | 0.60 | 0.60 | 1.00 | 1.00 | +| 34 | Interact with a moving agent | 0.60 | 0.20 | 0.53 | 0.00 | 0.53 | 0.20 | +| Average (Unit Tests) | 0.76 | 0.60 | 0.66 | 0.44 | 0.77 | 0.64 | + +4.2 Baseline Agent Models + +The baseline agents are described below, with model performance on Discovery tasks shown in Table 4, and performance on Unit Tests shown in Table 5. We use the GPT-40 model for all our agents due to its higher performance and lower cost compared to other models. For space we provide \ No newline at end of file diff --git a/olmocr/bench/sample_data/olmocr_pipeline/earnings_pg1_repeat1.md b/olmocr/bench/sample_data/olmocr_pipeline/earnings_pg1_repeat1.md new file mode 100644 index 0000000..f67b503 --- /dev/null +++ b/olmocr/bench/sample_data/olmocr_pipeline/earnings_pg1_repeat1.md @@ -0,0 +1,33 @@ +Recently Issued Accounting Pronouncements + +Recently Adopted Accounting Pronouncement + +In November 2023, the Financial Accounting Standards Board, or FASB, issued a new accounting standard requiring disclosures of significant expenses in operating segments. We adopted this standard in our fiscal year 2025 annual report. Refer to Note 16 of the Notes to the Consolidated Financial Statements in Part IV, Item 15 of this Annual Report on Form 10-K for further information. + +Recent Accounting Pronouncements Not Yet Adopted + +In December 2023, the FASB issued a new accounting standard which includes new and updated income tax disclosures, including disaggregation of information in the rate reconciliation and income taxes paid. We expect to adopt this standard in our fiscal year 2026 annual report. We do not expect the adoption of this standard to have a material impact on our Consolidated Financial Statements other than additional disclosures. + +In November 2024, the FASB issued a new accounting standard requiring disclosures of certain additional expense information on an annual and interim basis, including, among other items, the amounts of purchases of inventory, employee compensation, depreciation and intangible asset amortization included within each income statement expense caption, as applicable. We expect to adopt this standard in our fiscal year 2028 annual report. We do not expect the adoption of this standard to have a material impact on our Consolidated Financial Statements other than additional disclosures. + +Note 2 - Business Combination + +Termination of the Arm Share Purchase Agreement + +In February 2022, NVIDIA and SoftBank Group Corp, or SoftBank, announced the termination of the Share Purchase Agreement whereby NVIDIA would have acquired Arm from SoftBank. The parties agreed to terminate it due to significant regulatory challenges preventing the completion of the transaction. We recorded an acquisition termination cost of $1.4 billion in fiscal year 2023 reflecting the write-off of the prepayment provided at signing. + +Note 3 - Stock-Based Compensation + +Stock-based compensation expense is associated with RSUs, PSUs, market-based PSUs, and our ESPP. + +Consolidated Statements of Income include stock-based compensation expense, net of amounts capitalized into inventory and subsequently recognized to cost of revenue, as follows: + +| Year Ended | Jan 29, 2023 | Jan 28, 2024 | Jan 29, 2023 | +|------------|--------------|--------------|--------------| +| (In millions) | $138 | $141 | $138 | +| Cost of revenue | $178 | $141 | $138 | +| Research and development | 3,423 | 2,532 | 1,892 | +| Sales, general and administrative | 1,136 | 876 | 680 | +| Total | $4,737 | $3,549 | $2,710 | + +Stock-based compensation capitalized in inventories was not significant during fiscal years 2025, 2024, and 2023. \ No newline at end of file diff --git a/olmocr/bench/sample_data/olmocr_pipeline/headers_footers/ff0f0b22c55d8b90dd77d153f48e144fc9db_pg2_pg1_repeat1.md b/olmocr/bench/sample_data/olmocr_pipeline/headers_footers/ff0f0b22c55d8b90dd77d153f48e144fc9db_pg2_pg1_repeat1.md new file mode 100644 index 0000000..3d0d951 --- /dev/null +++ b/olmocr/bench/sample_data/olmocr_pipeline/headers_footers/ff0f0b22c55d8b90dd77d153f48e144fc9db_pg2_pg1_repeat1.md @@ -0,0 +1,23 @@ +Lassa Fever in Post-Conflict Sierra Leone + +Jeffrey G. Shaffer1, Donald S. Grant2, John S. Schieffelin3, Matt L. Boisen4,5, Augustine Goba2,9, Jessica N. Hartnett4, Danielle C. Levy4, Rachael E. Yenni4, Lina M. Moses4,6, Mohammed Fullah2, Mambo Momoh2, Mbalu Fonnie2, Richard Fonnie2, Lansana Kanneh2, Veronica J. Koroma2, Kande Kargbo2, Darin Ottomassathien9, Ivana J. Muncy5, Abigail B. Jones5, Megan M. Illic7, Peter C. Kulakosky8, Allyson M. Haislip4, Christopher M. Bishop4, Deborah H. Elliot3, Bethany L. Brown5, Hu Zhu9, Kathryn M. Hastie10, Kristian G. Andersen11,12, Stephen K. Gire11,12, Shervin Tabrizi11,12, Ridhi Tariyal12, Mathew Stremlau11,12, Alex Matschiner7, Darryl B. Sampey7, Jennifer S. Spence6, Robert W. Cross4,13, Joan B. Geisbert13, Onikepe A. Folarin14, Christian T. Happi14, Kelly R. Pitts5, F. Jon Geske5, Thomas W. Geisbert13, Erica Ollmann Saphire9,15, James E. Robinson3, Russell B. Wilson8, Pardis C. Sabeti10,11,16, Lee A. Henderson8, S. Humarr Khan25, Daniel G. Bausch61, Luis M. Branco8,17, Robert F. Garry4,12,17*, the Viral Hemorrhagic Fever Consortium† + +1 Department of Biostatistics and Bioinformatics, Tulane School of Public Health and Tropical Medicine, New Orleans, Louisiana, United States of America, 2 Lassa Fever Program, Kenema Government Hospital, Kenema, Sierra Leone, 3 Sections of Infectious Disease, Departments of Pediatrics and Internal Medicine, School of Medicine, Tulane University, New Orleans, Louisiana, United States of America, 4 Department of Microbiology and Immunology, Tulane University, New Orleans, Louisiana, United States of America, 5 Corgenix, Inc., Broomfield, Colorado, United States of America, 6 Department of Tropical Medicine, Tulane School of Public Health and Tropical Medicine, New Orleans, Louisiana, United States of America, 7 Biofactura Inc., Rockville, Maryland, United States of America, 8 Autoimmune Technologies, LLC, New Orleans, Louisiana, United States of America, 9 Vybion Inc., Ithaca, New York, United States of America, 10 Department of Immunology and Microbial Science, The Scripps Research Institute, La Jolla, California, United States of America, 11 FAS Center for Systems Biology, Department of Organismic and Evolutionary Biology, Harvard University, Cambridge, Massachusetts, United States of America, 12 Broad Institute, Cambridge, Massachusetts, United States of America, 13 Department of Microbiology and Immunology, University of Texas Medical Branch at Galveston, Galveston, Texas, United States of America, 14 Department of Biological Sciences, College of Natural Sciences, Redeemers University, Redemption City, Ogun State, Nigeria, 15 The Skaggs Institute for Chemical Biology, The Scripps Research Institute, La Jolla, California, United States of America, 16 Department of Immunology and Infectious Disease, Harvard School of Public Health, Boston, Massachusetts, United States of America, 17 Zalgen Labs, LLC, Germantown, Maryland, United States of America + +Abstract + +Background: Lassa fever (LF), an often-fatal hemorrhagic disease caused by Lassa virus (LASV), is a major public health threat in West Africa. When the violent civil conflict in Sierra Leone (1991 to 2002) ended, an international consortium assisted in restoration of the LF program at Kenema Government Hospital (KGH) in an area with the world’s highest incidence of the disease. + +Methodology/Principal Findings: Clinical and laboratory records of patients presenting to the KGH Lassa Ward in the post-conflict period were organized electronically. Recombinant antigen-based LF immunoassays were used to assess LASV antigenemia and LASV-specific antibodies in patients who met criteria for suspected LF. KGH has been reestablished as a center for LF treatment and research, with over 500 suspected cases now presenting yearly. Higher case fatality rates (CFRs) in LF patients were observed compared to studies conducted prior to the civil conflict. Different criteria for defining LF stages and differences in sensitivity of assays likely account for these differences. The highest incidence of LF in Sierra Leone was observed during the dry season. LF cases were observed in ten of Sierra Leone’s thirteen districts, with numerous cases from outside the traditional endemic zone. Deaths in patients presenting with LASV antigenemia were skewed towards individuals less than 29 years of age. Women self-reporting as pregnant were significantly overrepresented among LASV antigenemic patients. The CFR of ribavirin-treated patients presenting early in acute infection was lower than in untreated subjects. + +Conclusions/Significance: Lassa fever remains a major public health threat in Sierra Leone. Outreach activities should expand because LF may be more widespread in Sierra Leone than previously recognized. Enhanced case finding to ensure rapid diagnosis and treatment is imperative to reduce mortality. Even with ribavirin treatment, there was a high rate of fatalities underscoring the need to develop more effective and/or supplemental treatments for LF. + +Citation: Shaffer JG, Grant DS, Schieffelin JS, Boisen ML, Goba A, et al. (2014) Lassa Fever in Post-Conflict Sierra Leone. PLoS Negl Trop Dis 8(3): e2748. doi:10.1371/journal.pntd.0002748 + +Editor: Brian Bird, Centers for Disease Control and Prevention, United States of America + +Received August 10, 2013; Accepted February 3, 2014; Published March 20, 2014 + +Copyright: © 2014 Shaffer et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. + +Funding: This work was supported by National Institute of Allergy and Infectious Diseases grants/contracts AI067188, A082119, AI2008031, AI2009061, AI104216, AI104621, P20GM103501, HHSN272200900018C, HHSN272200900049C, HHSN272201000022C, and an Investigators in the Pathogenesis of Infectious Disease award from the Burroughs Wellcome Fund. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. \ No newline at end of file diff --git a/olmocr/bench/sample_data/olmocr_pipeline/headers_footers/ff1fc6a205ad039139ce566851b6b260c929_pg1_pg1_repeat1.md b/olmocr/bench/sample_data/olmocr_pipeline/headers_footers/ff1fc6a205ad039139ce566851b6b260c929_pg1_pg1_repeat1.md new file mode 100644 index 0000000..937fc8b --- /dev/null +++ b/olmocr/bench/sample_data/olmocr_pipeline/headers_footers/ff1fc6a205ad039139ce566851b6b260c929_pg1_pg1_repeat1.md @@ -0,0 +1,10 @@ +RTG Degradation Primer and Application to MMRTG + +Nuclear and Emerging Technology for Space (NETS) 2015 +February 23-26, 2015 +Abstract 5107 + +Presenting Author: Tom Hammel, Teledyne Energy Systems +Co-Authors: Russell Bennett, Teledyne Energy Systems +Robert Sievers, Teledyne Energy Systems +Bill Otting, Aerojet Rocketdyne \ No newline at end of file diff --git a/olmocr/bench/sample_data/olmocr_pipeline/headers_footers/ff3d6e051903fe5ca9bc172ece14964c5632_pg1_pg1_repeat1.md b/olmocr/bench/sample_data/olmocr_pipeline/headers_footers/ff3d6e051903fe5ca9bc172ece14964c5632_pg1_pg1_repeat1.md new file mode 100644 index 0000000..d4b30c0 --- /dev/null +++ b/olmocr/bench/sample_data/olmocr_pipeline/headers_footers/ff3d6e051903fe5ca9bc172ece14964c5632_pg1_pg1_repeat1.md @@ -0,0 +1,9 @@ +بررسی دیدگاه و نظرات کتابداران و اعضای هیئت علمی دانشگاه شیراز در بره گیری از فناوری شبکه‌های پی سیم در کتابخانه‌های دانشگاهی + +چکیده: نظر به اهمیت و کاربرد گسترده شبکه‌های پی سیم در محیط‌های دانشگاهی در کشورهای پیشرفته و استفاده از آن در خدمات کتابخانه‌ای، بهره‌گیری از این فناوری در کتابخانه‌های دانشگاهی کشور احساس می‌شود. این پژوهش با استفاده از روش پیام‌برداری، با هدف معرفی شبکه‌های پی سیم، به بررسی دیدگاه و نظرات کتابداران و اعضای هیئت علمی دانشگاه شیراز، در یک کارگیری از این شبکه‌ها در کتابخانه‌های دانشگاهی پرداخت. باقی‌مانده‌های تحقیق نشان داد که کتابداران تا مایل زیادی به استفاده از شبکه‌های پی سیم در امر خدمات کتابخانه‌های تظیم می‌باشند و قفسه‌خوانی و دسترسی به فهرست عمومی پوسته دارند. و گسترش گی پوشش و سیاست‌های پیشرفته و سیاست‌های پیشرفته در کتابخانه‌ها را خواستار شده‌اند. در حالی که اعضای هیئت علمی، ضرورت استفاده بیشتر از این شبکه‌ها در کل محیط دانشگاه و دسترسی به منابع کتابخانه‌ای از خارج از محیط کتابخانه را خواستار شده‌اند. در کل، با توجه به نتایج حاصل از این تحقیق می‌توان گفت که هر چند استفاده از شبکه‌های پی سیم در کتابخانه‌ها پیش‌بینی‌های زیادی و نوآوری‌های می‌باشد و هنوز در کشور ما چندان مورد توجه قرار نگرفته و ناشناخته مانده است، اما تماشای کتابداران، پژوهشگران و استادان به استفاده و کاربرد آن در محیط‌های دانشگاهی زیاد است. + +کلیدواژه‌ها: شبکه‌های پی سیم؛ کتابخانه‌های دانشگاهی؛ اعضای هیئت علمی؛ کتابداران؛ رایانه‌های قابل حمل؛ رایانه‌های دستی؛ منابع و خدمات کتابخانه‌ای؛ دانشگاه شیراز + +farbod4ever@gmail.com + +نویسنده رابطه: \ No newline at end of file diff --git a/olmocr/bench/sample_data/olmocr_pipeline/headers_footers/ff4f7dad78081cff727d19ab51c181d4a661_pg1_pg1_repeat1.md b/olmocr/bench/sample_data/olmocr_pipeline/headers_footers/ff4f7dad78081cff727d19ab51c181d4a661_pg1_pg1_repeat1.md new file mode 100644 index 0000000..989db55 --- /dev/null +++ b/olmocr/bench/sample_data/olmocr_pipeline/headers_footers/ff4f7dad78081cff727d19ab51c181d4a661_pg1_pg1_repeat1.md @@ -0,0 +1,11 @@ +Molecular markers of breast cancer metastasis +Weigelt, B. + +Citation for published version (APA): +Weigelt, B. (2005). Molecular markers of breast cancer metastasis + +General rights +It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons). + +Disclaimer/Complaints regulations +If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: http://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible. \ No newline at end of file diff --git a/olmocr/bench/sample_data/olmocr_pipeline/headers_footers/ff518b1240a66978f22035528ccb029450b5_pg2_pg1_repeat1.md b/olmocr/bench/sample_data/olmocr_pipeline/headers_footers/ff518b1240a66978f22035528ccb029450b5_pg2_pg1_repeat1.md new file mode 100644 index 0000000..eda6ee4 --- /dev/null +++ b/olmocr/bench/sample_data/olmocr_pipeline/headers_footers/ff518b1240a66978f22035528ccb029450b5_pg2_pg1_repeat1.md @@ -0,0 +1,15 @@ +Brief Notices + +Prophet of the Jubilee, translated and edited by Ronald D. Dennis (Religious Studies Center, Brigham Young University, 1997) + +In July 1846 in Rhydybont, Carmarthenshire, Wales, Dan Jones published the first issue of a monthly LDS periodical in the Welsh language on a press owned by John Jones, Dan’s brother, who was an ordained Congregational minister. The periodical, Prophwyd y Jubili (Prophet of the Jubilee), ran monthly thereafter through December 1848. Jones’s great-great-grandson Ronald Dennis has presented what he calls a “facsimile translation” (xxix) of the complete series, retaining original fonts, layout, and pagination, slightly enlarging font size for readability. Text and index are over seven hundred pages, and Geraint Bowen, former Archdruid of Wales, offers a superb introduction. + +Many articles in Prophet of the Jubilee rebut arguments of local anti-Mormons or apostates. Articles entitled “The ‘Hater of Deceit’ Proving Himself a False Prophet Again!!” and “The ‘Rev. W. R. Davies, from Dowlais,’ and His Cruel and Shameful Persecution Again!—Again!!” give a glimpse of the intense feelings between early Welsh Saints and their religious adversaries. Jones garnishes his numerous doctrinal treatises with occasional fiction and poetry, excerpts translated from the Millennial Star, the neighboring LDS periodical in England, and portions of articles on religious topics taken from European and U. S. newspapers. + +A brief summary of each article is provided at the beginning of the book, but after that the reader is left to plod through the text without annotations. While pagination is sure to confuse some readers, Prophet of the Jubilee opens up LDS historical documents that have been inaccessible to most English-speaking readers for 150 years. Here is a mass of interesting cultural and doctrinal history, as well as the voice of Dan Jones himself, one of the most prolific and persistent missionaries in the history of the Church. + +—Jed L. Woodworth + +Book of Mormon Authors: Their Words and Messages, by Roger R. Keller (Religious Studies Center, Brigham Young University, 1996) + +The statistical study of Book of Mormon texts is a well-traveled road in Book of Mormon scholarship. However, in Book of Mormon Authors, Roger Keller shows \ No newline at end of file diff --git a/olmocr/bench/sample_data/olmocr_pipeline/headers_footers/ffaac214730d2b8c2ec842e3618ccb9c4259_pg1_pg1_repeat1.md b/olmocr/bench/sample_data/olmocr_pipeline/headers_footers/ffaac214730d2b8c2ec842e3618ccb9c4259_pg1_pg1_repeat1.md new file mode 100644 index 0000000..e69de29 diff --git a/olmocr/bench/sample_data/olmocr_pipeline/headers_footers/fff590bed29a2854ac1f874dad5752ede1aa_pg1_pg1_repeat1.md b/olmocr/bench/sample_data/olmocr_pipeline/headers_footers/fff590bed29a2854ac1f874dad5752ede1aa_pg1_pg1_repeat1.md new file mode 100644 index 0000000..3e36218 --- /dev/null +++ b/olmocr/bench/sample_data/olmocr_pipeline/headers_footers/fff590bed29a2854ac1f874dad5752ede1aa_pg1_pg1_repeat1.md @@ -0,0 +1,20 @@ +User’s Manual + +Model 475 +DSP Gaussmeter + +Lake Shore Cryotronics, Inc. +575 McCorkle Blvd. +Westerville, Ohio 43082-8888 USA + +E-mail Addresses: +sales@lakeshore.com +service@lakeshore.com + +Visit Our Website At: +www.lakeshore.com + +Fax: (614) 891-1392 +Telephone: (614) 891-2243 + +Methods and apparatus disclosed and described herein have been developed solely on company funds of Lake Shore Cryotronics, Inc. No government or other contractual support or relationship whatsoever has existed which in any way affects or mitigates proprietary rights of Lake Shore Cryotronics, Inc. in these developments. Methods and apparatus disclosed herein may be subject to U.S. Patents existing or applied for. Lake Shore Cryotronics, Inc. reserves the right to add, improve, modify, or withdraw functions, design modifications, or products at any time without notice. Lake Shore shall not be liable for errors contained herein or for incidental or consequential damages in connection with furnishing, performance, or use of this material. \ No newline at end of file diff --git a/olmocr/bench/sample_data/olmocr_pipeline/lincoln_letter_pg1_repeat1.md b/olmocr/bench/sample_data/olmocr_pipeline/lincoln_letter_pg1_repeat1.md new file mode 100644 index 0000000..6c1dd7f --- /dev/null +++ b/olmocr/bench/sample_data/olmocr_pipeline/lincoln_letter_pg1_repeat1.md @@ -0,0 +1,9 @@ +Executive Mansion, + +Washington City, + +January 15th, 1864 + +Major General Hitchcock, Commissioner of Exchanges, is authorized and directed to offer Brigadier General Trimble, now a prisoner of war in Fort McHenry, in exchange for Major White, who is held as a prisoner at Richmond. He is also directed to send forward the offer of exchange by Henry M. Warfield, Esq. of Baltimore, under a flag of truce, and give him a pass to City Point. + +Abraham Lincoln \ No newline at end of file diff --git a/olmocr/bench/sample_data/olmocr_pipeline/math_2503_04086_pg1_repeat1.md b/olmocr/bench/sample_data/olmocr_pipeline/math_2503_04086_pg1_repeat1.md new file mode 100644 index 0000000..4c09c04 --- /dev/null +++ b/olmocr/bench/sample_data/olmocr_pipeline/math_2503_04086_pg1_repeat1.md @@ -0,0 +1,29 @@ +Proof. Let $S$ be the generating set associated with $D$ as described in Proposition 2.5. By the circulant diagonalization theorem, the spectrum of $G_R(D) = \Gamma(R, S)$ is the multiset $\{\lambda_g\}_{g \in R}$ where + +$$\lambda_g = \sum_{s \in S} \zeta_n^{\psi(gs)} = \sum_{i=1}^k \left[ \sum_{s, Rs = I_i} \zeta_n^{\psi(gs)} \right].$$ + +We remark that by Corollary 2.7, if $s \in R$ such that $Rs = I_i = Rx_i$ then $s$ has a unique representation of the form $s = ux_i$ where $u \in (R/\text{Ann}_R(x_i))^\times$ and $\hat{u}$ is a fixed lift of $u$ to $R^\times$. With this presentation, we can write + +$$\sum_{s, Rs = I_i} \zeta_n^{\psi(gs)} = \sum_{u \in (R/\text{Ann}_R(x_i))^\times} \zeta_n^{\psi(gux_i)} = \sum_{u \in (R/\text{Ann}_R(x_i))^\times} \zeta_n^{\psi_xi(gu)} = c(g, R/\text{Ann}_R(x_i)).$$ + +Here we recall that $\psi_xi$ is the induced linear functional on $R/\text{Ann}_R(x_i)$. We conclude that $\lambda_g = \sum_{i=1}^k c(g, R/\text{Ann}_R(x_i)).$ \hfill $\square$ + +The following corollary is simple yet important for our future work on perfect state transfers on gcd-graphs. + +**Corollary 4.17.** Suppose that $g' = ug$ for some $u \in R^\times$. Then $\lambda_g = \lambda_{g'}$. + +**Acknowledgements** + +We thank the Department of Mathematics and Computer Science at Lake Forest College for their generous financial support through an Overleaf subscription. We also thank Ján Mináč for his constant encouragement and support. + +**References** + +1. Reza Akhtar, Megan Boggess, Tiffany Jackson-Henderson, Isidora Jiménez, Rachel Karpman, Amanda Kinzel, and Dan Pritikin, *On the unitary Cayley graph of a finite ring*, Electron. J. Combin. 16 (2009), no. 1, Research Paper 117, 13 pages. +2. Milan Bašić, Aleksandar Ilić, and Aleksandar Stamenković, *Maximal diameter of integral circulant graphs*, Information and Computation 301 (2024), 105208. +3. Maria Chudnovsky, Michal Cizek, Logan Crew, Ján Mináč, Tung T. Nguyen, Sophie Spirkl, and Nguyễn Duy Tấn, *On prime Cayley graphs*, arXiv:2401.06062, to appear in Journal of Combinatorics (2024). +4. Thomas Honold, *Characterization of finite frobenius rings*, Archiv der Mathematik 76 (2001), no. 6, 406–415. +5. Irving Kaplansky, *Elementary divisors and modules*, Transactions of the American Mathematical Society 66 (1949), no. 2, 464–491. +6. Walter Klotz and Torsten Sander, *Some properties of unitary Cayley graphs*, The Electronic Journal of Combinatorics 14 (2007), no. 1, R45, 12 pages. +7. Erich Lamprecht, *Allgemeine theorie der Gaußschen Summen in endlichen kommutativen Ringen*, Mathematische Nachrichten 9 (1953), no. 3, 149–196. +8. Ján Mináč, Tung T Nguyen, and Nguyen Duy Tấn, *Isomorphic gcd-graphs over polynomial rings*, arXiv preprint arXiv:2411.01768 (2024). +9. ______, *On the gcd graphs over polynomial rings*, arXiv preprint arXiv:2409.01929 (2024). \ No newline at end of file diff --git a/olmocr/bench/sample_data/olmocr_pipeline/mathfuncs_colswitch_pg1_repeat1.md b/olmocr/bench/sample_data/olmocr_pipeline/mathfuncs_colswitch_pg1_repeat1.md new file mode 100644 index 0000000..3803af1 --- /dev/null +++ b/olmocr/bench/sample_data/olmocr_pipeline/mathfuncs_colswitch_pg1_repeat1.md @@ -0,0 +1,16 @@ +# The 20 Most Important Mathematical Equations + +A journey through the most elegant and influential formulas in mathematics + +| 1. Euler’s Identity | 3. The Fundamental Theorem of Calculus | +|---------------------|----------------------------------------| +| $ e^{i\pi} + 1 = 0 $ | $ \int_a^b f(x) \, dx = F(b) - F(a) $ | +| Connects five fundamental constants (e, i, π, 1, 0), revealing the profound relationship between exponential functions and trigonometry. | Establishes that differentiation and integration are inverse operations. If F is an antiderivative of f, the definite integral equals F(b) - F(a). Revolutionized mathematical problem-solving. | + +| 2. Pythagorean Theorem | 4. Maxwell’s Equations | +|------------------------|-----------------------| +| $ a^2 + b^2 = c^2 $ | $ \nabla \cdot \mathbf{E} = \frac{\rho}{\varepsilon_0} $ | +| In right triangles, the hypotenuse squared equals the sum of the squares of the other sides. Cornerstone of geometry with applications in navigation and architecture. | $ \nabla \cdot \mathbf{B} = 0 $ | +| | $ \nabla \times \mathbf{E} = -\frac{\partial \mathbf{B}}{\partial t} $ | +| | $ \nabla \times \mathbf{B} = \mu_0 \mathbf{J} + \mu_0 \varepsilon_0 \frac{\partial \mathbf{E}}{\partial t} $ | +| | Unified electricity and magnetism as manifestations of the same force. Describes electromagnetic field behavior, predicting waves traveling at light speed. Enabled technologies from radio to smartphones. | \ No newline at end of file diff --git a/olmocr/bench/sample_data/olmocr_pipeline/mathfuncs_pg1_repeat1.md b/olmocr/bench/sample_data/olmocr_pipeline/mathfuncs_pg1_repeat1.md new file mode 100644 index 0000000..7c60585 --- /dev/null +++ b/olmocr/bench/sample_data/olmocr_pipeline/mathfuncs_pg1_repeat1.md @@ -0,0 +1,19 @@ +# The 20 Most Important Mathematical Equations + +A journey through the most elegant and influential formulas in mathematics + +| 1. Euler's Identity | 2. Pythagorean Theorem | +|--------------------|------------------------| +| $ e^{i\pi} + 1 = 0 $ | $ a^2 + b^2 = c^2 $ | + +Connects five fundamental constants (e, i, π, 1, 0), revealing the profound relationship between exponential functions and trigonometry. + +In right triangles, the hypotenuse squared equals the sum of the squares of the other sides. Cornerstone of geometry with applications in navigation and architecture. + +| 3. The Fundamental Theorem of Calculus | 4. Maxwell's Equations | +|----------------------------------------|------------------------| +| $ \int_{a}^{b} f(x) \, dx = F(b) - F(a) $ | $ \nabla \cdot \mathbf{E} = \frac{Q}{\varepsilon_0} $ | + +Establishes that differentiation and integration are inverse operations. If $ F $ is an antiderivative of $ f $, the definite integral equals $ F(b) - F(a) $. Revolutionized mathematical problem-solving. + +Unified electricity and magnetism as manifestations of the same force. Describes electromagnetic field behavior, predicting waves traveling at light speed. Enabled technologies from radio to smartphones. \ No newline at end of file diff --git a/olmocr/bench/sample_data/olmocr_pipeline/mattsnotes_pg1_repeat1.md b/olmocr/bench/sample_data/olmocr_pipeline/mattsnotes_pg1_repeat1.md new file mode 100644 index 0000000..972717a --- /dev/null +++ b/olmocr/bench/sample_data/olmocr_pipeline/mattsnotes_pg1_repeat1.md @@ -0,0 +1,30 @@ +V-February Flow + +Data Components: + +Code: +The-Stack-V2 + +CodeText: +SE, whatever we've scraped + +WebText: +HQ DCLM + +DATA MIXES + +~85% Source Code +~10% CodeText +~5% Webtext + +~85% The-stack-V2 +~15% CodeText +~0% Webtext + +~100% Source Code + +Deepseek Coder + +StarCoder 2 + +Arctic \ No newline at end of file diff --git a/olmocr/bench/sample_data/olmocr_pipeline/mattsnotes_pg2_repeat1.md b/olmocr/bench/sample_data/olmocr_pipeline/mattsnotes_pg2_repeat1.md new file mode 100644 index 0000000..7f3c8a2 --- /dev/null +++ b/olmocr/bench/sample_data/olmocr_pipeline/mattsnotes_pg2_repeat1.md @@ -0,0 +1,18 @@ +P1: 100% Source code +P2: 80% code +20% language + +Code Data Recipe [StackCoder] +1) Order by Repo ✓ +2) Call Heuristic Filters ✗ +3) Group by Repo, lang → minhash ✓ +4) Pack into Repo-level docs □ +5) Select PL's □ + +6) Pack into FIM tokens ✗ + +✓: Eng Done +X: Eng definitely NOT done +D: so so easy + +Use Preprocessed code/text, web/text \ No newline at end of file diff --git a/olmocr/bench/sample_data/olmocr_pipeline/mattsnotes_pg3_repeat1.md b/olmocr/bench/sample_data/olmocr_pipeline/mattsnotes_pg3_repeat1.md new file mode 100644 index 0000000..41b22ee --- /dev/null +++ b/olmocr/bench/sample_data/olmocr_pipeline/mattsnotes_pg3_repeat1.md @@ -0,0 +1,9 @@ +ARCH + TRAINING + +- Pick Arch like OLMO-IB +- OR replicate a 3D model +- Follow standard LR flow + +Eval: + +Hacky nonsense for now \ No newline at end of file diff --git a/olmocr/bench/sample_data/olmocr_pipeline/multi_column_miss_pg1_repeat1.md b/olmocr/bench/sample_data/olmocr_pipeline/multi_column_miss_pg1_repeat1.md new file mode 100644 index 0000000..c2e8347 --- /dev/null +++ b/olmocr/bench/sample_data/olmocr_pipeline/multi_column_miss_pg1_repeat1.md @@ -0,0 +1,27 @@ +stakeholders has occurred in other nations, with groups and individuals refusing to risk being appropriated into the industry’s public relations ambitions. It now looks like that with vigilance, tobacco control advocates can easily foment disruption of its efforts to take its place alongside other industries—often with considerable social credit—in the hope that it might gain by association. + +Tobacco industry posturing about its corporate responsibility can never hide the ugly consequences of its ongoing efforts to “work with all relevant stakeholders for the preservation of opportunities for informed adults to consume tobacco products” (translation: “we will build alliances with others who want to profit from tobacco use, to do all we can to counteract effective tobacco control”). BAT has 15.4% and Philip Morris 16.4% of the global cigarette market. With 4.9 million smokers currently dying from tobacco use each year, and the industry unblinkingly concurring that its products are addictive, this leaves BAT to argue why it should not be held to be largely accountable for the annual deaths of some 754 600 smokers, and Philip Morris some 803 600 smokers. + +REFERENCES +1 British American Tobacco. Social Report. http://www.bat.com/204pp +2 Wroe D. Tobacco ad campaign angers MPs. The Age (Melbourne) 2004; May 17 http://www.theage.com.au/articles/2004/05/16/108464669771.html?oneclick=true +3 Hirschhorn N. Corporate social responsibility and the tobacco industry: hope or hype? Tobacco Control 2004;13:447–53. +4 Ethical Corporation Asia 2004. Conference website. http://www.ethicalcorp.com/asia2004/ +5 Chapman S, Shatenstein S. Extreme corporate makeover: tobacco companies, corporate responsibility and the corruption of “ethics”. Globalink petition. http://petition.globalink.org/view.php?code=extreme. +6 Mackay J, Erikson M. The tobacco atlas. Geneva: World Health Organization, 2002. + +INDUSTRY WATCH + +Corporate social responsibility and the tobacco industry: hope or hype? + +N Hirschhorn + +Corporate social responsibility (CSR) emerged from a realisation among transnational corporations of the need to account for and redress their adverse impact on society: specifically, on human rights, labour practices, and the environment. Two transnational tobacco companies have recently adopted CSR: Philip Morris, and British American Tobacco. This report explains the origins and theory behind CSR; examines internal company documents from Philip Morris showing the company’s deliberations on the matter, and the company’s perspective on its own behaviour; and reflects on whether marketing tobacco is antithetical to social responsibility. + +Over the past three decades increasing pressure from non-governmental organisations (NGOs), governments and the United Nations, has required transnational corporations (TNCs) to examine and redress the adverse impact their businesses have on society and the environment. Many have responded by taking up what is known as “corporate social responsibility” (CSR); only recently have two major cigarette companies followed suit: Philip Morris (PM) and British American Tobacco (BAT). This report first provides the context and development of CSR; then, from internal company documents, examines how PM came to its own version. This paper examines whether a tobacco company espousing CSR should be judged simply as a corporate entity along standards of business ethics, or as an irretrievably negative force in the realm of public health, thereby rendering CSR an oxymoron. + +CORPORATE SOCIAL RESPONSIBILITY: THE CONTEXT + +The term “corporate social responsibility” is in vogue at the moment but as a concept it is vague and means different things to different people. Some writers on CSR trace its American roots to the 19th century when large industries engaged in philanthropy and established great public institutions, a form of “noblesse oblige”. But the notion that corporations should be required to return more to society because of their impact on society was driven by pressures from the civil rights, peace, and environmental movements of the last half century. The unprecedented expansion of power and influence of TNCs over the past three decades has accelerated global trade and development, but also environmental damage and abuses of + +Abbreviations: ASH, Action on Smoking and Health; BAT, British American Tobacco; CERES, Coalition for Environmentally Responsible Economies; CSR, corporate social responsibility; DJSI, Dow Jones Sustainability Index; GCAC, Global Corporate Affairs Council; GRI, Global Reporting Initiative; MSA, Master Settlement Agreement; NGOs, non-governmental organisations; PM, Philip Morris; TNCs, transnational corporations; UNEP, United Nations Environment Program \ No newline at end of file diff --git a/olmocr/bench/sample_data/olmocr_pipeline/olmo2-pg4_pg1_repeat1.md b/olmocr/bench/sample_data/olmocr_pipeline/olmo2-pg4_pg1_repeat1.md new file mode 100644 index 0000000..623fb56 --- /dev/null +++ b/olmocr/bench/sample_data/olmocr_pipeline/olmo2-pg4_pg1_repeat1.md @@ -0,0 +1,24 @@ +Table 1 Composition of the pretraining data for OLMo 2. The OLMo 2 1124 Mix is composed of StarCoder (Li et al., 2023b; Kocetkov et al., 2022), peS2o (Soldaini and Lo, 2023), web text from DCLM (Li et al., 2024) and Wiki come from Dolma 1.7 (Soldaini et al., 2024). arXiv comes from Red-Pajama (Together AI, 2023), while OpenWebMath (Paster et al., 2023) and Algebraic Stack come from ProofPile II (Azerbayev et al., 2023). + +| Source | Type | Tokens | Words | Bytes | Docs | +|-------------------------------|-----------------------|---------|--------|--------|-------| +| DCLM-Baseline | Web pages | 3.71T | 3.32T | 21.32T | 2.95B | +| StarCoder | Code | 83.0B | 70.0B | 459B | 78.7M | +| filtered version from OLMoE Mix | Academic papers | 58.6B | 51.1B | 413B | 38.8M | +| peS2o from Dolma 1.7 | Math web pages | 12.2B | 11.1B | 47.2B | 2.89M | +| arXiv | Math proofs code | 11.8B | 10.8B | 44.0B | 2.83M | +| OpenWebMath | Encyclopedic | 3.7B | 3.16B | 16.2B | 6.17M | +| Wikipedia & Wikibooks from Dolma 1.7 | | | | | | +| Total | | 3.90T | 3.48T | 22.38T | 3.08B | + +2.1.1 Pretraining data: OLMo 2 Mix 1124 + +The mix used for this stage is shown in Table 1. It consists of approximately 3.9 trillion tokens, with over 95% derived from web data. We refer to this set as OLMo 2 Mix 1124. This is the same pretraining data used in OLMoE (Muennighoff et al., 2024). + +We combine data from DCLM (Li et al., 2024) and Dolma 1.7 (Soldaini et al., 2024). From DCLM, we use the “baseline 1.0” mix. From Dolma, we use the arXiv (Together AI, 2023), OpenWebMath (Paster et al., 2023), Algebraic Stack, peS2o (Soldaini and Lo, 2023), and Wikipedia subsets. arXiv, OpenWebMath, and Algebraic Stack were originally part of ProofPile II (Azerbayev et al., 2023). + +Finally, we include code from StarCoder (Li et al., 2023b), which is derived from permissively-licensed repositories from GitHub (Kocetkov et al., 2022). In an attempt to include higher quality code, we remove any document from a repository with fewer than 2 stars on GitHub. Further, through manual inspection of this source, we found it to contain documents encoded in binary format or containing mostly numerical content; to remove them, we discarded documents whose most frequent word constitutes over 30% of the document, or whose top-2 most frequent words constitute over 50% of the document. To mitigate possible training loss spikes, we remove documents with repeated sequences of 32 or more n-grams. We report details and show effectiveness of this intervention in Section §3.1. + +2.1.2 Mid-training data: Dolmino Mix 1124 + +After the initial pretraining stage on mostly web data, we further train with a mixture of web data that has been more restrictively filtered for quality and a collection of domain-specific high quality data, much of which is synthetic. The purpose of this mixture is to imbue the model with math-centric skills and provide focused exposure to STEM references and high quality text. We generate several variants of this mixture, with varying sizes, but generally refer to this mixture as Dolmino Mix 1124. The base sources from which Dolmino Mix 1124 is subsampled are described in Table 2. We refer the reader to Section §4 for a deep dive detailing our processes for experimenting and curating data for this mix. \ No newline at end of file diff --git a/olmocr/bench/sample_data/olmocr_pipeline/openstax_caculus_pg_273_pg1_repeat1.md b/olmocr/bench/sample_data/olmocr_pipeline/openstax_caculus_pg_273_pg1_repeat1.md new file mode 100644 index 0000000..dfcf39e --- /dev/null +++ b/olmocr/bench/sample_data/olmocr_pipeline/openstax_caculus_pg_273_pg1_repeat1.md @@ -0,0 +1,73 @@ +3.4 EXERCISES + +For the following exercises, the given functions represent the position of a particle traveling along a horizontal line. + +a. Find the velocity and acceleration functions. + +b. Determine the time intervals when the object is slowing down or speeding up. + +150. $ s(t) = 2t^3 - 3t^2 - 12t + 8 $ + +151. $ s(t) = 2t^3 - 15t^2 + 36t - 10 $ + +152. $ s(t) = \frac{t}{1 + t^2} $ + +153. A rocket is fired vertically upward from the ground. The distance $ s $ in feet that the rocket travels from the ground after $ t $ seconds is given by $ s(t) = -16t^2 + 560t $. + +a. Find the velocity of the rocket 3 seconds after being fired. + +b. Find the acceleration of the rocket 3 seconds after being fired. + +154. A ball is thrown downward with a speed of 8 ft/s from the top of a 64-foot-tall building. After $ t $ seconds, its height above the ground is given by $ s(t) = -16t^2 - 8t + 64 $. + +a. Determine how long it takes for the ball to hit the ground. + +b. Determine the velocity of the ball when it hits the ground. + +155. The position function $ s(t) = t^2 - 3t - 4 $ represents the position of the back of a car backing out of a driveway and then driving in a straight line, where $ s $ is in feet and $ t $ is in seconds. In this case, $ s(t) = 0 $ represents the time at which the back of the car is at the garage door, so $ s(0) = -4 $ is the starting position of the car, 4 feet inside the garage. + +a. Determine the velocity of the car when $ s(t) = 0 $. + +b. Determine the velocity of the car when $ s(t) = 14 $. + +156. The position of a hummingbird flying along a straight line in $ t $ seconds is given by $ s(t) = 3t^3 - 7t $ meters. + +a. Determine the velocity of the bird at $ t = 1 $ sec. + +b. Determine the acceleration of the bird at $ t = 1 $ sec. + +c. Determine the acceleration of the bird when the velocity equals 0. + +157. A potato is launched vertically upward with an initial velocity of 100 ft/s from a potato gun at the top of an 85-foot-tall building. The distance in feet that the potato travels from the ground after $ t $ seconds is given by $ s(t) = -16t^2 + 100t + 85 $. + +a. Find the velocity of the potato after 0.5 s and 5.75 s. + +b. Find the speed of the potato at 0.5 s and 5.75 s. + +c. Determine when the potato reaches its maximum height. + +d. Find the acceleration of the potato at 0.5 s and 1.5 s. + +e. Determine how long the potato is in the air. + +f. Determine the velocity of the potato upon hitting the ground. + +158. The position function $ s(t) = t^3 - 8t $ gives the position in miles of a freight train where east is the positive direction and $ t $ is measured in hours. + +a. Determine the direction the train is traveling when $ s(t) = 0 $. + +b. Determine the direction the train is traveling when $ s(t) = 0 $. + +c. Determine the time intervals when the train is slowing down or speeding up. + +159. The following graph shows the position $ y = s(t) $ of an object moving along a straight line. + +![Graph of position function](image) + +a. Use the graph of the position function to determine the time intervals when the velocity is positive, negative, or zero. + +b. Sketch the graph of the velocity function. + +c. Use the graph of the velocity function to determine the time intervals when the acceleration is positive, negative, or zero. + +d. Determine the time intervals when the object is speeding up or slowing down. \ No newline at end of file diff --git a/olmocr/bench/sample_data/olmocr_pipeline/small_page_size_pg1_repeat1.md b/olmocr/bench/sample_data/olmocr_pipeline/small_page_size_pg1_repeat1.md new file mode 100644 index 0000000..041a0fb --- /dev/null +++ b/olmocr/bench/sample_data/olmocr_pipeline/small_page_size_pg1_repeat1.md @@ -0,0 +1,21 @@ +any—was very trifling. Since the use of bones has, however, become general, the turnip crop has been, in many instances, ten-fold, and in few less than four or five-fold its former bulk. All the succeeding crops of grain and seeds have been amazingly increased, and, upon the four or five-shift system, there is no doubt the land will go on progressively improving, requiring a less quantity of bones annually, from its increased fertility and power." + +On light loams, the returns to the Doncaster Committee give bones a preference to farm-yard dung. And we learn that, upon the calcareous soil of the Yorkshire Wolds, heavy crops of turnips have been raised from 16 bushels per acre of bones, while in the same field, and under similar circumstances, but manured from the farm-yard at the rate of from 8 to 10 tons per acre, the turnips have been of the most inferior description. + +On peat soils, if previously drained and laid dry, their advantages are reported to be so striking, that from fifteen to twenty bushels of dust per acre, drilled, have been also found to very far surpass the ordinary dressing of stable-dung, and even of lime and pigeons'-dung. + +On gravels, the reports are meagre and contradictory, though perhaps reconcilable in principle, as it has been justly observed, that "a gravelly soil may embrace every variety of texture and quality, from the light dry sand to the water-logged yellow clay—preserving in each the necessary admixture of stones and grit." To wet gravel, their application has been found decidedly unfavourable. + +**ANALYSIS.** + +An examination of the component parts of soils, and of the power of bones, when applied to them as manure, would go far to explain the irregularity of their different effects upon various kinds of soil. Bone is known to consist of about equal parts of earthy and animal matter; the former chiefly composed of gypsum—which is of so indestructible a nature as to have been termed, by early chemists, the "earth of bones"—and a small portion of carbonate of lime; from which we may conclude that probably half the weight of bones is in the greater part consumed by plants as direct nourishment in their state of growth, and that the remainder is more gradually absorbed by the soil, as well also as by the plants; for lime, though in small amount, is always present, in greater or less quantity, in all vegetable substances. + +"The quantity of earthy matter varies according to the age of the animal; and, in like manner, the quantity of animal matter varies also in proportion to the condition of the animal. In the best kinds of bones for manure, viz., those from fat young animals, perhaps the following proportions may give an approximation to the relative quantities of each in 100 parts: + +| Earthy and saline matter | 40 | +| Cartilage and jelly | 40 | +| Fatty matter | 20 | + +The soft parts thus form, in the best bone, about sixty, and upon an average, perhaps, amount to fifty per cent., which are almost entirely constituted of the same elements of plants, and all of them, sooner or later, liable to be dissolved and absorbed by the roots. The cartilage, indeed, when the bones have been buried in a dry situation, is very indestructible; but when exposed to the action of air, water, soil, and vegetation, will probably pass into the state of jelly, and be dissolved, or otherwise decomposed, + +* Doncaster Report, p. 8. \ No newline at end of file diff --git a/olmocr/bench/sample_data/olmocr_pipeline/test-graphical-text_pg1_repeat1.md b/olmocr/bench/sample_data/olmocr_pipeline/test-graphical-text_pg1_repeat1.md new file mode 100644 index 0000000..9192afb --- /dev/null +++ b/olmocr/bench/sample_data/olmocr_pipeline/test-graphical-text_pg1_repeat1.md @@ -0,0 +1,3 @@ +THE POWER OF STORYTELLING +FOR LEADERS +ดร.วิทย์ สิทธิเวคิน \ No newline at end of file diff --git a/olmocr/bench/sample_data/pdfs/blank_book_pg1.pdf b/olmocr/bench/sample_data/pdfs/blank_book_pg1.pdf new file mode 100644 index 0000000..fecbd29 Binary files /dev/null and b/olmocr/bench/sample_data/pdfs/blank_book_pg1.pdf differ diff --git a/olmocr/bench/sample_data/pdfs/buildingnotes.pdf b/olmocr/bench/sample_data/pdfs/buildingnotes.pdf new file mode 100644 index 0000000..4f2348f Binary files /dev/null and b/olmocr/bench/sample_data/pdfs/buildingnotes.pdf differ diff --git a/olmocr/bench/sample_data/pdfs/discoverworld_crazy_table4.pdf b/olmocr/bench/sample_data/pdfs/discoverworld_crazy_table4.pdf new file mode 100644 index 0000000..9d5f15d Binary files /dev/null and b/olmocr/bench/sample_data/pdfs/discoverworld_crazy_table4.pdf differ diff --git a/olmocr/bench/sample_data/pdfs/earnings.pdf b/olmocr/bench/sample_data/pdfs/earnings.pdf new file mode 100644 index 0000000..3ebcf86 Binary files /dev/null and b/olmocr/bench/sample_data/pdfs/earnings.pdf differ diff --git a/olmocr/bench/sample_data/pdfs/headers_footers/ff0f0b22c55d8b90dd77d153f48e144fc9db_pg2.pdf b/olmocr/bench/sample_data/pdfs/headers_footers/ff0f0b22c55d8b90dd77d153f48e144fc9db_pg2.pdf new file mode 100644 index 0000000..2011a3a Binary files /dev/null and b/olmocr/bench/sample_data/pdfs/headers_footers/ff0f0b22c55d8b90dd77d153f48e144fc9db_pg2.pdf differ diff --git a/olmocr/bench/sample_data/pdfs/headers_footers/ff1fc6a205ad039139ce566851b6b260c929_pg1.pdf b/olmocr/bench/sample_data/pdfs/headers_footers/ff1fc6a205ad039139ce566851b6b260c929_pg1.pdf new file mode 100644 index 0000000..4284191 Binary files /dev/null and b/olmocr/bench/sample_data/pdfs/headers_footers/ff1fc6a205ad039139ce566851b6b260c929_pg1.pdf differ diff --git a/olmocr/bench/sample_data/pdfs/headers_footers/ff3d6e051903fe5ca9bc172ece14964c5632_pg1.pdf b/olmocr/bench/sample_data/pdfs/headers_footers/ff3d6e051903fe5ca9bc172ece14964c5632_pg1.pdf new file mode 100644 index 0000000..2e96b54 Binary files /dev/null and b/olmocr/bench/sample_data/pdfs/headers_footers/ff3d6e051903fe5ca9bc172ece14964c5632_pg1.pdf differ diff --git a/olmocr/bench/sample_data/pdfs/headers_footers/ff4f7dad78081cff727d19ab51c181d4a661_pg1.pdf b/olmocr/bench/sample_data/pdfs/headers_footers/ff4f7dad78081cff727d19ab51c181d4a661_pg1.pdf new file mode 100644 index 0000000..7c913bd Binary files /dev/null and b/olmocr/bench/sample_data/pdfs/headers_footers/ff4f7dad78081cff727d19ab51c181d4a661_pg1.pdf differ diff --git a/olmocr/bench/sample_data/pdfs/headers_footers/ff518b1240a66978f22035528ccb029450b5_pg2.pdf b/olmocr/bench/sample_data/pdfs/headers_footers/ff518b1240a66978f22035528ccb029450b5_pg2.pdf new file mode 100644 index 0000000..45e0903 Binary files /dev/null and b/olmocr/bench/sample_data/pdfs/headers_footers/ff518b1240a66978f22035528ccb029450b5_pg2.pdf differ diff --git a/olmocr/bench/sample_data/pdfs/headers_footers/ffaac214730d2b8c2ec842e3618ccb9c4259_pg1.pdf b/olmocr/bench/sample_data/pdfs/headers_footers/ffaac214730d2b8c2ec842e3618ccb9c4259_pg1.pdf new file mode 100644 index 0000000..bd5bc04 Binary files /dev/null and b/olmocr/bench/sample_data/pdfs/headers_footers/ffaac214730d2b8c2ec842e3618ccb9c4259_pg1.pdf differ diff --git a/olmocr/bench/sample_data/pdfs/headers_footers/fff590bed29a2854ac1f874dad5752ede1aa_pg1.pdf b/olmocr/bench/sample_data/pdfs/headers_footers/fff590bed29a2854ac1f874dad5752ede1aa_pg1.pdf new file mode 100644 index 0000000..ded2590 Binary files /dev/null and b/olmocr/bench/sample_data/pdfs/headers_footers/fff590bed29a2854ac1f874dad5752ede1aa_pg1.pdf differ diff --git a/olmocr/bench/sample_data/pdfs/lincoln_letter.pdf b/olmocr/bench/sample_data/pdfs/lincoln_letter.pdf new file mode 100644 index 0000000..615ea3c Binary files /dev/null and b/olmocr/bench/sample_data/pdfs/lincoln_letter.pdf differ diff --git a/olmocr/bench/sample_data/pdfs/math_2503_04086.pdf b/olmocr/bench/sample_data/pdfs/math_2503_04086.pdf new file mode 100644 index 0000000..ceaf665 Binary files /dev/null and b/olmocr/bench/sample_data/pdfs/math_2503_04086.pdf differ diff --git a/olmocr/bench/sample_data/pdfs/mathfuncs.pdf b/olmocr/bench/sample_data/pdfs/mathfuncs.pdf new file mode 100644 index 0000000..5dba28d Binary files /dev/null and b/olmocr/bench/sample_data/pdfs/mathfuncs.pdf differ diff --git a/olmocr/bench/sample_data/pdfs/mathfuncs_colswitch.pdf b/olmocr/bench/sample_data/pdfs/mathfuncs_colswitch.pdf new file mode 100644 index 0000000..371ccdf Binary files /dev/null and b/olmocr/bench/sample_data/pdfs/mathfuncs_colswitch.pdf differ diff --git a/olmocr/bench/sample_data/pdfs/mattsnotes.pdf b/olmocr/bench/sample_data/pdfs/mattsnotes.pdf new file mode 100644 index 0000000..b0821cc Binary files /dev/null and b/olmocr/bench/sample_data/pdfs/mattsnotes.pdf differ diff --git a/olmocr/bench/sample_data/pdfs/multi_column_miss.pdf b/olmocr/bench/sample_data/pdfs/multi_column_miss.pdf new file mode 100644 index 0000000..e85ae01 Binary files /dev/null and b/olmocr/bench/sample_data/pdfs/multi_column_miss.pdf differ diff --git a/olmocr/bench/sample_data/pdfs/olmo2-pg4.pdf b/olmocr/bench/sample_data/pdfs/olmo2-pg4.pdf new file mode 100644 index 0000000..a6ecd52 Binary files /dev/null and b/olmocr/bench/sample_data/pdfs/olmo2-pg4.pdf differ diff --git a/olmocr/bench/sample_data/pdfs/openstax_caculus_pg_273.pdf b/olmocr/bench/sample_data/pdfs/openstax_caculus_pg_273.pdf new file mode 100644 index 0000000..77310b9 Binary files /dev/null and b/olmocr/bench/sample_data/pdfs/openstax_caculus_pg_273.pdf differ diff --git a/olmocr/bench/sample_data/pdfs/small_page_size.pdf b/olmocr/bench/sample_data/pdfs/small_page_size.pdf new file mode 100644 index 0000000..c9f8806 Binary files /dev/null and b/olmocr/bench/sample_data/pdfs/small_page_size.pdf differ diff --git a/olmocr/bench/sample_data/pdfs/test-graphical-text.pdf b/olmocr/bench/sample_data/pdfs/test-graphical-text.pdf new file mode 100644 index 0000000..a120efc Binary files /dev/null and b/olmocr/bench/sample_data/pdfs/test-graphical-text.pdf differ diff --git a/olmocr/bench/scripts/convert_all.sh b/olmocr/bench/scripts/convert_all.sh new file mode 100755 index 0000000..bdccfc1 --- /dev/null +++ b/olmocr/bench/scripts/convert_all.sh @@ -0,0 +1,290 @@ +#!/bin/bash + +# Exit on error but allow the trap to execute +set -e + +# Global variables +SERVER_PID="" +BENCH_DIR="" # New variable to store the benchmark directory + +# Function to show usage information +show_usage() { + echo "Usage: $0 [--dir ]" + echo " --dir Specify the benchmark data directory (default: olmOCR-bench/bench_data/)" + exit 1 +} + +# Parse command line arguments +while [[ $# -gt 0 ]]; do + key="$1" + case $key in + --dir) + BENCH_DIR="$2" + shift 2 + ;; + -h|--help) + show_usage + ;; + *) + echo "Unknown option: $1" + show_usage + ;; + esac +done + +# Set default directory if not provided +if [ -z "$BENCH_DIR" ]; then + BENCH_DIR="olmOCR-bench/bench_data/" + echo "[INFO] Using default benchmark directory: $BENCH_DIR" +else + echo "[INFO] Using benchmark directory: $BENCH_DIR" +fi + +# Trap function to handle Ctrl+C (SIGINT) +cleanup() { + echo -e "\n[INFO] Received interrupt signal. Cleaning up..." + + # Find and kill any Python processes started by this script + echo "[INFO] Stopping any running Python processes" + pkill -P $$ python || true + + # Stop server if running + if [ -n "$SERVER_PID" ] && kill -0 "$SERVER_PID" 2>/dev/null; then + echo "[INFO] Stopping server (PID: $SERVER_PID)" + kill -TERM "$SERVER_PID" 2>/dev/null || true + wait "$SERVER_PID" 2>/dev/null || true + fi + + pkill vllm + + echo "[INFO] Cleanup complete. Exiting." + exit 1 +} + +# Set the trap for SIGINT (Ctrl+C) +trap cleanup SIGINT + +# Function to check if port 30000 is in use +check_port() { + port=30000 + echo "[INFO] Checking if port $port is available..." + + if command -v lsof >/dev/null 2>&1; then + # Linux/macOS + if lsof -i :$port >/dev/null 2>&1; then + echo "[ERROR] Port $port is already in use. Process details:" + lsof -i :$port + echo "[ERROR] Please stop the process using this port and try again." + echo " You can use: kill -9 " + return 1 + fi + elif command -v netstat >/dev/null 2>&1; then + # Windows/other systems with netstat + if netstat -an | grep -q ":$port "; then + echo "[ERROR] Port $port is already in use. Process details:" + if command -v findstr >/dev/null 2>&1; then + # Windows + netstat -ano | findstr ":$port" + echo "[ERROR] Please stop the process using this port and try again." + echo " You can use: taskkill /F /PID " + else + netstat -an | grep ":$port " + echo "[ERROR] Please stop the process using this port and try again." + fi + return 1 + fi + else + # Fallback method using nc if available + if command -v nc >/dev/null 2>&1; then + nc -z localhost $port >/dev/null 2>&1 + if [ $? -eq 0 ]; then + echo "[ERROR] Port $port is already in use, but cannot determine which process." + echo "[ERROR] Please ensure port $port is available before continuing." + return 1 + fi + else + echo "[WARNING] Cannot check if port $port is in use (neither lsof, netstat, nor nc available)." + echo "[WARNING] Continuing anyway, but this might fail if the port is already in use." + return 0 + fi + fi + + echo "[INFO] Port $port is available." + return 0 +} + +# Function to create conda environment if it doesn't exist +create_conda_env() { + env_name=$1 + python_version=$2 + + # Check if environment exists + if conda info --envs | grep -q "^$env_name "; then + echo "Environment $env_name already exists, using it." + else + echo "Creating conda environment: $env_name" + conda create -y -n $env_name python=$python_version + fi +} + +# Generic function to start a server (either vllm or sglang) +start_server() { + server_type=$1 + model_name=$2 + shift 2 # Remove server type and model name from the argument list + + echo "Starting $server_type server for model: $model_name" + echo "Additional arguments: $@" + + if [ "$server_type" = "sglang" ]; then + python -m sglang.launch_server --port 30000 --model "$model_name" "$@" & + elif [ "$server_type" = "vllm" ]; then + vllm serve $model_name --port 30000 "$@" & + else + echo "Unsupported server type: $server_type" + exit 1 + fi + SERVER_PID=$! + + # Check if the server process is running + if ! kill -0 $SERVER_PID 2>/dev/null; then + echo "Failed to start server process. Exiting." + exit 1 + fi + + # Wait for the server to be ready by checking the models endpoint + echo "Waiting for server to be ready..." + max_attempts=300 + attempt=0 + while [ $attempt -lt $max_attempts ]; do + if curl -s "http://localhost:30000/v1/models" -o /dev/null -w "%{http_code}" | grep -q "200"; then + echo "Server is ready!" + return 0 + fi + attempt=$((attempt + 1)) + echo "Waiting for server... attempt $attempt/$max_attempts" + sleep 2 + done + + echo "Server failed to become ready after multiple attempts. Exiting." + kill $SERVER_PID + SERVER_PID="" + exit 1 +} + +# Function to stop the server +stop_server() { + echo "Stopping server with PID: $SERVER_PID" + if [ -n "$SERVER_PID" ] && kill -0 "$SERVER_PID" 2>/dev/null; then + kill $SERVER_PID + wait $SERVER_PID 2>/dev/null || true + echo "Server stopped." + else + echo "No server to stop." + fi + SERVER_PID="" +} + +# Create and activate olmocr environment +create_conda_env "olmocr" "3.11" +source $(conda info --base)/etc/profile.d/conda.sh +source activate olmocr +pip install -e .[bench] + +# Run olmocr benchmarks, exactly as the pipeline.py does it +echo "Running olmocr benchmarks..." +python -m olmocr.bench.convert olmocr_pipeline --parallel 50 --dir "$BENCH_DIR" + +# Install marker-pdf and run benchmarks +echo "Installing marker-pdf and running benchmarks..." +pip install marker-pdf==1.6.1 +python -m olmocr.bench.convert marker --dir "$BENCH_DIR" + +# Install verovio and run benchmarks +# echo "Installing verovio and running benchmarks..." +# pip install verovio +# python -m olmocr.bench.convert gotocr + +# Run chatgpt benchmarks +echo "Running chatgpt benchmarks..." +python -m olmocr.bench.convert chatgpt --dir "$BENCH_DIR" --parallel 4 +#python -m olmocr.bench.convert chatgpt:name=chatgpt45:model=gpt-4.5-preview-2025-02-27 + +# Run gemini benchmarks +echo "Running gemini benchmarks..." +python -m olmocr.bench.convert gemini:name=gemini_flash2:model=gemini-2.0-flash --parallel 4 --dir "$BENCH_DIR" + +echo "Running mistral..." +pip install mistralai +python -m olmocr.bench.convert --dir "$BENCH_DIR" --parallel 4 mistral + +# Run raw server benchmarks with generic server function +# For each model, start server, run benchmark, then stop server + +# Check port availability at script start +check_port || exit 1 + +# olmocr_base_temp0_1 using sglang server +# start_server sglang "allenai/olmOCR-7B-0225-preview" --chat-template qwen2-vl --mem-fraction-static 0.7 +# python -m olmocr.bench.convert --dir "$BENCH_DIR" server:name=olmocr_base_temp0_0:model=allenai/olmOCR-7B-0225-preview:temperature=0.0:prompt_template=fine_tune:response_template=json --repeats 1 --parallel 50 +# python -m olmocr.bench.convert --dir "$BENCH_DIR" server:name=olmocr_base_temp0_1:model=allenai/olmOCR-7B-0225-preview:temperature=0.1:prompt_template=fine_tune:response_template=json --repeats 5 --parallel 50 +# python -m olmocr.bench.convert --dir "$BENCH_DIR" server:name=olmocr_base_temp0_2:model=allenai/olmOCR-7B-0225-preview:temperature=0.2:prompt_template=fine_tune:response_template=json --repeats 1 --parallel 50 +# python -m olmocr.bench.convert --dir "$BENCH_DIR" server:name=olmocr_base_temp0_3:model=allenai/olmOCR-7B-0225-preview:temperature=0.3:prompt_template=fine_tune:response_template=json --repeats 1 --parallel 50 +# python -m olmocr.bench.convert --dir "$BENCH_DIR" server:name=olmocr_base_temp0_4:model=allenai/olmOCR-7B-0225-preview:temperature=0.4:prompt_template=fine_tune:response_template=json --repeats 1 --parallel 50 +# python -m olmocr.bench.convert --dir "$BENCH_DIR" server:name=olmocr_base_temp0_5:model=allenai/olmOCR-7B-0225-preview:temperature=0.5:prompt_template=fine_tune:response_template=json --repeats 1 --parallel 50 +# python -m olmocr.bench.convert --dir "$BENCH_DIR" server:name=olmocr_base_temp0_6:model=allenai/olmOCR-7B-0225-preview:temperature=0.6:prompt_template=fine_tune:response_template=json --repeats 1 --parallel 50 +# python -m olmocr.bench.convert --dir "$BENCH_DIR" server:name=olmocr_base_temp0_7:model=allenai/olmOCR-7B-0225-preview:temperature=0.7:prompt_template=fine_tune:response_template=json --repeats 5 --parallel 50 + +# python -m olmocr.bench.convert server:name=olmocr_base_temp0_1:model=allenai/olmOCR-7B-0225-preview:temperature=0.1:prompt_template=fine_tune:response_template=json --repeats 5 --parallel 50 +# python -m olmocr.bench.convert server:name=olmocr_base_temp0_8:model=allenai/olmOCR-7B-0225-preview:temperature=0.8:prompt_template=fine_tune:response_template=json --repeats 5 --parallel 50 +# stop_server + +# start_server vllm "allenai/olmOCR-7B-0225-preview" +# python -m olmocr.bench.convert --dir "$BENCH_DIR" server:name=olmocr_base_temp_vllm0_1:model=allenai/olmOCR-7B-0225-preview:temperature=0.1:prompt_template=fine_tune:response_template=json --repeats 5 --parallel 50 +# python -m olmocr.bench.convert --dir "$BENCH_DIR" server:name=olmocr_base_temp_vllm0_7:model=allenai/olmOCR-7B-0225-preview:temperature=0.7:prompt_template=fine_tune:response_template=json --repeats 5 --parallel 50 + +# python -m olmocr.bench.convert server:name=olmocr_base_vllm_temp0_1:model=allenai/olmOCR-7B-0225-preview:temperature=0.1:prompt_template=fine_tune:response_template=json --repeats 5 --parallel 50 +# python -m olmocr.bench.convert server:name=olmocr_base_vllm_temp0_8:model=allenai/olmOCR-7B-0225-preview:temperature=0.8:prompt_template=fine_tune:response_template=json --repeats 5 --parallel 50 +# stop_server + +# Feel free to enable if you want, but qwen2 raw is pretty low scoring +# qwen2_vl_7b using sglang server +# start_server sglang "Qwen/Qwen2-VL-7B-Instruct" --chat-template qwen2-vl --mem-fraction-static 0.7 +# python -m olmocr.bench.convert server:name=qwen2_vl_7b:model=Qwen/Qwen2-VL-7B-Instruct:temperature=0.1:prompt_template=full:response_template=plain --repeats 5 --parallel 50 +# stop_server + +# qwen2.5 works best with vllm for now, in a fresh environment +create_conda_env "vllm" "3.11" +source activate vllm +pip install -e .[bench] +pip install --upgrade vllm==0.8.3 + + +start_server vllm "Qwen/Qwen2.5-VL-7B-Instruct" --max-model-len 8192 +python -m olmocr.bench.convert --dir "$BENCH_DIR" server:name=qwen25vl_prompt7:model=Qwen/Qwen2.5-VL-7B-Instruct:temperature=0.1:prompt_template=basic:response_template=plain --parallel 50 +stop_server + +start_server vllm "reducto/RolmOCR" --max-model-len 8192 +python -m olmocr.bench.convert --dir "$BENCH_DIR" rolmocr --parallel 50 +stop_server + +# TODO: Fix this, I was not able to get it to all install successfully +# Create and activate mineru environment +# create_conda_env "mineru" "3.11" +# source activate mineru + +# Install magic-pdf and run benchmarks +# echo "Installing magic-pdf and running mineru benchmarks..." +# pip install -U "magic-pdf[full]==1.2.2" --extra-index-url https://wheels.myhloli.com +# python -m pip install paddlepaddle==3.0.0rc1 -i https://www.paddlepaddle.org.cn/packages/stable/cpu/ +# pip install huggingface_hub Pillow paddleocr ultralytics doclayout-yolo pycocotools +# wget https://github.com/opendatalab/MinerU/raw/master/scripts/download_models_hf.py -O download_models_hf.py +# python download_models_hf.py +# python -m olmocr.bench.convert mineru + +# Final cleanup +if [ -n "$SERVER_PID" ] && kill -0 $SERVER_PID 2>/dev/null; then + stop_server +fi + +echo "All benchmarks completed successfully." \ No newline at end of file diff --git a/olmocr/bench/scripts/difference_viewer.py b/olmocr/bench/scripts/difference_viewer.py new file mode 100644 index 0000000..3babb7d --- /dev/null +++ b/olmocr/bench/scripts/difference_viewer.py @@ -0,0 +1,468 @@ +#!/usr/bin/env python3 +import argparse +import glob +import json +import os +import sys +from collections import defaultdict + +from runners.run_chatgpt import run_chatgpt +from runners.run_gemini import run_gemini + +from olmocr.data.renderpdf import render_pdf_to_base64png + + +def parse_rules_file(file_path): + """Parse the rules file and organize rules by PDF.""" + pdf_rules = defaultdict(list) + + with open(file_path, "r") as f: + for line in f: + line = line.strip() + if not line: + continue + try: + rule = json.loads(line) + if "pdf" in rule: + pdf_rules[rule["pdf"]].append(rule) + except json.JSONDecodeError: + print(f"Warning: Could not parse line as JSON: {line}") + + return pdf_rules + + +def get_model_outputs(pdf_path): + """Get outputs from both models for a given PDF.""" + try: + print(f"Attempting to process PDF: {pdf_path}") + print(f"File exists: {os.path.exists(pdf_path)}") + + chatgpt_output = run_chatgpt(pdf_path) + gemini_output = run_gemini(pdf_path) + return chatgpt_output, gemini_output + except Exception as e: + print(f"Error getting model outputs for {pdf_path}: {str(e)}") + return f"Error: {str(e)}", f"Error: {str(e)}" + + +def find_pdfs_in_directory(directory): + """Find all PDF files in the given directory.""" + if not os.path.isdir(directory): + print(f"Warning: {directory} is not a directory.") + return [] + pdf_files = [] + for ext in ["pdf", "PDF"]: + pattern = os.path.join(directory, f"*.{ext}") + pdf_files.extend(glob.glob(pattern)) + + print(f"Found {len(pdf_files)} PDF files in {directory}") + for pdf in pdf_files: + print(f" - {pdf}") + return pdf_files + + +def generate_html(pdf_rules, rules_file_path, pdfs_to_process=None): + """Generate the HTML page with PDF renderings and model outputs.""" + pdf_paths = [] + + if pdfs_to_process: + for pdf_item in pdfs_to_process: + if os.path.isdir(pdf_item): + pdf_paths.extend(find_pdfs_in_directory(pdf_item)) + elif os.path.isfile(pdf_item): + pdf_paths.append(pdf_item) + else: + print(f"Warning: {pdf_item} is neither a valid file nor directory") + else: + pdf_base_dir = os.path.join(os.path.dirname(rules_file_path), "pdfs") + pdf_paths = [os.path.join(pdf_base_dir, pdf_name) for pdf_name in list(pdf_rules.keys())[:10]] + pdf_paths = list(set(pdf_paths)) + print("Processing the following PDFs:") + for path in pdf_paths: + print(f" - {path} (exists: {os.path.exists(path)})") + + html = """ + + + + + + PDF Model Comparison + + + +

PDF Model Comparison

+ """ + + for pdf_path in pdf_paths: + if not os.path.exists(pdf_path) or not os.path.isfile(pdf_path): + print(f"Skipping non-existent or non-file path: {pdf_path}") + continue + pdf_name = os.path.basename(pdf_path) + pdf_id = pdf_name.replace(".", "-") + try: + print(f"Rendering PDF: {pdf_path}") + base64_img = render_pdf_to_base64png(pdf_path, 0) + img_html = f' ${pdf_name}$ ' + except Exception as e: + print(f"Error rendering PDF: {str(e)}") + img_html = f'

Error rendering PDF: {str(e)}

' + print(f"Getting model outputs for: {pdf_path}") + chatgpt_output, gemini_output = get_model_outputs(pdf_path) + + html += f""" +

+ {pdf_name} + +

+ {img_html} +

+ + + + + +

+ +

ChatGPT Output

{chatgpt_output}

Gemini Output

{gemini_output}

Differences

Loading differences...

+ """ + + html += """ +

+ + + + + """ + + return html + + +def main(): + parser = argparse.ArgumentParser(description="Generate an HTML visualization for comparing AI model outputs on PDFs.") + parser.add_argument("-r", "--rules_file", help="Rules file path", default="./sample_data/dataset.jsonl") + parser.add_argument("-o", "--output", help="Output HTML file path", default="pdf_model_comparison.html") + parser.add_argument("-p", "--pdfs", nargs="+", help="Specific PDF files or directories to process") + parser.add_argument("-l", "--limit", type=int, help="Limit the number of PDFs to process", default=10) + + args = parser.parse_args() + + if not os.path.exists(args.rules_file): + print(f"Error: Rules file not found: {args.rules_file}") + sys.exit(1) + if args.pdfs: + for pdf_path in args.pdfs: + if not os.path.exists(pdf_path): + print(f"WARNING: Path not found: {pdf_path}") + + pdf_rules = parse_rules_file(args.rules_file) + html = generate_html(pdf_rules, args.rules_file, args.pdfs) + + with open(args.output, "w") as f: + f.write(html) + + print(f"HTML visualization created: {args.output}") + print("Open this file in a web browser to view and rate model outputs.") + + +if __name__ == "__main__": + main() diff --git a/olmocr/bench/scripts/rotate_pdfs.py b/olmocr/bench/scripts/rotate_pdfs.py new file mode 100755 index 0000000..073a7b0 --- /dev/null +++ b/olmocr/bench/scripts/rotate_pdfs.py @@ -0,0 +1,82 @@ +#!/usr/bin/env python3 +import json +import os +import random +import shutil +from collections import defaultdict +from pathlib import Path + + +def main(): + # Set paths + bench_data_dir = Path("./olmOCR-bench/bench_data") + pdfs_dir = Path("./olmOCR-bench/bench_data/pdfs") + rotated_pdfs_dir = pdfs_dir / "rotated" + output_jsonl = Path("rotated.jsonl") + + # Create rotated directory if it doesn't exist + rotated_pdfs_dir.mkdir(parents=True, exist_ok=True) + + # Load all JSONL files and group by PDF + pdf_groups = defaultdict(list) + + print("Loading JSONL files...") + for jsonl_file in bench_data_dir.glob("*.jsonl"): + print(f" Reading {jsonl_file}") + with open(jsonl_file, "r") as f: + for line in f: + try: + data = json.loads(line.strip()) + if "pdf" in data: + pdf_groups[data["pdf"]].append(data) + except json.JSONDecodeError: + continue + + print(f"Found {len(pdf_groups)} unique PDF groups") + + # Randomly select 10% of PDF groups + num_to_select = max(1, int(len(pdf_groups) * 0.1)) + selected_pdfs = random.sample(list(pdf_groups.keys()), num_to_select) + + print(f"Selected {num_to_select} PDF groups (10% of total)") + + # Write selected entries to rotated.jsonl + print(f"Writing selected entries to {output_jsonl}") + with open(output_jsonl, "w") as f: + for pdf_name in selected_pdfs: + for entry in pdf_groups[pdf_name]: + f.write(json.dumps(entry) + "\n") + + # Copy corresponding PDF files + print("Copying PDF files to rotated directory...") + copied_count = 0 + missing_count = 0 + + for pdf_name in selected_pdfs: + # Try to find the PDF in subdirectories + pdf_found = False + print(pdf_name) + source_path = pdfs_dir / pdf_name + if source_path.exists(): + dest_path = rotated_pdfs_dir / os.path.basename(pdf_name) + print(f" Copying {source_path} -> {dest_path}") + shutil.copy2(source_path, dest_path) + copied_count += 1 + pdf_found = True + + if not pdf_found: + print(f" Warning: PDF not found: {pdf_name}") + missing_count += 1 + + print(f"\nSummary:") + print(f" Total PDF groups: {len(pdf_groups)}") + print(f" Selected groups: {num_to_select}") + print(f" PDFs copied: {copied_count}") + if missing_count > 0: + print(f" PDFs not found: {missing_count}") + print(f" Output JSONL: {output_jsonl}") + print(f" Rotated PDFs directory: {rotated_pdfs_dir}") + + +if __name__ == "__main__": + main() diff --git a/olmocr/bench/scripts/rotate_pdfs_random.sh b/olmocr/bench/scripts/rotate_pdfs_random.sh new file mode 100755 index 0000000..a3f636b --- /dev/null +++ b/olmocr/bench/scripts/rotate_pdfs_random.sh @@ -0,0 +1,67 @@ +#!/bin/bash + +# Directory containing PDFs to rotate +PDF_DIR="/home/ubuntu/olmocr/olmOCR-bench-0825/bench_data/pdfs/rotated" + +# Check if directory exists +if [ ! -d "$PDF_DIR" ]; then + echo "Error: Directory $PDF_DIR does not exist" + exit 1 +fi + +# Check if qpdf is installed (preferred for PDF rotation) +if ! command -v qpdf &> /dev/null; then + echo "qpdf is not installed. Installing..." + sudo apt-get update && sudo apt-get install -y qpdf +fi + +# Counter for processed files +total=0 +success=0 +failed=0 + +echo "Processing PDFs in $PDF_DIR" +echo "----------------------------------------" + +# Process each PDF file +for pdf_file in "$PDF_DIR"/*.pdf; do + # Check if any PDF files exist + if [ ! -f "$pdf_file" ]; then + echo "No PDF files found in $PDF_DIR" + exit 1 + fi + + # Get filename + filename=$(basename "$pdf_file") + + # Randomly select rotation angle (90, 180, or 270) + angles=(90 180 270) + rotation=${angles[$RANDOM % ${#angles[@]}]} + + echo "Rotating $filename by $rotation degrees..." + + # Create temporary file for rotated PDF + temp_file="${pdf_file}.tmp" + + # Rotate the PDF using qpdf + if qpdf "$pdf_file" "$temp_file" --rotate=+$rotation; then + # Replace original with rotated version + mv "$temp_file" "$pdf_file" + echo " ✓ Successfully rotated $filename by $rotation degrees" + ((success++)) + else + echo " ✗ Failed to rotate $filename" + rm -f "$temp_file" + ((failed++)) + fi + + ((total++)) +done + +echo "----------------------------------------" +echo "Summary:" +echo " Total PDFs processed: $total" +echo " Successfully rotated: $success" +if [ $failed -gt 0 ]; then + echo " Failed: $failed" +fi \ No newline at end of file diff --git a/olmocr/bench/scripts/run_difference.py b/olmocr/bench/scripts/run_difference.py new file mode 100644 index 0000000..e34464c --- /dev/null +++ b/olmocr/bench/scripts/run_difference.py @@ -0,0 +1,64 @@ +import os + +from openai import OpenAI +from runners.run_chatgpt import run_chatgpt +from runners.run_gemini import run_gemini + +from olmocr.data.renderpdf import render_pdf_to_base64png + + +def build_find_difference_prompt(base_text: str) -> str: + return ( + f"Below is an image of a document page, along with raw textual content previously extracted using different models." + f"Your goal is to carefully identify the differences between the extracted texts from both models and determine which one is more accurate by comparing them with the image." + f"Only return the differences and specify which model extracted the text with higher accuracy.\n" + f"Do not hallucinate.\n" + f"RAW_TEXT_START\n{base_text}\nRAW_TEXT_END" + ) + + +def combined_output(pdf_path: str) -> str: + chatgpt_output = run_chatgpt(pdf_path) + gemini_output = run_gemini(pdf_path) + return f"ChatGPT OUTPUT: \n" f"{chatgpt_output}\n\n" f"Gemini OUTPUT: \n" f"{gemini_output}" + + +def run_difference(pdf_path: str, page_num: int = 1, model: str = "gpt-4o-2024-08-06", temperature: float = 0.1) -> str: + """ + Convert page of a PDF file to markdown using GPT. + + This function renders the first page of the PDF to an image, runs OCR on that image, + and returns the OCR result as a markdown-formatted string. + + Args: + pdf_path (str): The local path to the PDF file. + page_num (int): Which page from document to pass. + model (str): Model used to process. + Temperature (float): Temperature used while utilizing the model. + + Returns: + str: The result in markdown format. + """ + # Convert the first page of the PDF to a base64-encoded PNG image. + image_base64 = render_pdf_to_base64png(pdf_path, page_num=page_num, target_longest_image_dim=2048) + anchor_text = combined_output(pdf_path) + client = OpenAI(api_key=os.getenv("OPENAI_API_KEY")) + + response = client.chat.completions.create( + model=model, + messages=[ + { + "role": "user", + "content": [ + {"type": "text", "text": build_find_difference_prompt(anchor_text)}, + {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{image_base64}"}}, + ], + } + ], + temperature=temperature, + max_tokens=3000, + ) + + raw_response = response.choices[0].message.content + + return raw_response diff --git a/olmocr/bench/scripts/screen_pdfs.py b/olmocr/bench/scripts/screen_pdfs.py new file mode 100644 index 0000000..7c40cef --- /dev/null +++ b/olmocr/bench/scripts/screen_pdfs.py @@ -0,0 +1,168 @@ +""" +Screen synthetic benchmark PDFs for sensitive/private content using OpenAI vision API. + +Flags documents where: + - #1 (resume/CV) is true, OR + - #2 (sensitive PII) is true AND #3 (public consumption) is false AND #4 (academic) is false +""" + +import argparse +import asyncio +import json +import sys +from pathlib import Path + +from openai import AsyncOpenAI +from tqdm.asyncio import tqdm + +sys.path.insert(0, str(Path(__file__).resolve().parent.parent)) +from olmocr.data.renderpdf import render_pdf_to_base64png + +BENCH_DATA = Path(__file__).parent / "bench_data" +PDFS_DIR = BENCH_DATA / "pdfs" + +SCREENING_PROMPT = """\ +Look at this document image and answer each question with ONLY "yes" or "no". + +1. Is this a resume or CV? +2. Does this document contain sensitive personally identifiable information (PII) such as social security numbers, personal addresses, phone numbers, dates of birth, or similar? +3. Is this document meant for public consumption or dissemination? +4. Is this document academic in nature (e.g. journal article, textbook, thesis, lecture notes)? + +Respond as JSON: {"is_resume": bool, "has_sensitive_pii": bool, "is_public": bool, "is_academic": bool} +""" + +MODEL = "gpt-5.4" +MAX_CONCURRENT = 5 + + +def should_flag(result: dict) -> bool: + """Return True if the document should be flagged for review.""" + if result.get("is_resume"): + return True + if result.get("has_sensitive_pii") and not result.get("is_public") and not result.get("is_academic"): + return True + return False + + +async def screen_pdf(client: AsyncOpenAI, pdf_path: str, model: str, semaphore: asyncio.Semaphore) -> dict: + """Render PDF and send screenshot to the API, returning the parsed screening result.""" + # Render using the same pdftoppm approach as the olmocr pipeline + image_base64 = await asyncio.to_thread(render_pdf_to_base64png, pdf_path, 1, 2048) + + async with semaphore: + response = await client.chat.completions.create( + model=model, + messages=[ + { + "role": "user", + "content": [ + { + "type": "image_url", + "image_url": {"url": f"data:image/png;base64,{image_base64}", "detail": "low"}, + }, + {"type": "text", "text": SCREENING_PROMPT}, + ], + } + ], + temperature=0, + max_completion_tokens=200, + ) + + text = response.choices[0].message.content.strip() + # Extract JSON from the response (handle markdown fences) + if "```" in text: + text = text.split("```")[1] + if text.startswith("json"): + text = text[4:] + parsed = json.loads(text) + # Normalize "yes"/"no" strings to booleans + for key in ("is_resume", "has_sensitive_pii", "is_public", "is_academic"): + if isinstance(parsed.get(key), str): + parsed[key] = parsed[key].lower().strip() == "yes" + return parsed + + +async def process_one(client: AsyncOpenAI, category: str, pdf_path: Path, model: str, semaphore: asyncio.Semaphore) -> dict: + """Process a single PDF and return the result entry.""" + rel = f"{category}/{pdf_path.name}" + try: + result = await screen_pdf(client, str(pdf_path), model, semaphore) + flag = should_flag(result) + return {"pdf": rel, "flag": flag, **result} + except Exception as e: + return {"pdf": rel, "flag": None, "error": str(e)} + + +async def main(): + parser = argparse.ArgumentParser(description="Screen synthetic benchmark PDFs for sensitive content") + parser.add_argument("--model", default=MODEL, help=f"OpenAI model to use (default: {MODEL})") + parser.add_argument("--output", default="flagged_pdfs.jsonl", help="Output file for flagged documents") + parser.add_argument("--all-results", default=None, help="Optional file to write ALL results (not just flagged)") + parser.add_argument("--max-concurrent", type=int, default=MAX_CONCURRENT, help=f"Max parallel API requests (default: {MAX_CONCURRENT})") + parser.add_argument("--dry-run", action="store_true", help="List PDFs that would be screened without calling API") + args = parser.parse_args() + + # Collect all synthetic_ folders + synthetic_dirs = sorted(p for p in PDFS_DIR.iterdir() if p.is_dir() and p.name.startswith("synthetic_")) + + # Build list of (category, pdf_path) pairs + pdf_files = [] + for d in synthetic_dirs: + for f in sorted(d.iterdir()): + if f.suffix.lower() == ".pdf": + pdf_files.append((d.name, f)) + + print(f"Found {len(pdf_files)} PDFs across {len(synthetic_dirs)} synthetic categories") + + if args.dry_run: + for cat, p in pdf_files: + print(f" {cat}/{p.name}") + return + + client = AsyncOpenAI() # uses OPENAI_API_KEY env var + semaphore = asyncio.Semaphore(args.max_concurrent) + + # Launch all tasks, bounded by semaphore + tasks = [process_one(client, category, pdf_path, args.model, semaphore) for category, pdf_path in pdf_files] + + all_results = [] + flagged = [] + flagged_count = 0 + error_count = 0 + + pbar = tqdm(total=len(tasks), desc="Screening PDFs", unit="pdf") + for coro in asyncio.as_completed(tasks): + entry = await coro + all_results.append(entry) + + if entry.get("error"): + error_count += 1 + pbar.set_postfix(flagged=flagged_count, errors=error_count) + elif entry["flag"]: + flagged.append(entry) + flagged_count += 1 + pbar.set_postfix(flagged=flagged_count, errors=error_count) + + pbar.update(1) + pbar.close() + + # Write flagged results + with open(args.output, "w") as f: + for entry in flagged: + f.write(json.dumps(entry) + "\n") + print(f"\nWrote {len(flagged)} flagged documents to {args.output}") + + if error_count: + print(f" ({error_count} errors encountered)") + + # Optionally write all results + if args.all_results: + with open(args.all_results, "w") as f: + for entry in all_results: + f.write(json.dumps(entry) + "\n") + print(f"Wrote {len(all_results)} total results to {args.all_results}") + + +if __name__ == "__main__": + asyncio.run(main()) diff --git a/olmocr/bench/scripts/split_table_tests.py b/olmocr/bench/scripts/split_table_tests.py new file mode 100644 index 0000000..6b69f64 --- /dev/null +++ b/olmocr/bench/scripts/split_table_tests.py @@ -0,0 +1,168 @@ +""" +Split table tests with multiple relationships into individual tests. + +For example, if you have a test case with cell=X, up=Y, down=Z, this script +will split it into two cases: one with cell=X, up=Y and another with cell=X, down=Z. + +Usage: + python split_table_tests.py input.jsonl output.jsonl +""" + +import argparse +import json +import sys +from pathlib import Path + +from olmocr.bench.tests import BasePDFTest, TableTest, TestType, load_tests + +# The relationship fields that can be split +RELATIONSHIP_FIELDS = ["up", "down", "left", "right", "top_heading", "left_heading"] + + +def base_test_to_dict(test: BasePDFTest) -> dict: + """ + Convert any BasePDFTest to a dict by extracting all non-None attributes. + """ + from dataclasses import fields, is_dataclass + + result = {} + if is_dataclass(test): + for field in fields(test): + value = getattr(test, field.name) + # Skip None values and empty strings for optional fields + if value is not None and value != "": + result[field.name] = value + return result + + +def test_to_dict_minimal(test: TableTest) -> dict: + """ + Convert a TableTest to a dict, removing empty relationship fields. + """ + result = { + "pdf": test.pdf, + "page": test.page, + "id": test.id, + "type": test.type, + "max_diffs": test.max_diffs, + "cell": test.cell, + } + + # Only include non-empty relationship fields + for field in RELATIONSHIP_FIELDS: + value = getattr(test, field) + if value and value.strip(): + result[field] = value + + # Include other optional fields if set + if test.checked is not None: + result["checked"] = test.checked + if test.url: + result["url"] = test.url + if test.ignore_markdown_tables: + result["ignore_markdown_tables"] = test.ignore_markdown_tables + + return result + + +def split_table_test(test: TableTest) -> list[dict]: + """ + Split a single TableTest with multiple relationships into separate test dicts. + + Each resulting test will have only one relationship field set (plus the cell). + + Args: + test: The TableTest to split + + Returns: + A list of dicts, one for each non-empty relationship + """ + # Find which relationship fields are set (non-empty) + active_relationships = [] + for field in RELATIONSHIP_FIELDS: + value = getattr(test, field) + if value and value.strip(): + active_relationships.append((field, value)) + + # If there's 0 or 1 relationship, no splitting needed - return as minimal dict + if len(active_relationships) <= 1: + return [test_to_dict_minimal(test)] + + # Create a new test dict for each relationship + split_tests = [] + for field, value in active_relationships: + result = { + "pdf": test.pdf, + "page": test.page, + "id": f"{test.id}_{field}", + "type": test.type, + "max_diffs": test.max_diffs, + "cell": test.cell, + field: value, # Only include the single relationship field + } + + # Include other optional fields if set + if test.checked is not None: + result["checked"] = test.checked + if test.url: + result["url"] = test.url + if test.ignore_markdown_tables: + result["ignore_markdown_tables"] = test.ignore_markdown_tables + + split_tests.append(result) + + return split_tests + + +def main(): + parser = argparse.ArgumentParser(description="Split table tests with multiple relationships into individual tests") + parser.add_argument("input_file", help="Input JSONL file containing tests") + parser.add_argument("output_file", help="Output JSONL file for split tests") + parser.add_argument("--keep-non-table", action="store_true", help="Keep non-table tests in the output (default: only output table tests)") + + args = parser.parse_args() + + input_path = Path(args.input_file) + output_path = Path(args.output_file) + + if not input_path.exists(): + print(f"Error: Input file {input_path} does not exist") + sys.exit(1) + + # Load all tests + print(f"Loading tests from {input_path}...") + tests = load_tests(str(input_path)) + print(f"Loaded {len(tests)} tests") + + # Process tests + output_tests = [] + table_tests_count = 0 + split_count = 0 + + for test in tests: + if test.type == TestType.TABLE.value: + table_tests_count += 1 + split_dicts = split_table_test(test) + if len(split_dicts) > 1: + split_count += 1 + print(f"Split test '{test.id}' into {len(split_dicts)} tests") + output_tests.extend(split_dicts) + elif args.keep_non_table: + # For non-table tests, just convert to dict (keep all fields) + output_tests.append(base_test_to_dict(test)) + + # Save output + print(f"\nSaving {len(output_tests)} tests to {output_path}...") + with open(output_path, "w") as f: + for test_dict in output_tests: + f.write(json.dumps(test_dict) + "\n") + + print("\nSummary:") + print(f" Input tests: {len(tests)}") + print(f" Table tests found: {table_tests_count}") + print(f" Table tests split: {split_count}") + print(f" Output tests: {len(output_tests)}") + + +if __name__ == "__main__": + main() diff --git a/olmocr/bench/scripts/url_matcher.py b/olmocr/bench/scripts/url_matcher.py new file mode 100755 index 0000000..3f92e2d --- /dev/null +++ b/olmocr/bench/scripts/url_matcher.py @@ -0,0 +1,123 @@ +#!/usr/bin/env python +import argparse +import glob +import json +import os + +from datasets import load_dataset + + +def extract_urls_from_jsonl(file_path): + """Extract URLs from a JSONL file.""" + urls = set() + url_to_data = {} + with open(file_path, "r", encoding="utf-8") as f: + for line in f: + try: + data = json.loads(line.strip()) + if "url" in data and data["url"]: + url = data["url"] + urls.add(url) + # Store minimal context for each URL + url_to_data[url] = {"id": data.get("id", ""), "type": data.get("type", ""), "page": data.get("page", "")} + except json.JSONDecodeError: + print(f"Warning: Could not parse JSON from line in {file_path}") + continue + return urls, url_to_data + + +def main(): + parser = argparse.ArgumentParser(description="Check for URL matches between local files and Hugging Face dataset") + parser.add_argument("--local-dir", default="/home/ubuntu/olmocr/olmOCR-bench/bench_data", help="Directory containing local JSONL files") + parser.add_argument("--output", default="url_matches.json", help="Output file for results") + args = parser.parse_args() + + # Step 1: Get all local JSONL files + local_jsonl_files = glob.glob(os.path.join(args.local_dir, "*.jsonl")) + print(f"Found {len(local_jsonl_files)} local JSONL files.") + + # Step 2: Extract URLs from local files + local_urls = {} + all_local_urls = set() + url_metadata = {} + + for file_path in local_jsonl_files: + file_name = os.path.basename(file_path) + urls, url_data = extract_urls_from_jsonl(file_path) + local_urls[file_name] = urls + all_local_urls.update(urls) + + # Store metadata with file information + for url, data in url_data.items(): + if url not in url_metadata: + url_metadata[url] = [] + url_metadata[url].append({"file": file_name, **data}) + + print(f"Extracted {len(all_local_urls)} unique URLs from local files.") + + # Step 3: Load Hugging Face dataset + print("Loading Hugging Face dataset...") + try: + dataset_documents = load_dataset("allenai/olmOCR-mix-0225", "00_documents") + dataset_books = load_dataset("allenai/olmOCR-mix-0225", "01_books") + + # Step 4: Extract URLs from Hugging Face dataset + hf_urls = set() + + for split in dataset_documents: + for item in dataset_documents[split]: + if "url" in item and item["url"]: + hf_urls.add(item["url"]) + + for split in dataset_books: + for item in dataset_books[split]: + if "url" in item and item["url"]: + hf_urls.add(item["url"]) + + print(f"Extracted {len(hf_urls)} unique URLs from Hugging Face dataset.") + + # Step 5: Find matches + matches = all_local_urls.intersection(hf_urls) + + # Step 6: Group matches by local file with metadata + matches_by_file = {} + match_details = [] + + for file_name, urls in local_urls.items(): + file_matches = urls.intersection(hf_urls) + if file_matches: + matches_by_file[file_name] = list(file_matches) + + # Add detailed metadata for each match + for url in file_matches: + if url in url_metadata: + for entry in url_metadata[url]: + match_details.append({"url": url, "metadata": entry}) + + # Print summary + print(f"Found {len(matches)} matching URLs between local files and Hugging Face dataset.") + + for file_name, file_matches in matches_by_file.items(): + match_percentage = (len(file_matches) / len(local_urls[file_name])) * 100 if local_urls[file_name] else 0 + print(f"{file_name}: {len(file_matches)}/{len(local_urls[file_name])} matches ({match_percentage:.2f}%)") + + # Save results + result = { + "total_local_urls": len(all_local_urls), + "total_hf_urls": len(hf_urls), + "total_matches": len(matches), + "matches_by_file": matches_by_file, + "match_details": match_details, + } + + with open(args.output, "w", encoding="utf-8") as f: + json.dump(result, f, indent=2) + + print(f"Results saved to {args.output}") + + except Exception as e: + print(f"Error loading or processing Hugging Face dataset: {e}") + + +if __name__ == "__main__": + main() diff --git a/olmocr/bench/scripts/workspace_to_bench.py b/olmocr/bench/scripts/workspace_to_bench.py new file mode 100644 index 0000000..efca0b0 --- /dev/null +++ b/olmocr/bench/scripts/workspace_to_bench.py @@ -0,0 +1,213 @@ +""" +Convert JSONL files to Markdown files and handle missing PDFs +Usage: + python workspace_to_benchmark.py localworkspace ./markdown_output --bench-path ../olmOCR-bench/ +""" + +import argparse +import json +import sys +from collections import defaultdict +from pathlib import Path + + +def load_jsonl_files(input_dir): + """Load all JSONL files from the input directory.""" + jsonl_files = list(Path(input_dir).glob("*.jsonl")) + if not jsonl_files: + print(f"No JSONL files found in {input_dir}") + return [] + + print(f"Found {len(jsonl_files)} JSONL files: {[f.name for f in jsonl_files]}") + return jsonl_files + + +def parse_jsonl_entries(jsonl_files): + """Parse all JSONL files and extract entries with text and metadata.""" + all_entries = [] + pdf_sources = set() + + for jsonl_file in jsonl_files: + print(f"Processing {jsonl_file.name}...") + + with open(jsonl_file, "r", encoding="utf-8") as f: + for line_num, line in enumerate(f, 1): + line = line.strip() + if not line: + continue + + entry = json.loads(line) + text = entry.get("text", "") + metadata = entry.get("metadata", {}) + source_file = metadata.get("Source-File", "") + + if source_file: + pdf_sources.add(source_file) + + all_entries.append({"text": text, "source_file": source_file, "metadata": metadata, "entry": entry}) + + print(f"Loaded {len(all_entries)} entries from JSONL files") + print(f"Found {len(pdf_sources)} unique PDF sources") + + return all_entries, pdf_sources + + +def get_subdir_and_pdf_name(source_file_path): + """Extract subdirectory and PDF filename from source file path.""" + if not source_file_path: + return None, None + + path_parts = Path(source_file_path).parts + + pdfs_index = path_parts.index("pdfs") + if pdfs_index + 1 < len(path_parts): + subdir = path_parts[pdfs_index + 1] + pdf_name = Path(source_file_path).stem + return subdir, pdf_name + + return None, None + + +def create_markdown_files(entries, output_dir, repeat_index=1): + """Create markdown files from JSONL entries in subdir/{pdf_name}.md format.""" + output_path = Path(output_dir) + + subdir_pdf_to_entries = defaultdict(list) + blank_files = 0 + + for entry in entries: + subdir, pdf_name = get_subdir_and_pdf_name(entry["source_file"]) + if subdir and pdf_name: + key = (subdir, pdf_name) + subdir_pdf_to_entries[key].append(entry) + + created_files = set() + + for (subdir, pdf_name), pdf_entries in subdir_pdf_to_entries.items(): + subdir_path = output_path / subdir + subdir_path.mkdir(parents=True, exist_ok=True) + + md_filename = f"{pdf_name}_pg1_repeat{repeat_index}.md" + md_filepath = subdir_path / md_filename + + assert len(pdf_entries) == 1, "Expecting just one entry mapping to each file, otherwise something is wrong" + file_text = pdf_entries[0]["text"] + + with open(md_filepath, "w", encoding="utf-8") as f: + f.write(file_text) + + if not file_text.strip(): + blank_files += 1 + + created_files.add((subdir, pdf_name)) + print(f"Created: {subdir}/{md_filename}_pg1_repeat{repeat_index}") + + print(f"Created {len(created_files)} markdown files from JSONL data") + print(f"{blank_files} of those had empty content") + return created_files + + +def find_missing_pdfs(pdf_sources, created_files, base_bench_path): + """Find PDFs that exist in directories but are missing from JSONL data.""" + subdirs = set() + + for source_file in pdf_sources: + if not source_file: + continue + + subdir, _ = get_subdir_and_pdf_name(source_file) + if subdir: + subdirs.add(subdir) + + print(f"Found PDF subdirectories: {sorted(subdirs)}") + + missing_pdfs = [] + + for subdir in subdirs: + pdf_dir = Path(base_bench_path) / "bench_data" / "pdfs" / subdir + + if not pdf_dir.exists(): + print(f"Warning: Directory {pdf_dir} does not exist") + continue + + pdf_files = list(pdf_dir.glob("*.pdf")) + print(f"Found {len(pdf_files)} PDF files in {subdir}/") + + for pdf_file in pdf_files: + pdf_name = pdf_file.stem + + if (subdir, pdf_name) not in created_files: + missing_pdfs.append({"pdf_name": pdf_name, "full_path": pdf_file, "subdir": subdir}) + + print(f"Found {len(missing_pdfs)} missing PDFs") + return missing_pdfs + + +def create_blank_markdown_files(missing_pdfs, output_dir, repeat_index=1): + """Create blank markdown files for missing PDFs in subdir/{pdf_name}.md format.""" + output_path = Path(output_dir) + + for missing_pdf in missing_pdfs: + subdir = missing_pdf["subdir"] + pdf_name = missing_pdf["pdf_name"] + + subdir_path = output_path / subdir + subdir_path.mkdir(parents=True, exist_ok=True) + + md_filename = f"{pdf_name}_pg1_repeat{repeat_index}.md" + md_filepath = subdir_path / md_filename + + content = "" + + with open(md_filepath, "w", encoding="utf-8") as f: + f.write(content) + + print(f"Created blank: {subdir}/{md_filename}_pg1_repeat{repeat_index}") + + print(f"Created {len(missing_pdfs)} blank markdown files for missing PDFs") + + +def main(): + parser = argparse.ArgumentParser(description="Convert JSONL files to Markdown and handle missing PDFs") + parser.add_argument("workspace_dir", help="Your workspace directory") + parser.add_argument("output_dir", nargs="?", default="./markdown_output", help="Output directory for markdown files (default: ./markdown_output)") + parser.add_argument("--bench-path", default="../olmOCR-bench", help="Path to olmOCR-bench directory (default: ../olmOCR-bench)") + parser.add_argument( + "--repeat-index", default=1, type=int, help="If you want to run multiple workspaces as different repeats to get a better average, set this" + ) + + args = parser.parse_args() + input_dir = args.workspace_dir + "/results" + input_dir = Path(input_dir) + output_dir = Path(args.output_dir) + bench_path = Path(args.bench_path) + + if not input_dir.exists(): + print(f"Error: Input directory {input_dir} does not exist") + sys.exit(1) + + jsonl_files = load_jsonl_files(input_dir) + if not jsonl_files: + sys.exit(1) + + entries, pdf_sources = parse_jsonl_entries(jsonl_files) + if not entries: + print("No entries found in JSONL files") + sys.exit(1) + + created_files = create_markdown_files(entries, output_dir) + + missing_pdfs = find_missing_pdfs(pdf_sources, created_files, bench_path) + + if missing_pdfs: + create_blank_markdown_files(missing_pdfs, output_dir) + + print("\nSummary:") + print(f"Created {len(created_files)} markdown files from JSONL data") + print(f"Created {len(missing_pdfs)} blank markdown files for missing PDFs") + print(f"Total markdown files: {len(created_files) + len(missing_pdfs)}") + print(f"Output directory: {output_dir.absolute()}") + + +if __name__ == "__main__": + main() diff --git a/olmocr/bench/table_parsing.py b/olmocr/bench/table_parsing.py new file mode 100644 index 0000000..2c92d12 --- /dev/null +++ b/olmocr/bench/table_parsing.py @@ -0,0 +1,444 @@ +from collections import defaultdict +from dataclasses import dataclass +from typing import Dict, List, Optional, Set, Tuple, Union + +from bs4 import BeautifulSoup + + +@dataclass(frozen=True) +class TableData: + """Class which holds table data as a graph of cells. Ex. you can access the value at any row, col. + + Cell texts are only ever present one time, so ex if on row 0, you have a colspan 1 and a colspan 2 column, + then text gets stored at (0,0) and (0,1) only, (0,2) is not present in the data + + However, you can also ask, given a row, col, which set of row,col pairs is "left", "right", "up", and "down" + from that one. There can be multiple values returned, because rowspans and colspans mean that you can have multiple cells in each direction. + + Further more, you can also query "top_heading" and "left_heading". Where we also mark cells that are considered "headings", ex. if they are in a thead + html tag. + + To get up/down/left/right simple relations, you can just query the data directly + + To get top_heading/left_heading relations you should use the methods provided by this class which walk the connection graph from a cell to get the headings + """ + + cell_text: Dict[tuple[int, int], str] # Stores map from row, col to cell text + heading_cells: Set[tuple[int, int]] # Contains the row, col pairs which are headings + + is_rectangular: bool # Set to false if you have any missing cells, compared to the maximal number of row/cols including spans > 1 + + up_relations: Dict[tuple[int, int], Set[tuple[int, int]]] + down_relations: Dict[tuple[int, int], Set[tuple[int, int]]] + left_relations: Dict[tuple[int, int], Set[tuple[int, int]]] + right_relations: Dict[tuple[int, int], Set[tuple[int, int]]] + + def _walk_heading_relations(self, start: tuple[int, int], relation: Dict[tuple[int, int], Set[tuple[int, int]]]) -> Set[tuple[int, int]]: + resulting_heading_cells = set() + resulting_end_cells = set() + + visited = set() + to_visit = {start} + + while len(to_visit) > 0: + cur = to_visit.pop() + visited.add(cur) + + if cur in self.heading_cells: + resulting_heading_cells.add(cur) + + if cur not in relation or len(relation[cur]) == 0: + resulting_end_cells.add(cur) + else: + to_visit |= relation[cur] + + resulting_heading_cells -= {start} + resulting_end_cells -= {start} + + if resulting_heading_cells: + return resulting_heading_cells + else: + return resulting_end_cells + + def top_heading_relations(self, start_row: int, start_col: int) -> Set[tuple[int, int]]: + return self._walk_heading_relations((start_row, start_col), self.up_relations) + + def left_heading_relations(self, start_row: int, start_col: int) -> Set[tuple[int, int]]: + return self._walk_heading_relations((start_row, start_col), self.left_relations) + + +def _safe_span_int(value: Optional[Union[str, int]], default: int = 1) -> int: + """Convert rowspan/colspan attributes to positive integers.""" + if value in (None, "", 0): + return default + try: + span = int(value) + except (TypeError, ValueError): + return default + if span <= 0: + return default + return span + + +def _build_table_data_from_specs(row_specs: List[List[Dict[str, Union[str, int, bool]]]]) -> Optional[TableData]: + """ + Build a TableData object from a list of row specifications. + + Each row specification is a list of dictionaries with keys: + - text: cell text content + - rowspan: integer rowspan (>= 1) + - colspan: integer colspan (>= 1) + - is_heading: bool indicating if the cell should be treated as a heading + """ + if not row_specs: + return None + + cell_text: Dict[Tuple[int, int], str] = {} + heading_cells: Set[Tuple[int, int]] = set() + cell_meta: Dict[Tuple[int, int], Dict[str, Union[int, bool]]] = {} + occupancy: List[List[Optional[Tuple[int, int]]]] = [] + active_rowspans: List[Optional[Tuple[Tuple[int, int], int]]] = [] + + for row_idx, cells in enumerate(row_specs): + row_entries: List[Optional[Tuple[int, int]]] = [] + col_index = 0 + spec_idx = 0 + total_specs = len(cells) + + while spec_idx < total_specs or col_index < len(active_rowspans): + if col_index < len(active_rowspans) and active_rowspans[col_index] is not None: + cell_id, remaining = active_rowspans[col_index] + row_entries.append(cell_id) + remaining -= 1 + active_rowspans[col_index] = (cell_id, remaining) if remaining > 0 else None + col_index += 1 + continue + + if spec_idx >= total_specs: + if col_index < len(active_rowspans): + row_entries.append(None) + col_index += 1 + continue + break + + spec = cells[spec_idx] + spec_idx += 1 + + text = spec.get("text", "") or "" + rowspan = spec.get("rowspan", 1) + colspan = spec.get("colspan", 1) + is_heading = bool(spec.get("is_heading", False)) + + rowspan = rowspan if isinstance(rowspan, int) else _safe_span_int(rowspan) + colspan = colspan if isinstance(colspan, int) else _safe_span_int(colspan) + rowspan = max(1, rowspan) + colspan = max(1, colspan) + + cell_id = (row_idx, col_index) + cell_text[cell_id] = text + if is_heading: + heading_cells.add(cell_id) + + cell_meta[cell_id] = { + "row": row_idx, + "col": col_index, + "rowspan": rowspan, + "colspan": colspan, + } + + required_len = col_index + colspan + if len(active_rowspans) < required_len: + active_rowspans.extend([None] * (required_len - len(active_rowspans))) + + for offset in range(colspan): + current_col = col_index + offset + row_entries.append(cell_id) + if rowspan > 1: + active_rowspans[current_col] = (cell_id, rowspan - 1) + else: + active_rowspans[current_col] = None + + col_index += colspan + + occupancy.append(row_entries) + + # Flush any remaining active rowspans into additional rows + while any(entry is not None for entry in active_rowspans): + row_entries: List[Optional[Tuple[int, int]]] = [] + for col_index, span_entry in enumerate(active_rowspans): + if span_entry is None: + row_entries.append(None) + continue + cell_id, remaining = span_entry + row_entries.append(cell_id) + remaining -= 1 + active_rowspans[col_index] = (cell_id, remaining) if remaining > 0 else None + occupancy.append(row_entries) + + if not cell_text: + return None + + # Normalize occupancy to a consistent width based on populated columns + valid_columns = {idx for row in occupancy for idx, value in enumerate(row) if value is not None} + if valid_columns: + table_width = max(valid_columns) + 1 + for row in occupancy: + if len(row) < table_width: + row.extend([None] * (table_width - len(row))) + elif len(row) > table_width: + del row[table_width:] + else: + return None + + table_height = len(occupancy) + + up_rel = defaultdict(set) + down_rel = defaultdict(set) + left_rel = defaultdict(set) + right_rel = defaultdict(set) + + for cell_id, meta in cell_meta.items(): + row_start = meta["row"] + col_start = meta["col"] + rowspan = meta["rowspan"] + colspan = meta["colspan"] + row_end = row_start + rowspan - 1 + col_end = col_start + colspan - 1 + + # Right relations + for row in range(row_start, row_end + 1): + for col in range(col_end + 1, table_width): + neighbor = occupancy[row][col] + if neighbor is None or neighbor == cell_id: + continue + right_rel[cell_id].add(neighbor) + break + + # Left relations + for row in range(row_start, row_end + 1): + for col in range(col_start - 1, -1, -1): + neighbor = occupancy[row][col] + if neighbor is None or neighbor == cell_id: + continue + left_rel[cell_id].add(neighbor) + break + + # Down relations + for col in range(col_start, col_end + 1): + for row in range(row_end + 1, table_height): + if col >= len(occupancy[row]): + continue + neighbor = occupancy[row][col] + if neighbor is None or neighbor == cell_id: + continue + down_rel[cell_id].add(neighbor) + break + + # Up relations + for col in range(col_start, col_end + 1): + for row in range(row_start - 1, -1, -1): + neighbor = occupancy[row][col] + if neighbor is None or neighbor == cell_id: + continue + up_rel[cell_id].add(neighbor) + break + + # Ensure every cell has an entry in relations dictionaries + up_relations = {cell_id: set(up_rel[cell_id]) for cell_id in cell_text} + down_relations = {cell_id: set(down_rel[cell_id]) for cell_id in cell_text} + left_relations = {cell_id: set(left_rel[cell_id]) for cell_id in cell_text} + right_relations = {cell_id: set(right_rel[cell_id]) for cell_id in cell_text} + + return TableData( + cell_text=cell_text, + heading_cells=heading_cells, + up_relations=up_relations, + down_relations=down_relations, + left_relations=left_relations, + right_relations=right_relations, + is_rectangular=not any(any(x is None for x in row) for row in occupancy), + ) + + +def parse_markdown_tables(md_content: str) -> List[TableData]: + """ + Extract and parse all markdown tables from the provided content. + Uses a direct approach to find and parse tables, which is more robust for tables + at the end of files or with irregular formatting. + + Args: + md_content: The markdown content containing tables + + Returns: + A list of TableData objects, each containing the table data and header information + """ + # Split the content into lines and process line by line + lines = md_content.strip().split("\n") + + parsed_tables = [] + current_table_lines = [] + in_table = False + + # Identify potential tables by looking for lines with pipe characters + for i, line in enumerate(lines): + # Check if this line has pipe characters (a table row indicator) + if "|" in line: + # If we weren't in a table before, start a new one + if not in_table: + in_table = True + current_table_lines = [line] + else: + # Continue adding to the current table + current_table_lines.append(line) + else: + # No pipes in this line, so if we were in a table, we've reached its end + if in_table: + # Process the completed table if it has at least 2 rows + if len(current_table_lines) >= 2: + table_data = _process_table_lines(current_table_lines) + if table_data and len(table_data) > 0: + row_specs: List[List[Dict[str, Union[str, int, bool]]]] = [] + for row_idx, row in enumerate(table_data): + row_specs.append( + [ + { + "text": cell, + "rowspan": 1, + "colspan": 1, + "is_heading": row_idx == 0 or col_idx == 0, + } + for col_idx, cell in enumerate(row) + ] + ) + + table = _build_table_data_from_specs(row_specs) + if table: + parsed_tables.append(table) + in_table = False + + # Process the last table if we're still tracking one at the end of the file + if in_table and len(current_table_lines) >= 2: + table_data = _process_table_lines(current_table_lines) + if table_data and len(table_data) > 0: + row_specs = [] + for row_idx, row in enumerate(table_data): + row_specs.append( + [ + { + "text": cell, + "rowspan": 1, + "colspan": 1, + "is_heading": row_idx == 0 or col_idx == 0, + } + for col_idx, cell in enumerate(row) + ] + ) + + table = _build_table_data_from_specs(row_specs) + if table: + parsed_tables.append(table) + + return parsed_tables + + +def _process_table_lines(table_lines: List[str]) -> List[List[str]]: + """ + Process a list of lines that potentially form a markdown table. + + Args: + table_lines: List of strings, each representing a line in a potential markdown table + + Returns: + A list of rows, each a list of cell values + """ + table_data = [] + separator_row_index = None + + # First, identify the separator row (the row with dashes) + for i, line in enumerate(table_lines): + # Check if this looks like a separator row (contains mostly dashes) + content_without_pipes = line.replace("|", "").strip() + if content_without_pipes and all(c in "- :" for c in content_without_pipes): + separator_row_index = i + break + + # Process each line, filtering out the separator row + for i, line in enumerate(table_lines): + # Skip the separator row + if i == separator_row_index: + continue + + # Skip lines that are entirely formatting + if line.strip() and all(c in "- :|" for c in line): + continue + + # Process the cells in this row + cells = [cell.strip() for cell in line.split("|")] + + # Remove empty cells at the beginning and end (caused by leading/trailing pipes) + if cells and cells[0] == "": + cells = cells[1:] + if cells and cells[-1] == "": + cells = cells[:-1] + + if cells: # Only add non-empty rows + table_data.append(cells) + + return table_data + + +def parse_html_tables(html_content: str) -> List[TableData]: + """ + Extract and parse all HTML tables from the provided content. + Identifies header rows and columns, and maps them properly handling rowspan/colspan. + + Args: + html_content: The HTML content containing tables + + Returns: + A list of TableData objects, each containing the table data and header information + """ + soup = BeautifulSoup(html_content, "html.parser") + tables = soup.find_all("table") + + parsed_tables = [] + + for table in tables: + rows = table.find_all("tr") + if not rows: + continue + + row_specs: List[List[Dict[str, Union[str, int, bool]]]] = [] + total_rows = len(rows) + + for row_idx, row in enumerate(rows): + cells = row.find_all(["th", "td"], recursive=False) + heading_context = row.find_parent("thead") is not None + + row_spec: List[Dict[str, Union[str, int, bool]]] = [] + for cell in cells: + for br in cell.find_all("br"): + br.replace_with("\n") + + text = cell.get_text(separator="").strip() + raw_rowspan = cell.get("rowspan") + raw_colspan = cell.get("colspan") + + rowspan = _safe_span_int(raw_rowspan, 1) + colspan = _safe_span_int(raw_colspan, 1) + + # HTML specifies rowspan=0 to extend to the end of the table section + if isinstance(raw_rowspan, str) and raw_rowspan.strip() == "0": + rowspan = max(1, total_rows - row_idx) + + is_heading = cell.name == "th" or heading_context + + row_spec.append({"text": text, "rowspan": rowspan, "colspan": colspan, "is_heading": is_heading}) + + row_specs.append(row_spec) + + table_data = _build_table_data_from_specs(row_specs) + if table_data: + parsed_tables.append(table_data) + + return parsed_tables diff --git a/olmocr/bench/templates/all_done.html b/olmocr/bench/templates/all_done.html new file mode 100644 index 0000000..9a8b6d1 --- /dev/null +++ b/olmocr/bench/templates/all_done.html @@ -0,0 +1,38 @@ + + + + + + All Tests Reviewed + + + +

All Tests Reviewed!

You have completed reviewing all tests in the dataset.

+ + \ No newline at end of file diff --git a/olmocr/bench/templates/all_done_latex.html b/olmocr/bench/templates/all_done_latex.html new file mode 100644 index 0000000..5784f94 --- /dev/null +++ b/olmocr/bench/templates/all_done_latex.html @@ -0,0 +1,54 @@ + + + + + + + All Done! + + + +

All Done! 🎉

You have reviewed all equations in the dataset.

+ + + +

+ + + \ No newline at end of file diff --git a/olmocr/bench/templates/review.html b/olmocr/bench/templates/review.html new file mode 100644 index 0000000..d3b2da5 --- /dev/null +++ b/olmocr/bench/templates/review.html @@ -0,0 +1,1443 @@ + + + + + + PDF Test Review + + + + + + + + +

PDF Test Review: {{ pdf_name }} ({{ pdf_index + 1 }}/{{ total_pdfs }})

+ +

Total Tests: {{ stats.total }}

Unchecked: {{ stats.null }}

Verified: {{ stats.verified }}

Rejected: {{ stats.rejected }}

Completion: {{ "%.1f"|format(stats.completion) }}%

+ + + +

+ + / + + + +

Loading PDF...

+ +

Tests ({{ tests|length }})

+ +

+ + {% for test in tests %} +

+ {{ test.type|upper }} +

+ + + +

+ +

+ {% if test.type == 'present' or test.type == 'absent' %} +

Text: {{ test.text }}

Case Sensitive: {{ test.case_sensitive }}

+ {% if test.first_n %}

First N: {{ test.first_n }}

{% endif %} + {% if test.last_n %}

Last N: {{ test.last_n }}

{% endif %} + {% elif test.type == 'order' %} +

Before: {{ test.before }}

After: {{ test.after }}

+ {% elif test.type == 'table' %} +

Cell: {{ test.cell }}

+ {% if test.up %}

Up: {{ test.up }}

{% endif %} + {% if test.down %}

Down: {{ test.down }}

{% endif %} + {% if test.left %}

Left: {{ test.left }}

{% endif %} + {% if test.right %}

Right: {{ test.right }}

{% endif %} + {% if test.top_heading %}

Top Heading: {{ test.top_heading }}

{% endif %} + {% if test.left_heading %}

Left Heading: {{ test.left_heading }}

{% endif %} + {% elif test.type == 'math' %} +

Math:

{{ test.math }}

+ {% endif %} +

+ Max Diffs: + + {{ test.max_diffs }} + +

Status: {{ test.checked or 'Not checked' }}

+ {% endfor %} +

+ +

Keyboard Shortcuts

↑/↓: Navigate tests

A: Approve and move to next

R: Reject and move to next

E: Edit current test

Q: Reject all tests & next PDF

N: Next PDF

P: Previous PDF

+ / -: Increase/decrease Max Diffs

?: Toggle this help

+ + + + \ No newline at end of file diff --git a/olmocr/bench/templates/review_latex.html b/olmocr/bench/templates/review_latex.html new file mode 100644 index 0000000..c0be69b --- /dev/null +++ b/olmocr/bench/templates/review_latex.html @@ -0,0 +1,361 @@ + + + + + + + + Equation Verification + + + + + + +

+ + + +

Equations ({{ tests|length }})

+ {% for test in tests %} + +

+ {{ test.text|safe }} +

+ + + + +

+ {% endfor %} +

+ + + + + \ No newline at end of file diff --git a/olmocr/bench/tests.py b/olmocr/bench/tests.py new file mode 100644 index 0000000..996eb9f --- /dev/null +++ b/olmocr/bench/tests.py @@ -0,0 +1,880 @@ +import json +import os +import re +import unicodedata +from concurrent.futures import ThreadPoolExecutor, as_completed +from dataclasses import asdict, dataclass +from enum import Enum +from typing import Dict, List, Optional, Tuple, Union + +from fuzzysearch import find_near_matches +from rapidfuzz import fuzz +from tqdm import tqdm + +from olmocr.repeatdetect import RepeatDetector + +from .katex.render import compare_rendered_equations, render_equation +from .table_parsing import parse_html_tables, parse_markdown_tables + +# Tell pytest these are not tests +__test__ = False + + +class TestType(str, Enum): + __test__ = False # Tell pytest this is not a test class + + BASELINE = "baseline" + PRESENT = "present" + ABSENT = "absent" + ORDER = "order" + TABLE = "table" + MATH = "math" + FORMAT = "format" + FOOTNOTE = "footnote" + + +class TestChecked(str, Enum): + VERIFIED = "verified" + REJECTED = "rejected" + + +class ValidationError(Exception): + """Exception raised for validation errors.""" + + pass + + +def normalize_text(md_content: str) -> str: + if md_content is None: + return None + + # Normalize
and
to newlines + md_content = re.sub(r"
", " ", md_content) + + # Remove markdown bold formatting (** or __ for bold) + md_content = re.sub(r"\*\*(.*?)\*\*", r"\1", md_content) + md_content = re.sub(r"__(.*?)__", r"\1", md_content) + md_content = re.sub(r"", "", md_content) # Remove tags if they exist + md_content = re.sub(r"", "", md_content) # Remove tags if they exist + + # Remove markdown italics formatting (* or _ for italics) + # Logic: The dot (.) in regex matches any character EXCEPT a newline. + # This automatically prevents matching **start \n\n end**. + # We use group \1 to ensure we match matching pairs (**...** or __...__). + md_content = re.sub(r"(\*\*|__)(.*?)\1", r"\2", md_content) # Bold + md_content = re.sub(r"(\*|_)(.*?)\1", r"\2", md_content) # Italics + + # Normalize whitespace in the md_content + md_content = re.sub(r"\s+", " ", md_content) + + # Convert down to a consistent unicode form, so é == e + accent, unicode forms + md_content = unicodedata.normalize("NFC", md_content) + + # Dictionary of characters to replace: keys are fancy characters, values are ASCII equivalents, unicode micro with greek mu comes up often enough too + replacements = {"‘": "'", "’": "'", "‚": "'", "“": '"', "”": '"', "„": '"', "＿": "_", "–": "-", "—": "-", "‑": "-", "‒": "-", "−": "-", "\u00b5": "\u03bc"} + + # Apply all replacements from the dictionary + for fancy_char, ascii_char in replacements.items(): + md_content = md_content.replace(fancy_char, ascii_char) + + return md_content + + +@dataclass(kw_only=True) +class BasePDFTest: + """ + Base class for all PDF test types. + + Attributes: + pdf: The PDF filename. + page: The page number for the test. + id: Unique identifier for the test. + type: The type of test. + threshold: A float between 0 and 1 representing the threshold for fuzzy matching. + """ + + pdf: str + page: int + id: str + type: str + max_diffs: int = 0 + checked: Optional[TestChecked] = None + url: Optional[str] = None + + def __post_init__(self): + if not self.pdf: + raise ValidationError("PDF filename cannot be empty") + if not self.id: + raise ValidationError("Test ID cannot be empty") + if not isinstance(self.max_diffs, int) or self.max_diffs < 0: + raise ValidationError("Max diffs must be positive number or 0") + if self.type not in {t.value for t in TestType}: + raise ValidationError(f"Invalid test type: {self.type}") + + def run(self, md_content: str) -> Tuple[bool, str]: + """ + Run the test on the provided markdown content. + + Args: + md_content: The content of the .md file. + + Returns: + A tuple (passed, explanation) where 'passed' is True if the test passes, + and 'explanation' provides details when the test fails. + """ + raise NotImplementedError("Subclasses must implement the run method") + + +@dataclass +class TextPresenceTest(BasePDFTest): + """ + Test to verify the presence or absence of specific text in a PDF. + + Attributes: + text: The text string to search for. + """ + + text: str + case_sensitive: bool = True + first_n: Optional[int] = None + last_n: Optional[int] = None + + def __post_init__(self): + super().__post_init__() + if self.type not in {TestType.PRESENT.value, TestType.ABSENT.value}: + raise ValidationError(f"Invalid type for TextPresenceTest: {self.type}") + self.text = normalize_text(self.text) + if not self.text.strip(): + raise ValidationError("Text field cannot be empty") + + def run(self, md_content: str) -> Tuple[bool, str]: + reference_query = self.text + + # Normalize whitespace in the md_content + md_content = normalize_text(md_content) + + if not self.case_sensitive: + reference_query = reference_query.lower() + md_content = md_content.lower() + + if self.first_n and self.last_n: + md_content = md_content[: self.first_n] + md_content[-self.last_n :] + elif self.first_n: + md_content = md_content[: self.first_n] + elif self.last_n: + md_content = md_content[-self.last_n :] + + # Threshold for fuzzy matching derived from max_diffs + threshold = 1.0 - (self.max_diffs / (len(reference_query) if len(reference_query) > 0 else 1)) + best_ratio = fuzz.partial_ratio(reference_query, md_content) / 100.0 + + if self.type == TestType.PRESENT.value: + if best_ratio >= threshold: + return True, "" + else: + msg = f"Expected '{reference_query[:40]}...' with threshold {threshold} " f"but best match ratio was {best_ratio:.3f}" + return False, msg + else: # ABSENT + if best_ratio < threshold: + return True, "" + else: + msg = f"Expected absence of '{reference_query[:40]}...' with threshold {threshold} " f"but best match ratio was {best_ratio:.3f}" + return False, msg + + +@dataclass +class TextOrderTest(BasePDFTest): + """ + Test to verify that one text appears before another in a PDF. + + Attributes: + before: The text expected to appear first. + after: The text expected to appear after the 'before' text. + """ + + before: str + after: str + + def __post_init__(self): + super().__post_init__() + if self.type != TestType.ORDER.value: + raise ValidationError(f"Invalid type for TextOrderTest: {self.type}") + self.before = normalize_text(self.before) + self.after = normalize_text(self.after) + if not self.before.strip(): + raise ValidationError("Before field cannot be empty") + if not self.after.strip(): + raise ValidationError("After field cannot be empty") + if self.max_diffs > len(self.before) // 2 or self.max_diffs > len(self.after) // 2: + raise ValidationError("Max diffs is too large for this test, greater than 50% of the search string") + + def run(self, md_content: str) -> Tuple[bool, str]: + md_content = normalize_text(md_content) + + before_matches = find_near_matches(self.before, md_content, max_l_dist=self.max_diffs) + after_matches = find_near_matches(self.after, md_content, max_l_dist=self.max_diffs) + + if not before_matches: + return False, f"'before' text '{self.before[:40]}...' not found with max_l_dist {self.max_diffs}" + if not after_matches: + return False, f"'after' text '{self.after[:40]}...' not found with max_l_dist {self.max_diffs}" + + for before_match in before_matches: + for after_match in after_matches: + if before_match.start < after_match.start: + return True, "" + return False, (f"Could not find a location where '{self.before[:40]}...' appears before " f"'{self.after[:40]}...'.") + + +@dataclass +class FormatTest(BasePDFTest): + """ + Test to verify that specific text appears with the correct formatting. + + Attributes: + text: The text to search for. + format: The expected format ("heading", "bold", or "italic"). + """ + + text: str + format: str + case_sensitive: bool = True + + def __post_init__(self): + super().__post_init__() + if self.type != TestType.FORMAT.value: + raise ValidationError(f"Invalid type for FormatTest: {self.type}") + self.text = normalize_text(self.text) + if not self.text.strip(): + raise ValidationError("Text field cannot be empty") + if self.format not in {"heading", "bold", "italic"}: + raise ValidationError(f"Invalid format type: {self.format}. Must be 'heading', 'bold', or 'italic'") + + def run(self, md_content: str) -> Tuple[bool, str]: + """ + Extract all text with the specified format and check if our text is among them. + """ + # Store the original content before any normalization for pattern matching + original_content = md_content + + # Extract formatted text based on the format type + formatted_texts = [] + + if self.format == "heading": + # Markdown headings (# through ######) + heading_patterns = [ + r"^#{1,6}\s+(.+?)$", # Standard markdown headings + ] + for pattern in heading_patterns: + matches = re.findall(pattern, original_content, re.MULTILINE) + formatted_texts.extend(matches) + + # HTML headings (
through

) + html_heading_pattern = r"]>(.?)" + matches = re.findall(html_heading_pattern, original_content, re.IGNORECASE | re.DOTALL) + formatted_texts.extend(matches) + + elif self.format == "bold": + # Markdown bold patterns + bold_patterns = [ + r"\\(.?)\\*", # text + r"(.?)", # text + ] + for pattern in bold_patterns: + matches = re.findall(pattern, original_content, re.DOTALL) + formatted_texts.extend(matches) + + # HTML bold patterns + html_bold_patterns = [r"]>(.?)", r"]>(.?)"] # text # text + for pattern in html_bold_patterns: + matches = re.findall(pattern, original_content, re.IGNORECASE | re.DOTALL) + formatted_texts.extend(matches) + + elif self.format == "italic": + # Markdown italic patterns - be careful not to match bold + # We need to match single or _ that are not part of ** or __ + italic_patterns = [ + r"(?]>(.?)", r"]>(.?)"] # text # text + for pattern in html_italic_patterns: + matches = re.findall(pattern, original_content, re.IGNORECASE | re.DOTALL) + formatted_texts.extend(matches) + + # Normalize all extracted formatted texts + normalized_formatted_texts = [normalize_text(text) for text in formatted_texts] + + # Normalize the search text + search_text = self.text + if not self.case_sensitive: + search_text = search_text.lower() + normalized_formatted_texts = [text.lower() for text in normalized_formatted_texts] + + # Check if the text appears in any of the formatted texts using fuzzy matching + threshold = 1.0 - (self.max_diffs / (len(search_text) if len(search_text) > 0 else 1)) + + for formatted_text in normalized_formatted_texts: + # Use partial_ratio for substring matching + ratio = fuzz.partial_ratio(search_text, formatted_text) / 100.0 + if ratio >= threshold: + return True, "" + + # If we didn't find the text with the specified format + found_formats = [] + if len(normalized_formatted_texts) > 0: + # Show a sample of what we did find with this format + sample = normalized_formatted_texts[:3] + sample_str = ", ".join([f"'{t[:20]}...'" if len(t) > 20 else f"'{t}'" for t in sample]) + found_formats.append(f"Found {self.format} text: {sample_str}") + else: + found_formats.append(f"No {self.format} formatted text found") + + return False, f"Text '{self.text[:40]}...' not found with {self.format} formatting. {'; '.join(found_formats)}" + + +@dataclass +class TableTest(BasePDFTest): + """ + Test to verify certain properties of a table are held, namely that some cells appear relative to other cells correctly + """ + + # This is the target cell, which must exist in at least one place in the table + cell: str + + # These properties say that the cell immediately up/down/left/right of the target cell has the string specified + up: str = "" + down: str = "" + left: str = "" + right: str = "" + + # These properties say that the cell all the way up, or all the way left of the target cell (ex. headings) has the string value specified + top_heading: str = "" + left_heading: str = "" + + ignore_markdown_tables: bool = False + + def __post_init(self): + super().post_init__() + if self.type != TestType.TABLE.value: + raise ValidationError(f"Invalid type for TableTest: {self.type}") + + # Normalize the search text too + self.cell = normalize_text(self.cell) + self.up = normalize_text(self.up) + self.down = normalize_text(self.down) + self.left = normalize_text(self.left) + self.right = normalize_text(self.right) + self.top_heading = normalize_text(self.top_heading) + self.left_heading = normalize_text(self.left_heading) + + def run(self, content: str) -> Tuple[bool, str]: + """ + Run the table test on provided content. + + Finds all tables (markdown and/or HTML based on content_type) and checks if any cell + matches the target cell and satisfies the specified relationships. + + Args: + content: The content containing tables (markdown or HTML) + + Returns: + A tuple (passed, explanation) where 'passed' is True if the test passes, + and 'explanation' provides details when the test fails. + """ + # Initialize variables to track tables and results + tables_to_check = [] + failed_reasons = [] + + # Threshold for fuzzy matching derived from max_diffs + threshold = 1.0 - (self.max_diffs / (len(self.cell) if len(self.cell) > 0 else 1)) + threshold = max(0.5, threshold) + + # Parse tables based on content_type + if not self.ignore_markdown_tables: + md_tables = parse_markdown_tables(content) + tables_to_check.extend(md_tables) + + html_tables = parse_html_tables(content) + tables_to_check.extend(html_tables) + + # If no tables found, return failure + if not tables_to_check: + return False, "No tables found in the content" + + # Check each table + for table_data in tables_to_check: + # Find all cells that match the target cell using fuzzy matching + matches = [] + for rowcol, cell_content in table_data.cell_text.items(): + similarity = fuzz.ratio(self.cell, normalize_text(cell_content)) / 100.0 + + if similarity >= threshold: + matches.append(rowcol) + + # If no matches found in this table, continue to the next table + if not matches: + continue + + # Check the relationships for each matching cell + for rowcol in matches: + all_relationships_satisfied = True + current_failed_reasons = [] + + def _check_relationship(comparison_str: str, relation_func): + nonlocal all_relationships_satisfied + cur_relation_satisified = False + best_similarity = 0 + best_similarity_text = None + + for rowcol_up in relation_func(rowcol): + test_cell = normalize_text(table_data.cell_text[rowcol_up]) + test_similarity = fuzz.ratio(comparison_str, test_cell) / 100.0 + if test_similarity > best_similarity: + best_similarity = test_similarity + best_similarity_text = test_cell + + if test_similarity >= max(0.5, 1.0 - (self.max_diffs / (len(comparison_str) if len(comparison_str) > 0 else 1))): + cur_relation_satisified = True + + if not cur_relation_satisified: + all_relationships_satisfied = False + current_failed_reasons.append( + f"Cell compared to '{best_similarity_text}' doesn't match expected '{comparison_str}' (best similarity: {best_similarity:.2f})" + ) + + # Check up relationship + if self.up: + _check_relationship(self.up, lambda rowcol: table_data.up_relations[rowcol]) + + if self.down: + _check_relationship(self.down, lambda rowcol: table_data.down_relations[rowcol]) + + if self.left: + _check_relationship(self.left, lambda rowcol: table_data.left_relations[rowcol]) + + if self.right: + _check_relationship(self.right, lambda rowcol: table_data.right_relations[rowcol]) + + if self.left_heading: + _check_relationship(self.left_heading, lambda rowcol: table_data.left_heading_relations(rowcol)) + + if self.top_heading: + _check_relationship(self.top_heading, lambda rowcol: table_data.top_heading_relations(rowcol)) + + # If all relationships are satisfied for this cell, the test passes + if all_relationships_satisfied: + return True, "" + else: + failed_reasons.extend(current_failed_reasons) + + # If we've gone through all tables and all matching cells and none satisfied all relationships + if not failed_reasons: + return False, f"No cell matching '{self.cell}' found in any table with threshold {threshold}" + else: + return False, f"Found cells matching '{self.cell}' but relationships were not satisfied: {'; '.join(failed_reasons)}" + + +@dataclass +class BaselineTest(BasePDFTest): + """ + This test makes sure that several baseline quality checks pass for the output generation. + + Namely, the output is not blank, not endlessly repeating, and contains characters of the proper + character sets. + + """ + + max_length: Optional[int] = None # Used to implement blank page checks + max_length_skips_image_alt_tags: bool = False + + max_repeats: int = 30 + check_disallowed_characters: bool = True + + def run(self, content: str) -> Tuple[bool, str]: + base_content_len = len("".join(c for c in content if c.isalnum()).strip()) + + # If this a blank page check, then it short circuits the rest of the checks + if self.max_length is not None: + if self.max_length_skips_image_alt_tags: + # Remove markdown image tags like ![alt text](image.png) from the text length count + content_for_length_check = re.sub(r"!\[.?\]$.?$", "", content) + base_content_len = len("".join(c for c in content_for_length_check if c.isalnum()).strip()) + + if base_content_len > self.max_length: + return False, f"{base_content_len} characters were output for a page we expected to be blank" + else: + return True, "" + + if base_content_len == 0: + return False, "The text contains no alpha numeric characters" + + # Makes sure that the content has no egregious repeated ngrams at the end, which indicate a degradation of quality + # Honestly, this test doesn't seem to catch anything at the moment, maybe it can be refactored to a "text-quality" + # test or something, that measures repetition, non-blanks, charsets, etc + d = RepeatDetector(max_ngram_size=5) + d.add_letters(content) + repeats = d.ngram_repeats() + + for index, count in enumerate(repeats): + if count > self.max_repeats: + return False, f"Text ends with {count} repeating {index+1}-grams, invalid" + + pattern = re.compile( + r"[" + r"\u4e00-\u9FFF" # CJK Unified Ideographs (Chinese characters) + r"\u3040-\u309F" # Hiragana (Japanese) + r"\u30A0-\u30FF" # Katakana (Japanese) + r"\U0001F600-\U0001F64F" # Emoticons (Emoji) + r"\U0001F300-\U0001F5FF" # Miscellaneous Symbols and Pictographs (Emoji) + r"\U0001F680-\U0001F6FF" # Transport and Map Symbols (Emoji) + r"\U0001F1E0-\U0001F1FF" # Regional Indicator Symbols (flags, Emoji) + r"]", + flags=re.UNICODE, + ) + + matches = pattern.findall(content) + if self.check_disallowed_characters and matches: + return False, f"Text contains disallowed characters {matches}" + + return True, "" + + +@dataclass +class MathTest(BasePDFTest): + math: str + + ignore_dollar_delimited: bool = False + + def __post_init(self): + super().post_init__() + if self.type != TestType.MATH.value: + raise ValidationError(f"Invalid type for MathTest: {self.type}") + if len(self.math.strip()) == 0: + raise ValidationError("Math test must have non-empty math expression") + + self.reference_render = render_equation(self.math) + + if self.reference_render is None: + raise ValidationError(f"Math equation {self.math} was not able to render") + + def run(self, content: str) -> Tuple[bool, str]: + # Store both the search pattern and the full pattern to replace + patterns = [ + (r"\\$(.+?)\\$", r"\\$(.+?)\\$"), # $...$ + (r"\\\[(.+?)\\\]", r"\\\[(.+?)\\\]"), # \[...\] + ] + + if not self.ignore_dollar_delimited: + patterns.extend( + [ + (r"\$\$(.+?)\$\$", r"\$\$(.+?)\$\$"), # $$...$$ + (r"\$(.+?)\$", r"\$(.+?)\$"), # $...$]) + ] + ) + + equations = [] + modified_content = content + + for search_pattern, replace_pattern in patterns: + # Find all matches for the current pattern + matches = re.findall(search_pattern, modified_content, re.DOTALL) + equations.extend([e.strip() for e in matches]) + + # Replace all instances of this pattern with empty strings + modified_content = re.sub(replace_pattern, "", modified_content, flags=re.DOTALL) + + # If an equation in the markdown exactly matches our math string, then that's good enough + # we don't have to do a more expensive comparison + if any(hyp == self.math for hyp in equations): + return True, "" + + # If not, then let's render the math equation itself and now compare to each hypothesis + # But, to speed things up, since rendering equations is hard, we sort the equations on the page + # by fuzzy similarity to the hypothesis + equations.sort(key=lambda x: -fuzz.ratio(x, self.math)) + for hypothesis in equations: + hypothesis_render = render_equation(hypothesis) + + if not hypothesis_render: + continue + + if compare_rendered_equations(self.reference_render, hypothesis_render): + return True, "" + + # self.reference_render.save(f"maths/{self.id}_ref.png", format="PNG") + # best_match_render.save(f"maths/{self.id}_hyp.png", format="PNG") + + return False, f"No match found for {self.math} anywhere in content" + + +@dataclass +class FootnoteTest(BasePDFTest): + """ + Test to verify that footnotes appear correctly on a page. + + Attributes: + marker: The footnote marker (e.g., "1", "2"). Must appear as superscript or [^marker]. Required. + appears_before_marker: Optional text that should appear before the marker (ignoring whitespace/non-alpha). + appears_after_marker: Optional text that should appear after the marker (ignoring whitespace/non-alpha). + """ + + marker: str + appears_before_marker: Optional[str] = None + appears_after_marker: Optional[str] = None + + def __post_init(self): + super().post_init__() + if self.type != TestType.FOOTNOTE.value: + raise ValidationError(f"Invalid type for FootnoteTest: {self.type}") + + # marker is required + if not self.marker: + raise ValidationError("marker field is required") + + # Validate marker doesn't contain whitespace + if " " in self.marker: + raise ValidationError("Marker cannot contain whitespace") + + # Normalize the optional text fields + if self.appears_before_marker: + self.appears_before_marker = normalize_text(self.appears_before_marker) + if not self.appears_before_marker.strip(): + raise ValidationError("appears_before_marker field cannot be empty if provided") + + if self.appears_after_marker: + self.appears_after_marker = normalize_text(self.appears_after_marker) + if not self.appears_after_marker.strip(): + raise ValidationError("appears_after_marker field cannot be empty if provided") + + def run(self, md_content: str) -> Tuple[bool, str]: + """ + Run the footnote test on provided markdown content. + + Args: + md_content: The markdown content to test. + + Returns: + A tuple (passed, explanation) where 'passed' is True if the test passes, + and 'explanation' provides details when the test fails. + """ + # Find all occurrences of the marker in various formats + marker_positions = [] + + # Check for markdown footnote reference [^marker] (but not definition [^marker]:) + markdown_pattern = rf"\[\^{re.escape(self.marker)}\](?!:)" + for match in re.finditer(markdown_pattern, md_content): + marker_positions.append({"start": match.start(), "end": match.end(), "type": "markdown"}) + + # Check for superscript HTML ^marker + html_sup_pattern = rf"]>{re.escape(self.marker)}" + for match in re.finditer(html_sup_pattern, md_content, re.IGNORECASE): + marker_positions.append({"start": match.start(), "end": match.end(), "type": "html"}) + + # Check for Unicode superscript characters (for common digits) + superscript_map = {"0": "⁰", "1": "¹", "2": "²", "3": "³", "4": "⁴", "5": "⁵", "6": "⁶", "7": "⁷", "8": "⁸", "9": "⁹"} + + # Convert marker to superscript if all characters are digits + if all(c in superscript_map for c in self.marker): + superscript_marker = "".join(superscript_map[c] for c in self.marker) + for match in re.finditer(re.escape(superscript_marker), md_content): + marker_positions.append({"start": match.start(), "end": match.end(), "type": "unicode"}) + + # If no markers found at all, fail + if not marker_positions: + return False, f"Footnote marker '{self.marker}' not found as [^{self.marker}], ^{{self.marker}}, or superscript" + + # If no additional checks needed, pass + if not self.appears_before_marker and not self.appears_after_marker: + return True, "" + + # Helper function to clean text for comparison (remove whitespace and non-alpha) + def clean_for_comparison(text): + # Remove all non-alphanumeric characters and normalize + return "".join(c for c in normalize_text(text) if c.isalnum()).lower() + + # Check appears_before_marker if provided + before_found = False if self.appears_before_marker else True + if self.appears_before_marker: + clean_target_before = clean_for_comparison(self.appears_before_marker) + threshold = 1.0 - (self.max_diffs / (len(self.appears_before_marker) if len(self.appears_before_marker) > 0 else 1)) + + for pos in marker_positions: + # Get text before this marker position + start_pos = max(0, pos["start"] - 200) # Look back up to 200 chars + text_before = md_content[start_pos : pos["start"]] + + # Clean the text before for comparison + clean_text_before = clean_for_comparison(text_before) + + # Check if appears_before_marker is at the end of this text (using fuzzy matching) + if clean_text_before: + # Use partial_ratio to check if target appears at the end + # We'll check the last portion that's roughly the size of our target + check_length = min(len(clean_text_before), len(clean_target_before) 2) + text_to_check = clean_text_before[-check_length:] if check_length > 0 else clean_text_before + + similarity = fuzz.partial_ratio(clean_target_before, text_to_check) / 100.0 + if similarity >= threshold: + before_found = True + break + + # Check appears_after_marker if provided + after_found = False if self.appears_after_marker else True + if self.appears_after_marker: + clean_target_after = clean_for_comparison(self.appears_after_marker) + threshold = 1.0 - (self.max_diffs / (len(self.appears_after_marker) if len(self.appears_after_marker) > 0 else 1)) + + for pos in marker_positions: + # Get text after this marker position + end_pos = min(len(md_content), pos["end"] + 200) # Look ahead up to 200 chars + text_after = md_content[pos["end"] : end_pos] + + # Clean the text after for comparison + clean_text_after = clean_for_comparison(text_after) + + # Check if appears_after_marker is at the beginning of this text (using fuzzy matching) + if clean_text_after: + # Use partial_ratio to check if target appears at the beginning + # We'll check the first portion that's roughly the size of our target + check_length = min(len(clean_text_after), len(clean_target_after) * 2) + text_to_check = clean_text_after[:check_length] if check_length > 0 else clean_text_after + + similarity = fuzz.partial_ratio(clean_target_after, text_to_check) / 100.0 + if similarity >= threshold: + after_found = True + break + + # Build failure message if needed + failures = [] + if self.appears_before_marker and not before_found: + failures.append(f"Text '{self.appears_before_marker[:40]}...' not found before any occurrence of marker '{self.marker}'") + if self.appears_after_marker and not after_found: + failures.append(f"Text '{self.appears_after_marker[:40]}...' not found after any occurrence of marker '{self.marker}'") + + if failures: + return False, "; ".join(failures) + else: + return True, "" + + +def load_single_test(data: Union[str, Dict]) -> BasePDFTest: + """ + Load a single test from a JSON line string or JSON object. + + Args: + data: Either a JSON string to parse or a dictionary containing test data. + + Returns: + A test object of the appropriate type. + + Raises: + ValidationError: If the test type is unknown or data is invalid. + json.JSONDecodeError: If the string cannot be parsed as JSON. + """ + # Handle JSON string input + if isinstance(data, str): + data = data.strip() + if not data: + raise ValueError("Empty string provided") + data = json.loads(data) + + # Process the test data + test_type = data.get("type") + if test_type in {TestType.PRESENT.value, TestType.ABSENT.value}: + test = TextPresenceTest(data) + elif test_type == TestType.ORDER.value: + test = TextOrderTest(data) + elif test_type == TestType.TABLE.value: + test = TableTest(data) + elif test_type == TestType.MATH.value: + test = MathTest(data) + elif test_type == TestType.BASELINE.value: + test = BaselineTest(data) + elif test_type == TestType.FORMAT.value: + test = FormatTest(data) + elif test_type == TestType.FOOTNOTE.value: + test = FootnoteTest(**data) + else: + raise ValidationError(f"Unknown test type: {test_type}") + + return test + + +def load_tests(jsonl_file: str) -> List[BasePDFTest]: + """ + Load tests from a JSONL file using parallel processing with a ThreadPoolExecutor. + + Args: + jsonl_file: Path to the JSONL file containing test definitions. + + Returns: + A list of test objects. + """ + + def process_line_with_number(line_tuple: Tuple[int, str]) -> Optional[Tuple[int, BasePDFTest]]: + """ + Process a single line from the JSONL file and return a tuple of (line_number, test object). + Returns None for empty lines. + """ + line_number, line = line_tuple + line = line.strip() + if not line: + return None + + try: + test = load_single_test(line) + return (line_number, test) + except json.JSONDecodeError as e: + print(f"Error parsing JSON on line {line_number}: {e}") + raise + except (ValidationError, KeyError) as e: + print(f"Error on line {line_number}: {e}") + raise + except Exception as e: + print(f"Unexpected error on line {line_number}: {e}") + raise + + tests = [] + + # Read all lines along with their line numbers. + with open(jsonl_file, "r") as f: + lines = list(enumerate(f, start=1)) + + # Use a ThreadPoolExecutor to process each line in parallel. + with ThreadPoolExecutor(max_workers=min(os.cpu_count() or 1, 64)) as executor: + # Submit all tasks concurrently. + futures = {executor.submit(process_line_with_number, item): item[0] for item in lines} + # Use tqdm to show progress as futures complete. + for future in tqdm(as_completed(futures), total=len(futures), desc="Loading tests"): + result = future.result() + if result is not None: + _, test = result + tests.append(test) + + # Check for duplicate test IDs after parallel processing. + unique_ids = set() + for test in tests: + if test.id in unique_ids: + raise ValidationError(f"Test with duplicate id {test.id} found, error loading tests.") + unique_ids.add(test.id) + + return tests + + +def save_tests(tests: List[BasePDFTest], jsonl_file: str) -> None: + """ + Save tests to a JSONL file using asdict for conversion. + + Args: + tests: A list of test objects. + jsonl_file: Path to the output JSONL file. + """ + with open(jsonl_file, "w") as file: + for test in tests: + file.write(json.dumps(asdict(test)) + "\n") diff --git a/olmocr/bench/utils.py b/olmocr/bench/utils.py new file mode 100644 index 0000000..5fe5f38 --- /dev/null +++ b/olmocr/bench/utils.py @@ -0,0 +1,194 @@ +from typing import List, Tuple + +import numpy as np + + +def calculate_bootstrap_ci(test_scores: List[float], n_bootstrap: int = 1000, ci_level: float = 0.95, splits: List[int] = None) -> Tuple[float, float]: + """ + Calculate bootstrap confidence interval for test scores, respecting category splits. + + Args: + test_scores: List of test scores (0.0 to 1.0 for each test) + n_bootstrap: Number of bootstrap samples to generate + ci_level: Confidence interval level (default: 0.95 for 95% CI) + splits: List of sizes for each category. If provided, resampling will be done + within each category independently, and the overall score will be the + average of per-category scores. If None, resampling is done across all tests. + + Returns: + Tuple of (lower_bound, upper_bound) representing the confidence interval + """ + if not test_scores: + return (0.0, 0.0) + + # Convert to numpy array for efficiency + scores = np.array(test_scores) + + # Simple case - no splits provided, use traditional bootstrap + if splits is None: + # Generate bootstrap samples + bootstrap_means = [] + for _ in range(n_bootstrap): + # Sample with replacement + sample = np.random.choice(scores, size=len(scores), replace=True) + bootstrap_means.append(np.mean(sample)) + else: + # Validate splits + if sum(splits) != len(scores): + raise ValueError(f"Sum of splits ({sum(splits)}) must equal length of test_scores ({len(scores)})") + + # Convert flat scores list to a list of category scores + category_scores = [] + start_idx = 0 + for split_size in splits: + category_scores.append(scores[start_idx : start_idx + split_size]) + start_idx += split_size + + # Generate bootstrap samples respecting category structure + bootstrap_means = [] + for _ in range(n_bootstrap): + # Sample within each category independently + category_means = [] + for cat_scores in category_scores: + if len(cat_scores) > 0: + # Sample with replacement within this category + cat_sample = np.random.choice(cat_scores, size=len(cat_scores), replace=True) + category_means.append(np.mean(cat_sample)) + + # Overall score is average of category means (if any categories have scores) + if category_means: + bootstrap_means.append(np.mean(category_means)) + + # Calculate confidence interval + alpha = (1 - ci_level) / 2 + lower_bound = np.percentile(bootstrap_means, alpha * 100) + upper_bound = np.percentile(bootstrap_means, (1 - alpha) * 100) + + return (lower_bound, upper_bound) + + +def perform_permutation_test( + scores_a: List[float], scores_b: List[float], n_permutations: int = 10000, splits_a: List[int] = None, splits_b: List[int] = None +) -> Tuple[float, float]: + """ + Perform a permutation test to determine if there's a significant difference + between two sets of test scores. + + Args: + scores_a: List of test scores for candidate A + scores_b: List of test scores for candidate B + n_permutations: Number of permutations to perform + splits_a: List of sizes for each category in scores_a + splits_b: List of sizes for each category in scores_b + + Returns: + Tuple of (observed_difference, p_value) + """ + if not scores_a or not scores_b: + return (0.0, 1.0) + + # Function to calculate mean of means with optional category splits + def mean_of_category_means(scores, splits=None): + if splits is None: + return np.mean(scores) + + category_means = [] + start_idx = 0 + for split_size in splits: + if split_size > 0: + category_scores = scores[start_idx : start_idx + split_size] + category_means.append(np.mean(category_scores)) + start_idx += split_size + + return np.mean(category_means) if category_means else 0.0 + + # Calculate observed difference in means using category structure if provided + mean_a = mean_of_category_means(scores_a, splits_a) + mean_b = mean_of_category_means(scores_b, splits_b) + observed_diff = mean_a - mean_b + + # If no splits are provided, fall back to traditional permutation test + if splits_a is None and splits_b is None: + # Combine all scores + combined = np.concatenate([scores_a, scores_b]) + n_a = len(scores_a) + + # Perform permutation test + count_greater_or_equal = 0 + for _ in range(n_permutations): + # Shuffle the combined array + np.random.shuffle(combined) + + # Split into two groups of original sizes + perm_a = combined[:n_a] + perm_b = combined[n_a:] + + # Calculate difference in means + perm_diff = np.mean(perm_a) - np.mean(perm_b) + + # Count how many permuted differences are >= to observed difference in absolute value + if abs(perm_diff) >= abs(observed_diff): + count_greater_or_equal += 1 + else: + # For category-based permutation test, we need to maintain category structure + # Validate that the splits match the score lengths + if splits_a is not None and sum(splits_a) != len(scores_a): + raise ValueError(f"Sum of splits_a ({sum(splits_a)}) must equal length of scores_a ({len(scores_a)})") + if splits_b is not None and sum(splits_b) != len(scores_b): + raise ValueError(f"Sum of splits_b ({sum(splits_b)}) must equal length of scores_b ({len(scores_b)})") + + # Create category structures + categories_a = [] + categories_b = [] + + if splits_a is not None: + start_idx = 0 + for split_size in splits_a: + categories_a.append(scores_a[start_idx : start_idx + split_size]) + start_idx += split_size + else: + # If no splits for A, treat all scores as one category + categories_a = [scores_a] + + if splits_b is not None: + start_idx = 0 + for split_size in splits_b: + categories_b.append(scores_b[start_idx : start_idx + split_size]) + start_idx += split_size + else: + # If no splits for B, treat all scores as one category + categories_b = [scores_b] + + # Perform permutation test maintaining category structure + count_greater_or_equal = 0 + for _ in range(n_permutations): + # For each category pair, shuffle and redistribute + perm_categories_a = [] + perm_categories_b = [] + + for cat_a, cat_b in zip(categories_a, categories_b): + # Combine and shuffle + combined = np.concatenate([cat_a, cat_b]) + np.random.shuffle(combined) + + # Redistribute maintaining original sizes + perm_categories_a.append(combined[: len(cat_a)]) + perm_categories_b.append(combined[len(cat_a) :]) + + # Flatten permuted categories + perm_a = np.concatenate(perm_categories_a) + perm_b = np.concatenate(perm_categories_b) + + # Calculate difference in means respecting category structure + perm_mean_a = mean_of_category_means(perm_a, splits_a) + perm_mean_b = mean_of_category_means(perm_b, splits_b) + perm_diff = perm_mean_a - perm_mean_b + + # Count how many permuted differences are >= to observed difference in absolute value + if abs(perm_diff) >= abs(observed_diff): + count_greater_or_equal += 1 + + # Calculate p-value + p_value = count_greater_or_equal / n_permutations + + return (observed_diff, p_value) diff --git a/olmocr/check.py b/olmocr/check.py new file mode 100644 index 0000000..e90c77f --- /dev/null +++ b/olmocr/check.py @@ -0,0 +1,47 @@ +import importlib.util +import logging +import subprocess +import sys + +logger = logging.getLogger(name) + + +def check_poppler_version(): + try: + result = subprocess.run(["pdftoppm", "-h"], stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True) + if result.returncode == 0 and result.stderr.startswith("pdftoppm"): + logger.info("pdftoppm is installed and working.") + else: + logger.error("pdftoppm is installed but returned an error.") + sys.exit(1) + except FileNotFoundError: + logger.error("pdftoppm is not installed.") + logger.error("Check the README in the https://github.com/allenai/olmocr/blob/main/README.md for installation instructions") + sys.exit(1) + + +def check_sglang_version(): + if importlib.util.find_spec("sglang") is None: + logger.error("Please make sure sglang is installed according to the latest instructions here: https://docs.sglang.ai/start/install.html") + logger.error("Sglang needs to be installed with a separate command in order to find all dependencies properly.") + sys.exit(1) + + +def check_torch_gpu_available(min_gpu_memory: int = 15 * 10243): + try: + import torch + except: + logger.error("Pytorch must be installed, visit https://pytorch.org/ for installation instructions") + raise + + try: + gpu_memory = torch.cuda.get_device_properties(0).total_memory + assert gpu_memory >= min_gpu_memory + except: + logger.error(f"Torch was not able to find a GPU with at least {min_gpu_memory // (1024 3)} GB of RAM.") + raise + + +if name == "main": + check_poppler_version() + check_sglang_version() diff --git a/olmocr/data/init.py b/olmocr/data/init.py new file mode 100644 index 0000000..e69de29 diff --git a/olmocr/data/build_openai_batch_from_olmocrmix.py b/olmocr/data/build_openai_batch_from_olmocrmix.py new file mode 100755 index 0000000..2f4a1c0 --- /dev/null +++ b/olmocr/data/build_openai_batch_from_olmocrmix.py @@ -0,0 +1,285 @@ +#!/usr/bin/env python3 +""" +Build OpenAI batch requests from OLMoCR-mix folder structure. + +This script processes the folder structure created by prepare_olmocrmix.py +and generates OpenAI batch API requests for processing PDFs. +""" + +import argparse +import json +from concurrent.futures import ThreadPoolExecutor, as_completed +from pathlib import Path +from typing import Any, Dict, Generator, Optional, Tuple + +from pypdf import PdfReader +from tqdm import tqdm + +from olmocr.data.renderpdf import ( + get_png_dimensions_from_base64, + render_pdf_to_base64png, +) +from olmocr.prompts.prompts import ( + build_openai_silver_data_prompt_v3_simple, + openai_response_format_schema, +) + +TARGET_IMAGE_DIM = 2048 +MAX_FILE_SIZE = 99 * 1024 * 1024 # 99MB in bytes + + +def validate_single_page_pdf(pdf_path: Path) -> bool: + """ + Validate that a PDF has exactly one page. + + Args: + pdf_path: Path to the PDF file + + Returns: + True if PDF has exactly one page, False otherwise + """ + try: + pdf = PdfReader(pdf_path) + return len(pdf.pages) == 1 + except Exception as e: + print(f"Error reading PDF {pdf_path}: {e}") + return False + + +def build_custom_id(pdf_path: Path, base_dir: Path) -> str: + """ + Build a custom ID for the request that can be used to recover the file later. + + The ID preserves the full path structure for easy recovery. + Example: extracted/document_id.pdf becomes "extracted/document_id" + + Args: + pdf_path: Full path to the PDF file + base_dir: Base directory containing the processed folder + + Returns: + Custom ID string that preserves path structure + """ + # Get relative path from base directory + rel_path = pdf_path.relative_to(base_dir) + # Remove .pdf extension but keep directory structure + path_without_ext = str(rel_path).replace(".pdf", "") + return path_without_ext + + +def process_single_pdf(pdf_path: Path, base_dir: Path) -> Optional[Tuple[Dict[str, Any], Path]]: + """ + Process a single PDF and return the batch request if valid. + + Args: + pdf_path: Path to the PDF file + base_dir: Base directory for building custom IDs + + Returns: + Tuple of (request dict, pdf_path) if successful, None otherwise + """ + # Validate PDF has single page + try: + pdf = PdfReader(pdf_path) + if len(pdf.pages) != 1: + return None + except Exception as e: + print(f"Error reading PDF {pdf_path}: {e}") + return None + + try: + # Render PDF to base64 image + image_base64 = render_pdf_to_base64png(str(pdf_path), page_num=1, target_longest_image_dim=TARGET_IMAGE_DIM) + + # Get image dimensions for the prompt + width, height = get_png_dimensions_from_base64(image_base64) + + # Build the prompt using v3 simple version + prompt = build_openai_silver_data_prompt_v3_simple(width, height) + + # Build custom ID + custom_id = build_custom_id(pdf_path, base_dir) + + # Build the request in OpenAI batch format + request = { + "custom_id": custom_id, + "method": "POST", + "url": "/v1/chat/completions", + "body": { + "model": "gpt-4.1", + "messages": [ + { + "role": "user", + "content": [ + {"type": "text", "text": prompt}, + {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{image_base64}"}}, + ], + } + ], + "temperature": 0.1, + "max_completion_tokens": 12000, + "response_format": openai_response_format_schema(), + }, + } + + return (request, pdf_path) + except Exception as e: + print(f"Error processing {pdf_path}: {e}") + return None + + +def find_pdf_files(input_dir: Path) -> Generator[Path, None, None]: + """ + Find all PDF files in the processed folder structure. + + The structure is expected to be: + processed_XX_subset_split/ + extracted/ + .pdf + + Or for hugging_face downloads: + hugging_face/ + pdf_tarballs/ + extracted/ + .pdf + + Args: + input_dir: Input directory path + + Yields: + Path objects for each PDF file found + """ + + for pdf_path in input_dir.rglob(".pdf"): + yield pdf_path + + +def process_pdfs_to_batch_requests(input_dir: Path, output_dir: Path, max_pdfs: int = None, num_workers: int = 8) -> int: + """ + Process PDFs and create batch request files using parallel processing. + + Args: + input_dir: Directory containing the processed folder structure + output_dir: Directory to save batch request files + max_pdfs: Maximum number of PDFs to process (None for all) + num_workers: Number of parallel workers for processing + + Returns: + Number of PDFs processed + """ + # Ensure output directory exists + output_dir.mkdir(parents=True, exist_ok=True) + + # Initialize file management + file_num = 0 + current_file_size = 0 + current_file_path = output_dir / f"batch_requests_{file_num:04d}.jsonl" + current_file = open(current_file_path, "w") + + pdfs_processed = 0 + pdfs_skipped = 0 + + # Find PDF files + pdf_files = list(find_pdf_files(input_dir)) + + # Limit files if max_pdfs is specified + if max_pdfs: + pdf_files = pdf_files[:max_pdfs] + + total_pdfs = len(pdf_files) + + print(f"Found {total_pdfs} PDF files to process") + print(f"Using {num_workers} parallel workers") + + # Process PDFs in parallel using ThreadPoolExecutor + with ThreadPoolExecutor(max_workers=num_workers) as executor: + # Submit all PDF processing tasks + future_to_pdf = {executor.submit(process_single_pdf, pdf_path, input_dir): pdf_path for pdf_path in pdf_files} + + # Process results as they complete + with tqdm(total=total_pdfs, desc="Processing PDFs") as pbar: + for future in as_completed(future_to_pdf): + pdf_path = future_to_pdf[future] + + try: + result = future.result() + + if result is None: + # PDF was skipped (multi-page or error) + pdfs_skipped += 1 + else: + request, _ = result + request_json = json.dumps(request) + request_size = len(request_json.encode("utf-8")) + + # Check if we need to start a new file + if current_file_size + request_size > MAX_FILE_SIZE: + current_file.close() + file_num += 1 + current_file_path = output_dir / f"batch_requests_{file_num:04d}.jsonl" + current_file = open(current_file_path, "w") + current_file_size = 0 + print(f"\nStarting new batch file: {current_file_path.name}") + + # Write the request (only in main thread) + current_file.write(request_json) + current_file.write("\n") + current_file_size += request_size + + pdfs_processed += 1 + + except Exception as e: + print(f"\nError with {pdf_path}: {e}") + pdfs_skipped += 1 + + pbar.update(1) + + # Close the last file + current_file.close() + + print(f"\nProcessing complete:") + print(f" - PDFs processed: {pdfs_processed}") + print(f" - PDFs skipped: {pdfs_skipped}") + print(f" - Batch files created: {file_num + 1}") + print(f" - Output directory: {output_dir}") + + return pdfs_processed + + +def main(): + parser = argparse.ArgumentParser(description="Build OpenAI batch requests from OLMoCR-mix folder structure") + parser.add_argument("--output_dir", type=str, default=None, help="Output directory for batch request files (default: input_dir/batch_requests)") + parser.add_argument("--max_pdfs", type=int, default=None, help="Maximum number of PDFs to process (default: all)") + parser.add_argument("--num_workers", type=int, default=8, help="Number of parallel workers for processing (default: 8)") + parser.add_argument( + "input_dir", + type=str, + help="Input directory containing processed folder structure (e.g., ~/olmOCR-mix-0225/processed_00_documents_eval_s2pdf or ~/olmOCR-mix-0225)", + ) + + args = parser.parse_args() + + # Convert paths to Path objects + input_dir = Path(args.input_dir).expanduser().resolve() + + if not input_dir.exists(): + print(f"Error: Input directory does not exist: {input_dir}") + return 1 + + # Set default output directory if not specified + if args.output_dir: + output_dir = Path(args.output_dir).expanduser().resolve() + else: + output_dir = input_dir / "batch_requests" + + print(f"Input directory: {input_dir}") + print(f"Output directory: {output_dir}") + + # Process PDFs + process_pdfs_to_batch_requests(input_dir=input_dir, output_dir=output_dir, max_pdfs=args.max_pdfs, num_workers=args.num_workers) + + return 0 + + +if name == "main": + exit(main()) diff --git a/olmocr/data/buildsilver.py b/olmocr/data/buildsilver.py new file mode 100644 index 0000000..db4f908 --- /dev/null +++ b/olmocr/data/buildsilver.py @@ -0,0 +1,250 @@ +import argparse +import glob +import json +import os +import random +from concurrent.futures import ProcessPoolExecutor, as_completed +from typing import Generator +from urllib.parse import urlparse + +import boto3 +from pypdf import PdfReader +from tqdm import tqdm + +from olmocr.data.renderpdf import render_pdf_to_base64png +from olmocr.filter import PdfFilter +from olmocr.prompts import ( + build_openai_silver_data_prompt, + openai_response_format_schema, +) +from olmocr.prompts.anchor import get_anchor_text + +TARGET_IMAGE_DIM = 2048 + + +pdf_filter = PdfFilter() + + +def build_page_query(local_pdf_path: str, pretty_pdf_path: str, page: int) -> dict: + image_base64 = render_pdf_to_base64png(local_pdf_path, page, TARGET_IMAGE_DIM) + anchor_text = get_anchor_text(local_pdf_path, page, pdf_engine="pdfreport") + + # Construct OpenAI Batch API request format# + # There are a few tricks to know when doing data processing with OpenAI's apis + # First off, use the batch query system, it's 1/2 the price and exactly the same performance + # Second off, use structured outputs. If your application is not an actual chatbot, use structured outputs! + # Even if the last 10 queries you ran with the regular chat api returned exactly what you wanted without extra "LLM fluff text", that doesn't mean this will hold across 1000's of queries + # Also, structured outputs let you cheat, because the order in which fields are in your schema, is the order in which the model will answer them, so you can have it answer some "preperatory" or "chain of thought" style questions first before going into the meat of your response, which is going to give better answers + # Check your prompt for typos, it makes a performance difference! + # Ask for logprobs, it's not any more expensive and you can use them later to help identify problematic responses + return { + "custom_id": f"{pretty_pdf_path}-{page}", + "method": "POST", + "url": "/v1/chat/completions", + "body": { + "model": "gpt-4o-2024-08-06", + "messages": [ + { + "role": "user", + "content": [ + {"type": "text", "text": build_openai_silver_data_prompt(anchor_text)}, + {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{image_base64}"}}, + ], + } + ], + "temperature": 0.1, + "max_tokens": 6000, + "logprobs": True, + "top_logprobs": 5, + "response_format": openai_response_format_schema(), + }, + } + + +def sample_pdf_pages(num_pages: int, first_n_pages: int, max_sample_pages: int) -> list: + if num_pages <= first_n_pages: + return list(range(1, num_pages + 1)) # Return all pages if fewer than first_n_pages + sample_pages = list(range(1, first_n_pages + 1)) # Always get the first_n_pages + remaining_pages = list(range(first_n_pages + 1, num_pages + 1)) + if remaining_pages: + sample_pages += random.sample(remaining_pages, min(max_sample_pages - first_n_pages, len(remaining_pages))) + return sample_pages + + +def fetch_s3_file(s3_url: str, local_path: str) -> str: + parsed = urlparse(s3_url) + bucket_name = parsed.netloc + key = parsed.path.lstrip("/") + + s3 = boto3.client("s3") + s3.download_file(bucket_name, key, local_path) + return local_path + + +def process_pdf(pdf_path: str, first_n_pages: int, max_sample_pages: int, no_filter: bool) -> Generator[dict, None, None]: + if pdf_path.startswith("s3://"): + local_pdf_path = os.path.join("/tmp", os.path.basename(pdf_path)) + fetch_s3_file(pdf_path, local_pdf_path) + else: + local_pdf_path = pdf_path + + if (not no_filter) and pdf_filter.filter_out_pdf(local_pdf_path): + print(f"Skipping {local_pdf_path} due to common filter") + return [] + + pretty_pdf_path = pdf_path + + pdf = PdfReader(local_pdf_path) + num_pages = len(pdf.pages) + + sample_pages = sample_pdf_pages(num_pages, first_n_pages, max_sample_pages) + + result = [] + for page in sample_pages: + try: + query = build_page_query(local_pdf_path, pretty_pdf_path, page) + result.append(query) + except Exception as e: + print(f"Error processing page {page} of {pdf_path}: {e}") + + return result + + +def main(): + parser = argparse.ArgumentParser(description="Sample PDFs and create requests for GPT-4o.") + parser.add_argument("--glob_path", type=str, help="Local or S3 path glob (e.g., .pdf or s3://bucket/pdfs/.pdf).") + parser.add_argument("--path_list", type=str, help="Path to a file containing paths to PDFs, one per line.") + parser.add_argument("--no_filter", action="store_true", help="Disables the basic spam/language filtering so that ALL pdfs listed are used") + parser.add_argument("--num_sample_docs", type=int, default=5000, help="Number of PDF documents to sample.") + parser.add_argument("--first_n_pages", type=int, default=0, help="Always sample the first N pages of each PDF.") + parser.add_argument("--max_sample_pages", type=int, default=15, help="Max number of pages to sample per PDF.") + parser.add_argument("--output", type=str, default="openai_batch_data", help="Output destination") + parser.add_argument("--reservoir_size", type=int, default=None, help="Size of the reservoir for sampling paths. Defaults to 10x num_sample_docs.") + args = parser.parse_args() + + # Set default reservoir_size if not provided + if args.reservoir_size is None: + args.reservoir_size = 10 args.num_sample_docs + + # Initialize reservoir sampling variables + pdf_paths = [] + n = 0 # Total number of items seen + + # Load PDF paths from glob or path_list using reservoir sampling + if args.glob_path: + if args.glob_path.startswith("s3://"): + # Handle S3 globbing using boto3 with pagination + parsed = urlparse(args.glob_path) + s3 = boto3.client("s3") + bucket_name = parsed.netloc + prefix = os.path.dirname(parsed.path.lstrip("/")) + "/" + paginator = s3.get_paginator("list_objects_v2") + page_iterator = paginator.paginate(Bucket=bucket_name, Prefix=prefix) + + for page in page_iterator: + for obj in page.get("Contents", []): + if obj["Key"].endswith(".pdf"): + n += 1 + path = f"s3://{bucket_name}/{obj['Key']}" + if len(pdf_paths) < args.reservoir_size: + pdf_paths.append(path) + else: + s = random.randint(1, n) + if s <= args.reservoir_size: + pdf_paths[s - 1] = path + else: + # Handle local globbing using glob.iglob() + for path in glob.iglob(args.glob_path, recursive=True): + n += 1 + if len(pdf_paths) < args.reservoir_size: + pdf_paths.append(path) + else: + s = random.randint(1, n) + if s <= args.reservoir_size: + pdf_paths[s - 1] = path + elif args.path_list: + with open(args.path_list, "r") as f: + for line in f: + n += 1 + path = line.strip() + if len(pdf_paths) < args.reservoir_size: + pdf_paths.append(path) + else: + s = random.randint(1, n) + if s <= args.reservoir_size: + pdf_paths[s - 1] = path + + # Shuffle the reservoir + random.shuffle(pdf_paths) + + print(f"Loaded and shuffled {len(pdf_paths)} paths to use.") + + # Rest of the code remains the same + cur_file_num = 0 + output_dir = args.output + max_file_size = 99 * 1024 * 1024 # 99MB in bytes + cur_file_size = 0 + cur_file_path = os.path.join(output_dir, f"output_{cur_file_num}.jsonl") + + # Ensure output directory exists + os.makedirs(output_dir, exist_ok=True) + + # Open the first file for writing + cur_file = open(cur_file_path, "w") + + # Counter to track PDFs that produce at least one output + pdfs_with_output = 0 + + # Using ThreadPoolExecutor to process files concurrently + with ProcessPoolExecutor() as executor: + futures = [] + + with tqdm(desc="Processing PDFs", leave=False, total=args.num_sample_docs) as pb: + for pdf_path in pdf_paths: + futures.append(executor.submit(process_pdf, pdf_path, args.first_n_pages, args.max_sample_pages, args.no_filter)) + + for future in as_completed(futures): + has_output = False # Track if the current PDF produces at least one request + try: + request_results = future.result() # Get the result from the thread + + for request_obj in request_results: + request_json = json.dumps(request_obj) + request_size = len(request_json.encode("utf-8")) # Calculate size in bytes + + # Check if the current request can fit in the current file + if cur_file_size + request_size > max_file_size: + # Close the current file and create a new one + cur_file.close() + cur_file_num += 1 + cur_file_path = os.path.join(output_dir, f"output_{cur_file_num}.jsonl") + cur_file = open(cur_file_path, "w") + cur_file_size = 0 # Reset file size + + # Write the JSON entry to the file + cur_file.write(request_json) + cur_file.write("\n") + cur_file_size += request_size + + has_output = True # At least one request object was generated + + if has_output: + pdfs_with_output += 1 + pb.update(1) + + if pdfs_with_output >= args.num_sample_docs: + executor.shutdown(cancel_futures=True) + break + + except Exception as e: + print(f"Error processing {pdf_path}: {str(e)}") + + # Close the last open file + cur_file.close() + + # Print or log the number of PDFs that resulted in at least one output + print(f"Number of sampled PDFs that produced at least one output: {pdfs_with_output}") + + +if name == "main": + main() diff --git a/olmocr/data/clean_olmocrmix.py b/olmocr/data/clean_olmocrmix.py new file mode 100755 index 0000000..61b00e0 --- /dev/null +++ b/olmocr/data/clean_olmocrmix.py @@ -0,0 +1,314 @@ +#!/usr/bin/env python3 +# Takes a dataset location in olmocr-mix format, (ex. a nested directory structure folder/subfolder/document.md with a corresponding folder/subfolder/document.pdf) +# Then, it will randomly shuffle these (with a fixed seed), and prompt chatgpt to clean up the transcription, and output a cleaned document +# Uses structured output to get a good result, then writes things back in the same format in a new root folder, preserving the original folder structure + +import argparse +import json +import os +import random +import sys +from concurrent.futures import ThreadPoolExecutor, as_completed +from dataclasses import dataclass +from pathlib import Path +from typing import Any, Dict, List, Tuple + +from openai import OpenAI +from pydantic import BaseModel, Field +from pypdf import PdfReader +from tqdm import tqdm + +from olmocr.data.renderpdf import render_pdf_to_base64png + + +# Structured output model for ChatGPT response +class CleanedDocument(BaseModel): + cleaned_text: str = Field(description="The cleaned and corrected version of the OCR transcription") + confidence_score: float = Field(description="Confidence score from 0 to 1 indicating how confident the model is in the cleaning", ge=0.0, le=1.0) + corrections_made: List[str] = Field(description="List of major corrections or improvements made to the text") + is_page_all_blank: bool = Field(description="Document consists entirely of blank page, or only headers/footers that would otherwise be removed") + primary_language: str = Field(default="en", description="Primary language of the document (ISO 639-1 code, e.g. 'en' for English, 'es' for Spanish)") + is_rotation_valid: bool = Field(default=True, description="Whether the page orientation/rotation appears correct") + rotation_correction: int = Field(default=0, description="Degrees of rotation needed to correct orientation (0, 90, 180, or 270)") + is_table: bool = Field(default=False, description="Whether the page primarily contains a table") + is_diagram: bool = Field(default=False, description="Whether the page primarily contains a diagram or figure") + + +@dataclass +class DocumentPair: + md_path: Path + pdf_path: Path + relative_path: Path # Relative path from root for preserving structure + + +def parse_args(): + parser = argparse.ArgumentParser(description="Clean OCR transcriptions using ChatGPT with visual PDF context") + parser.add_argument("input_dir", help="Input directory containing olmocr-mix format data (MD files with corresponding PDFs)") + parser.add_argument("output_dir", help="Output directory for cleaned documents (preserves folder structure)") + parser.add_argument( + "--openai-api-key", help="OpenAI API key (can also be set via OPENAI_API_KEY environment variable)", default=os.getenv("OPENAI_API_KEY") + ) + parser.add_argument("--model", default="gpt-4o-2024-08-06", help="OpenAI model to use (default: gpt-4o-mini)") + parser.add_argument("--seed", type=int, default=42, help="Random seed for shuffling documents (default: 42)") + parser.add_argument("--batch-size", type=int, default=10, help="Number of documents to process in parallel (default: 10)") + parser.add_argument("--max-documents", type=int, help="Maximum number of documents to process (useful for testing)") + parser.add_argument("--skip-existing", action="store_true", help="Skip documents that already have cleaned versions in the output directory") + parser.add_argument("--verbose", action="store_true", help="Enable verbose output") + return parser.parse_args() + + +def check_single_page_pdf(pdf_path: Path) -> bool: + """Check if a PDF has exactly one page.""" + try: + with open(pdf_path, "rb") as pdf_file: + pdf_reader = PdfReader(pdf_file) + return len(pdf_reader.pages) == 1 + except Exception as e: + print(f"Error checking PDF {pdf_path}: {e}") + return False + + +def find_document_pairs(input_dir: Path, verbose: bool = False) -> List[DocumentPair]: + """Find all MD files with corresponding PDF files.""" + pairs = [] + skipped_no_pdf = 0 + + for md_path in input_dir.rglob("*.md"): + # Check for corresponding PDF + pdf_path = md_path.with_suffix(".pdf") + if not pdf_path.exists(): + if verbose: + print(f"Warning: No PDF found for {md_path}") + skipped_no_pdf += 1 + continue + + relative_path = md_path.relative_to(input_dir) + pairs.append(DocumentPair(md_path, pdf_path, relative_path)) + + if skipped_no_pdf > 0: + print(f"Skipped {skipped_no_pdf} files without PDFs") + + return pairs + + +def render_single_page_pdf(pdf_path: Path) -> str: + """Render a single-page PDF to base64 PNG image.""" + try: + # Use render_pdf_to_base64png with target_longest_image_dim=2048 + base64_png = render_pdf_to_base64png(str(pdf_path), 1, target_longest_image_dim=2048) # Always page 1 since we validated it's a single-page PDF + return base64_png + except Exception as e: + raise RuntimeError(f"Could not render PDF {pdf_path}: {e}") + + +def clean_document_with_chatgpt(client: OpenAI, model: str, md_content: str, pdf_image: str, verbose: bool = False) -> CleanedDocument: + """Use ChatGPT to clean the OCR transcription with PDF context.""" + + # Prepare the messages + messages: List[Dict[str, Any]] = [ + { + "role": "system", + "content": ( + "You are an expert at cleaning and correcting OCR transcriptions. " + "You will be given an OCR transcription and an image of the original PDF page. " + "Your task is to:\n" + "1. Correct formatting issues.\n" + "2. Preserve the exact spelling of words from the original document.\n" + "3. Remove any original transcriber's marks and notes, usually indicated by [ and ] symbols.\n" + "4. Fix word breaks and line breaks\n" + "5. Ensure mathematical formulas and special characters are correct\n" + "6. If there are any figures or charts, label them with the following markdown syntax ![Alt text describing the contents of the figure](page_startx_starty_width_height.png)\n" + "7. Maintain the semantic structure of the document\n" + "8. Remove any headers or footers that are not semantically relevant to the main document contents, ex page numbers, document classifications, etc.\n" + "9. Convert tables into HTML format. Keep the syntax simple, but use

Text Evaluation Review

+ + Toggle diff +

+ +

+ + Reveal Gold/Eval +

Votes

+ {% for entry in entries %} +

+ +

Alignment: {{ entry.alignment }}

+ {{ entry.s3_path }} (Page {{ entry.page }}) + + +

+ + + +

+ +

+ {% endfor %} +

+ """ + pattern = re.compile( + r' found for this key in the HTML + continue + + left_method = entries_dict[entry_id]["left_metadata"] + right_method = entries_dict[entry_id]["right_metadata"] + + if left_method == right_method: + # Same "method" on both sides => skip + continue + + if vote == "left": + # left_method beats right_method + pair = tuple(sorted([left_method, right_method])) + if pair[0] == left_method: + comparisons[pair][0] += 1 + else: + comparisons[pair][1] += 1 + + elif vote == "right": + # right_method beats left_method + pair = tuple(sorted([left_method, right_method])) + if pair[0] == right_method: + comparisons[pair][0] += 1 + else: + comparisons[pair][1] += 1 + + else: + # "both_good", "both_bad", "invalid_pdf", etc. -> not counted + pass + + return comparisons + + +def elo_update(ratingA, ratingB, scoreA, scoreB, k=32): + """ + Perform a single ELO update for a match between A and B. + - ratingA, ratingB are current ELO ratings of A and B. + - scoreA, scoreB in {0 or 1} (1 if the player is the winner, 0 if loser). + - Returns (new_ratingA, new_ratingB). + """ + # Expected scores for each player + expectedA = 1 / (1 + 10 ** ((ratingB - ratingA) / 400)) + expectedB = 1 / (1 + 10 ** ((ratingA - ratingB) / 400)) + + new_ratingA = ratingA + k * (scoreA - expectedA) + new_ratingB = ratingB + k * (scoreB - expectedB) + return new_ratingA, new_ratingB + + +def compute_elo_arena(comparisons, k=32, initial_rating=1500): + """ + Given the aggregated comparisons dict: + comparisons[(A, B)] = [A_wins, B_wins] + + 1) Collect all unique methods. + 2) Initialize them to initial_rating (1500). + 3) For each pair (A, B), apply A_wins times the scenario + "A beats B" -> ELO update + B_wins times the scenario "B beats A" -> ELO update + + Because we don't have a strict order of matches, we just + apply them in some consistent but arbitrary order. + + Returns a dict { method_name: final_elo_rating } + """ + # 1) Collect all unique methods + methods = set() + for A, B in comparisons.keys(): + methods.add(A) + methods.add(B) + + # 2) Initialize ratings + ratings = {m: float(initial_rating) for m in methods} + + # 3) Walk through each pair + for (A, B), (A_wins, B_wins) in comparisons.items(): + for _ in range(A_wins): + # A beats B + oldA = ratings[A] + oldB = ratings[B] + newA, newB = elo_update(oldA, oldB, 1, 0, k=k) + ratings[A] = newA + ratings[B] = newB + + for _ in range(B_wins): + # B beats A + oldA = ratings[A] + oldB = ratings[B] + newA, newB = elo_update(oldA, oldB, 0, 1, k=k) + ratings[A] = newA + ratings[B] = newB + + return ratings + + +def make_report(urls): + """ + Main function that: + - Fetches each HTML page + - Extracts presignedGetUrl + - Fetches the JSON datastore + - Parses .entry blocks for metadata + - Produces an overall "win rate" report for each method vs method + - Produces an ELO arena result for each method + """ + # Aggregate all entries from all URLs into a single dict + # so each entry_id is unique across all pages (they usually are). + master_entries_dict = {} + master_datastore = {} + + for url in urls: + try: + html = fetch_review_page_html(url) + except Exception as e: + print(f"Error fetching HTML from {url}: {e}") + continue + + # Extract the presignedGetUrl + presigned_url = extract_presigned_url(html) + if not presigned_url: + print(f"Warning: Could not find presignedGetUrl in {url}") + continue + + # Fetch the datastore + datastore = fetch_presigned_datastore(presigned_url) + + # Parse the HTML for entry metadata + entries_dict = parse_entry_metadata(html) + + # Merge into master + for k, v in entries_dict.items(): + master_entries_dict[k] = v + for k, v in datastore.items(): + master_datastore[k] = v + + # Now build the comparison report + comparisons = build_comparison_report(master_entries_dict, master_datastore) + + print("=== Pairwise Win/Loss Report ===") + if not comparisons: + print("No head-to-head comparisons found (did not find left/right votes).") + return + + # Print out each matchup + for (A, B), (A_wins, B_wins) in comparisons.items(): + total = A_wins + B_wins + A_rate = A_wins / total * 100 if total else 0 + B_rate = B_wins / total * 100 if total else 0 + print(f"{A} vs {B}: " f"{A} wins={A_wins} ({A_rate:.1f}%), " f"{B} wins={B_wins} ({B_rate:.1f}%)") + + # -- ADDED: Write the same data to scoreelo.csv + with open("scoreelo.csv", "w", newline="", encoding="utf-8") as csvfile: + writer = csv.writer(csvfile) + writer.writerow(["MethodA", "MethodB", "A_wins", "B_wins", "A_rate(%)", "B_rate(%)"]) + for (A, B), (A_wins, B_wins) in comparisons.items(): + total = A_wins + B_wins + A_rate = A_wins / total * 100 if total else 0 + B_rate = B_wins / total * 100 if total else 0 + writer.writerow([A, B, A_wins, B_wins, f"{A_rate:.1f}", f"{B_rate:.1f}"]) + + # ==== ELO Arena ==== + elo_ratings = compute_elo_arena(comparisons, k=32, initial_rating=1500) + + # Sort methods by final rating descending + sorted_ratings = sorted(elo_ratings.items(), key=lambda x: x[1], reverse=True) + + print("\n=== ELO Arena Results ===") + for method, rating in sorted_ratings: + print(f"{method}: {rating:.2f}") + + +if __name__ == "__main__": + # Example usage + urls = [ + "https://jakep-tinyhost.s3.amazonaws.com/review_page_0-ff70abb8f517.html?AWSAccessKeyId=AKIASHLPW4FEVZOPGK46&Signature=NarEyyCfvusCh%2FHdB47VfHOnnBs%3D&Expires=1738359221", + "https://jakep-tinyhost.s3.amazonaws.com/review_page_1-0800f9af46cf.html?AWSAccessKeyId=AKIASHLPW4FEVZOPGK46&Signature=ncTWAu5rSndBJJsU26HRYDaK6i8%3D&Expires=1738359222", + "https://jakep-tinyhost.s3.amazonaws.com/review_page_10-f7081f6ca6f9.html?AWSAccessKeyId=AKIASHLPW4FEVZOPGK46&Signature=gYX8yjGyYshRqXGgdsX17%2Fdi9Ig%3D&Expires=1738359223", + "https://jakep-tinyhost.s3.amazonaws.com/review_page_11-355dc69335bc.html?AWSAccessKeyId=AKIASHLPW4FEVZOPGK46&Signature=7%2Bc5qoa8Tbk06z0VcvJiIIVAz9M%3D&Expires=1738359224", + "https://jakep-tinyhost.s3.amazonaws.com/review_page_12-95fce9bf0c18.html?AWSAccessKeyId=AKIASHLPW4FEVZOPGK46&Signature=fw4PBo0LnxikmLZ8xH%2BGD%2F%2BhXMU%3D&Expires=1738359225", + "https://jakep-tinyhost.s3.amazonaws.com/review_page_13-f88f7d7482bf.html?AWSAccessKeyId=AKIASHLPW4FEVZOPGK46&Signature=yXkQp9oFDtroKgiO50EwpYdGLcA%3D&Expires=1738359226", + "https://jakep-tinyhost.s3.amazonaws.com/review_page_14-8ac0b974bfd5.html?AWSAccessKeyId=AKIASHLPW4FEVZOPGK46&Signature=EgZTpj1%2FdzMBUgd%2BX4pVZ1Sp%2FrA%3D&Expires=1738359226", + "https://jakep-tinyhost.s3.amazonaws.com/review_page_15-e3136188de5c.html?AWSAccessKeyId=AKIASHLPW4FEVZOPGK46&Signature=YKhAv4unNIlRcerQAaHN4kjc4qI%3D&Expires=1738359227", + "https://jakep-tinyhost.s3.amazonaws.com/review_page_16-2c5abde50d49.html?AWSAccessKeyId=AKIASHLPW4FEVZOPGK46&Signature=Mj8%2BK5ISKzAYQFeYvmzTgCPcRwA%3D&Expires=1738359228", + "https://jakep-tinyhost.s3.amazonaws.com/review_page_17-f13132a4cdcc.html?AWSAccessKeyId=AKIASHLPW4FEVZOPGK46&Signature=%2FHuzw2cjJ4oFm91UXojPnGzYi8Q%3D&Expires=1738359229", + "https://jakep-tinyhost.s3.amazonaws.com/review_page_18-25070f2aa05e.html?AWSAccessKeyId=AKIASHLPW4FEVZOPGK46&Signature=ctd%2BUIM%2FxryJm%2FcwA%2BRZ%2FbRzBp8%3D&Expires=1738359230", + "https://jakep-tinyhost.s3.amazonaws.com/review_page_19-d436ee434162.html?AWSAccessKeyId=AKIASHLPW4FEVZOPGK46&Signature=jVdFKobIoHlbTQ7zziG%2BXiIQ0Fo%3D&Expires=1738359230", + "https://jakep-tinyhost.s3.amazonaws.com/review_page_2-a5ece743fd31.html?AWSAccessKeyId=AKIASHLPW4FEVZOPGK46&Signature=K8hIrjWtvo4SLVQrOB8TiXLgNJk%3D&Expires=1738359231", + "https://jakep-tinyhost.s3.amazonaws.com/review_page_3-9ce03af05f51.html?AWSAccessKeyId=AKIASHLPW4FEVZOPGK46&Signature=T0fLGSH%2Bv%2F19veqbxnLxoSf7gVA%3D&Expires=1738359232", + "https://jakep-tinyhost.s3.amazonaws.com/review_page_4-94eec18f8027.html?AWSAccessKeyId=AKIASHLPW4FEVZOPGK46&Signature=u2R1LundKpfnAUCcD%2BdGHA6uIR0%3D&Expires=1738359233", + "https://jakep-tinyhost.s3.amazonaws.com/review_page_5-377d0a7d8f5a.html?AWSAccessKeyId=AKIASHLPW4FEVZOPGK46&Signature=5R38ZQAR9ew5x%2BRmMVQbTqbfVh0%3D&Expires=1738359234", + "https://jakep-tinyhost.s3.amazonaws.com/review_page_6-537b22646a26.html?AWSAccessKeyId=AKIASHLPW4FEVZOPGK46&Signature=PLOELum1qzOXW8Cm5rfZphlFeMw%3D&Expires=1738359235", + "https://jakep-tinyhost.s3.amazonaws.com/review_page_7-a4a7dcb08f20.html?AWSAccessKeyId=AKIASHLPW4FEVZOPGK46&Signature=DxPHukGXEpPrEPL6TF9QBKPE1Xg%3D&Expires=1738359236", + "https://jakep-tinyhost.s3.amazonaws.com/review_page_8-48a71c829863.html?AWSAccessKeyId=AKIASHLPW4FEVZOPGK46&Signature=TjEINKj69HdmXsKY59k4f3PieeM%3D&Expires=1738359237", + "https://jakep-tinyhost.s3.amazonaws.com/review_page_9-8557438928c3.html?AWSAccessKeyId=AKIASHLPW4FEVZOPGK46&Signature=F7sQxw5A%2FDOcOaa%2FQSeqepH0PQc%3D&Expires=1738359238", + ] + # import tinyhost + + # print(tinyhost.tinyhost(urls)) + + make_report(urls) diff --git a/scripts/hf_local_test.py b/scripts/hf_local_test.py new file mode 100644 index 0000000..c71e0e1 --- /dev/null +++ b/scripts/hf_local_test.py @@ -0,0 +1,62 @@ +import base64 +import urllib.request +from io import BytesIO + +import torch +from PIL import Image +from transformers import AutoProcessor, Qwen2_5_VLForConditionalGeneration + +from olmocr.data.renderpdf import render_pdf_to_base64png +from olmocr.prompts import build_no_anchoring_v4_yaml_prompt + +if __name__ == "__main__": + # Initialize the model + model = Qwen2_5_VLForConditionalGeneration.from_pretrained("allenai/olmOCR-7B-1025", torch_dtype=torch.bfloat16).eval() + processor = AutoProcessor.from_pretrained("Qwen/Qwen2.5-VL-7B-Instruct") + device = torch.device("cuda" if torch.cuda.is_available() else "cpu") + model.to(device) + + # Grab a sample PDF + urllib.request.urlretrieve("https://olmocr.allenai.org/papers/olmocr.pdf", "./paper.pdf") + + # Render page 1 to an image + image_base64 = render_pdf_to_base64png("./paper.pdf", 1, target_longest_image_dim=1288) + + # Build the full prompt + messages = [ + { + "role": "user", + "content": [ + {"type": "text", "text": build_no_anchoring_v4_yaml_prompt()}, + {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{image_base64}"}}, + ], + } + ] + + # Apply the chat template and processor + text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True) + main_image = Image.open(BytesIO(base64.b64decode(image_base64))) + + inputs = processor( + text=[text], + images=[main_image], + padding=True, + return_tensors="pt", + ) + inputs = {key: value.to(device) for (key, value) in inputs.items()} + + # Generate the output + output = model.generate( + **inputs, + temperature=0.1, + max_new_tokens=50, + num_return_sequences=1, + do_sample=True, + ) + + # Decode the output + prompt_length = inputs["input_ids"].shape[1] + new_tokens = output[:, prompt_length:] + text_output = processor.tokenizer.batch_decode(new_tokens, skip_special_tokens=True) + + print(text_output) diff --git a/scripts/infinigram_count.py b/scripts/infinigram_count.py new file mode 100644 index 0000000..b0eb5f2 --- /dev/null +++ b/scripts/infinigram_count.py @@ -0,0 +1,173 @@ +#!/usr/bin/env python3 +import argparse +import json +import random +import re +import time + +import boto3 +import requests +from tqdm import tqdm +from transformers import AutoTokenizer + +# Allowed characters: alphanumeric, space, and basic punctuation ".,!?()" +ALLOWED_RE = re.compile(r"^[A-Za-z0-9\.,!?() ]+$") + + +def get_random_line_from_s3(bucket, key): + """ + Reads an S3 object line-by-line and returns a random line using reservoir sampling. + """ + s3 = boto3.client("s3") + response = s3.get_object(Bucket=bucket, Key=key) + random_line = None + count = 0 + for line in response["Body"].iter_lines(): + if not line: + continue + line_str = line.decode("utf-8") + count += 1 + if random.randint(1, count) == 1: + random_line = line_str + return random_line + + +def query_infinigram(ngram, index="v4_rpj_llama_s4", retries=3): + """ + Sends a count query to the infini-gram API for the given n-gram. + Retries a few times in case of network issues. + """ + url = "https://api.infini-gram.io/" + payload = { + "index": index, + "query_type": "count", + "query": ngram, + } + for i in range(retries): + try: + response = requests.post(url, json=payload, timeout=10) + if response.status_code == 200: + result = response.json() + if "count" in result: + return result["count"] + except Exception: # type: ignore + time.sleep(1) + return 0 + + +def process_document(doc, tokenizer, ngram_size, num_samples, index="v4_rpj_llama_s4"): + """ + Tokenizes the document using the Llama2 tokenizer and samples random n-grams. + Each n-gram is chosen such that: + 1. It starts on a word-split boundary (using the offset mapping and a check on the preceding character). + 2. Its decoded string contains only alphanumeric characters, spaces, and the punctuation marks ".,!?()". + + Each valid n-gram is then queried using the infini-gram API. + The function returns the document id, the number of matching n-grams (i.e. API count > 0), + the total number of valid n-grams sampled, and a list of tuples (flag, ngram_string). + """ + text = doc.get("text", "") + doc_id = doc.get("id", "Unknown") + # Get tokenized representation with offset mapping to determine word boundaries. + tokenized = tokenizer(text, add_special_tokens=False, return_offsets_mapping=True) + token_ids = tokenized["input_ids"] + # offsets = tokenized["offset_mapping"] + + if len(token_ids) < ngram_size: + return doc_id, 0, 0, [] + + # Determine valid starting indices based on word-split boundaries. + valid_positions = [] + # for i in range(len(token_ids) - ngram_size + 1): + # start_offset = offsets[i][0] + # if start_offset == 0 or (start_offset > 0 and text[start_offset - 1] == " "): + # valid_positions.append(i) + + if not valid_positions: + # Fallback: if no valid positions are found, use all possible positions. + valid_positions = list(range(len(token_ids) - ngram_size + 1)) + + valid_ngram_details = [] + attempts = 0 + max_attempts = num_samples * 10 # Limit to prevent infinite loops. + while len(valid_ngram_details) < num_samples and attempts < max_attempts: + idx = random.choice(valid_positions) + ngram_token_ids = token_ids[idx : idx + ngram_size] + ngram_str = tokenizer.decode(ngram_token_ids, clean_up_tokenization_spaces=True) + # Only accept n-grams that contain only allowed characters. + if ALLOWED_RE.fullmatch(ngram_str) and len(ngram_str.strip()) > ngram_size * 3: + count = query_infinigram(ngram_str, index=index) + flag = "YES" if count > 0 else "NO" + valid_ngram_details.append((flag, ngram_str)) + attempts += 1 + + match_count = sum(1 for flag, _ in valid_ngram_details if flag == "YES") + sample_count = len(valid_ngram_details) + return doc_id, match_count, sample_count, valid_ngram_details + + +def main(): + parser = argparse.ArgumentParser(description="Infini-gram n-gram matching script with Llama2 tokenization.") + parser.add_argument("N", type=int, help="Number of random .jsonl files to process") + parser.add_argument("s3_path", type=str, help="S3 path to a prefix containing .jsonl files (e.g., s3://my-bucket/my-prefix/)") + parser.add_argument("--index", type=str, default="v4_dolma-v1_7_llama", help="Infini-gram index to use (default: v4_rpj_llama_s4)") + parser.add_argument("--ngram_size", type=int, default=10, help="Size of the n-gram to sample (default: 10)") + parser.add_argument("--num_ngrams", type=int, default=100, help="Number of random n-grams to sample from each document (default: 100)") + args = parser.parse_args() + + if not args.s3_path.startswith("s3://"): + print("Error: s3_path must start with 's3://'") + return + path_without_scheme = args.s3_path[5:] + parts = path_without_scheme.split("/", 1) + bucket = parts[0] + prefix = parts[1] if len(parts) > 1 else "" + + print("Listing .jsonl files from S3...") + s3 = boto3.client("s3") + response = s3.list_objects_v2(Bucket=bucket, Prefix=prefix) + files = [obj["Key"] for obj in response.get("Contents", []) if obj["Key"].endswith(".jsonl")] + if not files: + print("No .jsonl files found in the given prefix.") + return + + if args.N > len(files): + print(f"Requested {args.N} files, but only found {len(files)}. Processing all available files.") + args.N = len(files) + random_files = random.sample(files, args.N) + + print("Loading Llama2 tokenizer...") + tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-7b-chat-hf") + + total_matches = 0 + total_ngrams_sampled = 0 + + for key in tqdm(random_files, desc="Processing files"): + line = get_random_line_from_s3(bucket, key) + if not line: + print(f"Skipping {key}: No valid lines found.") + continue + try: + doc = json.loads(line) + except Exception as e: + print(f"Error parsing JSON in {key}: {e}") + continue + doc_id, match_count, sample_count, details = process_document(doc, tokenizer, args.ngram_size, args.num_ngrams, index=args.index) + + # Print per-document n-gram summary + print(f"\nDocument ID: {doc_id}") + for flag, ngram in details: + # Print the flag in a fixed-width field (4 characters) followed by the n-gram representation. + print(f"{flag:4} {repr(ngram)}") + percentage = (match_count / sample_count * 100) if sample_count else 0 + print(f"Matched n-grams: {match_count}/{sample_count} ({percentage:.2f}%)") + + total_matches += match_count + total_ngrams_sampled += sample_count + + overall_percentage = (total_matches / total_ngrams_sampled * 100) if total_ngrams_sampled else 0 + print(f"\nTotal matched n-grams: {total_matches}/{total_ngrams_sampled} ({overall_percentage:.2f}%)") + + +if __name__ == "__main__": + main() diff --git a/scripts/jsonl_to_markdown.py b/scripts/jsonl_to_markdown.py new file mode 100644 index 0000000..ca488f1 --- /dev/null +++ b/scripts/jsonl_to_markdown.py @@ -0,0 +1,56 @@ +import json +import os +import sys + + +# This is a simple script to convert JSONL files to Markdown format. +# It reads each line of the JSONL file, extracts the 'text' field, +# and saves it as a Markdown file with the line number as the filename. +# The script also handles potential JSON decoding errors and prints relevant messages. +def jsonl_to_markdown(input_file, output_dir): + """ + Reads a JSONL file, extracts the 'text' field from each line, and saves it as a Markdown file. + + Args: + input_file (str): Path to the input JSONL file. + output_dir (str): Directory to save the Markdown files. + """ + if not os.path.exists(output_dir): + os.makedirs(output_dir) + + with open(input_file, "r", encoding="utf-8") as file: + for i, line in enumerate(file): + try: + # Parse the JSON line + data = json.loads(line) + text_content = data.get("text", "") + + # Save to a Markdown file + output_file = os.path.join(output_dir, f"line_{i + 1}.md") + with open(output_file, "w", encoding="utf-8") as md_file: + md_file.write(text_content) + + print(f"Extracted and saved line {i + 1} to {output_file}") + except json.JSONDecodeError as e: + print(f"Error decoding JSON on line {i + 1}: {e}") + except Exception as e: + print(f"Unexpected error on line {i + 1}: {e}") + + +# Example usage +# input_jsonl_file = "/path/to/test.jsonl" # Replace with the actual path to your JSONL file +# output_directory = "/path/to/output_markdown" # Replace with the desired output directory +# jsonl_to_markdown(input_jsonl_file, output_directory) + +# This is the main entrypoint to use the script from the command line. +# It takes two arguments: the input JSONL file and the output directory. +# The script will create the output directory if it does not exist. +if __name__ == "__main__": + if len(sys.argv) != 3: + print("Usage: python jsonl_to_markdown.py ") + sys.exit(1) + + input_file = sys.argv[1] + output_dir = sys.argv[2] + + jsonl_to_markdown(input_file, output_dir) diff --git a/scripts/movedolmadocs_to_md.py b/scripts/movedolmadocs_to_md.py new file mode 100644 index 0000000..352e5b5 --- /dev/null +++ b/scripts/movedolmadocs_to_md.py @@ -0,0 +1,103 @@ +#!/usr/bin/env python3 +import argparse +import json +import os +from urllib.parse import urlparse + +import boto3 + + +def parse_args(): + parser = argparse.ArgumentParser(description="Read JSONL files from an S3 prefix, extract text, and write to local .md files.") + parser.add_argument( + "--s3-prefix", + default="s3://ai2-oe-data/jakep/pdfworkspaces/pdelfin_testset/results/", + help="S3 prefix containing the JSONL files (default: s3://ai2-oe-data/jakep/pdfworkspaces/pdelfin_testset/results/)", + ) + parser.add_argument("--output-dir", default="output_md", help="Local directory to store output .md files (default: output_md)") + return parser.parse_args() + + +def main(): + args = parse_args() + + # Parse the s3-prefix into bucket and prefix + parsed_s3 = urlparse(args.s3_prefix) + # e.g. netloc = 'ai2-oe-data', path = '/jakep/pdfworkspaces/pdelfin_testset/results/' + bucket_name = parsed_s3.netloc + # Remove leading '/' from parsed_s3.path + prefix = parsed_s3.path.lstrip("/") + + # Ensure local output directory exists + os.makedirs(args.output_dir, exist_ok=True) + + # Initialize S3 client + s3 = boto3.client("s3") + + # List all objects under the prefix + paginator = s3.get_paginator("list_objects_v2") + pages = paginator.paginate(Bucket=bucket_name, Prefix=prefix) + + for page in pages: + if "Contents" not in page: + continue + + for obj in page["Contents"]: + key = obj["Key"] + # Skip non-jsonl files + if not key.endswith(".jsonl"): + continue + + print(f"Processing S3 object: s3://{bucket_name}/{key}") + + # Read the S3 object + s3_object = s3.get_object(Bucket=bucket_name, Key=key) + # s3_object['Body'] is a StreamingBody, so we can read it line-by-line + body_stream = s3_object["Body"].iter_lines() + + for line in body_stream: + if not line.strip(): + continue + + try: + record = json.loads(line) + except json.JSONDecodeError: + print("Warning: Failed to decode JSON line.") + continue + + # Extract text + text_content = record.get("text", "") + if not text_content.strip(): + # If there's no text, skip + continue + + # Derive the output filename based on the "Source-File" metadata + metadata = record.get("metadata", {}) + source_file = metadata.get("Source-File", "") + + # Example: source_file = 's3://ai2-oe-data/jakep/pdfdata/pdelfin_testset/fcffd2dd327d4e58d3c6d1d22ba62531c863_page8.pdf' + # We want to end up with: 'fcffd2dd327d4e58d3c6d1d22ba62531c863_page8_pdelf.md' + + # 1) Extract just the filename from the path + # 2) Remove '.pdf' + # 3) Append '_pdelf.md' + source_filename = os.path.basename(source_file) # e.g. 'fcffd2dd327d4e58d3c6d1d22ba62531c863_page8.pdf' + if source_filename.lower().endswith(".pdf"): + source_filename = source_filename[:-4] # remove .pdf + + output_filename = f"{source_filename}_pdelf.md" + output_path = os.path.join(args.output_dir, output_filename) + + # Append the text to the corresponding file + # If you want to overwrite instead, change mode to 'w' + with open(output_path, "a", encoding="utf-8") as f: + f.write(text_content + "\n") + + # Optional: Print or log what you've written + # print(f"Appended text to {output_path}") + + print("Done processing all JSONL files.") + + +if __name__ == "__main__": + main() diff --git a/scripts/parse_with_pdfminer.py b/scripts/parse_with_pdfminer.py new file mode 100644 index 0000000..31c600f --- /dev/null +++ b/scripts/parse_with_pdfminer.py @@ -0,0 +1,61 @@ +from pdfminer.high_level import extract_pages +from pdfminer.layout import LTChar + + +def extract_chars_with_transforms(pdf_path, page_num=0): + """ + Extract characters with transformation data for a specific page in a PDF. + + Args: + pdf_path (str): Path to the PDF file + page_num (int): Page number to extract (0-indexed) + """ + print(f"Analyzing PDF: {pdf_path}, Page: {page_num + 1}") + char_count = 0 + + # Extract only the specified page + for i, page_layout in enumerate(extract_pages(pdf_path)): + if i == page_num: + print(f"Processing page {page_num + 1}") + + # Recursively process all elements + def process_element(element, level=0): + nonlocal char_count + indent = " " * level + + if isinstance(element, LTChar): + char = element.get_text() + matrix = element.matrix + font = element.fontname if hasattr(element, "fontname") else "Unknown" + size = element.size if hasattr(element, "size") else "Unknown" + + print(f"{indent}Character: '{char}'") + print(f"{indent}Transform Matrix: {matrix}") + print(f"{indent}Font: {font}, Size: {size}") + print(f"{indent}{'-' * 30}") + char_count += 1 + + # For container elements, process their children + if hasattr(element, "_objs"): + for obj in element._objs: + process_element(obj, level + 1) + + # Process all elements in the page + for element in page_layout: + process_element(element) + + break # Stop after processing the requested page + + print(f"\nTotal characters extracted: {char_count}") + + if char_count == 0: + print("No characters were extracted. This could mean:") + print(f"1. Page {page_num + 1} doesn't exist or is empty") + print("2. The PDF contains scanned images rather than text") + print("3. The text is embedded in a way PDFMiner can't extract") + + +# Usage + +pdf_path = "/Users/kylel/Downloads/olmOCR_Technical_Report_COLM_2025.pdf" +extract_chars_with_transforms(pdf_path) diff --git a/scripts/pii/autoscan_dolmadocs.py b/scripts/pii/autoscan_dolmadocs.py new file mode 100644 index 0000000..e80fbca --- /dev/null +++ b/scripts/pii/autoscan_dolmadocs.py @@ -0,0 +1,592 @@ +import argparse +import json +import os +import random +import tempfile +from concurrent.futures import ThreadPoolExecutor +from enum import Enum +from pathlib import Path +from typing import Any, Dict, List, Optional + +import boto3 +import pydantic +from openai import OpenAI +from tqdm import tqdm + +from olmocr.data.renderpdf import render_pdf_to_base64png +from olmocr.s3_utils import get_s3_bytes, parse_s3_path + +LanguageCode = Enum( + "LanguageCode", + { + "en": "English", + "zh": "Chinese", + "hi": "Hindi", + "es": "Spanish", + "fr": "French", + "ar": "Arabic", + "bn": "Bengali", + "ru": "Russian", + "pt": "Portuguese", + "ur": "Urdu", + "id": "Indonesian", + "de": "German", + "ja": "Japanese", + "sw": "Swahili", + "mr": "Marathi", + "te": "Telugu", + "tr": "Turkish", + "vi": "Vietnamese", + "ta": "Tamil", + "ko": "Korean", + "other": "Other", + }, +) + + +class PIIAnnotation(pydantic.BaseModel): + """Structured model for PII annotations returned by ChatGPT""" + + document_description: str + language_code: LanguageCode + cannot_read: bool + inappropriate_content: bool + is_public_document: bool + + # PII identifiers + contains_names: bool + contains_email_addresses: bool + contains_phone_numbers: bool + + # PII that must co-occur with identifiers + contains_addresses: bool + contains_biographical_info: bool # DOB, gender, etc. + contains_location_info: bool + contains_employment_info: bool + contains_education_info: bool + contains_medical_info: bool + + # Always sensitive PII + contains_government_ids: bool + contains_financial_info: bool + contains_biometric_data: bool + contains_login_info: bool + + other_pii: str + + @property + def has_pii(self) -> bool: + """Check if the document contains any PII""" + pii_fields = [ + self.contains_names, + self.contains_email_addresses, + self.contains_phone_numbers, + self.contains_addresses, + self.contains_biographical_info, + self.contains_location_info, + self.contains_employment_info, + self.contains_education_info, + self.contains_medical_info, + self.contains_government_ids, + self.contains_financial_info, + self.contains_biometric_data, + self.contains_login_info, + ] + return any(pii_fields) or bool(self.other_pii.strip()) + + def get_pii_types(self) -> List[str]: + """Get a list of all PII types found in the document""" + pii_types = [] + + if self.contains_names: + pii_types.append("names") + if self.contains_email_addresses: + pii_types.append("email") + if self.contains_phone_numbers: + pii_types.append("phone") + if self.contains_addresses: + pii_types.append("addresses") + if self.contains_biographical_info: + pii_types.append("biographical") + if self.contains_location_info: + pii_types.append("location") + if self.contains_employment_info: + pii_types.append("employment") + if self.contains_education_info: + pii_types.append("education") + if self.contains_medical_info: + pii_types.append("medical") + if self.contains_government_ids: + pii_types.append("government-id") + if self.contains_financial_info: + pii_types.append("financial") + if self.contains_biometric_data: + pii_types.append("biometric") + if self.contains_login_info: + pii_types.append("login-info") + if self.other_pii.strip(): + pii_types.append("other") + + return pii_types + + +def parse_args(): + parser = argparse.ArgumentParser(description="Automatically scan OLMO OCR workspace results using ChatGPT") + parser.add_argument("workspace", help="OLMO OCR workspace path (s3://bucket/workspace)") + parser.add_argument("--pages_per_run", type=int, default=30, help="Number of pages per run") + parser.add_argument("--pdf_profile", help="AWS profile for accessing PDFs") + parser.add_argument("--output_dir", default="dolma_samples", help="Directory to save output files") + parser.add_argument("--max_workers", type=int, default=4, help="Maximum number of worker threads") + parser.add_argument("--openai_api_key", help="OpenAI API key (or set OPENAI_API_KEY env var)") + parser.add_argument("--openai_model", default="gpt-4.1", help="OpenAI model to use") + return parser.parse_args() + + +def list_result_files(s3_client, workspace_path): + """List all JSON result files in the workspace results directory.""" + bucket, prefix = parse_s3_path(workspace_path) + results_prefix = os.path.join(prefix, "results").rstrip("/") + "/" + + all_files = [] + paginator = s3_client.get_paginator("list_objects_v2") + for page in paginator.paginate(Bucket=bucket, Prefix=results_prefix): + if "Contents" in page: + all_files.extend([f"s3://{bucket}/{obj['Key']}" for obj in page["Contents"] if obj["Key"].endswith(".jsonl") or obj["Key"].endswith(".json")]) + + # if len(all_files) > 1000: + # break + + return all_files + + +def get_random_pages(s3_client, result_files, count=30): + """Get random pages from the result files.""" + random_pages = [] + + # Try to collect the requested number of pages + attempts = 0 + max_attempts = count * 3 # Allow extra attempts to handle potential failures + + while len(random_pages) < count and attempts < max_attempts: + attempts += 1 + + # Pick a random result file + if not result_files: + print("No result files found!") + break + + result_file = random.choice(result_files) + + try: + # Get the content of the file + content = get_s3_bytes(s3_client, result_file) + lines = content.decode("utf-8").strip().split("\n") + + if not lines: + continue + + # Pick a random line (which contains a complete document) + line = random.choice(lines) + doc = json.loads(line) + + # A Dolma document has "text", "metadata", and "attributes" fields + if "text" not in doc or "metadata" not in doc or "attributes" not in doc: + print(f"Document in {result_file} is not a valid Dolma document") + continue + + # Get the original PDF path from metadata + pdf_path = doc["metadata"].get("Source-File") + if not pdf_path: + continue + + # Get page spans from attributes + page_spans = doc["attributes"].get("pdf_page_numbers", []) + if not page_spans: + continue + + # Pick a random page span + page_span = random.choice(page_spans) + if len(page_span) >= 3: + # Page spans are [start_pos, end_pos, page_num] + page_num = page_span[2] + + # Extract text for this page + start_pos, end_pos = page_span[0], page_span[1] + page_text = doc["text"][start_pos:end_pos].strip() + + # Include the text snippet with the page info + random_pages.append((pdf_path, page_num, page_text, result_file)) + + if len(random_pages) >= count: + break + + except Exception as e: + print(f"Error processing {result_file}: {e}") + continue + + print(f"Found {len(random_pages)} random pages from Dolma documents") + return random_pages + + +def chatgpt_analyze_page(pdf_path: str, page_num: int, pdf_s3_client, openai_api_key: str, openai_model: str) -> Optional[PIIAnnotation]: + """Analyze a page using the ChatGPT vision model with structured outputs.""" + try: + # Download PDF to temp file and render to image + bucket, key = parse_s3_path(pdf_path) + with tempfile.NamedTemporaryFile(suffix=".pdf", delete=False) as temp_file: + pdf_data = pdf_s3_client.get_object(Bucket=bucket, Key=key)["Body"].read() + temp_file.write(pdf_data) + temp_file_path = temp_file.name + + # Render PDF to base64 image + base64_image = render_pdf_to_base64png(temp_file_path, page_num, target_longest_image_dim=2048) + + # Clean up temp file + os.unlink(temp_file_path) + + # Create OpenAI client + client = OpenAI(api_key=openai_api_key) + + # Prepare the user message with all instructions + user_message = """ +You are a document analyzer that identifies Personally Identifiable Information (PII) in documents. +Your task is to analyze the provided document image and determine: +1. Whether the document is intended for public release or dissemination (e.g., research paper, public report, etc.) +2. If the document contains any PII + +For PII identification, follow these specific guidelines: + +IDENTIFIERS FOR PII: +The following are considered identifiers that can make information PII: +- Names (full names, first names, last names, nicknames) +- Email addresses +- Phone numbers + +PII THAT MUST CO-OCCUR WITH AN IDENTIFIER: +The following types of information should ONLY be marked as PII if they occur ALONGSIDE an identifier (commonly, a person's name): +- Addresses (street address, postal code, etc.) +- Biographical Information (date of birth, place of birth, gender, sexual orientation, race, ethnicity, citizenship/immigration status, religion) +- Location Information (geolocations, specific coordinates) +- Employment Information (job titles, workplace names, employment history) +- Education Information (school names, degrees, transcripts) +- Medical Information (health records, diagnoses, genetic or neural data) + +PII THAT OCCURS EVEN WITHOUT AN IDENTIFIER: +The following should ALWAYS be marked as PII even if they do not occur alongside an identifier: +- Government IDs (Social Security Numbers, passport numbers, driver's license numbers, tax IDs) +- Financial Information (credit card numbers, bank account/routing numbers) +- Biometric Data (fingerprints, retina scans, facial recognition data, voice signatures) +- Login information (ONLY mark as PII when a username, password, and login location are present together) + +If the document is a form, then only consider fields which are filled out with specific values as potential PII. +If this page does not itself contain PII, but references documents (such as curriculum vitae, personal statements) that typically contain PII, then do not mark it as PII. +Only consider actual occurrences of the PII within the document shown. +""" + + # Use the chat completions API with the custom schema + completion = client.beta.chat.completions.parse( + model=openai_model, + messages=[ + { + "role": "user", + "content": [{"type": "text", "text": user_message}, {"type": "image_url", "image_url": {"url": f"data:image/webp;base64,{base64_image}"}}], + } + ], + response_format=PIIAnnotation, + max_tokens=1000, + ) + + return completion.choices[0].message.parsed + + except Exception as e: + print(f"Error analyzing page {pdf_path} (page {page_num}): {e}") + return None + + +def create_presigned_url(s3_client, pdf_path, expiration=3600 * 24 * 7): + """Create a presigned URL for the given S3 path.""" + try: + bucket, key = parse_s3_path(pdf_path) + url = s3_client.generate_presigned_url("get_object", Params={"Bucket": bucket, "Key": key}, ExpiresIn=expiration) + return url + except Exception as e: + print(f"Error creating presigned URL for {pdf_path}: {e}") + return None + + +def process_pages(random_pages, pdf_s3_client, openai_api_key, openai_model, max_workers): + """Process multiple pages in parallel using ThreadPoolExecutor.""" + results = [] + + # First generate presigned URLs for all PDFs + print("Generating presigned URLs for PDFs...") + presigned_urls = {} + for pdf_path, page_num, _, _ in random_pages: + if pdf_path not in presigned_urls and pdf_path.startswith("s3://"): + url = create_presigned_url(pdf_s3_client, pdf_path) + if url: + presigned_urls[pdf_path] = url + + with ThreadPoolExecutor(max_workers=max_workers) as executor: + futures = {} + + # Submit all tasks + for pdf_path, page_num, page_text, result_file in tqdm(random_pages, desc="Submitting pages for analysis"): + future = executor.submit(chatgpt_analyze_page, pdf_path, page_num, pdf_s3_client, openai_api_key, openai_model) + futures[future] = (pdf_path, page_num, page_text, result_file) + + # Process results as they complete + for future in tqdm(futures, desc="Processing results"): + pdf_path, page_num, page_text, result_file = futures[future] + try: + annotation = future.result() + if annotation: + # Get presigned URL with page number + presigned_url = None + if pdf_path in presigned_urls: + presigned_url = f"{presigned_urls[pdf_path]}#page={page_num}" + + results.append((pdf_path, page_num, page_text, result_file, annotation, presigned_url)) + else: + print(f"Failed to get annotation for {pdf_path} (page {page_num})") + except Exception as e: + print(f"Error processing {pdf_path} (page {page_num}): {e}") + + return results + + +def categorize_results(all_results): + """Categorize results for reporting.""" + categories = { + "public_document": [], + "private_document": [], + "cannot_read": [], + "report_content": [], + "no_annotation": [], + } + + for pdf_path, page_num, page_text, result_file, annotation, presigned_url in all_results: + if annotation.cannot_read or annotation.language_code != LanguageCode.en: + categories["cannot_read"].append({"pdf_path": pdf_path, "pdf_page": page_num, "result_file": result_file, "presigned_url": presigned_url}) + elif annotation.inappropriate_content: + categories["report_content"].append({"pdf_path": pdf_path, "pdf_page": page_num, "result_file": result_file, "presigned_url": presigned_url}) + elif annotation.is_public_document: + categories["public_document"].append( + { + "pdf_path": pdf_path, + "pdf_page": page_num, + "result_file": result_file, + "pii_types": annotation.get_pii_types(), + "has_pii": annotation.has_pii, + "description": annotation.other_pii, + "presigned_url": presigned_url, + } + ) + else: + # Private document + categories["private_document"].append( + { + "pdf_path": pdf_path, + "pdf_page": page_num, + "result_file": result_file, + "pii_types": annotation.get_pii_types(), + "has_pii": annotation.has_pii, + "description": annotation.other_pii, + "presigned_url": presigned_url, + } + ) + + return categories + + +def print_annotation_report(annotation_results: Dict[str, List[Dict[str, Any]]]): + """Print a summary report of annotations.""" + total_pages = sum(len(items) for items in annotation_results.values()) + + print("\n" + "=" * 80) + print(f"ANNOTATION REPORT - Total Pages: {total_pages}") + print("=" * 80) + + # Count pages with PII in public documents + public_with_pii = [page for page in annotation_results["public_document"] if page.get("has_pii", False)] + public_without_pii = [page for page in annotation_results["public_document"] if not page.get("has_pii", False)] + + # Count pages with PII in private documents + private_with_pii = [page for page in annotation_results["private_document"] if page.get("has_pii", False)] + private_without_pii = [page for page in annotation_results["private_document"] if not page.get("has_pii", False)] + + # Print summary statistics + print("\nSummary:") + print( + f" Public documents (total): {len(annotation_results['public_document'])} ({len(annotation_results['public_document'])/total_pages*100:.1f}% of all pages)" + ) + print(f" - With PII: {len(public_with_pii)} ({len(public_with_pii)/max(1, len(annotation_results['public_document']))*100:.1f}% of public docs)") + print( + f" - Without PII: {len(public_without_pii)} ({len(public_without_pii)/max(1, len(annotation_results['public_document']))*100:.1f}% of public docs)" + ) + + print( + f" Private documents (total): {len(annotation_results['private_document'])} ({len(annotation_results['private_document'])/total_pages*100:.1f}% of all pages)" + ) + print(f" - With PII: {len(private_with_pii)} ({len(private_with_pii)/max(1, len(annotation_results['private_document']))*100:.1f}% of private docs)") + print( + f" - Without PII: {len(private_without_pii)} ({len(private_without_pii)/max(1, len(annotation_results['private_document']))*100:.1f}% of private docs)" + ) + + print(f" Unreadable pages: {len(annotation_results['cannot_read'])} ({len(annotation_results['cannot_read'])/total_pages*100:.1f}%)") + print(f" Pages with reported content: {len(annotation_results['report_content'])} ({len(annotation_results['report_content'])/total_pages*100:.1f}%)") + print(f" Pages without annotation: {len(annotation_results['no_annotation'])} ({len(annotation_results['no_annotation'])/total_pages*100:.1f}%)") + + # Analyze PII types in private documents + if private_with_pii: + # Categorize the PII types for clearer reporting + pii_categories = { + "Identifiers": ["names", "email", "phone"], + "PII requiring identifiers": ["addresses", "biographical", "location", "employment", "education", "medical"], + "Always sensitive PII": ["government-id", "financial", "biometric", "login-info"], + } + + # Dictionary to track all PII counts + pii_counts_private = {} + for page in private_with_pii: + for pii_type in page.get("pii_types", []): + pii_counts_private[pii_type] = pii_counts_private.get(pii_type, 0) + 1 + + # Print categorized PII counts + print("\nPII Types in Private Documents:") + + # Print each category + for category, pii_types in pii_categories.items(): + print(f"\n {category}:") + for pii_type in pii_types: + count = pii_counts_private.get(pii_type, 0) + if count > 0: + print(f" - {pii_type}: {count} ({count/len(private_with_pii)*100:.1f}%)") + + # Print any other PII types not in our categories (like "other") + other_pii = [pii_type for pii_type in pii_counts_private.keys() if not any(pii_type in types for types in pii_categories.values())] + if other_pii: + print("\n Other PII types:") + for pii_type in other_pii: + count = pii_counts_private.get(pii_type, 0) + print(f" - {pii_type}: {count} ({count/len(private_with_pii)*100:.1f}%)") + + # Print detailed report for private documents with PII + if private_with_pii: + print("\nDetailed Report - Private Documents with PII:") + print("-" * 80) + for i, item in enumerate(private_with_pii, 1): + pdf_path = item["pdf_path"] + pdf_page = item["pdf_page"] + presigned_url = item.get("presigned_url") + + print(f"{i}. PDF: {pdf_path}") + print(f" Page: {pdf_page}") + if presigned_url: + print(f" Presigned URL: {presigned_url}") + print(f" PII Types: {', '.join(item['pii_types'])}") + if item.get("description"): + print(f" Description: {item['description']}") + print("-" * 80) + + # Print links to unreadable pages + # if annotation_results["cannot_read"]: + # print("\nUnreadable Pages:") + # print("-" * 80) + # for i, item in enumerate(annotation_results["cannot_read"], 1): + # pdf_path = item["pdf_path"] + # pdf_page = item["pdf_page"] + # presigned_url = item.get("presigned_url") + + # print(f"{i}. PDF: {pdf_path}") + # print(f" Page: {pdf_page}") + # if presigned_url: + # print(f" Presigned URL: {presigned_url}") + # print("-" * 80) + + # Print links to inappropriate content + if annotation_results["report_content"]: + print("\nReported Content:") + print("-" * 80) + for i, item in enumerate(annotation_results["report_content"], 1): + pdf_path = item["pdf_path"] + pdf_page = item["pdf_page"] + presigned_url = item.get("presigned_url") + + print(f"{i}. PDF: {pdf_path}") + print(f" Page: {pdf_page}") + if presigned_url: + print(f" Presigned URL: {presigned_url}") + print("-" * 80) + + print("\nReport complete.") + + +def save_results(results, output_dir): + """Save the results to a JSON file.""" + output_path = Path(output_dir) / "autoscan_results.json" + + # Convert results to serializable format + serializable_results = [] + for pdf_path, page_num, page_text, result_file, annotation, presigned_url in results: + serializable_results.append( + { + "pdf_path": pdf_path, + "page_num": page_num, + "page_text": page_text, + "result_file": result_file, + "annotation": annotation.dict(), + "presigned_url": presigned_url, + } + ) + + with open(output_path, "w") as f: + json.dump(serializable_results, f, indent=2, default=lambda o: o.value if isinstance(o, Enum) else o) + + print(f"Results saved to {output_path}") + + +def main(): + args = parse_args() + + # Get OpenAI API key from args or environment + openai_api_key = args.openai_api_key or os.environ.get("OPENAI_API_KEY") + if not openai_api_key: + raise ValueError("OpenAI API key must be provided via --openai_api_key or OPENAI_API_KEY environment variable") + + # Set up S3 clients + s3_client = boto3.client("s3") + + # Set up PDF S3 client with profile if specified + if args.pdf_profile: + pdf_session = boto3.Session(profile_name=args.pdf_profile) + pdf_s3_client = pdf_session.client("s3") + else: + pdf_s3_client = s3_client + + # Create output directory + output_dir = Path(args.output_dir) + output_dir.mkdir(exist_ok=True, parents=True) + + # List all result files + print(f"Listing result files in {args.workspace}/results...") + result_files = list_result_files(s3_client, args.workspace) + print(f"Found {len(result_files)} result files") + + # Get random pages + random_pages = get_random_pages(s3_client, result_files, args.pages_per_run) + + # Process pages with ChatGPT + print(f"Processing {len(random_pages)} pages with ChatGPT...") + all_results = process_pages(random_pages, pdf_s3_client, openai_api_key, args.openai_model, args.max_workers) + + # Save results + save_results(all_results, args.output_dir) + + # Categorize and report results + categorized_results = categorize_results(all_results) + print_annotation_report(categorized_results) + + +if __name__ == "__main__": + main() diff --git a/scripts/pii/chatgpt_tag_dolmadocs_v1.py b/scripts/pii/chatgpt_tag_dolmadocs_v1.py new file mode 100644 index 0000000..d42a69f --- /dev/null +++ b/scripts/pii/chatgpt_tag_dolmadocs_v1.py @@ -0,0 +1,655 @@ +#!/usr/bin/env python3 +""" +Rich Autoscan Dolma Documents with ChatGPT Vision + +This script combines the functionality of autoscan_dolmadocs.py and rich_tagging_pipeline.py: +1. Uses ChatGPT Vision API to analyze PDF pages for PII +2. Creates attribute folders mirroring the document structure +3. Simplifies with direct ThreadPoolExecutor usage instead of work queue system +""" + +import argparse +import gzip +import json +import logging +import os +import tempfile +from concurrent.futures import ThreadPoolExecutor +from enum import Enum +from pathlib import Path +from typing import List, Optional + +import boto3 +import pydantic +import zstandard as zstd +from openai import OpenAI +from tqdm import tqdm + +from olmocr.data.renderpdf import render_pdf_to_base64png +from olmocr.s3_utils import get_s3_bytes, parse_s3_path + +# Initialize logger +logger = logging.getLogger(__name__) +logger.setLevel(logging.DEBUG) +logger.propagate = False + +file_handler = logging.FileHandler("rich-autoscan-debug.log", mode="a") +file_handler.setLevel(logging.DEBUG) +file_handler.setFormatter(logging.Formatter("%(asctime)s - %(name)s - %(levelname)s - %(message)s")) + +console_handler = logging.StreamHandler() +console_handler.setLevel(logging.INFO) +console_handler.setFormatter(logging.Formatter("%(asctime)s - %(name)s - %(levelname)s - %(message)s")) + +# Add handlers to the logger +logger.addHandler(file_handler) +logger.addHandler(console_handler) + + +LanguageCode = Enum( + "LanguageCode", + { + "en": "English", + "zh": "Chinese", + "hi": "Hindi", + "es": "Spanish", + "fr": "French", + "ar": "Arabic", + "bn": "Bengali", + "ru": "Russian", + "pt": "Portuguese", + "ur": "Urdu", + "id": "Indonesian", + "de": "German", + "ja": "Japanese", + "sw": "Swahili", + "mr": "Marathi", + "te": "Telugu", + "tr": "Turkish", + "vi": "Vietnamese", + "ta": "Tamil", + "ko": "Korean", + "other": "Other", + }, +) + + +class PIIAnnotation(pydantic.BaseModel): + """Structured model for PII annotations returned by ChatGPT""" + + document_description: str + language_code: LanguageCode + cannot_read: bool + inappropriate_content: bool + is_public_document: bool + + # PII identifiers + contains_names: bool + contains_email_addresses: bool + contains_phone_numbers: bool + + # PII that must co-occur with identifiers + contains_addresses: bool + contains_biographical_info: bool # DOB, gender, etc. + contains_location_info: bool + contains_employment_info: bool + contains_education_info: bool + contains_medical_info: bool + + # Always sensitive PII + contains_government_ids: bool + contains_financial_info: bool + contains_biometric_data: bool + contains_login_info: bool + + other_pii: str + + @property + def has_pii(self) -> bool: + """Check if the document contains any PII""" + pii_fields = [ + self.contains_names, + self.contains_email_addresses, + self.contains_phone_numbers, + self.contains_addresses, + self.contains_biographical_info, + self.contains_location_info, + self.contains_employment_info, + self.contains_education_info, + self.contains_medical_info, + self.contains_government_ids, + self.contains_financial_info, + self.contains_biometric_data, + self.contains_login_info, + ] + return any(pii_fields) or bool(self.other_pii.strip()) + + def get_pii_types(self) -> List[str]: + """Get a list of all PII types found in the document""" + pii_types = [] + + if self.contains_names: + pii_types.append("names") + if self.contains_email_addresses: + pii_types.append("email") + if self.contains_phone_numbers: + pii_types.append("phone") + if self.contains_addresses: + pii_types.append("addresses") + if self.contains_biographical_info: + pii_types.append("biographical") + if self.contains_location_info: + pii_types.append("location") + if self.contains_employment_info: + pii_types.append("employment") + if self.contains_education_info: + pii_types.append("education") + if self.contains_medical_info: + pii_types.append("medical") + if self.contains_government_ids: + pii_types.append("government-id") + if self.contains_financial_info: + pii_types.append("financial") + if self.contains_biometric_data: + pii_types.append("biometric") + if self.contains_login_info: + pii_types.append("login-info") + if self.other_pii.strip(): + pii_types.append("other") + + return pii_types + + +def parse_args(): + parser = argparse.ArgumentParser(description="Rich Autoscan for OLMO OCR workspace using ChatGPT Vision") + parser.add_argument("workspace", help="OLMO OCR workspace path (s3://bucket/workspace)") + parser.add_argument("--workspace_profile", help="AWS profile for accessing workspace (documents and attributes)") + parser.add_argument("--pdf_profile", help="AWS profile for accessing PDFs (can be different from workspace profile)") + parser.add_argument("--output_dir", default="dolma_samples", help="Directory to save output files") + parser.add_argument("--max_workers", type=int, default=16, help="Maximum number of worker threads") + parser.add_argument("--openai_api_key", help="OpenAI API key (or set OPENAI_API_KEY env var)") + parser.add_argument("--openai_model", default="gpt-4.1", help="OpenAI model to use") + parser.add_argument("--attribute_name", default="chatgpt_pii_vision", help="Path to use for attribute naming") + parser.add_argument("--batch_size", type=int, default=1000, help="Number of documents to process in each batch") + return parser.parse_args() + + +def list_result_files(s3_client, workspace_path): + """List all JSONL files in the workspace documents directory.""" + bucket, prefix = parse_s3_path(workspace_path) + documents_prefix = os.path.join(prefix, "documents").rstrip("/") + "/" + + all_files = [] + paginator = s3_client.get_paginator("list_objects_v2") + + logger.info(f"Listing files from s3://{bucket}/{documents_prefix}") + for page in paginator.paginate(Bucket=bucket, Prefix=documents_prefix): + if "Contents" in page: + all_files.extend( + [ + f"s3://{bucket}/{obj['Key']}" + for obj in page["Contents"] + if ( + obj["Key"].endswith(".jsonl") + or obj["Key"].endswith(".json") + or obj["Key"].endswith(".jsonl.gz") + or obj["Key"].endswith(".jsonl.zst") + or obj["Key"].endswith(".jsonl.ztd") + or obj["Key"].endswith(".jsonl.zstd") + ) + ] + ) + + if len(all_files) % 100 == 0: + logger.info(f"Found {len(all_files)} files so far...") + + logger.info(f"Total files found: {len(all_files)}") + return all_files + + +def load_document_file(s3_client, file_path): + """Load a single document file and return its contents.""" + try: + # Fetch raw bytes (S3 or local) + if file_path.startswith("s3://"): + raw = get_s3_bytes(s3_client, file_path) + else: + with open(file_path, "rb") as f: + raw = f.read() + + # Decompress if needed + if file_path.endswith(".gz"): + file_bytes = gzip.decompress(raw) + elif file_path.endswith(".zst") or file_path.endswith(".ztd") or file_path.endswith(".zstd"): + dctx = zstd.ZstdDecompressor() + file_bytes = dctx.decompress(raw, max_output_size=1_000_000_000) + else: + file_bytes = raw + + # Return the decoded lines + return file_bytes.decode("utf-8").strip().split("\n") + except Exception as e: + logger.error(f"Error loading file {file_path}: {e}") + return [] + + +def get_document_info_from_line(line, file_path, line_index): + """Extract document information from a single line.""" + try: + doc = json.loads(line) + + # A Dolma document has "text", "metadata", and "attributes" fields + if "text" not in doc or "metadata" not in doc or "attributes" not in doc: + logger.warning(f"Document in {file_path} line {line_index} is not a valid Dolma document") + return None + + # Get the original PDF path from metadata + pdf_path = doc["metadata"].get("Source-File") + if not pdf_path: + return None + + # Get page spans from attributes + page_spans = doc["attributes"].get("pdf_page_numbers", []) + if not page_spans: + return None + + # Just use the first page for each document + if page_spans: + page_span = page_spans[0] # Just get the first page + if len(page_span) >= 3: + # Page spans are [start_pos, end_pos, page_num] + page_num = page_span[2] + + # Extract text for this page + start_pos, end_pos = page_span[0], page_span[1] + page_text = doc["text"][start_pos:end_pos].strip() + + # Return the information + return { + "pdf_path": pdf_path, + "page_num": page_num, + "page_text": page_text, + "start_pos": start_pos, + "end_pos": end_pos, + "doc_id": doc["id"], + "source_file": file_path, + "line_index": line_index, + } + + return None + except json.JSONDecodeError: + logger.warning(f"Invalid JSON in {file_path} line {line_index}") + return None + except Exception as e: + logger.warning(f"Error processing document in {file_path} line {line_index}: {e}") + return None + + +def get_all_pages(s3_client, document_files): + """Get all pages from the document files for processing, preserving file and line order.""" + file_contents = {} + + # First, collect all file paths and their document info + for file_path in tqdm(document_files, desc="Loading document files"): + lines = load_document_file(s3_client, file_path) + if not lines: + logger.warning(f"Empty or invalid file: {file_path}") + continue + + # Parse each line for document info + documents = [] + for i, line in enumerate(lines): + doc_info = get_document_info_from_line(line, file_path, i) + # Always add an entry for each line, even if None, to preserve line alignment + documents.append(doc_info) + + # Store all documents for this file + file_contents[file_path] = documents + logger.info(f"Loaded {len(documents)} documents from {file_path}") + + logger.info(f"Loaded documents from {len(file_contents)} files") + return file_contents + + +def chatgpt_analyze_page(pdf_path: str, page_num: int, pdf_s3_client, openai_api_key: str, openai_model: str) -> Optional[PIIAnnotation]: + """Analyze a page using the ChatGPT vision model with structured outputs.""" + try: + # Download PDF to temp file and render to image + bucket, key = parse_s3_path(pdf_path) + with tempfile.NamedTemporaryFile(suffix=".pdf", delete=False) as temp_file: + pdf_data = pdf_s3_client.get_object(Bucket=bucket, Key=key)["Body"].read() + temp_file.write(pdf_data) + temp_file_path = temp_file.name + + # Render PDF to base64 image + base64_image = render_pdf_to_base64png(temp_file_path, page_num, target_longest_image_dim=2048) + + # Clean up temp file + os.unlink(temp_file_path) + + # Create OpenAI client + client = OpenAI(api_key=openai_api_key) + + # Prepare the user message with all instructions - Keep exact prompt from autoscan_dolmadocs.py + user_message = """ +You are a document analyzer that identifies Personally Identifiable Information (PII) in documents. +Your task is to analyze the provided document image and determine: +1. Whether the document is intended for public release or dissemination (e.g., research paper, public report, etc.) +2. If the document contains any PII + +For PII identification, follow these specific guidelines: + +IDENTIFIERS FOR PII: +The following are considered identifiers that can make information PII: +- Names (full names, first names, last names, nicknames) +- Email addresses +- Phone numbers + +PII THAT MUST CO-OCCUR WITH AN IDENTIFIER: +The following types of information should ONLY be marked as PII if they occur ALONGSIDE an identifier (commonly, a person's name): +- Addresses (street address, postal code, etc.) +- Biographical Information (date of birth, place of birth, gender, sexual orientation, race, ethnicity, citizenship/immigration status, religion) +- Location Information (geolocations, specific coordinates) +- Employment Information (job titles, workplace names, employment history) +- Education Information (school names, degrees, transcripts) +- Medical Information (health records, diagnoses, genetic or neural data) + +PII THAT OCCURS EVEN WITHOUT AN IDENTIFIER: +The following should ALWAYS be marked as PII even if they do not occur alongside an identifier: +- Government IDs (Social Security Numbers, passport numbers, driver's license numbers, tax IDs) +- Financial Information (credit card numbers, bank account/routing numbers) +- Biometric Data (fingerprints, retina scans, facial recognition data, voice signatures) +- Login information (ONLY mark as PII when a username, password, and login location are present together) + +If the document is a form, then only consider fields which are filled out with specific values as potential PII. +If this page does not itself contain PII, but references documents (such as curriculum vitae, personal statements) that typically contain PII, then do not mark it as PII. +Only consider actual occurrences of the PII within the document shown. +""" + + # Use the chat completions API with the custom schema + completion = client.beta.chat.completions.parse( + model=openai_model, + messages=[ + { + "role": "user", + "content": [{"type": "text", "text": user_message}, {"type": "image_url", "image_url": {"url": f"data:image/webp;base64,{base64_image}"}}], + } + ], + response_format=PIIAnnotation, + max_tokens=1000, + ) + + return completion.choices[0].message.parsed + + except Exception as e: + logger.error(f"Error analyzing page {pdf_path} (page {page_num}): {e}") + return None + + +def process_single_page(args, doc_info, pdf_s3_client=None): + """Process a single document and generate attribute data.""" + # Skip if document info is None + if doc_info is None: + return None + + # Extract info from the document info + pdf_path = doc_info["pdf_path"] + page_num = doc_info["page_num"] + start_pos = doc_info["start_pos"] + end_pos = doc_info["end_pos"] + doc_id = doc_info["doc_id"] + source_file = doc_info["source_file"] + line_index = doc_info["line_index"] + + # Use provided PDF S3 client if given, otherwise create one + if pdf_s3_client is None: + if args.pdf_profile: + pdf_session = boto3.Session(profile_name=args.pdf_profile) + pdf_s3_client = pdf_session.client("s3") + else: + pdf_s3_client = boto3.client("s3") + + # Get OpenAI API key + openai_api_key = args.openai_api_key or os.environ.get("OPENAI_API_KEY") + if not openai_api_key: + raise ValueError("OpenAI API key must be provided via --openai_api_key or OPENAI_API_KEY environment variable") + + # Analyze page with ChatGPT + annotation = chatgpt_analyze_page(pdf_path, page_num, pdf_s3_client, openai_api_key, args.openai_model) + + if not annotation: + logger.warning(f"No annotation for {pdf_path} page {page_num}") + return { + "id": doc_id, + "line_index": line_index, + "attributes": None, + "source_file": source_file, + } + + # Generate attribute key names using model name + model_prefix = args.openai_model.replace("/", "_").replace("-", "_").replace(".", "_") + language_key_name = f"{model_prefix}_language" + contains_pii_key_name = f"{model_prefix}_contains_pii" + + # Initialize result attributes with all PIIAnnotation fields + result_attributes = { + contains_pii_key_name: [[start_pos, end_pos, annotation.has_pii]], + language_key_name: [[start_pos, end_pos, annotation.language_code.name]], + f"{model_prefix}_is_public_document": [[start_pos, end_pos, annotation.is_public_document]], + f"{model_prefix}_contains_names": [[start_pos, end_pos, annotation.contains_names]], + f"{model_prefix}_contains_email_addresses": [[start_pos, end_pos, annotation.contains_email_addresses]], + f"{model_prefix}_contains_phone_numbers": [[start_pos, end_pos, annotation.contains_phone_numbers]], + f"{model_prefix}_contains_addresses": [[start_pos, end_pos, annotation.contains_addresses]], + f"{model_prefix}_contains_biographical_info": [[start_pos, end_pos, annotation.contains_biographical_info]], + f"{model_prefix}_contains_location_info": [[start_pos, end_pos, annotation.contains_location_info]], + f"{model_prefix}_contains_employment_info": [[start_pos, end_pos, annotation.contains_employment_info]], + f"{model_prefix}_contains_education_info": [[start_pos, end_pos, annotation.contains_education_info]], + f"{model_prefix}_contains_medical_info": [[start_pos, end_pos, annotation.contains_medical_info]], + f"{model_prefix}_contains_government_ids": [[start_pos, end_pos, annotation.contains_government_ids]], + f"{model_prefix}_contains_financial_info": [[start_pos, end_pos, annotation.contains_financial_info]], + f"{model_prefix}_contains_biometric_data": [[start_pos, end_pos, annotation.contains_biometric_data]], + f"{model_prefix}_contains_login_info": [[start_pos, end_pos, annotation.contains_login_info]], + f"{model_prefix}_other_pii": [[start_pos, end_pos, annotation.other_pii]], + } + + # Return document ID, line index, and attributes + return { + "id": doc_id, + "line_index": line_index, + "attributes": result_attributes, + "source_file": source_file, + } + + +def write_attribute_file(args, processed_docs, file_documents, workspace_s3): + """Write attribute results to the appropriate files, preserving exact line order.""" + # Group results by source file and organize by line index + results_by_file = {} + for result in processed_docs: + if result is None: + continue + + source_file = result["source_file"] + if source_file not in results_by_file: + results_by_file[source_file] = {} + + # Store by line index to preserve order + results_by_file[source_file][result["line_index"]] = {"id": result["id"], "attributes": result["attributes"]} + + # Process each source file + for source_file, file_results_dict in results_by_file.items(): + try: + # 1. Build the relative path that mirrors documents/… + if source_file.startswith("s3://"): + _, key = parse_s3_path(source_file) + _, docs_prefix = parse_s3_path(args.workspace) + rel_path = key[len(os.path.join(docs_prefix, "documents/")) :] + else: + docs_root = os.path.join(args.workspace, "documents") + rel_path = os.path.relpath(source_file, docs_root) + + # 2. Create ordered attribute entries in exact same order as source file + file_entries = [] + # Get the original documents to ensure we have ALL lines in order + original_docs = file_documents[source_file] + + # Create attribute entries for every line + for i, doc_info in enumerate(original_docs): + if i in file_results_dict and file_results_dict[i]["attributes"] is not None: + # We have a processed result for this line + file_entries.append(file_results_dict[i]) + elif doc_info is not None: + # We have document info but no processed attributes (processing failed) + # Create an empty attributes entry with the correct ID + file_entries.append({"id": doc_info["doc_id"], "attributes": {}}) + else: + # This line in the source file was invalid or not a document + # Create a placeholder with a generated ID + placeholder_id = f"placeholder_{source_file}_{i}" + file_entries.append({"id": placeholder_id, "attributes": {}}) + + # 3. Create output JSONL + out_rel = os.path.join("attributes", args.attribute_name, rel_path) + out_jsonl = "\n".join(json.dumps(entry) for entry in file_entries) + "\n" + + # 4. Preserve compression type + if rel_path.endswith(".gz"): + payload = gzip.compress(out_jsonl.encode("utf-8")) + elif rel_path.endswith((".zst", ".ztd", ".zstd")): + payload = zstd.ZstdCompressor().compress(out_jsonl.encode("utf-8")) + else: + payload = out_jsonl.encode("utf-8") + + # 5. Write to args.workspace (local or S3) + if args.workspace.startswith("s3://"): + bucket, prefix = parse_s3_path(args.workspace) + key = os.path.join(prefix, out_rel) + workspace_s3.put_object(Bucket=bucket, Key=key, Body=payload) + logger.info(f"Wrote {len(file_entries)} attribute entries to s3://{bucket}/{key}") + else: + out_path = os.path.join(args.workspace, out_rel) + os.makedirs(os.path.dirname(out_path), exist_ok=True) + with open(out_path, "wb") as fh: + fh.write(payload) + logger.info(f"Wrote {len(file_entries)} attribute entries to {out_path}") + + except Exception as e: + logger.error(f"Error writing attributes for {source_file}: {e}") + continue + + +def save_results(results, output_dir): + """Save the full results to a JSON file for analysis.""" + output_path = Path(output_dir) / "rich_autoscan_results.json" + + # Convert results to serializable format + serializable_results = [] + for result in results: + if result is None: + continue + serializable_results.append(result) + + with open(output_path, "w") as f: + json.dump(serializable_results, f, indent=2, default=lambda o: o.value if isinstance(o, Enum) else o) + + print(f"Results saved to {output_path}") + + +def main(): + args = parse_args() + + # Set up S3 clients with appropriate profiles + if args.workspace_profile: + workspace_session = boto3.Session(profile_name=args.workspace_profile) + workspace_s3 = workspace_session.client("s3") + logger.info(f"Using AWS profile '{args.workspace_profile}' for workspace access") + else: + workspace_s3 = boto3.client("s3") + logger.info("Using default AWS credentials for workspace access") + + if args.pdf_profile: + pdf_session = boto3.Session(profile_name=args.pdf_profile) + pdf_s3 = pdf_session.client("s3") + logger.info(f"Using AWS profile '{args.pdf_profile}' for PDF access") + else: + # If no PDF profile specified, use the workspace profile or default + if args.workspace_profile: + pdf_s3 = workspace_s3 + logger.info(f"Using workspace profile '{args.workspace_profile}' for PDF access") + else: + pdf_s3 = boto3.client("s3") + logger.info("Using default AWS credentials for PDF access") + + # Create output directory + output_dir = Path(args.output_dir) + output_dir.mkdir(exist_ok=True, parents=True) + + # List all document files + logger.info(f"Listing document files in {args.workspace}/documents...") + document_files = list_result_files(workspace_s3, args.workspace) + logger.info(f"Found {len(document_files)} document files") + + # Load all document files and their contents, organized by file + logger.info("Loading all document files...") + file_documents = get_all_pages(workspace_s3, document_files) + + # Process each file individually + for file_index, (file_path, documents) in enumerate(file_documents.items()): + logger.info(f"Processing file {file_index+1}/{len(file_documents)}: {file_path}") + + # Only process documents that have valid information + valid_docs = [] + for doc in documents: + if doc is not None: + valid_docs.append(doc) + + # Skip if no valid documents + if not valid_docs: + logger.warning(f"No valid documents in {file_path}") + continue + + # Process in batches to manage memory and API rate limits + total_docs = len(valid_docs) + logger.info(f"Found {total_docs} valid documents to process in {file_path}") + + # Process in batches (process by document, but maintain file coherence) + all_results = [] + for i in range(0, total_docs, args.batch_size): + batch = valid_docs[i : i + args.batch_size] + batch_num = i // args.batch_size + 1 + total_batches = (total_docs + args.batch_size - 1) // args.batch_size + + logger.info(f"Processing batch {batch_num}/{total_batches} of {file_path} ({len(batch)} documents)...") + results = [] + + with ThreadPoolExecutor(max_workers=args.max_workers) as executor: + futures = [] + # Process documents in parallel but within the same file + for doc_info in batch: + futures.append(executor.submit(process_single_page, args, doc_info, pdf_s3)) + + for future in tqdm(futures, desc=f"Processing batch {batch_num} of {file_path}"): + try: + result = future.result() + results.append(result) + except Exception as e: + logger.error(f"Error processing document: {e}") + + # Save results for this batch + batch_output_dir = os.path.join(args.output_dir, f"file_{file_index+1}_batch_{batch_num}") + os.makedirs(batch_output_dir, exist_ok=True) + save_results(results, batch_output_dir) + + # Collect all results for this file + all_results.extend(results) + logger.info(f"Completed batch {batch_num}/{total_batches} of {file_path}") + + # Write attributes for the entire file, maintaining line order + write_attribute_file(args, all_results, file_documents, workspace_s3) + logger.info(f"Completed processing file {file_index+1}/{len(file_documents)}: {file_path}") + + logger.info(f"Processing complete - processed {len(file_documents)} files") + + +if __name__ == "__main__": + main() diff --git a/scripts/pii/chatgpt_tag_dolmadocs_v2.py b/scripts/pii/chatgpt_tag_dolmadocs_v2.py new file mode 100644 index 0000000..805aea0 --- /dev/null +++ b/scripts/pii/chatgpt_tag_dolmadocs_v2.py @@ -0,0 +1,622 @@ +#!/usr/bin/env python3 +""" +Document Classification for Dolma Documents with ChatGPT Vision + +This script uses ChatGPT Vision API to classify PDF documents into various categories: +1. Uses ChatGPT Vision API to analyze PDF pages and determine their document type +2. Classifies documents into categories like academic papers, textbooks, news articles, etc. +3. Creates attribute folders mirroring the document structure with classification results +""" + +import argparse +import gzip +import json +import logging +import os +import tempfile +from concurrent.futures import ThreadPoolExecutor +from enum import Enum +from pathlib import Path +from typing import List, Optional + +import boto3 +import pydantic +import zstandard as zstd +from openai import OpenAI +from tqdm import tqdm + +from olmocr.data.renderpdf import render_pdf_to_base64png +from olmocr.s3_utils import get_s3_bytes, parse_s3_path + +# Initialize logger +logger = logging.getLogger(__name__) +logger.setLevel(logging.DEBUG) +logger.propagate = False + +file_handler = logging.FileHandler("rich-autoscan-debug.log", mode="a") +file_handler.setLevel(logging.DEBUG) +file_handler.setFormatter(logging.Formatter("%(asctime)s - %(name)s - %(levelname)s - %(message)s")) + +console_handler = logging.StreamHandler() +console_handler.setLevel(logging.INFO) +console_handler.setFormatter(logging.Formatter("%(asctime)s - %(name)s - %(levelname)s - %(message)s")) + +# Add handlers to the logger +logger.addHandler(file_handler) +logger.addHandler(console_handler) + + +LanguageCode = Enum( + "LanguageCode", + { + "en": "English", + "zh": "Chinese", + "hi": "Hindi", + "es": "Spanish", + "fr": "French", + "ar": "Arabic", + "bn": "Bengali", + "ru": "Russian", + "pt": "Portuguese", + "ur": "Urdu", + "id": "Indonesian", + "de": "German", + "ja": "Japanese", + "sw": "Swahili", + "mr": "Marathi", + "te": "Telugu", + "tr": "Turkish", + "vi": "Vietnamese", + "ta": "Tamil", + "ko": "Korean", + "other": "Other", + }, +) + + +class DocumentClassification(pydantic.BaseModel): + """Structured model for document classification returned by ChatGPT""" + + document_description: str + language_code: LanguageCode + cannot_read: bool + inappropriate_content: bool + + # Document type categories + is_academic_paper: bool + is_textbook: bool + is_news_article: bool + is_test_or_quiz: bool + is_homework_assignment: bool + is_class_syllabus: bool + is_meeting_minutes: bool + is_legal_contract: bool + is_form: bool + is_correspondence_or_letter: bool + is_public_order: bool + is_court_notice: bool + + # Additional context or notes + classification_notes: str + + @property + def get_document_types(self) -> List[str]: + """Get a list of all document types identified""" + doc_types = [] + + if self.is_academic_paper: + doc_types.append("academic_paper") + if self.is_textbook: + doc_types.append("textbook") + if self.is_news_article: + doc_types.append("news_article") + if self.is_test_or_quiz: + doc_types.append("test_or_quiz") + if self.is_homework_assignment: + doc_types.append("homework_assignment") + if self.is_class_syllabus: + doc_types.append("class_syllabus") + if self.is_meeting_minutes: + doc_types.append("meeting_minutes") + if self.is_legal_contract: + doc_types.append("legal_contract") + if self.is_form: + doc_types.append("form") + if self.is_correspondence_or_letter: + doc_types.append("correspondence_or_letter") + if self.is_public_order: + doc_types.append("public_order") + if self.is_court_notice: + doc_types.append("court_notice") + + return doc_types + + +def parse_args(): + parser = argparse.ArgumentParser(description="Document Classification for OLMO OCR workspace using ChatGPT Vision") + parser.add_argument("workspace", help="OLMO OCR workspace path (s3://bucket/workspace)") + parser.add_argument("--workspace_profile", help="AWS profile for accessing workspace (documents and attributes)") + parser.add_argument("--pdf_profile", help="AWS profile for accessing PDFs (can be different from workspace profile)") + parser.add_argument("--output_dir", default="dolma_samples", help="Directory to save output files") + parser.add_argument("--max_workers", type=int, default=16, help="Maximum number of worker threads") + parser.add_argument("--openai_api_key", help="OpenAI API key (or set OPENAI_API_KEY env var)") + parser.add_argument("--openai_model", default="gpt-4.1", help="OpenAI model to use") + parser.add_argument("--attribute_name", default="chatgpt_doc_classification", help="Path to use for attribute naming") + parser.add_argument("--batch_size", type=int, default=1000, help="Number of documents to process in each batch") + return parser.parse_args() + + +def list_result_files(s3_client, workspace_path): + """List all JSONL files in the workspace documents directory.""" + bucket, prefix = parse_s3_path(workspace_path) + documents_prefix = os.path.join(prefix, "documents").rstrip("/") + "/" + + all_files = [] + paginator = s3_client.get_paginator("list_objects_v2") + + logger.info(f"Listing files from s3://{bucket}/{documents_prefix}") + for page in paginator.paginate(Bucket=bucket, Prefix=documents_prefix): + if "Contents" in page: + all_files.extend( + [ + f"s3://{bucket}/{obj['Key']}" + for obj in page["Contents"] + if ( + obj["Key"].endswith(".jsonl") + or obj["Key"].endswith(".json") + or obj["Key"].endswith(".jsonl.gz") + or obj["Key"].endswith(".jsonl.zst") + or obj["Key"].endswith(".jsonl.ztd") + or obj["Key"].endswith(".jsonl.zstd") + ) + ] + ) + + if len(all_files) % 100 == 0: + logger.info(f"Found {len(all_files)} files so far...") + + logger.info(f"Total files found: {len(all_files)}") + return all_files + + +def load_document_file(s3_client, file_path): + """Load a single document file and return its contents.""" + try: + # Fetch raw bytes (S3 or local) + if file_path.startswith("s3://"): + raw = get_s3_bytes(s3_client, file_path) + else: + with open(file_path, "rb") as f: + raw = f.read() + + # Decompress if needed + if file_path.endswith(".gz"): + file_bytes = gzip.decompress(raw) + elif file_path.endswith(".zst") or file_path.endswith(".ztd") or file_path.endswith(".zstd"): + dctx = zstd.ZstdDecompressor() + file_bytes = dctx.decompress(raw, max_output_size=1_000_000_000) + else: + file_bytes = raw + + # Return the decoded lines + return file_bytes.decode("utf-8").strip().split("\n") + except Exception as e: + logger.error(f"Error loading file {file_path}: {e}") + return [] + + +def get_document_info_from_line(line, file_path, line_index): + """Extract document information from a single line.""" + try: + doc = json.loads(line) + + # A Dolma document has "text", "metadata", and "attributes" fields + if "text" not in doc or "metadata" not in doc or "attributes" not in doc: + logger.warning(f"Document in {file_path} line {line_index} is not a valid Dolma document") + return None + + # Get the original PDF path from metadata + pdf_path = doc["metadata"].get("Source-File") + if not pdf_path: + return None + + # Get page spans from attributes + page_spans = doc["attributes"].get("pdf_page_numbers", []) + if not page_spans: + return None + + # Just use the first page for each document + if page_spans: + page_span = page_spans[0] # Just get the first page + if len(page_span) >= 3: + # Page spans are [start_pos, end_pos, page_num] + page_num = page_span[2] + + # Extract text for this page + start_pos, end_pos = page_span[0], page_span[1] + page_text = doc["text"][start_pos:end_pos].strip() + + # Return the information + return { + "pdf_path": pdf_path, + "page_num": page_num, + "page_text": page_text, + "start_pos": start_pos, + "end_pos": end_pos, + "doc_id": doc["id"], + "source_file": file_path, + "line_index": line_index, + } + + return None + except json.JSONDecodeError: + logger.warning(f"Invalid JSON in {file_path} line {line_index}") + return None + except Exception as e: + logger.warning(f"Error processing document in {file_path} line {line_index}: {e}") + return None + + +def get_all_pages(s3_client, document_files): + """Get all pages from the document files for processing, preserving file and line order.""" + file_contents = {} + + # First, collect all file paths and their document info + for file_path in tqdm(document_files, desc="Loading document files"): + lines = load_document_file(s3_client, file_path) + if not lines: + logger.warning(f"Empty or invalid file: {file_path}") + continue + + # Parse each line for document info + documents = [] + for i, line in enumerate(lines): + doc_info = get_document_info_from_line(line, file_path, i) + # Always add an entry for each line, even if None, to preserve line alignment + documents.append(doc_info) + + # Store all documents for this file + file_contents[file_path] = documents + logger.info(f"Loaded {len(documents)} documents from {file_path}") + + logger.info(f"Loaded documents from {len(file_contents)} files") + return file_contents + + +def chatgpt_analyze_page(pdf_path: str, page_num: int, pdf_s3_client, openai_api_key: str, openai_model: str) -> Optional[DocumentClassification]: + """Analyze a page using the ChatGPT vision model with structured outputs for document classification.""" + try: + # Download PDF to temp file and render to image + bucket, key = parse_s3_path(pdf_path) + with tempfile.NamedTemporaryFile(suffix=".pdf", delete=False) as temp_file: + pdf_data = pdf_s3_client.get_object(Bucket=bucket, Key=key)["Body"].read() + temp_file.write(pdf_data) + temp_file_path = temp_file.name + + # Render PDF to base64 image + base64_image = render_pdf_to_base64png(temp_file_path, page_num, target_longest_image_dim=2048) + + # Clean up temp file + os.unlink(temp_file_path) + + # Create OpenAI client + client = OpenAI(api_key=openai_api_key) + + # Prepare the user message with document classification instructions + user_message = """ +You are a document classifier that analyzes and categorizes documents. +Your task is to analyze the provided document image and determine if it belongs to any of the following categories: + +DOCUMENT CATEGORIES: +- Academic Papers: Scholarly articles, research papers, academic journals +- Textbooks: Educational books, reference materials, instructional texts +- News Articles: Newspaper or magazine articles, news reports, press releases +- Tests or Quizzes: Exams, tests, quizzes, assessment materials +- Homework Assignments: Student assignments, problem sets, learning exercises +- Class Syllabus: Course outlines, class schedules, curriculum documents +- Meeting Minutes: Records of meetings, proceedings, official summaries +- Legal Contracts: Agreements, contracts, legal documents +- Forms: Official forms, applications, questionnaires +- Correspondence or Letters: Letters, emails, communications directed to a person, company, or the public +- Public Orders: Government orders, official announcements, public notices +- Court Notices: Legal notices, court summons, judicial documents + +For each category, provide a TRUE or FALSE determination based on your analysis. + +Guidelines: +1. Examine the document content, layout, formatting, and any other visual indicators +2. Be specific and consider the primary purpose of the document +3. A document may belong to multiple categories if it has multiple purposes +4. If you cannot read the document clearly, indicate this in your response +5. If the document contains inappropriate content, note this in your response +6. Provide a brief explanation of your classification in the notes section + +Only classify the document based on what you can observe in the provided image, without making assumptions about content you cannot see. +""" + + # Use the chat completions API with the custom schema + completion = client.beta.chat.completions.parse( + model=openai_model, + messages=[ + { + "role": "user", + "content": [{"type": "text", "text": user_message}, {"type": "image_url", "image_url": {"url": f"data:image/webp;base64,{base64_image}"}}], + } + ], + response_format=DocumentClassification, + max_tokens=1000, + ) + + return completion.choices[0].message.parsed + + except Exception as e: + logger.error(f"Error analyzing page {pdf_path} (page {page_num}): {e}") + return None + + +def process_single_page(args, doc_info, pdf_s3_client=None): + """Process a single document and generate document classification attribute data.""" + # Skip if document info is None + if doc_info is None: + return None + + # Extract info from the document info + pdf_path = doc_info["pdf_path"] + page_num = doc_info["page_num"] + start_pos = doc_info["start_pos"] + end_pos = doc_info["end_pos"] + doc_id = doc_info["doc_id"] + source_file = doc_info["source_file"] + line_index = doc_info["line_index"] + + # Use provided PDF S3 client if given, otherwise create one + if pdf_s3_client is None: + if args.pdf_profile: + pdf_session = boto3.Session(profile_name=args.pdf_profile) + pdf_s3_client = pdf_session.client("s3") + else: + pdf_s3_client = boto3.client("s3") + + # Get OpenAI API key + openai_api_key = args.openai_api_key or os.environ.get("OPENAI_API_KEY") + if not openai_api_key: + raise ValueError("OpenAI API key must be provided via --openai_api_key or OPENAI_API_KEY environment variable") + + # Analyze page with ChatGPT + classification = chatgpt_analyze_page(pdf_path, page_num, pdf_s3_client, openai_api_key, args.openai_model) + + if not classification: + logger.warning(f"No classification for {pdf_path} page {page_num}") + return { + "id": doc_id, + "line_index": line_index, + "attributes": None, + "source_file": source_file, + } + + # Generate attribute key names using model name + model_prefix = args.openai_model.replace("/", "_").replace("-", "_").replace(".", "_") + language_key_name = f"{model_prefix}_language" + + # Initialize result attributes with all DocumentClassification fields + result_attributes = { + language_key_name: [[start_pos, end_pos, classification.language_code.name]], + f"{model_prefix}_document_types": [[start_pos, end_pos, ",".join(classification.get_document_types)]], + f"{model_prefix}_is_academic_paper": [[start_pos, end_pos, classification.is_academic_paper]], + f"{model_prefix}_is_textbook": [[start_pos, end_pos, classification.is_textbook]], + f"{model_prefix}_is_news_article": [[start_pos, end_pos, classification.is_news_article]], + f"{model_prefix}_is_test_or_quiz": [[start_pos, end_pos, classification.is_test_or_quiz]], + f"{model_prefix}_is_homework_assignment": [[start_pos, end_pos, classification.is_homework_assignment]], + f"{model_prefix}_is_class_syllabus": [[start_pos, end_pos, classification.is_class_syllabus]], + f"{model_prefix}_is_meeting_minutes": [[start_pos, end_pos, classification.is_meeting_minutes]], + f"{model_prefix}_is_legal_contract": [[start_pos, end_pos, classification.is_legal_contract]], + f"{model_prefix}_is_form": [[start_pos, end_pos, classification.is_form]], + f"{model_prefix}_is_correspondence_or_letter": [[start_pos, end_pos, classification.is_correspondence_or_letter]], + f"{model_prefix}_is_public_order": [[start_pos, end_pos, classification.is_public_order]], + f"{model_prefix}_is_court_notice": [[start_pos, end_pos, classification.is_court_notice]], + f"{model_prefix}_classification_notes": [[start_pos, end_pos, classification.classification_notes]], + f"{model_prefix}_cannot_read": [[start_pos, end_pos, classification.cannot_read]], + f"{model_prefix}_inappropriate_content": [[start_pos, end_pos, classification.inappropriate_content]], + } + + # Return document ID, line index, and attributes + return { + "id": doc_id, + "line_index": line_index, + "attributes": result_attributes, + "source_file": source_file, + } + + +def write_attribute_file(args, processed_docs, file_documents, workspace_s3): + """Write attribute results to the appropriate files, preserving exact line order.""" + # Group results by source file and organize by line index + results_by_file = {} + for result in processed_docs: + if result is None: + continue + + source_file = result["source_file"] + if source_file not in results_by_file: + results_by_file[source_file] = {} + + # Store by line index to preserve order + results_by_file[source_file][result["line_index"]] = {"id": result["id"], "attributes": result["attributes"]} + + # Process each source file + for source_file, file_results_dict in results_by_file.items(): + try: + # 1. Build the relative path that mirrors documents/… + if source_file.startswith("s3://"): + _, key = parse_s3_path(source_file) + _, docs_prefix = parse_s3_path(args.workspace) + rel_path = key[len(os.path.join(docs_prefix, "documents/")) :] + else: + docs_root = os.path.join(args.workspace, "documents") + rel_path = os.path.relpath(source_file, docs_root) + + # 2. Create ordered attribute entries in exact same order as source file + file_entries = [] + # Get the original documents to ensure we have ALL lines in order + original_docs = file_documents[source_file] + + # Create attribute entries for every line + for i, doc_info in enumerate(original_docs): + if i in file_results_dict and file_results_dict[i]["attributes"] is not None: + # We have a processed result for this line + file_entries.append(file_results_dict[i]) + elif doc_info is not None: + # We have document info but no processed attributes (processing failed) + # Create an empty attributes entry with the correct ID + file_entries.append({"id": doc_info["doc_id"], "attributes": {}}) + else: + # This line in the source file was invalid or not a document + # Create a placeholder with a generated ID + placeholder_id = f"placeholder_{source_file}_{i}" + file_entries.append({"id": placeholder_id, "attributes": {}}) + + # 3. Create output JSONL + out_rel = os.path.join("attributes", args.attribute_name, rel_path) + out_jsonl = "\n".join(json.dumps(entry) for entry in file_entries) + "\n" + + # 4. Preserve compression type + if rel_path.endswith(".gz"): + payload = gzip.compress(out_jsonl.encode("utf-8")) + elif rel_path.endswith((".zst", ".ztd", ".zstd")): + payload = zstd.ZstdCompressor().compress(out_jsonl.encode("utf-8")) + else: + payload = out_jsonl.encode("utf-8") + + # 5. Write to args.workspace (local or S3) + if args.workspace.startswith("s3://"): + bucket, prefix = parse_s3_path(args.workspace) + key = os.path.join(prefix, out_rel) + workspace_s3.put_object(Bucket=bucket, Key=key, Body=payload) + logger.info(f"Wrote {len(file_entries)} attribute entries to s3://{bucket}/{key}") + else: + out_path = os.path.join(args.workspace, out_rel) + os.makedirs(os.path.dirname(out_path), exist_ok=True) + with open(out_path, "wb") as fh: + fh.write(payload) + logger.info(f"Wrote {len(file_entries)} attribute entries to {out_path}") + + except Exception as e: + logger.error(f"Error writing attributes for {source_file}: {e}") + continue + + +def save_results(results, output_dir): + """Save the full results to a JSON file for analysis.""" + output_path = Path(output_dir) / "rich_autoscan_results.json" + + # Convert results to serializable format + serializable_results = [] + for result in results: + if result is None: + continue + serializable_results.append(result) + + with open(output_path, "w") as f: + json.dump(serializable_results, f, indent=2, default=lambda o: o.value if isinstance(o, Enum) else o) + + print(f"Results saved to {output_path}") + + +def main(): + args = parse_args() + + # Set up S3 clients with appropriate profiles + if args.workspace_profile: + workspace_session = boto3.Session(profile_name=args.workspace_profile) + workspace_s3 = workspace_session.client("s3") + logger.info(f"Using AWS profile '{args.workspace_profile}' for workspace access") + else: + workspace_s3 = boto3.client("s3") + logger.info("Using default AWS credentials for workspace access") + + if args.pdf_profile: + pdf_session = boto3.Session(profile_name=args.pdf_profile) + pdf_s3 = pdf_session.client("s3") + logger.info(f"Using AWS profile '{args.pdf_profile}' for PDF access") + else: + # If no PDF profile specified, use the workspace profile or default + if args.workspace_profile: + pdf_s3 = workspace_s3 + logger.info(f"Using workspace profile '{args.workspace_profile}' for PDF access") + else: + pdf_s3 = boto3.client("s3") + logger.info("Using default AWS credentials for PDF access") + + # Create output directory + output_dir = Path(args.output_dir) + output_dir.mkdir(exist_ok=True, parents=True) + + # List all document files + logger.info(f"Listing document files in {args.workspace}/documents...") + document_files = list_result_files(workspace_s3, args.workspace) + logger.info(f"Found {len(document_files)} document files") + + # Load all document files and their contents, organized by file + logger.info("Loading all document files...") + file_documents = get_all_pages(workspace_s3, document_files) + + # Process each file individually + for file_index, (file_path, documents) in enumerate(file_documents.items()): + logger.info(f"Processing file {file_index+1}/{len(file_documents)}: {file_path}") + + # Only process documents that have valid information + valid_docs = [] + for doc in documents: + if doc is not None: + valid_docs.append(doc) + + # Skip if no valid documents + if not valid_docs: + logger.warning(f"No valid documents in {file_path}") + continue + + # Process in batches to manage memory and API rate limits + total_docs = len(valid_docs) + logger.info(f"Found {total_docs} valid documents to process in {file_path}") + + # Process in batches (process by document, but maintain file coherence) + all_results = [] + for i in range(0, total_docs, args.batch_size): + batch = valid_docs[i : i + args.batch_size] + batch_num = i // args.batch_size + 1 + total_batches = (total_docs + args.batch_size - 1) // args.batch_size + + logger.info(f"Processing batch {batch_num}/{total_batches} of {file_path} ({len(batch)} documents)...") + results = [] + + with ThreadPoolExecutor(max_workers=args.max_workers) as executor: + futures = [] + # Process documents in parallel but within the same file + for doc_info in batch: + futures.append(executor.submit(process_single_page, args, doc_info, pdf_s3)) + + for future in tqdm(futures, desc=f"Processing batch {batch_num} of {file_path}"): + try: + result = future.result() + results.append(result) + except Exception as e: + logger.error(f"Error processing document: {e}") + + # Save results for this batch + batch_output_dir = os.path.join(args.output_dir, f"file_{file_index+1}_batch_{batch_num}") + os.makedirs(batch_output_dir, exist_ok=True) + save_results(results, batch_output_dir) + + # Collect all results for this file + all_results.extend(results) + logger.info(f"Completed batch {batch_num}/{total_batches} of {file_path}") + + # Write attributes for the entire file, maintaining line order + write_attribute_file(args, all_results, file_documents, workspace_s3) + logger.info(f"Completed processing file {file_index+1}/{len(file_documents)}: {file_path}") + + logger.info(f"Processing complete - processed {len(file_documents)} files") + + +if __name__ == "__main__": + main() diff --git a/scripts/pii/check_qual.sh b/scripts/pii/check_qual.sh new file mode 100755 index 0000000..8ec9c22 --- /dev/null +++ b/scripts/pii/check_qual.sh @@ -0,0 +1,24 @@ +#!/bin/bash + +set -e + +python scripts/pii_rule_comparison.py \ + --docs-folder /home/ubuntu/s2pdf_dedupe_minhash_v1_with_no_pii/documents \ + --ref-rule "ft_lang_id_en_doc_v2__ft_lang_id_en_doc_v2__en:avg>0.5 and \ + fineweb_edu_fasttext_gt2__fineweb_edu_fasttext_gt2__score:avg>0.001 and \ + avg_fraction_numbers_in_line_v1__avg_fraction_numbers_in_line_v1__avg_fraction_numbers_in_line_ratio:avg<0.2 and \ + pipe_delimited_lines_v1__pipe_delimited_lines_v1__pipe_delimited_lines_ratio:avg<0.3 \ + " \ + --hyp-rule "ft_lang_id_en_doc_v2__ft_lang_id_en_doc_v2__en:avg>0.5 and \ + fineweb_edu_fasttext_gt2__fineweb_edu_fasttext_gt2__score:avg>0.001 and \ + avg_fraction_numbers_in_line_v1__avg_fraction_numbers_in_line_v1__avg_fraction_numbers_in_line_ratio:avg<0.2 and \ + pipe_delimited_lines_v1__pipe_delimited_lines_v1__pipe_delimited_lines_ratio:avg<0.4 \ + " \ + --output-dir results/pii_detection \ + + +# Run1, langid, pipes and numbers +# Prompt, boilerplate, reference, prose, table classification -> train fasttext +# 50k docs to train fast text + +tinyhost results/pii_detection/* \ No newline at end of file diff --git a/scripts/pii/pii_rule_comparison.py b/scripts/pii/pii_rule_comparison.py new file mode 100644 index 0000000..38bd361 --- /dev/null +++ b/scripts/pii/pii_rule_comparison.py @@ -0,0 +1,2152 @@ +#!/usr/bin/env python3 +""" +Compare PII Detection Rules and Calculate IoU + +This script processes documents and their attributes from S3 or local storage, +applies different rules for PII detection, and calculates the +Intersection over Union (IoU) to measure how well they overlap. + +How it works: +1. Documents are stored in one location (--docs-folder) +2. Attributes are automatically found in ../attributes/ relative to the documents folder +3. The script merges documents with all available attributes by matching filenames and document IDs +4. PII detection rules are applied to the merged documents +5. IoU and other metrics are calculated to compare the results + +Expected folder structure: +- s3://bucket/path/documents/ - Contains the main document JSONL files +- s3://bucket/path/attributes/ - Contains attributes that can be matched with documents by ID + +Document and attribute matching: +- Files are matched by basename (example.jsonl in documents matches example.jsonl in attributes) +- Within each file, documents are matched by their "id" field +- When a match is found, attributes from the attribute file are merged into the document + +Example usage: +python pii_rule_comparison.py \ + --docs-folder s3://bucket/path/documents \ + --ref-rule "gpt_4_1_contains_pii:any" \ + --hyp-rule "gpt_4_1_contains_email_addresses:any" \ + --output-file iou_results.json \ + --recursive + +Rule expression syntax: +- Simple rule: "attribute_name:rule_type" where rule_type is "any" or "all" +- Boolean expressions: "not rule1:any and rule2:all" +- Parentheses for grouping: "(rule1:any or rule2:any) and not rule3:all" +""" + +import argparse +import base64 +import gzip +import html as pyhtml +import io +import json +import logging +import os +from collections import defaultdict +from enum import Enum, auto +from io import BytesIO +from pathlib import Path + +import boto3 +import numpy as np +import zstandard as zstd +from matplotlib.figure import Figure + +# Initialize logger +logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(name)s - %(levelname)s - %(message)s") +logger = logging.getLogger(__name__) + + +# Define token types for the rule expression parser +class TokenType(Enum): + RULE = auto() + AND = auto() + OR = auto() + NOT = auto() + LPAREN = auto() + RPAREN = auto() + EOF = auto() + + +class Token: + """Token for rule expression parsing""" + + def __init__(self, type, value=None): + self.type = type + self.value = value + + def __repr__(self): + if self.value: + return f"Token({self.type}, {self.value})" + return f"Token({self.type})" + + +class ExpressionNode: + """Base class for expression tree nodes""" + + pass + + +class RuleNode(ExpressionNode): + """Leaf node representing a single rule""" + + def __init__(self, attribute_name, rule_type): + self.attribute_name = attribute_name + self.rule_type = rule_type + + def __repr__(self): + return f"Rule({self.attribute_name}:{self.rule_type})" + + +class NotNode(ExpressionNode): + """Unary NOT operation node""" + + def __init__(self, operand): + self.operand = operand + + def __repr__(self): + return f"NOT({self.operand})" + + +class BinaryNode(ExpressionNode): + """Binary operation (AND/OR) node""" + + def __init__(self, left, right, operator): + self.left = left + self.right = right + self.operator = operator + + def __repr__(self): + return f"{self.operator}({self.left}, {self.right})" + + +def parse_args(): + parser = argparse.ArgumentParser(description="Compare PII detection rules and calculate IoU") + parser.add_argument("--docs-folder", required=True, help="Documents folder path containing JSONL files (local or s3://)") + parser.add_argument("--attr-folder", help="Attributes folder path (if different from standard ../attributes/ location)") + parser.add_argument( + "--ref-rule", + required=True, + help="""Reference rule expression. Can be a simple rule in format 'attribute_name:rule_type', + where rule_type is 'any' or 'all'. Or a boolean expression like + 'not rule1:any and rule2:all' or '(rule1:any or rule2:any) and not rule3:all'""", + ) + parser.add_argument( + "--hyp-rule", + required=True, + help="""Hypothesis rule expression. Can be a simple rule in format 'attribute_name:rule_type', + where rule_type is 'any' or 'all'. Or a boolean expression like + 'not rule1:any and rule2:all' or '(rule1:any or rule2:any) and not rule3:all'""", + ) + parser.add_argument("--output-dir", default="results", help="Directory to save HTML result files") + parser.add_argument("--aws-profile", help="AWS profile for S3 access") + parser.add_argument("--recursive", action="store_true", help="Recursively process folder structure") + parser.add_argument("--debug", action="store_true", help="Enable debug logging for more detailed output") + parser.add_argument("--disable-plots", action="store_true", help="Disable CDF plots generation") + parser.add_argument("--max-plots", type=int, default=200, help="Maximum number of CDF plots to generate (default: 200)") + return parser.parse_args() + + +def parse_s3_path(s3_path): + """Parse S3 path into bucket and prefix.""" + parts = s3_path.replace("s3://", "").split("/", 1) + bucket = parts[0] + prefix = parts[1] if len(parts) > 1 else "" + return bucket, prefix + + +def get_attributes_folder(docs_folder, attr_folder=None): + """ + Determine the attributes folder path based on the documents folder. + + Args: + docs_folder: Path to the documents folder + attr_folder: Manually specified attributes folder (optional) + + Returns: + Path to the attributes folder + """ + if attr_folder: + return attr_folder + + # If no attributes folder specified, derive it from the documents folder + if docs_folder.startswith("s3://"): + # For S3 paths + bucket, prefix = parse_s3_path(docs_folder) + + # Remove trailing slashes for consistent path handling + prefix = prefix.rstrip("/") + + # Check if the documents folder is in a 'documents' directory + if prefix.endswith("/documents"): + # Replace /documents with /attributes + attr_prefix = prefix[: -len("/documents")] + "/attributes" + else: + # Otherwise, add a parent level and include 'attributes' + path_parts = prefix.split("/") + # Remove the last part (assumed to be the documents directory name) + path_parts.pop() + # Add 'attributes' + path_parts.append("attributes") + attr_prefix = "/".join(path_parts) + + return f"s3://{bucket}/{attr_prefix}" + else: + # For local paths + docs_path = Path(docs_folder) + + # Check if the documents folder is in a 'documents' directory + if docs_path.name == "documents": + # Replace /documents with /attributes + attr_path = docs_path.parent / "attributes" + else: + # Otherwise, add a parent level and include 'attributes' + attr_path = docs_path.parent / "attributes" + + return str(attr_path) + + +def get_s3_bytes(s3_client, s3_path): + """Get bytes from S3 object.""" + bucket, key = parse_s3_path(s3_path) + response = s3_client.get_object(Bucket=bucket, Key=key) + return response["Body"].read() + + +def list_jsonl_files(path, s3_client=None, recursive=False): + """List all JSONL files in the given path, locally or in S3.""" + jsonl_files = [] + + if path.startswith("s3://"): + bucket, prefix = parse_s3_path(path) + prefix = prefix.rstrip("/") + "/" + + # List objects in S3 bucket with given prefix + paginator = s3_client.get_paginator("list_objects_v2") + for page in paginator.paginate(Bucket=bucket, Prefix=prefix): + if "Contents" in page: + for obj in page["Contents"]: + key = obj["Key"] + if ( + key.endswith(".jsonl") + or key.endswith(".json") + or key.endswith(".jsonl.gz") + or key.endswith(".jsonl.zst") + or key.endswith(".jsonl.ztd") + or key.endswith(".jsonl.zstd") + ): + jsonl_files.append(f"s3://{bucket}/{key}") + else: + # Local file system + path_obj = Path(path) + if recursive: + for file_path in path_obj.rglob("*"): + if ( + file_path.name.endswith(".jsonl") + or file_path.name.endswith(".json") + or file_path.name.endswith(".jsonl.gz") + or file_path.name.endswith(".jsonl.zst") + or file_path.name.endswith(".jsonl.ztd") + or file_path.name.endswith(".jsonl.zstd") + ): + jsonl_files.append(str(file_path)) + else: + for file_path in path_obj.glob("*"): + if ( + file_path.name.endswith(".jsonl") + or file_path.name.endswith(".json") + or file_path.name.endswith(".jsonl.gz") + or file_path.name.endswith(".jsonl.zst") + or file_path.name.endswith(".jsonl.ztd") + or file_path.name.endswith(".jsonl.zstd") + ): + jsonl_files.append(str(file_path)) + + return jsonl_files + + +def load_jsonl_file(file_path, s3_client=None): + """Load and decompress a JSONL file, either from local or S3.""" + try: + # Get file content + if file_path.startswith("s3://"): + if s3_client is None: + raise ValueError("S3 client is required for S3 paths") + raw_data = get_s3_bytes(s3_client, file_path) + else: + with open(file_path, "rb") as f: + raw_data = f.read() + + # Decompress if needed + if file_path.endswith(".gz"): + decompressed = gzip.decompress(raw_data) + elif file_path.endswith((".zst", ".ztd", ".zstd")): + try: + # First try with standard decompression + dctx = zstd.ZstdDecompressor() + decompressed = dctx.decompress(raw_data) + except zstd.ZstdError as e: + # If that fails, try with stream decompression + logger.warning(f"Standard zstd decompression failed for {file_path}, trying stream decompression: {e}") + + try: + # Try with content-size not required + dctx = zstd.ZstdDecompressor(max_window_size=2147483648) # Use a large window size + decompressor = dctx.stream_reader(io.BytesIO(raw_data)) + decompressed = decompressor.read() + except Exception as inner_e: + # If both methods fail, try with chunking + logger.warning(f"Stream decompression also failed, trying chunked reading: {inner_e}") + + # Chunked reading approach + buffer = io.BytesIO() + dctx = zstd.ZstdDecompressor(max_window_size=2147483648) + with dctx.stream_reader(io.BytesIO(raw_data)) as reader: + while True: + chunk = reader.read(16384) # Read in 16KB chunks + if not chunk: + break + buffer.write(chunk) + + buffer.seek(0) + decompressed = buffer.read() + else: + decompressed = raw_data + + # Parse JSON lines + lines = decompressed.decode("utf-8").strip().split("\n") + return [json.loads(line) for line in lines if line.strip()] + + except Exception as e: + logger.error(f"Error loading file {file_path}: {e}") + return [] + + +def load_documents_and_attributes(docs_folder, attr_folder, s3_client=None, recursive=False): + """ + Load documents and merge them with their attributes from all subdirectories. + + Args: + docs_folder: Path to the documents folder + attr_folder: Path to the attributes folder + s3_client: S3 client for S3 paths + recursive: Whether to process folders recursively + + Returns: + List of documents with their attributes merged in + """ + try: + # List all document files + logger.info(f"Finding document files in: {docs_folder}") + doc_files = list_jsonl_files(docs_folder, s3_client, recursive) + logger.info(f"Found {len(doc_files)} document files") + + if not doc_files: + logger.warning(f"No document files found in {docs_folder}. Check the path and permissions.") + return [] + + # Get all attribute subdirectories if it's an S3 path + attr_subdirs = [] + if attr_folder.startswith("s3://"): + bucket, attr_prefix = parse_s3_path(attr_folder) + attr_prefix = attr_prefix.rstrip("/") + "/" + + # List top-level directories in the attributes folder + logger.info(f"Finding attribute subdirectories in: {attr_folder}") + + # Using delimiter parameter to list "directories" in S3 + paginator = s3_client.get_paginator("list_objects_v2") + for page in paginator.paginate(Bucket=bucket, Prefix=attr_prefix, Delimiter="/"): + if "CommonPrefixes" in page: + for prefix in page["CommonPrefixes"]: + subdir = f"s3://{bucket}/{prefix['Prefix']}" + attr_subdirs.append(subdir) + logger.info(f"Found attribute subdirectory: {subdir}") + + # If no subdirectories, use the main folder + if not attr_subdirs: + attr_subdirs = [attr_folder] + logger.info(f"No subdirectories found, using main attribute folder: {attr_folder}") + else: + # For local paths + attr_path = Path(attr_folder) + if attr_path.exists() and attr_path.is_dir(): + # Get subdirectories + subdirs = [str(d) for d in attr_path.iterdir() if d.is_dir()] + if subdirs: + attr_subdirs = subdirs + logger.info(f"Found {len(attr_subdirs)} attribute subdirectories") + else: + attr_subdirs = [attr_folder] + logger.info(f"No subdirectories found, using main attribute folder: {attr_folder}") + else: + logger.warning(f"Attributes folder not found or not a directory: {attr_folder}") + attr_subdirs = [] + + # Load and merge documents with all attributes from all subdirectories + merged_docs = [] + docs_by_id = {} + total_attr_files = 0 + + # First, load all document files and create a document-by-ID mapping + for doc_path in doc_files: + try: + if doc_path.startswith("s3://"): + _, doc_key = parse_s3_path(doc_path) + basename = os.path.basename(doc_key) + else: + basename = os.path.basename(doc_path) + + # Load documents + docs = load_jsonl_file(doc_path, s3_client) + if not docs: + logger.warning(f"No documents loaded from {basename} (path: {doc_path})") + continue + + logger.info(f"Loaded {len(docs)} documents from {basename}") + + # Add to the merged documents list and create ID mapping + for doc in docs: + if "id" in doc: + # If the document already exists, use the one with attributes if possible + doc_id = doc["id"] + if doc_id in docs_by_id: + if "attributes" not in doc and "attributes" in docs_by_id[doc_id]: + # Keep the existing document that has attributes + continue + + # Initialize attributes if needed + if "attributes" not in doc: + doc["attributes"] = {} + + # Add to the mapping + docs_by_id[doc_id] = doc + else: + # No ID, can't match with attributes + if "attributes" not in doc: + doc["attributes"] = {} + merged_docs.append(doc) + except Exception as e: + logger.error(f"Error processing document file {doc_path}: {e}") + continue + + logger.info(f"Loaded {len(docs_by_id)} unique documents with IDs") + + # Now process each attribute subdirectory + for subdir in attr_subdirs: + try: + logger.info(f"Processing attribute directory: {subdir}") + attr_files = list_jsonl_files(subdir, s3_client, recursive) + total_attr_files += len(attr_files) + logger.info(f"Found {len(attr_files)} attribute files in {subdir}") + + # Create a mapping from document basename to attribute file path + attr_file_map = {} + for attr_path in attr_files: + if attr_path.startswith("s3://"): + _, attr_key = parse_s3_path(attr_path) + basename = os.path.basename(attr_key) + else: + basename = os.path.basename(attr_path) + + attr_file_map[basename] = attr_path + + # Go through the document files again to find matching attributes + for doc_path in doc_files: + try: + if doc_path.startswith("s3://"): + _, doc_key = parse_s3_path(doc_path) + basename = os.path.basename(doc_key) + else: + basename = os.path.basename(doc_path) + + # Find matching attribute file + if basename in attr_file_map: + attr_path = attr_file_map[basename] + attrs = load_jsonl_file(attr_path, s3_client) + + if not attrs: + logger.warning(f"No attributes loaded from {os.path.basename(attr_path)} (path: {attr_path})") + continue + + logger.info(f"Loaded {len(attrs)} attributes from {os.path.basename(attr_path)}") + + # Create a mapping from document ID to attributes + attr_by_id = {attr["id"]: attr for attr in attrs if "id" in attr} + + # Count documents with matched attributes + docs_matched_in_file = 0 + + # Merge attributes into documents by ID + for doc_id, doc in docs_by_id.items(): + if doc_id in attr_by_id: + docs_matched_in_file += 1 + + # If attributes document has attributes field, merge them + if "attributes" in attr_by_id[doc_id]: + doc["attributes"].update(attr_by_id[doc_id]["attributes"]) + + logger.info(f"Matched attributes for {docs_matched_in_file} documents from {basename} in {subdir}") + + except Exception as e: + logger.error(f"Error processing attribute file {attr_path}: {e}") + continue + except Exception as e: + logger.error(f"Error processing attribute subdirectory {subdir}: {e}") + continue + + # Convert the dictionary to a list for return + merged_docs.extend(docs_by_id.values()) + + logger.info(f"Total documents processed: {len(merged_docs)}") + logger.info(f"Total attribute files processed: {total_attr_files}") + logger.info(f"Total attribute subdirectories processed: {len(attr_subdirs)}") + + return merged_docs + except Exception as e: + logger.error(f"Error in load_documents_and_attributes: {e}") + raise + + +def apply_rule(doc, rule): + """ + Apply a rule to determine if a document meets the PII criteria. + + Args: + doc: The document JSON object + rule: Either a tuple (attribute_name, rule_type) for simple rules, + or an ExpressionNode for complex boolean expressions + + Returns: + True if the document matches the rule, False otherwise + """ + # Handle simple rule + if not is_complex_expression(rule): + return apply_simple_rule(doc, rule[0], rule[1]) + + # Handle complex expression + return evaluate_expression(doc, rule) + + +def calculate_attribute_aggregate(doc, attribute_name, operation_type): + """ + Calculate an aggregate value for a numeric attribute. + + Args: + doc: The document JSON object + attribute_name: The attribute field to aggregate (e.g., "pii_tagging_ratio") + operation_type: The type of aggregation to perform (e.g., "avg") + + Returns: + The aggregated value, or None if calculation is not possible + """ + # Check if document has attributes + if "attributes" not in doc or not doc["attributes"]: + logger.debug(f"Document {doc.get('id', 'unknown')} has no attributes") + return None + + attributes = doc["attributes"] + + # Check if the specific attribute exists + if attribute_name not in attributes: + logger.debug(f"Document {doc.get('id', 'unknown')} doesn't have attribute: {attribute_name}") + return None + + if not attributes[attribute_name]: + logger.debug(f"Document {doc.get('id', 'unknown')} has empty attribute: {attribute_name}") + return None + + # Extract the numeric values from the attribute spans + # Each span is formatted as [start_pos, end_pos, value] + values = [span[2] for span in attributes[attribute_name] if len(span) >= 3 and span[2] is not None] + + if not values: + logger.debug(f"Document {doc.get('id', 'unknown')} has no valid values for attribute: {attribute_name}") + return None + + # Convert all values to float + try: + numeric_values = [float(value) for value in values] + except (ValueError, TypeError): + logger.debug(f"Document {doc.get('id', 'unknown')} has non-numeric values for attribute: {attribute_name}") + return None + + # Perform the aggregation + if operation_type == "avg": + if not numeric_values: + return None + return sum(numeric_values) / len(numeric_values) + # Add more aggregation types here as needed + else: + raise ValueError(f"Unknown operation type: {operation_type}") + + +def apply_simple_rule(doc, attribute_name, rule_type): + """ + Apply a simple rule to determine if a document meets the PII criteria. + + Args: + doc: The document JSON object + attribute_name: The attribute field to check (e.g., "gpt_4_1_contains_pii") + rule_type: 'any' for any true value, 'all' for all true values, + or a string containing an operation and comparison (e.g., 'avg>0.3') + + Returns: + True if the document matches the rule, False otherwise + """ + # Check if document has attributes + if "attributes" not in doc or not doc["attributes"]: + logger.debug(f"Document {doc.get('id', 'unknown')} has no attributes") + return False + + attributes = doc["attributes"] + + # Check if the specific attribute exists + if attribute_name not in attributes: + logger.debug(f"Document {doc.get('id', 'unknown')} doesn't have attribute: {attribute_name}") + return False + + if not attributes[attribute_name]: + logger.debug(f"Document {doc.get('id', 'unknown')} has empty attribute: {attribute_name}") + return False + + # Handle numeric comparison rules (e.g., 'avg>0.3') + if any(op in rule_type for op in [">", "<", ">=", "<=", "=="]): + # Parse the rule type into operation and comparison + operation_parts = rule_type.split(">") + if len(operation_parts) == 2: + operation_type, threshold = operation_parts + comparison_op = ">" + else: + operation_parts = rule_type.split("<") + if len(operation_parts) == 2: + operation_type, threshold = operation_parts + comparison_op = "<" + else: + operation_parts = rule_type.split(">=") + if len(operation_parts) == 2: + operation_type, threshold = operation_parts + comparison_op = ">=" + else: + operation_parts = rule_type.split("<=") + if len(operation_parts) == 2: + operation_type, threshold = operation_parts + comparison_op = "<=" + else: + operation_parts = rule_type.split("==") + if len(operation_parts) == 2: + operation_type, threshold = operation_parts + comparison_op = "==" + else: + raise ValueError(f"Invalid rule type: {rule_type}") + + # Convert threshold to float + try: + threshold = float(threshold) + except ValueError: + raise ValueError(f"Invalid threshold value: {threshold}") + + # Calculate the aggregate value + aggregate_value = calculate_attribute_aggregate(doc, attribute_name, operation_type) + if aggregate_value is None: + logger.debug(f"Document {doc.get('id', 'unknown')} has no valid aggregate value for attribute: {attribute_name}") + return False + + # Apply the comparison + if comparison_op == ">": + result = aggregate_value > threshold + elif comparison_op == "<": + result = aggregate_value < threshold + elif comparison_op == ">=": + result = aggregate_value >= threshold + elif comparison_op == "<=": + result = aggregate_value <= threshold + elif comparison_op == "==": + result = aggregate_value == threshold + else: + raise ValueError(f"Invalid comparison operator: {comparison_op}") + + if result: + logger.debug(f"Document {doc.get('id', 'unknown')} matched numeric rule '{attribute_name}:{rule_type}' with value {aggregate_value}") + return result + + # Handle boolean rules (any/all) + if rule_type in ["any", "all"]: + # Extract the boolean values from the attribute spans + # Each span is formatted as [start_pos, end_pos, value] + values = [span[2] for span in attributes[attribute_name] if len(span) >= 3 and span[2] is not None] + + if not values: + logger.debug(f"Document {doc.get('id', 'unknown')} has no valid values for attribute: {attribute_name}") + return False + + # Apply the rule + if rule_type == "any": + result = any(values) + if result: + logger.debug(f"Document {doc.get('id', 'unknown')} matched rule '{attribute_name}:{rule_type}' (found True in {len(values)} values)") + return result + elif rule_type == "all": + result = all(values) + if result: + logger.debug(f"Document {doc.get('id', 'unknown')} matched rule '{attribute_name}:{rule_type}' (all {len(values)} values are True)") + return result + + raise ValueError(f"Unknown rule type: {rule_type}") + + +def evaluate_expression(doc, expr): + """ + Evaluate a boolean expression on a document. + + Args: + doc: The document JSON object + expr: An ExpressionNode representing a boolean expression + + Returns: + True if the document matches the expression, False otherwise + """ + if isinstance(expr, RuleNode): + # Base case: evaluate a leaf rule node + return apply_simple_rule(doc, expr.attribute_name, expr.rule_type) + + elif isinstance(expr, NotNode): + # NOT operator + return not evaluate_expression(doc, expr.operand) + + elif isinstance(expr, BinaryNode): + # Binary operators (AND/OR) + if expr.operator == "AND": + # Short-circuit AND evaluation + return evaluate_expression(doc, expr.left) and evaluate_expression(doc, expr.right) + elif expr.operator == "OR": + # Short-circuit OR evaluation + return evaluate_expression(doc, expr.left) or evaluate_expression(doc, expr.right) + + # Should not reach here if the expression tree is well-formed + raise ValueError(f"Invalid expression node type: {type(expr)}") + + +def tokenize_expression(expression): + """ + Tokenize a rule expression string into a list of tokens. + + Args: + expression: A string containing a boolean rule expression + (e.g., "not rule1:any and rule2:all") + + Returns: + A list of Token objects + """ + tokens = [] + i = 0 + expression = expression.strip() + + while i < len(expression): + char = expression[i] + + # Skip whitespace + if char.isspace(): + i += 1 + continue + + # Handle parentheses + elif char == "(": + tokens.append(Token(TokenType.LPAREN)) + i += 1 + elif char == ")": + tokens.append(Token(TokenType.RPAREN)) + i += 1 + + # Handle operators + elif i + 2 < len(expression) and expression[i : i + 3].lower() == "and": + # Check if it's a standalone 'and' and not part of a word + if (i == 0 or expression[i - 1].isspace() or expression[i - 1] in "()") and ( + i + 3 >= len(expression) or expression[i + 3].isspace() or expression[i + 3] in "()" + ): + tokens.append(Token(TokenType.AND)) + i += 3 + else: + # It's part of an attribute name + rule_start = i + while i < len(expression) and not expression[i].isspace() and expression[i] not in "()": + if i + 1 < len(expression) and expression[i] == ":": + break + i += 1 + + # Process rule if we found a colon + if i < len(expression) and expression[i] == ":": + rule_end = i + i += 1 # Skip the colon + + # Find the rule type + type_start = i + while i < len(expression) and not expression[i].isspace() and expression[i] not in "()": + i += 1 + + rule_name = expression[rule_start:rule_end] + rule_type = expression[type_start:i] + + tokens.append(Token(TokenType.RULE, (rule_name, rule_type))) + else: + raise ValueError(f"Invalid rule format at position {rule_start}") + + elif i + 1 < len(expression) and expression[i : i + 2].lower() == "or": + # Check if it's a standalone 'or' and not part of a word + if (i == 0 or expression[i - 1].isspace() or expression[i - 1] in "()") and ( + i + 2 >= len(expression) or expression[i + 2].isspace() or expression[i + 2] in "()" + ): + tokens.append(Token(TokenType.OR)) + i += 2 + else: + # Part of an attribute name + rule_start = i + while i < len(expression) and not expression[i].isspace() and expression[i] not in "()": + if i + 1 < len(expression) and expression[i] == ":": + break + i += 1 + + # Process rule if we found a colon + if i < len(expression) and expression[i] == ":": + rule_end = i + i += 1 # Skip the colon + + # Find the rule type + type_start = i + while i < len(expression) and not expression[i].isspace() and expression[i] not in "()": + i += 1 + + rule_name = expression[rule_start:rule_end] + rule_type = expression[type_start:i] + + tokens.append(Token(TokenType.RULE, (rule_name, rule_type))) + else: + raise ValueError(f"Invalid rule format at position {rule_start}") + + elif i + 2 < len(expression) and expression[i : i + 3].lower() == "not": + # Check if it's a standalone 'not' and not part of a word + if (i == 0 or expression[i - 1].isspace() or expression[i - 1] in "()") and ( + i + 3 >= len(expression) or expression[i + 3].isspace() or expression[i + 3] in "()" + ): + tokens.append(Token(TokenType.NOT)) + i += 3 + else: + # Part of an attribute name + rule_start = i + while i < len(expression) and not expression[i].isspace() and expression[i] not in "()": + if i + 1 < len(expression) and expression[i] == ":": + break + i += 1 + + # Process rule if we found a colon + if i < len(expression) and expression[i] == ":": + rule_end = i + i += 1 # Skip the colon + + # Find the rule type + type_start = i + while i < len(expression) and not expression[i].isspace() and expression[i] not in "()": + i += 1 + + rule_name = expression[rule_start:rule_end] + rule_type = expression[type_start:i] + + tokens.append(Token(TokenType.RULE, (rule_name, rule_type))) + else: + raise ValueError(f"Invalid rule format at position {rule_start}") + + # Handle rule (attribute:type) + else: + rule_start = i + while i < len(expression) and not expression[i].isspace() and expression[i] not in "()": + if i + 1 < len(expression) and expression[i] == ":": + break + i += 1 + + # Process rule if we found a colon + if i < len(expression) and expression[i] == ":": + rule_end = i + i += 1 # Skip the colon + + # Find the rule type + type_start = i + while i < len(expression) and not expression[i].isspace() and expression[i] not in "()": + i += 1 + + rule_name = expression[rule_start:rule_end] + rule_type = expression[type_start:i] + + tokens.append(Token(TokenType.RULE, (rule_name, rule_type))) + else: + raise ValueError(f"Invalid rule format at position {rule_start}") + + tokens.append(Token(TokenType.EOF)) + return tokens + + +class Parser: + """ + Parser for boolean rule expressions. + Implements a recursive descent parser for expressions with the following grammar: + + expression → or_expr + or_expr → and_expr ("or" and_expr)* + and_expr → unary_expr ("and" unary_expr)* + unary_expr → "not" unary_expr | primary + primary → rule | "(" expression ")" + rule → ATTRIBUTE ":" RULE_TYPE + """ + + def __init__(self, tokens): + self.tokens = tokens + self.current = 0 + + def parse(self): + """Parse the tokens into an expression tree.""" + return self.expression() + + def expression(self): + """Parse an expression (top level).""" + return self.or_expr() + + def or_expr(self): + """Parse an OR expression.""" + expr = self.and_expr() + + while self.match(TokenType.OR): + right = self.and_expr() + expr = BinaryNode(expr, right, "OR") + + return expr + + def and_expr(self): + """Parse an AND expression.""" + expr = self.unary_expr() + + while self.match(TokenType.AND): + right = self.unary_expr() + expr = BinaryNode(expr, right, "AND") + + return expr + + def unary_expr(self): + """Parse a unary expression (NOT).""" + if self.match(TokenType.NOT): + operand = self.unary_expr() + return NotNode(operand) + + return self.primary() + + def primary(self): + """Parse a primary expression (rule or parenthesized expression).""" + if self.match(TokenType.RULE): + rule_tuple = self.previous().value + attribute_name, rule_type = rule_tuple + + # Validate rule type + if rule_type not in ["any", "all"] and not any(op in rule_type for op in [">", "<", ">=", "<=", "=="]): + raise ValueError(f"Invalid rule type: {rule_type}. Supported types: 'any', 'all', or numeric comparison (e.g., 'avg>0.3')") + + return RuleNode(attribute_name, rule_type) + + if self.match(TokenType.LPAREN): + expr = self.expression() + self.consume(TokenType.RPAREN, "Expected ')' after expression.") + return expr + + raise ValueError(f"Expected rule or '(' at position {self.current}") + + def match(self, *types): + """Check if the current token matches any of the given types.""" + for type in types: + if self.check(type): + self.advance() + return True + + return False + + def check(self, type): + """Check if the current token is of the given type without advancing.""" + if self.is_at_end(): + return False + return self.peek().type == type + + def advance(self): + """Advance to the next token and return the previous one.""" + if not self.is_at_end(): + self.current += 1 + return self.previous() + + def consume(self, type, message): + """Consume the current token if it matches the expected type.""" + if self.check(type): + return self.advance() + + raise ValueError(f"{message} at position {self.current}") + + def is_at_end(self): + """Check if we've reached the end of the tokens.""" + return self.peek().type == TokenType.EOF + + def peek(self): + """Return the current token without advancing.""" + return self.tokens[self.current] + + def previous(self): + """Return the previous token.""" + return self.tokens[self.current - 1] + + +def parse_rule(rule_string): + """ + Parse a rule string into an expression tree or a simple attribute-rule_type tuple. + + Args: + rule_string: A string containing a rule or boolean expression of rules + + Returns: + Either a tuple (attribute_name, rule_type) for simple rules, + or an ExpressionNode for complex boolean expressions + """ + # Check if this is a simple rule + if ( + "and" not in rule_string.lower() + and "or" not in rule_string.lower() + and "not" not in rule_string.lower() + and "(" not in rule_string + and ")" not in rule_string + ): + # Simple rule format: attribute_name:rule_type + parts = rule_string.split(":", 1) + if len(parts) != 2: + raise ValueError(f"Invalid rule format: {rule_string}. Expected format: 'attribute_name:rule_type'") + + attribute_name, rule_type = parts + + # Check for numeric comparison rule_type + if any(op in rule_type for op in [">", "<", ">=", "<=", "=="]): + # This is a numeric comparison rule - we'll validate it in apply_simple_rule + return attribute_name, rule_type + elif rule_type not in ["any", "all"]: + raise ValueError(f"Invalid rule type: {rule_type}. Supported types: 'any', 'all', or numeric comparison (e.g., 'avg>0.3')") + + return attribute_name, rule_type + else: + # Complex rule expression + try: + tokens = tokenize_expression(rule_string) + parser = Parser(tokens) + return parser.parse() + except Exception as e: + raise ValueError(f"Error parsing expression '{rule_string}': {e}") + + +def is_complex_expression(rule): + """Check if the rule is a complex boolean expression.""" + return isinstance(rule, ExpressionNode) + + +def calculate_iou(ref_ids, hyp_ids): + """Calculate Intersection over Union of two sets of document IDs.""" + ref_set = set(ref_ids) + hyp_set = set(hyp_ids) + + intersection = ref_set.intersection(hyp_set) + union = ref_set.union(hyp_set) + + if not union: + return 0.0 + + return len(intersection) / len(union) + + +def collect_rule_stats(expression, doc): + """ + Collect statistics for all rules within a complex expression. + + Args: + expression: A rule expression (either a tuple or ExpressionNode) + doc: The document to analyze + + Returns: + A dictionary with rule statistics + """ + rule_stats = defaultdict(int) + + # Handle simple rule + if not is_complex_expression(expression): + attribute_name, rule_type = expression + # Only process if document has this attribute + if "attributes" in doc and doc["attributes"] and attribute_name in doc["attributes"] and doc["attributes"][attribute_name]: + # The rule name will be the key for the statistics + rule_name = f"{attribute_name}:{rule_type}" + + # Count entries in the attribute + entries = doc["attributes"][attribute_name] + rule_stats[f"{rule_name}_total_entries"] += len(entries) + + # Count positive values + for span in entries: + if len(span) >= 3 and span[2] is True: + rule_stats[f"{rule_name}_positive_entries"] += 1 + + # Check if document matches the rule + if apply_simple_rule(doc, attribute_name, rule_type): + rule_stats[f"{rule_name}_matched_docs"] += 1 + + return rule_stats + + # For complex expressions, traverse the expression tree + if isinstance(expression, RuleNode): + # Base case: leaf node is a simple rule + attribute_name, rule_type = expression.attribute_name, expression.rule_type + if "attributes" in doc and doc["attributes"] and attribute_name in doc["attributes"] and doc["attributes"][attribute_name]: + # The rule name will be the key for the statistics + rule_name = f"{attribute_name}:{rule_type}" + + # Count entries in the attribute + entries = doc["attributes"][attribute_name] + rule_stats[f"{rule_name}_total_entries"] += len(entries) + + # Count positive values + for span in entries: + if len(span) >= 3 and span[2] is True: + rule_stats[f"{rule_name}_positive_entries"] += 1 + + # Check if document matches the rule + if apply_simple_rule(doc, attribute_name, rule_type): + rule_stats[f"{rule_name}_matched_docs"] += 1 + + elif isinstance(expression, NotNode): + # Get stats from the operand + operand_stats = collect_rule_stats(expression.operand, doc) + # Merge with current stats + for key, value in operand_stats.items(): + rule_stats[key] += value + + elif isinstance(expression, BinaryNode): + # Get stats from both sides + left_stats = collect_rule_stats(expression.left, doc) + right_stats = collect_rule_stats(expression.right, doc) + + # Merge with current stats + for key, value in left_stats.items(): + rule_stats[key] += value + for key, value in right_stats.items(): + rule_stats[key] += value + + return rule_stats + + +def get_expression_summary(expression): + """ + Generate a string representation of a rule expression. + + Args: + expression: A rule expression (either a tuple or ExpressionNode) + + Returns: + A string representation of the expression + """ + if not is_complex_expression(expression): + return f"{expression[0]}:{expression[1]}" + + if isinstance(expression, RuleNode): + return f"{expression.attribute_name}:{expression.rule_type}" + + elif isinstance(expression, NotNode): + return f"not {get_expression_summary(expression.operand)}" + + elif isinstance(expression, BinaryNode): + left_summary = get_expression_summary(expression.left) + right_summary = get_expression_summary(expression.right) + return f"({left_summary} {expression.operator.lower()} {right_summary})" + + return str(expression) + + +def compare_documents(ref_docs, hyp_docs, ref_rule, hyp_rule): + """ + Compare two sets of documents using the specified rules and calculate IoU. + + Args: + ref_docs: List of reference documents + hyp_docs: List of hypothesis documents + ref_rule: Rule expression for reference (tuple or ExpressionNode) + hyp_rule: Rule expression for hypothesis (tuple or ExpressionNode) + + Returns: + Dictionary with comparison results + """ + # Extract document IDs and create ID-to-document maps + ref_id_to_doc = {doc["id"]: doc for doc in ref_docs if "id" in doc} + hyp_id_to_doc = {doc["id"]: doc for doc in hyp_docs if "id" in doc} + + # Get common document IDs + common_ids = set(ref_id_to_doc.keys()).intersection(set(hyp_id_to_doc.keys())) + + # Apply rules to each document + ref_matches = set() + hyp_matches = set() + + # Track rule statistics + ref_rule_stats = defaultdict(int) + hyp_rule_stats = defaultdict(int) + + for doc_id in common_ids: + ref_doc = ref_id_to_doc[doc_id] + hyp_doc = hyp_id_to_doc[doc_id] + + # Collect statistics for all rules in the expressions + doc_ref_rule_stats = collect_rule_stats(ref_rule, ref_doc) + doc_hyp_rule_stats = collect_rule_stats(hyp_rule, hyp_doc) + + # Merge with overall stats + for key, value in doc_ref_rule_stats.items(): + ref_rule_stats[key] += value + for key, value in doc_hyp_rule_stats.items(): + hyp_rule_stats[key] += value + + # Check if document matches the rule expressions + if apply_rule(ref_doc, ref_rule): + ref_matches.add(doc_id) + ref_rule_stats["expression_matched_docs"] += 1 + + if apply_rule(hyp_doc, hyp_rule): + hyp_matches.add(doc_id) + hyp_rule_stats["expression_matched_docs"] += 1 + + # Calculate IoU + iou = calculate_iou(ref_matches, hyp_matches) + + # Collect detailed statistics + tp = len(ref_matches.intersection(hyp_matches)) + fp = len(hyp_matches - ref_matches) + fn = len(ref_matches - hyp_matches) + + precision = tp / (tp + fp) if (tp + fp) > 0 else 0 + recall = tp / (tp + fn) if (tp + fn) > 0 else 0 + f1 = 2 * precision * recall / (precision + recall) if (precision + recall) > 0 else 0 + + # Generate string representations of the expressions + ref_rule_str = get_expression_summary(ref_rule) + hyp_rule_str = get_expression_summary(hyp_rule) + + return { + "total_docs": len(common_ids), + "ref_rule": ref_rule_str, + "hyp_rule": hyp_rule_str, + "ref_matches": len(ref_matches), + "hyp_matches": len(hyp_matches), + "intersection": tp, + "union": tp + fp + fn, + "true_positives": tp, + "false_positives": fp, + "false_negatives": fn, + "precision": precision, + "recall": recall, + "f1": f1, + "iou": iou, + "ref_rule_stats": dict(ref_rule_stats), + "hyp_rule_stats": dict(hyp_rule_stats), + } + + +def format_rule_stats(rule_stats): + """Format rule statistics for display.""" + # Group the statistics by rule name + grouped_stats = defaultdict(dict) + + # Process regular rule stats (format: "{rule_name}_{stat_type}") + for key, value in rule_stats.items(): + if key == "expression_matched_docs": + # Special case for the overall expression match count + continue + + # Extract rule name and stat type + if "_total_entries" in key: + rule_name = key.replace("_total_entries", "") + grouped_stats[rule_name]["total_entries"] = value + elif "_positive_entries" in key: + rule_name = key.replace("_positive_entries", "") + grouped_stats[rule_name]["positive_entries"] = value + elif "_matched_docs" in key: + rule_name = key.replace("_matched_docs", "") + grouped_stats[rule_name]["matched_docs"] = value + + # Format the grouped statistics as a list of strings + formatted_stats = [] + for rule_name, stats in grouped_stats.items(): + formatted_stats.append( + f" {rule_name}:\n" + f" - Total Entries: {stats.get('total_entries', 0)}\n" + f" - Positive Entries: {stats.get('positive_entries', 0)}\n" + f" - Matched Documents: {stats.get('matched_docs', 0)}" + ) + + # Add the expression matched count if available + if "expression_matched_docs" in rule_stats: + formatted_stats.append(f" Overall Expression Matched Documents: {rule_stats['expression_matched_docs']}") + + return "\n".join(formatted_stats) + + +def collect_numeric_attributes(documents): + """ + Collect all numeric attribute values from documents. + + Args: + documents: List of documents with attributes + + Returns: + Dictionary mapping attribute names to lists of numeric values + """ + numeric_attributes = defaultdict(list) + + for doc in documents: + if "attributes" not in doc or not doc["attributes"]: + continue + + for attr_name, attr_values in doc["attributes"].items(): + if not attr_values: + continue + + # Try to extract numeric values from the attribute spans + # Each span is formatted as [start_pos, end_pos, value] + for span in attr_values: + if len(span) >= 3 and span[2] is not None: + try: + # Convert to float if it's a numeric value + value = float(span[2]) + numeric_attributes[attr_name].append(value) + except (ValueError, TypeError): + # Not a numeric value, skip + pass + + # Filter out attributes with no or too few numeric values + return {k: v for k, v in numeric_attributes.items() if len(v) > 5} + + +def generate_cdf_plot(values, attribute_name): + """ + Generate a CDF plot for the given numeric values. + + Args: + values: List of numeric values + attribute_name: Name of the attribute (for plot title) + + Returns: + Base64-encoded PNG image of the plot or None if there's an error + """ + try: + # Ensure we have enough data points + if len(values) < 5: + logger.warning(f"Not enough data points to generate CDF for {attribute_name}") + return None + + # Remove any NaN or infinite values + values = np.array([v for v in values if np.isfinite(v)]) + if len(values) < 5: + logger.warning(f"Not enough finite data points to generate CDF for {attribute_name}") + return None + + # Handle extreme values by removing outliers (optional) + # if len(values) > 30: # Only apply if we have enough data points + # q1, q3 = np.percentile(values, [25, 75]) + # iqr = q3 - q1 + # lower_bound = q1 - 3 * iqr + # upper_bound = q3 + 3 * iqr + # values = values[(values >= lower_bound) & (values <= upper_bound)] + + # Sort values for CDF calculation + values = np.sort(values) + + # Create a Figure object (no interactive display) + fig = Figure(figsize=(10, 6)) + ax = fig.add_subplot(1, 1, 1) + + # Calculate CDF (y-values are 0 to 1 for cumulative probability) + y = np.arange(1, len(values) + 1) / len(values) + + # Plot the CDF + ax.plot(values, y, "b-", linewidth=2) + ax.grid(True, linestyle="--", alpha=0.7) + + # Add labels and title + ax.set_xlabel("Value", fontsize=12) + ax.set_ylabel("Cumulative Probability", fontsize=12) + ax.set_title(f"CDF of {attribute_name}", fontsize=14) + + # Ensure the y-axis goes from 0 to 1 for probability + ax.set_ylim(0, 1.05) + + # Add some statistics to the plot + if len(values) > 0: + mean_val = np.mean(values) + median_val = np.median(values) + min_val = np.min(values) + max_val = np.max(values) + stats_text = f"n={len(values)}\nmin={min_val:.2f}\nmax={max_val:.2f}\nmean={mean_val:.2f}\nmedian={median_val:.2f}" + ax.text(0.02, 0.98, stats_text, transform=ax.transAxes, verticalalignment="top", bbox=dict(boxstyle="round", facecolor="white", alpha=0.8)) + + # Make layout tight + fig.tight_layout() + + # Convert to base64 for embedding in HTML + buf = BytesIO() + fig.savefig(buf, format="png", dpi=100) + buf.seek(0) + img_base64 = base64.b64encode(buf.getvalue()).decode("utf-8") + + return img_base64 + + except Exception as e: + logger.error(f"Error generating CDF plot for {attribute_name}: {e}") + return None + + +def generate_attribute_plots_html(numeric_attributes, max_plots=20): + """ + Generate HTML section with CDF plots for all numeric attributes. + + Args: + numeric_attributes: Dictionary mapping attribute names to lists of numeric values + max_plots: Maximum number of plots to generate + + Returns: + HTML string with embedded CDF plots + """ + if not numeric_attributes: + return "" + + html = """ +

Numeric Attribute Distributions

+ """ + + plot_count = 0 + + # Sort attributes by number of values (most values first) + sorted_attrs = sorted(numeric_attributes.items(), key=lambda x: len(x[1]), reverse=True) + + for attr_name, values in sorted_attrs: + if len(values) < 10: # Skip attributes with too few values for meaningful plots + continue + + if plot_count >= max_plots: + logger.info(f"Limiting CDF plots to {max_plots} attributes to avoid performance issues") + break + + # Generate the CDF plot + img_base64 = generate_cdf_plot(values, attr_name) + + # Only add to HTML if plot generation was successful + if img_base64: + html += f""" +

{attr_name}

+ $CDF plot for {attr_name}$ +

Number of values: {len(values)}

+ """ + plot_count += 1 + + if plot_count == 0: + return "" # Don't add the section if no plots were generated + + html += """ +

+ """ + + return html + + +def generate_html_report(docs, title, summary, output_path): + """ + Generate an HTML report file with document texts + + Args: + docs: List of documents to include in the report + title: Title of the report + summary: Summary statistics to include at the top + output_path: Path to save the HTML file + + Returns: + None + """ + # Create header with CSS styling + html = f""" + + + + {title} + + + + + + + +

Summary

{summary}

+ + +

+""" + + # Add each document with a unique ID + for i, doc in enumerate(docs, 1): + doc_id = doc.get("id", f"unknown_{i}") + # Get document text, falling back to JSON representation if not available + doc_text = doc.get("text", json.dumps(doc, indent=2)) + + # The first document gets the "selected" class + selected_class = " selected" if i == 1 else "" + + html += f""" +

Document ID: {doc_id}

{pyhtml.escape(doc_text)}

+""" + + # Add JavaScript for keyboard navigation and summary toggle + html += """ +

+ + + + +""" + + # Create directory if it doesn't exist + os.makedirs(os.path.dirname(os.path.abspath(output_path)), exist_ok=True) + + # Write HTML to file + with open(output_path, "w", encoding="utf-8") as f: + f.write(html) + + logger.info(f"Generated HTML report: {output_path}") + + +def main(): + global args + args = parse_args() + + # Set up logging based on arguments + if args.debug: + logger.setLevel(logging.DEBUG) + logger.debug("Debug logging enabled") + + # Set up S3 client if needed + s3_client = None + if args.docs_folder.startswith("s3://") or (args.attr_folder and args.attr_folder.startswith("s3://")): + session = boto3.Session(profile_name=args.aws_profile) if args.aws_profile else boto3.Session() + s3_client = session.client("s3") + + # Parse the rules + logger.info(f"Parsing reference rule expression: {args.ref_rule}") + ref_rule = parse_rule(args.ref_rule) + + logger.info(f"Parsing hypothesis rule expression: {args.hyp_rule}") + hyp_rule = parse_rule(args.hyp_rule) + + # Generate string representations of the expressions + ref_rule_str = get_expression_summary(ref_rule) + hyp_rule_str = get_expression_summary(hyp_rule) + + logger.info(f"Reference rule parsed as: {ref_rule_str}") + logger.info(f"Hypothesis rule parsed as: {hyp_rule_str}") + + # Determine attributes folder + attr_folder = get_attributes_folder(args.docs_folder, args.attr_folder) + logger.info(f"Using attributes folder: {attr_folder}") + + # Load documents and merge with attributes from all subdirectories + logger.info("Loading documents and merging with all attributes...") + all_docs = load_documents_and_attributes(args.docs_folder, attr_folder, s3_client, args.recursive) + + # Create output directory if it doesn't exist + os.makedirs(args.output_dir, exist_ok=True) + + # Use the same documents for both reference and hypothesis evaluation + # since we've loaded all attributes into each document + ref_docs = all_docs + hyp_docs = all_docs + + # Compare the documents + logger.info("Comparing documents using reference and hypothesis rules...") + comparison_result = compare_documents(ref_docs, hyp_docs, ref_rule, hyp_rule) + + # Get document IDs for each category + ref_matches = set() + hyp_matches = set() + + # Create mappings from document IDs to documents + doc_map = {doc["id"]: doc for doc in all_docs if "id" in doc} + + # Find documents that match the reference and hypothesis rules + for doc_id, doc in doc_map.items(): + if apply_rule(doc, ref_rule): + ref_matches.add(doc_id) + if apply_rule(doc, hyp_rule): + hyp_matches.add(doc_id) + + # Calculate document sets for each category + true_positives_ids = ref_matches.intersection(hyp_matches) + true_negatives_ids = set(doc_map.keys()) - ref_matches - hyp_matches + false_positives_ids = hyp_matches - ref_matches + false_negatives_ids = ref_matches - hyp_matches + + # Create document lists for each category + true_positives = [doc_map[doc_id] for doc_id in true_positives_ids] + true_negatives = [doc_map[doc_id] for doc_id in true_negatives_ids] + false_positives = [doc_map[doc_id] for doc_id in false_positives_ids] + false_negatives = [doc_map[doc_id] for doc_id in false_negatives_ids] + + # Calculate metrics + tp = len(true_positives) + tn = len(true_negatives) + fp = len(false_positives) + fn = len(false_negatives) + + precision = tp / (tp + fp) if (tp + fp) > 0 else 0 + recall = tp / (tp + fn) if (tp + fn) > 0 else 0 + f1 = 2 * precision * recall / (precision + recall) if (precision + recall) > 0 else 0 + iou = tp / (tp + fp + fn) if (tp + fp + fn) > 0 else 0 + + # Prepare overall statistics + overall_stats = { + "total_docs": len(doc_map), + "ref_matches": len(ref_matches), + "hyp_matches": len(hyp_matches), + "true_positives": tp, + "true_negatives": tn, + "false_positives": fp, + "false_negatives": fn, + "precision": precision, + "recall": recall, + "f1": f1, + "iou": iou, + "ref_rule_stats": comparison_result["ref_rule_stats"], + "hyp_rule_stats": comparison_result["hyp_rule_stats"], + } + + # Prepare summary + summary = f"""Reference Rule: {args.ref_rule} +Hypothesis Rule: {args.hyp_rule} + +Total Documents: {overall_stats['total_docs']} +Reference Matches: {overall_stats['ref_matches']} +Hypothesis Matches: {overall_stats['hyp_matches']} + +True Positives: {tp} +True Negatives: {tn} +False Positives: {fp} +False Negatives: {fn} + +Precision: {precision:.4f} +Recall: {recall:.4f} +F1 Score: {f1:.4f} +IoU: {iou:.4f} +""" + + # Generate HTML reports for each category + logger.info("Generating HTML reports...") + + # True Positives + generate_html_report( + true_positives[:1000], + "True Positives - Documents matching both Reference and Hypothesis Rules", + summary, + os.path.join(args.output_dir, "true_positives.html"), + ) + + # True Negatives + generate_html_report( + true_negatives[:1000], "True Negatives - Documents not matching either Rule", summary, os.path.join(args.output_dir, "true_negatives.html") + ) + + # False Positives + generate_html_report( + false_positives[:1000], + "False Positives - Documents matching Hypothesis but not Reference Rule", + summary, + os.path.join(args.output_dir, "false_positives.html"), + ) + + # False Negatives + generate_html_report( + false_negatives[:1000], + "False Negatives - Documents matching Reference but not Hypothesis Rule", + summary, + os.path.join(args.output_dir, "false_negatives.html"), + ) + + # Collect numeric attributes and generate CDF plots if not disabled + attribute_plots_html = "" + if not args.disable_plots: + logger.info("Collecting numeric attributes for CDF plots...") + numeric_attributes = collect_numeric_attributes(all_docs) + + if numeric_attributes: + logger.info(f"Found {len(numeric_attributes)} numeric attributes suitable for CDF plots") + # Generate CDF plots HTML with the specified maximum number of plots + attribute_plots_html = generate_attribute_plots_html(numeric_attributes, args.max_plots) + else: + logger.info("No numeric attributes found for CDF plots") + else: + logger.info("CDF plot generation disabled by --disable-plots flag") + + # Generate index.html file that links to all reports + index_html = f""" + + + + PII Rule Comparison Results + + + +

PII Rule Comparison Results

Summary

{summary}

Result Categories

True Positives: {tp}

Documents that match both the reference and hypothesis rules.

+ View True Positives +

True Negatives: {tn}

Documents that don't match either the reference or hypothesis rules.

+ View True Negatives +

False Positives: {fp}

Documents that match the hypothesis rule but not the reference rule.

+ View False Positives +

False Negatives: {fn}

Documents that match the reference rule but not the hypothesis rule.

+ View False Negatives +

+ + {attribute_plots_html} + + +""" + + with open(os.path.join(args.output_dir, "index.html"), "w", encoding="utf-8") as f: + f.write(index_html) + + # Print summary + logger.info("\n--- COMPARISON SUMMARY ---") + logger.info(f"Documents Folder: {args.docs_folder}") + logger.info(f"Attributes Folder: {attr_folder}") + logger.info(f"Reference Rule Expression: {args.ref_rule}") + logger.info(f" Parsed as: {ref_rule_str}") + logger.info(f"Hypothesis Rule Expression: {args.hyp_rule}") + logger.info(f" Parsed as: {hyp_rule_str}") + logger.info(f"Total Documents: {overall_stats['total_docs']}") + + # Print rule statistics + logger.info("\n--- RULE MATCH STATISTICS ---") + + logger.info("\nReference Rules:") + logger.info(format_rule_stats(overall_stats["ref_rule_stats"])) + + logger.info("\nHypothesis Rules:") + logger.info(format_rule_stats(overall_stats["hyp_rule_stats"])) + + # Print comparison metrics + logger.info("\n--- COMPARISON METRICS ---") + logger.info(f"True Positives: {tp}") + logger.info(f"True Negatives: {tn}") + logger.info(f"False Positives: {fp}") + logger.info(f"False Negatives: {fn}") + logger.info(f"Precision: {precision:.4f}") + logger.info(f"Recall: {recall:.4f}") + logger.info(f"F1 Score: {f1:.4f}") + logger.info(f"IoU: {iou:.4f}") + + # Output all available attributes that have been loaded + logger.info("\n--- AVAILABLE ATTRIBUTES ---") + all_attributes = set() + for doc in all_docs: + if "attributes" in doc and doc["attributes"]: + all_attributes.update(doc["attributes"].keys()) + + if all_attributes: + logger.info(f"Found {len(all_attributes)} unique attributes:") + for attr in sorted(all_attributes): + logger.info(f" - {attr}") + else: + logger.info("No attributes found in any documents.") + + logger.info(f"\nResults saved to: {args.output_dir}/index.html") + + +if __name__ == "__main__": + main() + +# Example commands with actual S3 paths: +""" +# Example for AI2 OE data with resume detection: +python scripts/pii_rule_comparison.py \ + --docs-folder s3://ai2-oe-data/jakep/s2pdf_dedupe_minhash_v1_mini/documents/ \ + --ref-rule "gpt_4_1_contains_pii:any and not gpt_4_1_is_public_document:all" \ + --hyp-rule "google_gemma-3-4b-it_is_resume_cv:any" \ + --output-dir results/resume_detection \ + --recursive \ + --debug + +# Example for PII detection comparison: +python scripts/pii_rule_comparison.py \ + --docs-folder s3://allenai-dolma/documents/v1.5 \ + --ref-rule "contains_pii:any" \ + --hyp-rule "(contains_email_addresses:any or contains_phone_numbers:any) and not false_positive:any" \ + --output-dir results/pii_detection \ + --recursive \ + --aws-profile dolma + +# Example with custom attributes folder: +python scripts/pii_rule_comparison.py \ + --docs-folder s3://bucket/path/documents \ + --attr-folder s3://bucket/custom/location/attributes \ + --ref-rule "gpt_4_1_contains_pii:any" \ + --hyp-rule "custom_model_pii_detection:any" \ + --output-dir results/custom_comparison \ + --recursive +""" diff --git a/scripts/pii/rich_tagging_pipeline.py b/scripts/pii/rich_tagging_pipeline.py new file mode 100644 index 0000000..079a09e --- /dev/null +++ b/scripts/pii/rich_tagging_pipeline.py @@ -0,0 +1,956 @@ +#!/usr/bin/env python3 +""" +Tagging pipeline for Dolma JSONL datasets. + +For each .jsonl, .jsonl.gz, or .jsonl.ztd file under the dataset/documents folder, +this script issues a model prompt completion +collects the yes/no answers, and writes corresponding Dolma attributes JSONL files under +scratch/attributes/, mirroring the input structure. +""" + +import argparse +import asyncio +import atexit +import gzip +import json +import logging +import os +import re +import sys +import time +from typing import Optional +from urllib.parse import urlparse + +import boto3 +import httpx +import zstandard as zstd +from huggingface_hub import snapshot_download +from pydantic import BaseModel, Field, ValidationError + +from olmocr.check import check_torch_gpu_available +from olmocr.metrics import MetricsKeeper +from olmocr.s3_utils import ( + download_directory, + expand_s3_glob, + get_s3_bytes_with_backoff, + parse_s3_path, +) +from olmocr.version import VERSION +from olmocr.work_queue import LocalBackend, S3Backend, WorkQueue + +# Initialize logger +logger = logging.getLogger(__name__) +logger.setLevel(logging.DEBUG) +logger.propagate = False + +server_logger = logging.getLogger("vllm") +server_logger.propagate = False + +file_handler = logging.FileHandler("olmocr-pipeline-debug.log", mode="a") +file_handler.setLevel(logging.DEBUG) +file_handler.setFormatter(logging.Formatter("%(asctime)s - %(name)s - %(levelname)s - %(message)s")) + +console_handler = logging.StreamHandler() +console_handler.setLevel(logging.INFO) +console_handler.setFormatter(logging.Formatter("%(asctime)s - %(name)s - %(levelname)s - %(message)s")) + +# Add handlers to the logger +logger.addHandler(file_handler) +logger.addHandler(console_handler) +server_logger.addHandler(file_handler) + + +# Default port; overridden by --port +SERVER_PORT = 30024 + +# Global variables for token statistics +metrics = MetricsKeeper(window=60 * 5) + + +class PIIClassification(BaseModel): + primary_language: str = Field(..., description="Primary language as a two-letter code") + document_type: str = Field(..., description="Basic summary of document type classification") + is_resume_cv: Optional[bool] = Field(None, description="True if the document is a page from a resume or cv") + + is_academic_paper: Optional[bool] = None + is_textbook: Optional[bool] = None + is_news_article: Optional[bool] = None + is_test_or_quiz: Optional[bool] = None + is_homework_assignment: Optional[bool] = None + is_class_syllabus: Optional[bool] = None + is_meeting_minutes: Optional[bool] = None + is_legal_contract: Optional[bool] = None + is_form: Optional[bool] = None + is_correspondence_or_letter: Optional[bool] = None + is_public_order: Optional[bool] = None + is_court_notice: Optional[bool] = None + is_completion_certificate: Optional[bool] = None + + contains_pii: Optional[bool] = Field(None, description="True if document contains PII") + + +class RichPIIClassification(BaseModel): + primary_language: str = Field(..., description="Primary language as a two-letter code") + document_type: str = Field(..., description="Basic summary of document type classification") + + is_public_document: bool = Field(False, description="True if the document is meant for public dissemination") + + contains_pii_government_id: Optional[bool] = Field(None) + contains_pii_financial_info: Optional[bool] = Field(None) + contains_pii_biometric_data: Optional[bool] = Field(None) + contains_pii_login_info: Optional[bool] = Field(None) + + contains_identifier_name: Optional[bool] = Field(None, description="True if the document contains a name") + contains_identifier_email: Optional[bool] = Field(None, description="True if the document contains an email address") + contains_identifier_phone_number: Optional[bool] = Field(None, description="True if the document contains phone numbers") + + contains_identifier_with_address: Optional[bool] = Field(None) + contains_identifier_with_biographical_info: Optional[bool] = Field(None) + contains_identifier_with_location_info: Optional[bool] = Field(None) + contains_identifier_with_employment_info: Optional[bool] = Field(None) + contains_identifier_with_education_info: Optional[bool] = Field(None) + contains_identifier_with_medical_info: Optional[bool] = Field(None) + + def contains_any_pii(self) -> bool: + if self.is_public_document: + return False + + if self.contains_pii_government_id or self.contains_pii_financial_info or self.contains_pii_biometric_data or self.contains_pii_login_info: + return True + + if self.contains_identifier_name or self.contains_identifier_email or self.contains_identifier_phone_number: + return ( + self.contains_identifier_with_address + or self.contains_identifier_with_biographical_info + or self.contains_identifier_with_location_info + or self.contains_identifier_with_employment_info + or self.contains_identifier_with_education_info + or self.contains_identifier_with_medical_info + ) + else: + return False + + +async def _process_single_page(page_text: str) -> RichPIIClassification: + """Helper function to process a single document or page.""" + + rich_prompt = """You are a document analyzer that identifies Personally Identifiable Information (PII) in documents. +Your task is to analyze the document provided below and determine: +1. Whether the document is intended for public release or dissemination (e.g., research paper, public report, etc.) +2. If the document contains any PII + +For PII identification, follow these specific guidelines: + +PII THAT OCCURS EVEN WITHOUT AN IDENTIFIER: +The following should ALWAYS be marked as PII even if they do not occur alongside an identifier: +- Government IDs (Social Security Numbers, passport numbers, driver's license numbers, tax IDs) +- Financial Information (credit card numbers, bank account/routing numbers) +- Biometric Data (fingerprints, retina scans, facial recognition data, voice signatures) +- Login information (ONLY mark as PII when a username, password, and login location are present together) + +IDENTIFIERS FOR PII: +The following are considered identifiers that can make information PII: +- Names (full names, first names, last names, nicknames) +- Email addresses +- Phone numbers + +PII THAT MUST CO-OCCUR WITH AN IDENTIFIER: +The following types of information should ONLY be marked as PII if they occur ALONGSIDE an identifier (commonly, a person's name): +- Addresses (street address, postal code, etc.) +- Biographical Information (date of birth, place of birth, gender, sexual orientation, race, ethnicity, citizenship/immigration status, religion) +- Location Information (geolocations, specific coordinates) +- Employment Information (job titles, workplace names, employment history) +- Education Information (school names, degrees, transcripts) +- Medical Information (health records, diagnoses, genetic or neural data) + +If the document is a form, then ONLY consider fields which are filled out with specific values as potential PII. An empty form that asks for PII is not to be marked as containing PII. +If this page does not itself contain PII, but references documents (such as curriculum vitae, personal statements) that typically contain PII, then do not mark it as PII. +Only consider actual occurrences of the PII within the document shown. +""" + + prompt_end = "Answer as a JSON object with the following schema {'primary_language': str, 'document_type': str, 'is_public_document': bool, 'contains_pii_government_id': bool, 'contains_pii_financial_info': bool, 'contains_pii_biometric_data': bool, 'contains_pii_login_info': bool, 'contains_identifier_name': bool, 'contains_identifier_email': bool, 'contains_identifier_phone_number': bool, 'contains_identifier_with_address': bool, 'contains_identifier_with_biographical_info': bool, 'contains_identifier_with_location_info': bool, 'contains_identifier_with_employment_info': bool, 'contains_identifier_with_education_info': bool, 'contains_identifier_with_medical_info': bool}" + + query = { + "model": "google/gemma-3-12b-it", + "messages": [ + { + "role": "user", + "content": [ + { + "type": "text", + "text": (f"{rich_prompt}\n-----DOCUMENT_START-----\n{page_text}\n-----DOCUMENT_END-----\n{prompt_end}"), + } + ], + } + ], + "max_tokens": 300, + "temperature": 0.0, + "response_format": {"type": "json_schema", "json_schema": {"name": "RichPIIClassification", "schema": RichPIIClassification.model_json_schema()}}, + } + + url = f"http://localhost:{SERVER_PORT}/v1/chat/completions" + + # ---------- HTTP call --------------------------------------------------- + try: + status, body = await apost(url, json_data=query) + except Exception as e: + logger.warning(f"Server network error: {e!s}") + metrics.add_metrics(server_errors=1) + return RichPIIClassification(primary_language="en", document_type="unknown", is_public_document=False) + + metrics.add_metrics(server_requests=1) + + if status != 200: + logger.warning(f"Server HTTP {status}: {body[:250]!r}") + metrics.add_metrics(server_errors=1) + return RichPIIClassification(primary_language="en", document_type="unknown", is_public_document=False) + + # ---------- Parse base JSON -------------------------------------------- + try: + base = json.loads(body) + except json.JSONDecodeError: + logger.warning(f"Server response is not valid JSON: {body[:250]!r}") + metrics.add_metrics(server_errors=1) + return RichPIIClassification(primary_language="en", document_type="unknown", is_public_document=False) + + # Token accounting if available + if "usage" in base: + metrics.add_metrics( + server_input_tokens=base["usage"].get("prompt_tokens", 0), + server_output_tokens=base["usage"].get("completion_tokens", 0), + ) + + # ---------- Extract the model message ---------------------------------- + try: + content = base["choices"][0]["message"].get("content") + except (KeyError, IndexError, AttributeError) as e: + logger.warning(f"Missing fields in Server response: {e!s}") + metrics.add_metrics(server_errors=1) + return RichPIIClassification(primary_language="en", document_type="unknown", is_public_document=False) + + if not isinstance(content, str): + logger.warning("Server `content` is not a string; treating as error.") + metrics.add_metrics(server_errors=1) + return RichPIIClassification(primary_language="en", document_type="unknown", is_public_document=False) + + try: + pii_classification: RichPIIClassification = RichPIIClassification.model_validate_json(content) + return pii_classification + except ValidationError as e: + logger.warning(f"Unable to parse pii classification object: {e!s}") + metrics.add_metrics(server_errors=1) + return RichPIIClassification(primary_language="en", document_type="unknown", is_public_document=False) + + +# Manual simple implementation of HTTP Post +# It feels strange perhaps, but httpx and aiohttp are very complex beasts +# Ex. the sessionpool in httpcore has 4 different locks in it, and I've noticed +# that at the scale of 100M+ requests, that they deadlock in different strange ways +async def apost(url, json_data): + parsed_url = urlparse(url) + host = parsed_url.hostname + port = parsed_url.port or 80 + path = parsed_url.path or "/" + + writer = None + try: + reader, writer = await asyncio.open_connection(host, port) + + json_payload = json.dumps(json_data) + request = ( + f"POST {path} HTTP/1.1\r\n" + f"Host: {host}\r\n" + f"Content-Type: application/json\r\n" + f"Content-Length: {len(json_payload)}\r\n" + f"Connection: close\r\n\r\n" + f"{json_payload}" + ) + writer.write(request.encode()) + await writer.drain() + + # Read status line + status_line = await reader.readline() + if not status_line: + raise ConnectionError("No response from server") + status_parts = status_line.decode().strip().split(" ", 2) + if len(status_parts) < 2: + raise ValueError(f"Malformed status line: {status_line.decode().strip()}") + status_code = int(status_parts[1]) + + # Read headers + headers = {} + while True: + line = await reader.readline() + if line in (b"\r\n", b"\n", b""): + break + key, _, value = line.decode().partition(":") + headers[key.strip().lower()] = value.strip() + + # Read response body + if "content-length" in headers: + body_length = int(headers["content-length"]) + response_body = await reader.readexactly(body_length) + else: + raise ConnectionError("Anything other than fixed content length responses are not implemented yet") + + return status_code, response_body + except Exception as e: + # Pass through errors + raise e + finally: + # But just make sure to close the socket on your way out + if writer is not None: + try: + writer.close() + await writer.wait_closed() + except: + pass + + +async def process_dolma_document(args, dolma_doc, sem): + """ + Query model to detect PII, enforcing a JSON schema. + + Resilient to: + • Transport / HTTP errors + • Missing or malformed fields in the response + • Non-string or None `content` + • Bad JSON in the model's answer + + Always returns: (doc_id, contains_pii: bool, text_length: int) + """ + text = dolma_doc.get("text", "") or "" + + # Generate attribute key names using model name + model_prefix = args.model.replace("/", "_") + language_key_name = f"{model_prefix}_language" + contains_pii_key_name = f"{model_prefix}_contains_pii" + + # Initialize result attributes with all RichPIIClassification attributes + result_attributes = { + contains_pii_key_name: [], + language_key_name: [], + f"{model_prefix}_is_public_document": [], + f"{model_prefix}_contains_pii_government_id": [], + f"{model_prefix}_contains_pii_financial_info": [], + f"{model_prefix}_contains_pii_biometric_data": [], + f"{model_prefix}_contains_pii_login_info": [], + f"{model_prefix}_contains_identifier_name": [], + f"{model_prefix}_contains_identifier_email": [], + f"{model_prefix}_contains_identifier_phone_number": [], + f"{model_prefix}_contains_identifier_with_address": [], + f"{model_prefix}_contains_identifier_with_biographical_info": [], + f"{model_prefix}_contains_identifier_with_location_info": [], + f"{model_prefix}_contains_identifier_with_employment_info": [], + f"{model_prefix}_contains_identifier_with_education_info": [], + f"{model_prefix}_contains_identifier_with_medical_info": [], + } + + # If pdf_page_numbers is present, split the text and process each page separately + if "attributes" in dolma_doc and "pdf_page_numbers" in dolma_doc["attributes"]: + page_numbers = dolma_doc["attributes"]["pdf_page_numbers"] + + # Filter pages down to actual real content + selected_page_numbers = [tuple(p) for p in page_numbers if p[0] < p[1]] + + # Select just the first page + selected_page_numbers = selected_page_numbers[:1] + + # Original select first + 2 more code + # first_page_number = selected_page_numbers[0] + + # # Sample 3 pages max per document, but always include the first page, it's a good signal for CV classification + # random.shuffle(selected_page_numbers) + # selected_page_numbers = selected_page_numbers[:3] + + # if first_page_number not in selected_page_numbers: + # selected_page_numbers[0] = first_page_number + + for start_pos, end_pos, page_num in page_numbers: + if (start_pos, end_pos, page_num) in selected_page_numbers: + page_text = text[start_pos:end_pos] + + # Process each page with the semaphore to limit concurrent requests + async with sem: + pii_class = await _process_single_page(page_text) + + # Add all attributes from RichPIIClassification + result_attributes[contains_pii_key_name].append([start_pos, end_pos, pii_class.contains_any_pii()]) + result_attributes[language_key_name].append([start_pos, end_pos, pii_class.primary_language]) + result_attributes[f"{model_prefix}_is_public_document"].append([start_pos, end_pos, pii_class.is_public_document]) + result_attributes[f"{model_prefix}_contains_pii_government_id"].append([start_pos, end_pos, pii_class.contains_pii_government_id]) + result_attributes[f"{model_prefix}_contains_pii_financial_info"].append([start_pos, end_pos, pii_class.contains_pii_financial_info]) + result_attributes[f"{model_prefix}_contains_pii_biometric_data"].append([start_pos, end_pos, pii_class.contains_pii_biometric_data]) + result_attributes[f"{model_prefix}_contains_pii_login_info"].append([start_pos, end_pos, pii_class.contains_pii_login_info]) + result_attributes[f"{model_prefix}_contains_identifier_name"].append([start_pos, end_pos, pii_class.contains_identifier_name]) + result_attributes[f"{model_prefix}_contains_identifier_email"].append([start_pos, end_pos, pii_class.contains_identifier_email]) + result_attributes[f"{model_prefix}_contains_identifier_phone_number"].append([start_pos, end_pos, pii_class.contains_identifier_phone_number]) + result_attributes[f"{model_prefix}_contains_identifier_with_address"].append([start_pos, end_pos, pii_class.contains_identifier_with_address]) + result_attributes[f"{model_prefix}_contains_identifier_with_biographical_info"].append( + [start_pos, end_pos, pii_class.contains_identifier_with_biographical_info] + ) + result_attributes[f"{model_prefix}_contains_identifier_with_location_info"].append( + [start_pos, end_pos, pii_class.contains_identifier_with_location_info] + ) + result_attributes[f"{model_prefix}_contains_identifier_with_employment_info"].append( + [start_pos, end_pos, pii_class.contains_identifier_with_employment_info] + ) + result_attributes[f"{model_prefix}_contains_identifier_with_education_info"].append( + [start_pos, end_pos, pii_class.contains_identifier_with_education_info] + ) + result_attributes[f"{model_prefix}_contains_identifier_with_medical_info"].append( + [start_pos, end_pos, pii_class.contains_identifier_with_medical_info] + ) + else: + # For pages we don't process, set all attributes to None + for attr_key in result_attributes: + result_attributes[attr_key].append([start_pos, end_pos, None]) + + return result_attributes + else: + # Document without pdf_page_numbers - sample first N characters + SAMPLE_LENGTH = 1000 + sample_text = text[:SAMPLE_LENGTH] + text_length = len(text) + span_end = min(SAMPLE_LENGTH, text_length) + + async with sem: + pii_class = await _process_single_page(sample_text) + + # Add all attributes from RichPIIClassification + result_attributes[contains_pii_key_name].append([0, span_end, pii_class.contains_any_pii()]) + result_attributes[language_key_name].append([0, span_end, pii_class.primary_language]) + result_attributes[f"{model_prefix}_is_public_document"].append([0, span_end, pii_class.is_public_document]) + result_attributes[f"{model_prefix}_contains_pii_government_id"].append([0, span_end, pii_class.contains_pii_government_id]) + result_attributes[f"{model_prefix}_contains_pii_financial_info"].append([0, span_end, pii_class.contains_pii_financial_info]) + result_attributes[f"{model_prefix}_contains_pii_biometric_data"].append([0, span_end, pii_class.contains_pii_biometric_data]) + result_attributes[f"{model_prefix}_contains_pii_login_info"].append([0, span_end, pii_class.contains_pii_login_info]) + result_attributes[f"{model_prefix}_contains_identifier_name"].append([0, span_end, pii_class.contains_identifier_name]) + result_attributes[f"{model_prefix}_contains_identifier_email"].append([0, span_end, pii_class.contains_identifier_email]) + result_attributes[f"{model_prefix}_contains_identifier_phone_number"].append([0, span_end, pii_class.contains_identifier_phone_number]) + result_attributes[f"{model_prefix}_contains_identifier_with_address"].append([0, span_end, pii_class.contains_identifier_with_address]) + result_attributes[f"{model_prefix}_contains_identifier_with_biographical_info"].append( + [0, span_end, pii_class.contains_identifier_with_biographical_info] + ) + result_attributes[f"{model_prefix}_contains_identifier_with_location_info"].append([0, span_end, pii_class.contains_identifier_with_location_info]) + result_attributes[f"{model_prefix}_contains_identifier_with_employment_info"].append([0, span_end, pii_class.contains_identifier_with_employment_info]) + result_attributes[f"{model_prefix}_contains_identifier_with_education_info"].append([0, span_end, pii_class.contains_identifier_with_education_info]) + result_attributes[f"{model_prefix}_contains_identifier_with_medical_info"].append([0, span_end, pii_class.contains_identifier_with_medical_info]) + + # If the document is longer than the sample, add null spans for the rest + if text_length > SAMPLE_LENGTH: + for attr_key in result_attributes: + result_attributes[attr_key].append([span_end, text_length, None]) + + return result_attributes + + +async def process_file(args, worker_id: int, file_uri: str): + """ + Download a JSONL file, query model per record, and collect attributes. + """ + # Fetch raw bytes (S3 or local) + if file_uri.startswith("s3://"): + raw = await asyncio.to_thread(get_s3_bytes_with_backoff, dataset_s3, file_uri) + else: + with open(file_uri, "rb") as f: + raw = f.read() + + # Decompress if needed + if file_uri.endswith(".gz"): + file_bytes = gzip.decompress(raw) + elif file_uri.endswith(".ztd") or file_uri.endswith(".zst") or file_uri.endswith(".zstd"): + dctx = zstd.ZstdDecompressor() + file_bytes = dctx.stream_reader(raw).read() + else: + file_bytes = raw + + lines = [ln for ln in file_bytes.decode("utf-8").split("\n") if ln] + page_tasks = {} + + # Send all records in parallel, max N queued at a time + sem = asyncio.Semaphore(args.parallel_requests) + + async with asyncio.TaskGroup() as tg: + for line in lines: + dolma_doc = json.loads(line) + task = tg.create_task(process_dolma_document(args, dolma_doc, sem)) + page_tasks[dolma_doc["id"]] = (task, dolma_doc) + + logger.info(f"Finished taskgroup with {len(page_tasks)} items for {file_uri}") + + # Collect results and build attributes + attributes = [] + for doc_id, (task, dolma_doc) in page_tasks.items(): + doc_attributes = task.result() + + attributes.append({"id": doc_id, "attributes": doc_attributes}) + + return attributes + + +async def worker(args, work_queue: WorkQueue, semaphore: asyncio.Semaphore, worker_id: int): + """ + Pop work-items off the queue, run PII tagging, write the attributes file + next to the dataset (keeping the original compression), mark the item done, + and drop an empty sentinel file in /results/. + """ + while True: + await semaphore.acquire() + work_item = await work_queue.get_work() + + if work_item is None: + logger.info(f"Worker {worker_id} exiting – queue empty") + semaphore.release() + break + + file_uri = work_item.work_paths[0] + logger.info(f"Worker {worker_id} processing {file_uri}") + + try: + # ------------------------------------------------------------------ + # Run the per-file pipeline + # ------------------------------------------------------------------ + attributes = await process_file(args, worker_id, file_uri) + + # 1. Build the relative path that mirrors documents/… + if file_uri.startswith("s3://"): + _, key = parse_s3_path(file_uri) + _, docs_prefix = parse_s3_path(args.dataset) + rel_path = key[len(os.path.join(docs_prefix, "documents/")) :] + else: + docs_root = os.path.join(args.dataset, "documents") + rel_path = os.path.relpath(file_uri, docs_root) + + out_rel = os.path.join("attributes", args.attribute_name, rel_path) + out_jsonl = "\n".join(json.dumps(x) for x in attributes) + "\n" + + # 2. Preserve compression type + if rel_path.endswith(".gz"): + payload = gzip.compress(out_jsonl.encode("utf-8")) + elif rel_path.endswith((".zst", ".ztd")): + payload = zstd.ZstdCompressor().compress(out_jsonl.encode("utf-8")) + else: + payload = out_jsonl.encode("utf-8") + + # 3. Write to args.dataset (local or S3) + if args.dataset.startswith("s3://"): + bucket, prefix = parse_s3_path(args.dataset) + key = os.path.join(prefix, out_rel) + workspace_s3.put_object(Bucket=bucket, Key=key, Body=payload) + else: + out_path = os.path.join(args.dataset, out_rel) + os.makedirs(os.path.dirname(out_path), exist_ok=True) + with open(out_path, "wb") as fh: + fh.write(payload) + + # 4. Mark queue item done + await work_queue.mark_done(work_item) + + # 5. Drop empty sentinel file in /results/ + sentinel_rel = os.path.join("results", f"output_{work_item.hash}.jsonl") + if args.scratch.startswith("s3://"): + bkt, pfx = parse_s3_path(args.scratch) + key = os.path.join(pfx, sentinel_rel) + workspace_s3.put_object(Bucket=bkt, Key=key, Body=b"") + else: + sentinel_path = os.path.join(args.scratch, sentinel_rel) + os.makedirs(os.path.dirname(sentinel_path), exist_ok=True) + open(sentinel_path, "w").close() + + except Exception as exc: + logger.exception(f"Worker {worker_id} exception: {exc!s}") + finally: + semaphore.release() + + +async def server_task(model_name_or_path, args, semaphore): + # Check GPU memory, lower mem devices need a bit less KV cache space because the VLM takes additional memory + # mem_fraction_arg = ["--mem-fraction-static", "0.80"] + + cmd = [ + "vllm", + "serve", + model_name_or_path, + "--port", + str(SERVER_PORT), + "--uvicorn-log-level", + "warning", + "--disable-log-requests", + "--max-model-len", + "16000", + # "--hf_overrides", "{\"architectures\": [\"Gemma3ForCausalLM\"]}", + "--limit-mm-per-prompt", + '{"image":0, "video": 0}', # Disabling video encoder saves RAM that you can put towards the KV cache, thanks @charitarthchugh + ] + + proc = await asyncio.create_subprocess_exec( + *cmd, + stdout=asyncio.subprocess.PIPE, + stderr=asyncio.subprocess.PIPE, + ) + + # Ensure the subprocess is terminated on exit + def _kill_proc(): + proc.terminate() + + atexit.register(_kill_proc) + + # Shared variables between tasks + last_running_req, last_queue_req = 0, 0 + server_printed_ready_message = False + last_semaphore_release = time.time() + + async def process_line(line): + nonlocal last_running_req, last_queue_req, last_semaphore_release, server_printed_ready_message + server_logger.info(line) + + # if the server hasn't initialized yet, log all the lines to the main logger also, so that the user + # can see any warnings/errors more easily + if not server_printed_ready_message: + logger.info(line) + + if not server_printed_ready_message and "The server is fired up and ready to roll!" in line: + server_printed_ready_message = True + last_semaphore_release = time.time() + + match = re.search(r"Running: (\d+) reqs", line) + if match: + last_running_req = int(match.group(1)) + + match = re.search(r"Waiting: (\d+) reqs", line) + if match: + last_queue_req = int(match.group(1)) + logger.info(f"running req: {last_running_req} queue req: {last_queue_req}") + + async def read_stream(stream): + while True: + line = await stream.readline() + if not line: + break + try: + line = line.decode("utf-8").rstrip() + await process_line(line) + except Exception as ex: + logger.warning(f"Got {ex} when reading log line from inference server, skipping") + + async def timeout_handler(): + nonlocal last_running_req, last_queue_req, last_semaphore_release + try: + while True: + await asyncio.sleep(1) + if server_printed_ready_message and last_queue_req == 0 and time.time() - last_semaphore_release > 30 and semaphore.locked(): + semaphore.release() + last_semaphore_release = time.time() + logger.info("Semaphore released, allowing a worker to proceed.") + except asyncio.CancelledError: + pass # Clean up if the task is cancelled + + # Start tasks to read stdout, stderr, and handle timeout logic + stdout_task = asyncio.create_task(read_stream(proc.stdout)) + stderr_task = asyncio.create_task(read_stream(proc.stderr)) + timeout_task = asyncio.create_task(timeout_handler()) + + try: + await proc.wait() + except asyncio.CancelledError: + logger.info("Got cancellation request for server") + proc.terminate() + raise + + timeout_task.cancel() + await asyncio.gather(stdout_task, stderr_task, timeout_task, return_exceptions=True) + + +async def server_host(model_name_or_path, args, semaphore): + MAX_RETRIES = 5 + retry = 0 + + while retry < MAX_RETRIES: + await server_task(model_name_or_path, args, semaphore) + logger.warning("Server task ended") + retry += 1 + + if retry >= MAX_RETRIES: + logger.error(f"Ended up starting the server more than {retry} times, cancelling pipeline") + logger.error("") + logger.error("Please make sure vllm is installed according to the latest instructions for 0.8.4") + sys.exit(1) + + +async def check_server_ready(): + max_attempts = 600 + delay_sec = 1 + url = f"http://localhost:{SERVER_PORT}/v1/models" + + for attempt in range(1, max_attempts + 1): + try: + async with httpx.AsyncClient() as session: + response = await session.get(url) + + if response.status_code == 200: + logger.info("server is ready.") + return + else: + logger.info(f"Attempt {attempt}: Unexpected status code {response.status_code}") + except Exception: + logger.warning(f"Attempt {attempt}: Please wait for model server to become ready...") + + await asyncio.sleep(delay_sec) + + raise Exception("model server did not become ready after waiting.") + + +async def download_model(model_name_or_path: str): + if model_name_or_path.startswith("s3://") or model_name_or_path.startswith("gs://") or model_name_or_path.startswith("weka://"): + logger.info(f"Downloading model directory from '{model_name_or_path}'") + model_cache_dir = os.path.join(os.path.expanduser("~"), ".cache", "olmocr", "model") + download_directory([model_name_or_path], model_cache_dir) + return model_cache_dir + elif os.path.isabs(model_name_or_path) and os.path.isdir(model_name_or_path): + logger.info(f"Using local model path at '{model_name_or_path}'") + return model_name_or_path + else: + logger.info(f"Downloading model with hugging face '{model_name_or_path}'") + snapshot_download(repo_id=model_name_or_path) + return model_name_or_path + + +async def metrics_reporter(work_queue): + while True: + # Leading newlines preserve table formatting in logs + logger.info(f"Queue remaining: {work_queue.size}") + logger.info("\n" + str(metrics)) + await asyncio.sleep(10) + + +def submit_beaker_job(args): + from beaker import ( # type: ignore + Beaker, + BeakerConstraints, + BeakerEnvVar, + BeakerExperimentSpec, + BeakerImageSource, + BeakerJobPriority, + BeakerResultSpec, + BeakerRetrySpec, + BeakerTaskContext, + BeakerTaskResources, + BeakerTaskSpec, + ) + from beaker.exceptions import BeakerSecretNotFound + + Beaker.TIMEOUT = 60 + b = Beaker.from_env(default_workspace=args.beaker_workspace) + owner = b.user_name + beaker_image = f"jakep/olmocr-inference-{VERSION}" + + task_name = f"olmocr-richtagging-{os.path.basename(args.scratch.rstrip('/'))}" + + # Take out --beaker flag so the workers will just run things + args_list = [arg for arg in sys.argv[1:] if arg != "--beaker"] + + # Take out the --pdfs [arg] or --pdfs=[arg], since the queue is populated locally + args_list = [arg for i, arg in enumerate(args_list) if not (arg.startswith("--pdfs") or (i > 0 and args_list[i - 1] == "--pdfs"))] + + try: + b.secret.get(f"{owner}-WEKA_ACCESS_KEY_ID") + b.secret.get(f"{owner}-WEKA_SECRET_ACCESS_KEY") + b.secret.get(f"{owner}-AWS_CREDENTIALS_FILE") + except BeakerSecretNotFound: + print( + f"Expected beaker secrets for accessing Weka and S3 are not found. Are you okay to write those to your beaker workspace {args.beaker_workspace}? [y/n]" + ) + + if input().strip().lower() != "y": + print("Exiting...") + sys.exit(1) + + b.secret.write(f"{owner}-WEKA_ACCESS_KEY_ID", os.environ.get("WEKA_ACCESS_KEY_ID", "")) + b.secret.write(f"{owner}-WEKA_SECRET_ACCESS_KEY", os.environ.get("WEKA_SECRET_ACCESS_KEY", "")) + b.secret.write( + f"{owner}-AWS_CREDENTIALS_FILE", + open(os.path.join(os.path.expanduser("~"), ".aws", "credentials")).read(), + ) + + env_var_secrets = [ + BeakerEnvVar(name="WEKA_ACCESS_KEY_ID", secret=f"{owner}-WEKA_ACCESS_KEY_ID"), + BeakerEnvVar(name="WEKA_SECRET_ACCESS_KEY", secret=f"{owner}-WEKA_SECRET_ACCESS_KEY"), + BeakerEnvVar(name="AWS_CREDENTIALS_FILE", secret=f"{owner}-AWS_CREDENTIALS_FILE"), + ] + + try: + b.secret.get("OLMOCR_PREVIEW_HF_TOKEN") + env_var_secrets.append(BeakerEnvVar(name="HF_TOKEN", secret="OLMOCR_PREVIEW_HF_TOKEN")) + except BeakerSecretNotFound: + pass + + try: + b.secret.get("OE_DATA_GCS_SA_KEY") + env_var_secrets.append(BeakerEnvVar(name="GOOGLE_APPLICATION_CREDENTIALS_FILE", secret="OE_DATA_GCS_SA_KEY")) + except BeakerSecretNotFound: + print("Input the olmo-gcs SA key if you would like to load weights from gcs (end with a double newline):") + lines = [] + prev_empty = False + for line in iter(input, None): + if not line and prev_empty: + break + prev_empty = not line + lines.append(line) + gcs_sa_key = "\n".join(lines[:-1]).strip() # Remove the last empty line + if gcs_sa_key: + b.secret.write("OE_DATA_GCS_SA_KEY", gcs_sa_key) + env_var_secrets.append(BeakerEnvVar(name="GOOGLE_APPLICATION_CREDENTIALS_FILE", secret="OE_DATA_GCS_SA_KEY")) + + # Create the experiment spec + experiment_spec = BeakerExperimentSpec( + budget="ai2/oe-base", + description=task_name, + tasks=[ + BeakerTaskSpec( + name=task_name, + propagate_failure=False, + propagate_preemption=False, + replicas=args.beaker_gpus, + context=BeakerTaskContext( + priority=BeakerJobPriority[args.beaker_priority], + preemptible=True, + ), + image=BeakerImageSource(beaker=beaker_image), + command=["python", "scripts/pii/rich_tagging_pipeline.py"] + args_list, + env_vars=[ + BeakerEnvVar(name="BEAKER_JOB_NAME", value=task_name), + BeakerEnvVar(name="OWNER", value=owner), + ] + + env_var_secrets, + resources=BeakerTaskResources(gpu_count=1, memory="125GB"), + constraints=BeakerConstraints(cluster=args.beaker_cluster if isinstance(args.beaker_cluster, list) else [args.beaker_cluster]), + result=BeakerResultSpec(path="/noop-results"), + ) + ], + retry=BeakerRetrySpec(allowed_task_retries=10), + ) + + workload = b.experiment.create(spec=experiment_spec) + + print(f"Experiment URL: https://beaker.org/ex/{workload.experiment.id}") + + +async def main(): + parser = argparse.ArgumentParser(description="Tagging pipeline for Dolma JSONL dataset") + parser.add_argument("dataset", help="Dolma dataset root (local or s3://) with documents/ folder") + parser.add_argument("scratch", help="Scratch workspace (local dir or s3://)") + parser.add_argument("--workers", type=int, default=4, help="Number of concurrent workers") + parser.add_argument("--parallel_requests", type=int, default=800, help="Max number of parallel requests to send to model") + parser.add_argument("--model", default="google/gemma-3-12b-it", help="Model path or name, hugging face or local path format") + parser.add_argument("--attribute_name", default="model_rich_pii_tagging", help="Path to use for attribute naming") + + # Beaker/job running stuff + parser.add_argument("--beaker", action="store_true", help="Submit this job to beaker instead of running locally") + parser.add_argument("--beaker_workspace", help="Beaker workspace to submit to", default="ai2/olmocr") + parser.add_argument( + "--beaker_cluster", + help="Beaker clusters you want to run on", + default=["ai2/jupiter", "ai2/ceres", "ai2/neptune", "ai2/saturn"], + ) + parser.add_argument("--beaker_gpus", type=int, default=1, help="Number of gpu replicas to run") + parser.add_argument("--beaker_priority", type=str, default="normal", help="Beaker priority level for the job") + + parser.add_argument("--port", type=int, default=30024, help="Port for Model server") + args = parser.parse_args() + + global SERVER_PORT, workspace_s3, dataset_s3 + SERVER_PORT = args.port + workspace_s3 = boto3.client("s3") + dataset_s3 = boto3.client("s3") + + # setup the job to work in beaker environment, load secrets, adjust logging, etc. + if "BEAKER_JOB_NAME" in os.environ: + cred_path = os.path.join(os.path.expanduser("~"), ".aws", "credentials") + os.makedirs(os.path.dirname(cred_path), exist_ok=True) + with open(cred_path, "w") as f: + f.write(os.environ.get("AWS_CREDENTIALS_FILE", "")) + cred_path = os.path.join(os.path.expanduser("~"), ".gcs", "credentials") + os.makedirs(os.path.dirname(cred_path), exist_ok=True) + with open(cred_path, "w") as f: + f.write(os.environ.get("GOOGLE_APPLICATION_CREDENTIALS_FILE", "")) + os.environ["GOOGLE_APPLICATION_CREDENTIALS"] = cred_path + workspace_s3 = boto3.client("s3") + dataset_s3 = boto3.client("s3") + + # Wait a little bit so that not all beaker jobs in a task start at the same time and download the model at the same time + replica_count = int(os.environ.get("BEAKER_REPLICA_COUNT", "1")) + interval = 10 if (replica_count - 1) * 10 <= 30 else 30 / max(1, replica_count - 1) + sleep_time = int(os.environ.get("BEAKER_REPLICA_RANK", "0")) * interval + logger.info(f"Beaker job sleeping for {sleep_time} seconds to stagger model downloads") + await asyncio.sleep(sleep_time) + + # Initialize work queue + if args.scratch.startswith("s3://"): + work_queue = WorkQueue(S3Backend(workspace_s3, args.scratch)) + else: + work_queue = WorkQueue(LocalBackend(args.scratch)) + + # Discover input files + files = set() + if args.dataset.startswith("s3://"): + pattern = args.dataset.rstrip("/") + "/documents/*.jsonl*" + matched = expand_s3_glob(dataset_s3, pattern) + files = set(matched.keys()) + else: + docs_dir = os.path.join(args.dataset, "documents") + for root, _, fns in os.walk(docs_dir): + for fn in fns: + if fn.endswith((".jsonl", ".jsonl.gz", ".jsonl.ztd")): + files.add(os.path.join(root, fn)) + + # Populate the work queue if needed + await work_queue.populate_queue(list(files), items_per_group=1) + + if args.beaker: + submit_beaker_job(args) + return + + # If you get this far, then you are doing inference and need a GPU + check_torch_gpu_available() + + logger.info(f"Starting pipeline with PID {os.getpid()}") + + # Download the model before you do anything else + model_name_or_path = await download_model(args.model) + + # Initialize the work queue + qsize = await work_queue.initialize_queue() + + if qsize == 0: + logger.info("No work to do, exiting") + return + + # Create a semaphore to control worker access + # We only allow one worker to move forward with requests, until the server has no more requests in its queue + # This lets us get full utilization by having many workers, but also to be outputting dolma docs as soon as possible + # As soon as one worker is no longer saturating the gpu, the next one can start sending requests + semaphore = asyncio.Semaphore(1) + + model_server = asyncio.create_task(server_host(model_name_or_path, args, semaphore)) + + await check_server_ready() + + metrics_task = asyncio.create_task(metrics_reporter(work_queue)) + + # Create worker tasks to process the queue concurrently. + worker_tasks = [] + for i in range(args.workers): + task = asyncio.create_task(worker(args, work_queue, semaphore, worker_id=i)) + worker_tasks.append(task) + + # Wait for all worker tasks to finish + await asyncio.gather(*worker_tasks) + + model_server.cancel() + metrics_task.cancel() + logger.info("Work done") + + +if __name__ == "__main__": + asyncio.run(main()) diff --git a/scripts/pii/run_tagging_pipeline.sh b/scripts/pii/run_tagging_pipeline.sh new file mode 100644 index 0000000..d42779d --- /dev/null +++ b/scripts/pii/run_tagging_pipeline.sh @@ -0,0 +1,7 @@ +#!/bin/bash + +set -e + +gantry run --gpus 1 --workspace ai2/olmocr --beaker-image ai2/pytorch2.5.1-cuda12.1-python3.11 --cluster ai2/jupiter-cirrascale-2 --budget ai2/oe-base --priority normal --env-secret AWS_CREDENTIALS_FILE=jakep-AWS_CREDENTIALS_FILE --env-secret HF_TOKEN=jake-HF_TOKEN --allow-dirty -- /bin/bash -c "pip install -e .[gpu] --find-links https://flashinfer.ai/whl/cu124/torch2.4/flashinfer/ && pip install --upgrade sglang==0.4.5.post3 transformers==4.51.3 && python scripts/tagging_pipeline.py s3://ai2-oe-data/jakep/s2pdf_dedupe_minhash_v1_mini s3://ai2-oe-data/jakep/s2pdf_dedupe_minhash_v1_mini_scratch" + +gantry run --gpus 1 --workspace ai2/olmocr --beaker-image ai2/pytorch2.5.1-cuda12.1-python3.11 --cluster ai2/jupiter-cirrascale-2 --budget ai2/oe-base --priority normal --env-secret AWS_CREDENTIALS_FILE=jakep-AWS_CREDENTIALS_FILE --env-secret HF_TOKEN=jake-HF_TOKEN --allow-dirty -- /bin/bash -c "pip install -e .[gpu,bench] --find-links https://flashinfer.ai/whl/cu124/torch2.4/flashinfer/ && huggingface-cli download allenai/olmOCR-bench --repo-type dataset --local-dir ./olmOCR-bench && olmocr/bench/scripts/convert_all.sh" \ No newline at end of file diff --git a/scripts/pii/tagging_pipeline.py b/scripts/pii/tagging_pipeline.py new file mode 100644 index 0000000..b9b0c3e --- /dev/null +++ b/scripts/pii/tagging_pipeline.py @@ -0,0 +1,788 @@ +#!/usr/bin/env python3 +""" +Tagging pipeline for Dolma JSONL datasets. + +For each .jsonl, .jsonl.gz, or .jsonl.ztd file under the dataset/documents folder, +this script issues a model prompt completion +collects the yes/no answers, and writes corresponding Dolma attributes JSONL files under +scratch/attributes/, mirroring the input structure. +""" + +import argparse +import asyncio +import atexit +import gzip +import json +import logging +import os +import re +import sys +import time +from typing import Optional +from urllib.parse import urlparse + +import boto3 +import httpx +import zstandard as zstd +from huggingface_hub import snapshot_download +from pydantic import BaseModel, Field, ValidationError + +from olmocr.check import check_torch_gpu_available +from olmocr.metrics import MetricsKeeper +from olmocr.s3_utils import ( + download_directory, + expand_s3_glob, + get_s3_bytes_with_backoff, + parse_s3_path, +) +from olmocr.version import VERSION +from olmocr.work_queue import LocalBackend, S3Backend, WorkQueue + +# Initialize logger +logger = logging.getLogger(__name__) +logger.setLevel(logging.DEBUG) +logger.propagate = False + +server_logger = logging.getLogger("vllm") +server_logger.propagate = False + +file_handler = logging.FileHandler("olmocr-pipeline-debug.log", mode="a") +file_handler.setLevel(logging.DEBUG) +file_handler.setFormatter(logging.Formatter("%(asctime)s - %(name)s - %(levelname)s - %(message)s")) + +console_handler = logging.StreamHandler() +console_handler.setLevel(logging.INFO) +console_handler.setFormatter(logging.Formatter("%(asctime)s - %(name)s - %(levelname)s - %(message)s")) + +# Add handlers to the logger +logger.addHandler(file_handler) +logger.addHandler(console_handler) +server_logger.addHandler(file_handler) + + +# Default port; overridden by --port +SERVER_PORT = 30024 + +# Global variables for token statistics +metrics = MetricsKeeper(window=60 * 5) + + +class PIIClassification(BaseModel): + primary_language: str = Field(..., description="Primary language as a two-letter code") + document_type: str = Field(..., description="Basic summary of document type classification") + is_resume_cv: Optional[bool] = Field(None, description="True if the document is a page from a resume or cv") + + is_academic_paper: Optional[bool] = None + is_textbook: Optional[bool] = None + is_news_article: Optional[bool] = None + is_test_or_quiz: Optional[bool] = None + is_homework_assignment: Optional[bool] = None + is_class_syllabus: Optional[bool] = None + is_meeting_minutes: Optional[bool] = None + is_legal_contract: Optional[bool] = None + is_form: Optional[bool] = None + is_correspondence_or_letter: Optional[bool] = None + is_public_order: Optional[bool] = None + is_court_notice: Optional[bool] = None + is_completion_certificate: Optional[bool] = None + + contains_pii: Optional[bool] = Field(None, description="True if document contains PII") + + +async def _process_single_page(page_text: str) -> PIIClassification: + """Helper function to process a single document or page.""" + text = page_text + + query = { + "model": "google/gemma-3-4b-it", + "messages": [ + { + "role": "user", + "content": [ + { + "type": "text", + "text": ( + f"{text}\n\n-----------\n" + "Given the text above, determine what type of document it is. Answer in JSON. The format of your json object should be {'primary_language': str, 'document_type': str, 'is_resume_cv': bool, 'is_academic_paper': bool, 'is_textbook': bool, 'is_news_article': bool, 'is_test_or_quiz': bool, 'is_homework_assignment': bool, 'is_class_syllabus': bool, 'is_meeting_minutes': bool, 'is_legal_contract': bool, 'is_form': bool, 'is_correspondence_or_letter': bool, 'is_public_order': bool, 'is_court_notice': bool, 'is_completion_certificate': bool, 'contains_pii': bool}" + ), + } + ], + } + ], + "max_tokens": 400, + "temperature": 0.0, + "response_format": {"type": "json_schema", "json_schema": {"name": "PIIClassification", "schema": PIIClassification.model_json_schema()}}, + } + + url = f"http://localhost:{SERVER_PORT}/v1/chat/completions" + + # ---------- HTTP call --------------------------------------------------- + try: + status, body = await apost(url, json_data=query) + except Exception as e: + logger.warning(f"Server network error: {e!s}") + metrics.add_metrics(server_errors=1) + return PIIClassification(primary_language="en", document_type="unknown", is_resume_cv=None, contains_pii=None) + + metrics.add_metrics(server_requests=1) + + if status != 200: + logger.warning(f"Server HTTP {status}: {body[:250]!r}") + metrics.add_metrics(server_errors=1) + return PIIClassification(primary_language="en", document_type="unknown", is_resume_cv=None, contains_pii=None) + + # ---------- Parse base JSON -------------------------------------------- + try: + base = json.loads(body) + except json.JSONDecodeError: + logger.warning(f"Server response is not valid JSON: {body[:250]!r}") + metrics.add_metrics(server_errors=1) + return PIIClassification(primary_language="en", document_type="unknown", is_resume_cv=None, contains_pii=None) + + # Token accounting if available + if "usage" in base: + metrics.add_metrics( + server_input_tokens=base["usage"].get("prompt_tokens", 0), + server_output_tokens=base["usage"].get("completion_tokens", 0), + ) + + # ---------- Extract the model message ---------------------------------- + try: + content = base["choices"][0]["message"].get("content") + except (KeyError, IndexError, AttributeError) as e: + logger.warning(f"Missing fields in Server response: {e!s}") + metrics.add_metrics(server_errors=1) + return PIIClassification(primary_language="en", document_type="unknown", is_resume_cv=None, contains_pii=None) + + if not isinstance(content, str): + logger.warning("Server `content` is not a string; treating as error.") + metrics.add_metrics(server_errors=1) + return PIIClassification(primary_language="en", document_type="unknown", is_resume_cv=None, contains_pii=None) + + try: + pii_classification: PIIClassification = PIIClassification.model_validate_json(content) + return pii_classification + except ValidationError as e: + logger.warning(f"Unable to parse pii classification object: {e!s}") + metrics.add_metrics(server_errors=1) + return PIIClassification(primary_language="en", document_type="unknown", is_resume_cv=None, contains_pii=None) + + +# Manual simple implementation of HTTP Post +# It feels strange perhaps, but httpx and aiohttp are very complex beasts +# Ex. the sessionpool in httpcore has 4 different locks in it, and I've noticed +# that at the scale of 100M+ requests, that they deadlock in different strange ways +async def apost(url, json_data): + parsed_url = urlparse(url) + host = parsed_url.hostname + port = parsed_url.port or 80 + path = parsed_url.path or "/" + + writer = None + try: + reader, writer = await asyncio.open_connection(host, port) + + json_payload = json.dumps(json_data) + request = ( + f"POST {path} HTTP/1.1\r\n" + f"Host: {host}\r\n" + f"Content-Type: application/json\r\n" + f"Content-Length: {len(json_payload)}\r\n" + f"Connection: close\r\n\r\n" + f"{json_payload}" + ) + writer.write(request.encode()) + await writer.drain() + + # Read status line + status_line = await reader.readline() + if not status_line: + raise ConnectionError("No response from server") + status_parts = status_line.decode().strip().split(" ", 2) + if len(status_parts) < 2: + raise ValueError(f"Malformed status line: {status_line.decode().strip()}") + status_code = int(status_parts[1]) + + # Read headers + headers = {} + while True: + line = await reader.readline() + if line in (b"\r\n", b"\n", b""): + break + key, _, value = line.decode().partition(":") + headers[key.strip().lower()] = value.strip() + + # Read response body + if "content-length" in headers: + body_length = int(headers["content-length"]) + response_body = await reader.readexactly(body_length) + else: + raise ConnectionError("Anything other than fixed content length responses are not implemented yet") + + return status_code, response_body + except Exception as e: + # Pass through errors + raise e + finally: + # But just make sure to close the socket on your way out + if writer is not None: + try: + writer.close() + await writer.wait_closed() + except: + pass + + +async def process_dolma_document(args, dolma_doc, sem): + """ + Query model to detect PII, enforcing a JSON schema. + + Resilient to: + • Transport / HTTP errors + • Missing or malformed fields in the response + • Non-string or None `content` + • Bad JSON in the model's answer + + Always returns: (doc_id, contains_pii: bool, text_length: int) + """ + text = dolma_doc.get("text", "") or "" + + # Create keys for all fields in PIIClassification + prefix = args.model.replace("/", "_") + "_v2tag_" + result_attributes = {} + + # Initialize attribute lists for all PIIClassification fields + for field_name in PIIClassification.model_fields: + key_name = f"{prefix}_{field_name}" + result_attributes[key_name] = [] + + # Take first 5000 characters of the document for classification + sample_text = text[:5000] + text_length = len(text) + span_end = min(5000, text_length) + + # Process the sample with the semaphore to limit concurrent requests + async with sem: + pii_class = await _process_single_page(sample_text) + + # Add all classification attributes to results + for field_name in PIIClassification.model_fields: + key_name = f"{prefix}_{field_name}" + attribute_value = getattr(pii_class, field_name) + + # Create a span from 0 to min(5000, len(text)) with the attribute value + result_attributes[key_name].append([0, span_end, attribute_value]) + + # If the document is longer than 5000 characters, add a null span for the rest + if text_length > 5000: + result_attributes[key_name].append([span_end, text_length, None]) + + return result_attributes + + +async def process_file(args, worker_id: int, file_uri: str): + """ + Download a JSONL file, query model per record, and collect attributes. + """ + # Fetch raw bytes (S3 or local) + if file_uri.startswith("s3://"): + raw = await asyncio.to_thread(get_s3_bytes_with_backoff, dataset_s3, file_uri) + else: + with open(file_uri, "rb") as f: + raw = f.read() + + # Decompress if needed + if file_uri.endswith(".gz"): + file_bytes = gzip.decompress(raw) + elif file_uri.endswith(".ztd") or file_uri.endswith(".zst") or file_uri.endswith(".zstd"): + dctx = zstd.ZstdDecompressor() + file_bytes = dctx.stream_reader(raw).read() + else: + file_bytes = raw + + lines = [ln for ln in file_bytes.decode("utf-8").split("\n") if ln] + page_tasks = {} + + # Send all records in parallel, max N queued at a time + sem = asyncio.Semaphore(args.parallel_requests) + + async with asyncio.TaskGroup() as tg: + for line in lines: + dolma_doc = json.loads(line) + task = tg.create_task(process_dolma_document(args, dolma_doc, sem)) + page_tasks[dolma_doc["id"]] = (task, dolma_doc) + + logger.info(f"Finished taskgroup with {len(page_tasks)} items for {file_uri}") + + # Collect results and build attributes + attributes = [] + for doc_id, (task, dolma_doc) in page_tasks.items(): + doc_attributes = task.result() + + attributes.append({"id": doc_id, "attributes": doc_attributes}) + + return attributes + + +async def worker(args, work_queue: WorkQueue, semaphore: asyncio.Semaphore, worker_id: int): + """ + Pop work-items off the queue, run PII tagging, write the attributes file + next to the dataset (keeping the original compression), mark the item done, + and drop an empty sentinel file in /results/. + """ + while True: + await semaphore.acquire() + work_item = await work_queue.get_work() + + if work_item is None: + logger.info(f"Worker {worker_id} exiting – queue empty") + semaphore.release() + break + + file_uri = work_item.work_paths[0] + logger.info(f"Worker {worker_id} processing {file_uri}") + + try: + # ------------------------------------------------------------------ + # Run the per-file pipeline + # ------------------------------------------------------------------ + attributes = await process_file(args, worker_id, file_uri) + + # 1. Build the relative path that mirrors documents/… + if file_uri.startswith("s3://"): + _, key = parse_s3_path(file_uri) + _, docs_prefix = parse_s3_path(args.dataset) + rel_path = key[len(os.path.join(docs_prefix, "documents/")) :] + else: + docs_root = os.path.join(args.dataset, "documents") + rel_path = os.path.relpath(file_uri, docs_root) + + out_rel = os.path.join("attributes", args.attribute_name, rel_path) + out_jsonl = "\n".join(json.dumps(x) for x in attributes) + "\n" + + # 2. Preserve compression type + if rel_path.endswith(".gz"): + payload = gzip.compress(out_jsonl.encode("utf-8")) + elif rel_path.endswith((".zst", ".ztd")): + payload = zstd.ZstdCompressor().compress(out_jsonl.encode("utf-8")) + else: + payload = out_jsonl.encode("utf-8") + + # 3. Write to args.dataset (local or S3) + if args.dataset.startswith("s3://"): + bucket, prefix = parse_s3_path(args.dataset) + key = os.path.join(prefix, out_rel) + workspace_s3.put_object(Bucket=bucket, Key=key, Body=payload) + else: + out_path = os.path.join(args.dataset, out_rel) + os.makedirs(os.path.dirname(out_path), exist_ok=True) + with open(out_path, "wb") as fh: + fh.write(payload) + + # 4. Mark queue item done + await work_queue.mark_done(work_item) + + # 5. Drop empty sentinel file in /results/ + sentinel_rel = os.path.join("results", f"output_{work_item.hash}.jsonl") + if args.scratch.startswith("s3://"): + bkt, pfx = parse_s3_path(args.scratch) + key = os.path.join(pfx, sentinel_rel) + workspace_s3.put_object(Bucket=bkt, Key=key, Body=b"") + else: + sentinel_path = os.path.join(args.scratch, sentinel_rel) + os.makedirs(os.path.dirname(sentinel_path), exist_ok=True) + open(sentinel_path, "w").close() + + except Exception as exc: + logger.exception(f"Worker {worker_id} exception: {exc!s}") + finally: + semaphore.release() + + +async def server_task(model_name_or_path, args, semaphore): + # Check GPU memory, lower mem devices need a bit less KV cache space because the VLM takes additional memory + # mem_fraction_arg = ["--mem-fraction-static", "0.80"] + + cmd = [ + "vllm", + "serve", + model_name_or_path, + "--port", + str(SERVER_PORT), + "--uvicorn-log-level", + "warning", + "--disable-log-requests", + "--limit-mm-per-prompt", + '{"image":0, "video": 0}', # Disabling video encoder saves RAM that you can put towards the KV cache, thanks @charitarthchugh + ] + + proc = await asyncio.create_subprocess_exec( + *cmd, + stdout=asyncio.subprocess.PIPE, + stderr=asyncio.subprocess.PIPE, + ) + + # Ensure the subprocess is terminated on exit + def _kill_proc(): + proc.terminate() + + atexit.register(_kill_proc) + + # Shared variables between tasks + last_running_req, last_queue_req = 0, 0 + server_printed_ready_message = False + last_semaphore_release = time.time() + + async def process_line(line): + nonlocal last_running_req, last_queue_req, last_semaphore_release, server_printed_ready_message + server_logger.info(line) + + # if the server hasn't initialized yet, log all the lines to the main logger also, so that the user + # can see any warnings/errors more easily + if not server_printed_ready_message: + logger.info(line) + + if not server_printed_ready_message and "The server is fired up and ready to roll!" in line: + server_printed_ready_message = True + last_semaphore_release = time.time() + + match = re.search(r"Running: (\d+) reqs", line) + if match: + last_running_req = int(match.group(1)) + + match = re.search(r"Waiting: (\d+) reqs", line) + if match: + last_queue_req = int(match.group(1)) + logger.info(f"running req: {last_running_req} queue req: {last_queue_req}") + + async def read_stream(stream): + while True: + line = await stream.readline() + if not line: + break + try: + line = line.decode("utf-8").rstrip() + await process_line(line) + except Exception as ex: + logger.warning(f"Got {ex} when reading log line from inference server, skipping") + + async def timeout_handler(): + nonlocal last_running_req, last_queue_req, last_semaphore_release + try: + while True: + await asyncio.sleep(1) + if server_printed_ready_message and last_queue_req == 0 and time.time() - last_semaphore_release > 30 and semaphore.locked(): + semaphore.release() + last_semaphore_release = time.time() + logger.info("Semaphore released, allowing a worker to proceed.") + except asyncio.CancelledError: + pass # Clean up if the task is cancelled + + # Start tasks to read stdout, stderr, and handle timeout logic + stdout_task = asyncio.create_task(read_stream(proc.stdout)) + stderr_task = asyncio.create_task(read_stream(proc.stderr)) + timeout_task = asyncio.create_task(timeout_handler()) + + try: + await proc.wait() + except asyncio.CancelledError: + logger.info("Got cancellation request for server") + proc.terminate() + raise + + timeout_task.cancel() + await asyncio.gather(stdout_task, stderr_task, timeout_task, return_exceptions=True) + + +async def server_host(model_name_or_path, args, semaphore): + MAX_RETRIES = 5 + retry = 0 + + while retry < MAX_RETRIES: + await server_task(model_name_or_path, args, semaphore) + logger.warning("Server task ended") + retry += 1 + + if retry >= MAX_RETRIES: + logger.error(f"Ended up starting the server more than {retry} times, cancelling pipeline") + logger.error("") + logger.error("Please make sure vllm is installed according to the latest instructions for 0.8.4") + sys.exit(1) + + +async def check_server_ready(): + max_attempts = 300 + delay_sec = 1 + url = f"http://localhost:{SERVER_PORT}/v1/models" + + for attempt in range(1, max_attempts + 1): + try: + async with httpx.AsyncClient() as session: + response = await session.get(url) + + if response.status_code == 200: + logger.info("server is ready.") + return + else: + logger.info(f"Attempt {attempt}: Unexpected status code {response.status_code}") + except Exception: + logger.warning(f"Attempt {attempt}: Please wait for model server to become ready...") + + await asyncio.sleep(delay_sec) + + raise Exception("model server did not become ready after waiting.") + + +async def download_model(model_name_or_path: str): + if model_name_or_path.startswith("s3://") or model_name_or_path.startswith("gs://") or model_name_or_path.startswith("weka://"): + logger.info(f"Downloading model directory from '{model_name_or_path}'") + model_cache_dir = os.path.join(os.path.expanduser("~"), ".cache", "olmocr", "model") + download_directory([model_name_or_path], model_cache_dir) + return model_cache_dir + elif os.path.isabs(model_name_or_path) and os.path.isdir(model_name_or_path): + logger.info(f"Using local model path at '{model_name_or_path}'") + return model_name_or_path + else: + logger.info(f"Downloading model with hugging face '{model_name_or_path}'") + snapshot_download(repo_id=model_name_or_path) + return model_name_or_path + + +async def metrics_reporter(work_queue): + while True: + # Leading newlines preserve table formatting in logs + logger.info(f"Queue remaining: {work_queue.size}") + logger.info("\n" + str(metrics)) + await asyncio.sleep(10) + + +def submit_beaker_job(args): + from beaker import ( # type: ignore + Beaker, + BeakerConstraints, + BeakerEnvVar, + BeakerExperimentSpec, + BeakerImageSource, + BeakerJobPriority, + BeakerResultSpec, + BeakerRetrySpec, + BeakerTaskContext, + BeakerTaskResources, + BeakerTaskSpec, + ) + from beaker.exceptions import BeakerSecretNotFound + + Beaker.TIMEOUT = 60 + b = Beaker.from_env(default_workspace=args.beaker_workspace) + owner = b.user_name + beaker_image = f"jakep/olmocr-inference-{VERSION}" + + task_name = f"olmocr-tagging-{os.path.basename(args.scratch.rstrip('/'))}" + + # Take out --beaker flag so the workers will just run things + args_list = [arg for arg in sys.argv[1:] if arg != "--beaker"] + + # Take out the --pdfs [arg] or --pdfs=[arg], since the queue is populated locally + args_list = [arg for i, arg in enumerate(args_list) if not (arg.startswith("--pdfs") or (i > 0 and args_list[i - 1] == "--pdfs"))] + + try: + b.secret.get(f"{owner}-WEKA_ACCESS_KEY_ID") + b.secret.get(f"{owner}-WEKA_SECRET_ACCESS_KEY") + b.secret.get(f"{owner}-AWS_CREDENTIALS_FILE") + except BeakerSecretNotFound: + print( + f"Expected beaker secrets for accessing Weka and S3 are not found. Are you okay to write those to your beaker workspace {args.beaker_workspace}? [y/n]" + ) + + if input().strip().lower() != "y": + print("Exiting...") + sys.exit(1) + + b.secret.write(f"{owner}-WEKA_ACCESS_KEY_ID", os.environ.get("WEKA_ACCESS_KEY_ID", "")) + b.secret.write(f"{owner}-WEKA_SECRET_ACCESS_KEY", os.environ.get("WEKA_SECRET_ACCESS_KEY", "")) + b.secret.write( + f"{owner}-AWS_CREDENTIALS_FILE", + open(os.path.join(os.path.expanduser("~"), ".aws", "credentials")).read(), + ) + + env_var_secrets = [ + BeakerEnvVar(name="WEKA_ACCESS_KEY_ID", secret=f"{owner}-WEKA_ACCESS_KEY_ID"), + BeakerEnvVar(name="WEKA_SECRET_ACCESS_KEY", secret=f"{owner}-WEKA_SECRET_ACCESS_KEY"), + BeakerEnvVar(name="AWS_CREDENTIALS_FILE", secret=f"{owner}-AWS_CREDENTIALS_FILE"), + ] + + try: + b.secret.get("OLMOCR_PREVIEW_HF_TOKEN") + env_var_secrets.append(BeakerEnvVar(name="HF_TOKEN", secret="OLMOCR_PREVIEW_HF_TOKEN")) + except BeakerSecretNotFound: + pass + + try: + b.secret.get("OE_DATA_GCS_SA_KEY") + env_var_secrets.append(BeakerEnvVar(name="GOOGLE_APPLICATION_CREDENTIALS_FILE", secret="OE_DATA_GCS_SA_KEY")) + except BeakerSecretNotFound: + print("Input the olmo-gcs SA key if you would like to load weights from gcs (end with a double newline):") + lines = [] + prev_empty = False + for line in iter(input, None): + if not line and prev_empty: + break + prev_empty = not line + lines.append(line) + gcs_sa_key = "\n".join(lines[:-1]).strip() # Remove the last empty line + if gcs_sa_key: + b.secret.write("OE_DATA_GCS_SA_KEY", gcs_sa_key) + env_var_secrets.append(BeakerEnvVar(name="GOOGLE_APPLICATION_CREDENTIALS_FILE", secret="OE_DATA_GCS_SA_KEY")) + + # Create the experiment spec + experiment_spec = BeakerExperimentSpec( + budget="ai2/oe-base", + description=task_name, + tasks=[ + BeakerTaskSpec( + name=task_name, + propagate_failure=False, + propagate_preemption=False, + replicas=args.beaker_gpus, + context=BeakerTaskContext( + priority=BeakerJobPriority[args.beaker_priority], + preemptible=True, + ), + image=BeakerImageSource(beaker=beaker_image), + command=["python", "scripts/pii/tagging_pipeline.py"] + args_list, + env_vars=[ + BeakerEnvVar(name="BEAKER_JOB_NAME", value=task_name), + BeakerEnvVar(name="OWNER", value=owner), + ] + + env_var_secrets, + resources=BeakerTaskResources(gpu_count=1, memory="125GB"), + constraints=BeakerConstraints(cluster=args.beaker_cluster if isinstance(args.beaker_cluster, list) else [args.beaker_cluster]), + result=BeakerResultSpec(path="/noop-results"), + ) + ], + retry=BeakerRetrySpec(allowed_task_retries=10), + ) + + workload = b.experiment.create(spec=experiment_spec) + + print(f"Experiment URL: https://beaker.org/ex/{workload.experiment.id}") + + +async def main(): + parser = argparse.ArgumentParser(description="Tagging pipeline for Dolma JSONL dataset") + parser.add_argument("dataset", help="Dolma dataset root (local or s3://) with documents/ folder") + parser.add_argument("scratch", help="Scratch workspace (local dir or s3://)") + parser.add_argument("--workers", type=int, default=4, help="Number of concurrent workers") + parser.add_argument("--parallel_requests", type=int, default=800, help="Max number of parallel requests to send to model") + parser.add_argument("--model", default="google/gemma-3-4b-it", help="Model path or name, hugging face or local path format") + parser.add_argument("--attribute_name", default="model_pii_tagging", help="Path to use for attribute naming") + + # Beaker/job running stuff + parser.add_argument("--beaker", action="store_true", help="Submit this job to beaker instead of running locally") + parser.add_argument("--beaker_workspace", help="Beaker workspace to submit to", default="ai2/olmocr") + parser.add_argument( + "--beaker_cluster", + help="Beaker clusters you want to run on", + default=["ai2/jupiter", "ai2/ceres", "ai2/neptune", "ai2/saturn"], + ) + parser.add_argument("--beaker_gpus", type=int, default=1, help="Number of gpu replicas to run") + parser.add_argument("--beaker_priority", type=str, default="normal", help="Beaker priority level for the job") + + parser.add_argument("--port", type=int, default=30024, help="Port for Model server") + args = parser.parse_args() + + global SERVER_PORT, workspace_s3, dataset_s3 + SERVER_PORT = args.port + workspace_s3 = boto3.client("s3") + dataset_s3 = boto3.client("s3") + + # setup the job to work in beaker environment, load secrets, adjust logging, etc. + if "BEAKER_JOB_NAME" in os.environ: + cred_path = os.path.join(os.path.expanduser("~"), ".aws", "credentials") + os.makedirs(os.path.dirname(cred_path), exist_ok=True) + with open(cred_path, "w") as f: + f.write(os.environ.get("AWS_CREDENTIALS_FILE", "")) + cred_path = os.path.join(os.path.expanduser("~"), ".gcs", "credentials") + os.makedirs(os.path.dirname(cred_path), exist_ok=True) + with open(cred_path, "w") as f: + f.write(os.environ.get("GOOGLE_APPLICATION_CREDENTIALS_FILE", "")) + os.environ["GOOGLE_APPLICATION_CREDENTIALS"] = cred_path + workspace_s3 = boto3.client("s3") + dataset_s3 = boto3.client("s3") + + # Wait a little bit so that not all beaker jobs in a task start at the same time and download the model at the same time + replica_count = int(os.environ.get("BEAKER_REPLICA_COUNT", "1")) + interval = 10 if (replica_count - 1) * 10 <= 30 else 30 / max(1, replica_count - 1) + sleep_time = int(os.environ.get("BEAKER_REPLICA_RANK", "0")) * interval + logger.info(f"Beaker job sleeping for {sleep_time} seconds to stagger model downloads") + await asyncio.sleep(sleep_time) + + # Initialize work queue + if args.scratch.startswith("s3://"): + work_queue = WorkQueue(S3Backend(workspace_s3, args.scratch)) + else: + work_queue = WorkQueue(LocalBackend(args.scratch)) + + # Discover input files + files = set() + if args.dataset.startswith("s3://"): + pattern = args.dataset.rstrip("/") + "/documents/*.jsonl*" + matched = expand_s3_glob(dataset_s3, pattern) + files = set(matched.keys()) + else: + docs_dir = os.path.join(args.dataset, "documents") + for root, _, fns in os.walk(docs_dir): + for fn in fns: + if fn.endswith((".jsonl", ".jsonl.gz", ".jsonl.ztd")): + files.add(os.path.join(root, fn)) + + # Populate the work queue if needed + await work_queue.populate_queue(list(files), items_per_group=1) + + if args.beaker: + submit_beaker_job(args) + return + + # If you get this far, then you are doing inference and need a GPU + check_torch_gpu_available() + + logger.info(f"Starting pipeline with PID {os.getpid()}") + + # Download the model before you do anything else + model_name_or_path = await download_model(args.model) + + # Initialize the work queue + qsize = await work_queue.initialize_queue() + + if qsize == 0: + logger.info("No work to do, exiting") + return + + # Create a semaphore to control worker access + # We only allow one worker to move forward with requests, until the server has no more requests in its queue + # This lets us get full utilization by having many workers, but also to be outputting dolma docs as soon as possible + # As soon as one worker is no longer saturating the gpu, the next one can start sending requests + semaphore = asyncio.Semaphore(1) + + model_server = asyncio.create_task(server_host(model_name_or_path, args, semaphore)) + + await check_server_ready() + + metrics_task = asyncio.create_task(metrics_reporter(work_queue)) + + # Create worker tasks to process the queue concurrently. + worker_tasks = [] + for i in range(args.workers): + task = asyncio.create_task(worker(args, work_queue, semaphore, worker_id=i)) + worker_tasks.append(task) + + # Wait for all worker tasks to finish + await asyncio.gather(*worker_tasks) + + model_server.cancel() + metrics_task.cancel() + logger.info("Work done") + + +if __name__ == "__main__": + asyncio.run(main()) diff --git a/scripts/plots/ocr_pareto.pdf b/scripts/plots/ocr_pareto.pdf new file mode 100644 index 0000000..9653f18 Binary files /dev/null and b/scripts/plots/ocr_pareto.pdf differ diff --git a/scripts/plots/ocr_pareto.png b/scripts/plots/ocr_pareto.png new file mode 100644 index 0000000..f61d5e3 Binary files /dev/null and b/scripts/plots/ocr_pareto.png differ diff --git a/scripts/plots/olmocr2_timeline.png b/scripts/plots/olmocr2_timeline.png new file mode 100644 index 0000000..0bfffc7 Binary files /dev/null and b/scripts/plots/olmocr2_timeline.png differ diff --git a/scripts/plots/pareto_plot.py b/scripts/plots/pareto_plot.py new file mode 100644 index 0000000..0605095 --- /dev/null +++ b/scripts/plots/pareto_plot.py @@ -0,0 +1,306 @@ +""" +Plot for OCR performance vs cost Pareto frontier figure for NeurIPS paper. + +Invocation: + python scripts/pareto_plot.py . +""" + +import argparse +import os +from dataclasses import dataclass +from typing import Literal + +import matplotlib.pyplot as plt +import matplotlib.ticker as ticker +import pandas as pd +from matplotlib import font_manager + +# Parse arguments +ap = argparse.ArgumentParser() +ap.add_argument("output_dir", type=str, help="Path to the output directory") +ap.add_argument( + "--font-path", + type=str, + help="Path to the font file", + default=None, +) +args = ap.parse_args() + +# Add custom font if provided +if args.font_path: + font_manager.fontManager.addfont(args.font_path) + plt.rcParams["font.family"] = "Manrope" + plt.rcParams["font.weight"] = "medium" + +# Ensure output directory exists +os.makedirs(args.output_dir, exist_ok=True) +OUTPUT_PATHS = [f"{args.output_dir}/ocr_pareto.pdf", f"{args.output_dir}/ocr_pareto.png"] +# Define column names +MODEL_COLUMN_NAME = "Model" +CATEGORY_COLUMN_NAME = "Category" +COST_COLUMN_NAME = "Cost_Per_Million" +PERF_COLUMN_NAME = "Performance" +COLOR_COLUMN_NAME = "Color" +OFFSET_COLUMN_NAME = "Label_Offset" +MARKER_COLUMN_NAME = "Marker" +# Define colors +DARK_BLUE = "#093235" +DARK_GREEN = "#255457" +LIGHT_GREEN = "#6FE0BA" +LIGHT_PINK = "#F697C4" +DARK_PINK = "#F0529C" +YELLOW = "#fff500" +ORANGE = "#f65834" +DARK_TEAL = "#0a3235" +OFF_WHITE = "#faf2e9" +TEAL = "#105257" +PURPLE = "#b11be8" +GREEN = "#0fcb8c" + + +# Dataclass for model data +@dataclass(frozen=True) +class ModelData: + name: str + cost_per_million: float + performance: float + category: str + label_offset: tuple[float, float] + + +def cost_per_million_by_token(gpu: Literal["a100", "h100", "l40s"], tokens_sec: float, tokens_per_page: float = 750) -> float: + """ + Calculate cost per million pages based on GPU type and token throughput. + + Args: + gpu: GPU type ("a100", "h100", or "l40s") + tokens_sec: Number of tokens processed per second + tokens_per_page: Average number of tokens per page (default: 750) + + Returns: + Cost per million pages in USD + """ + # GPU hourly costs in USD from https://www.runpod.io/pricing Nov 3 2025 + gpu_costs = { + "a100": 1.39, + "h100": 2.69, + "l40s": 0.79, + } + + cost_per_hour = gpu_costs[gpu] + + # Calculate pages per hour + # tokens per hour = tokens_sec * 3600 + # pages per hour = tokens per hour / tokens_per_page + tokens_per_hour = tokens_sec * 3600 + pages_per_hour = tokens_per_hour / tokens_per_page + + # Calculate cost per million pages + cost_per_million = (cost_per_hour / pages_per_hour) * 1_000_000 + + return cost_per_million + + +def cost_per_million_by_page(gpu: Literal["a100", "h100", "l40s"], pages_sec: float) -> float: + """ + Calculate cost per million pages based on GPU type and page throughput. + + Args: + gpu: GPU type ("a100", "h100", or "l40s") + pages_sec: Number of pages processed per second + + Returns: + Cost per million pages in USD + """ + # GPU hourly costs in USD from https://www.runpod.io/pricing Nov 3 2025 + gpu_costs = { + "a100": 1.39, + "h100": 2.69, + "l40s": 0.79, + } + + cost_per_hour = gpu_costs[gpu] + + # Calculate pages per hour + pages_per_hour = pages_sec * 3600 + + # Calculate cost per million pages + cost_per_million = (cost_per_hour / pages_per_hour) * 1_000_000 + + return cost_per_million + + +# All model data in one place for easy editing +MODEL_DATA = [ + # Perf data from olmocr paper + # ModelData(name="GPT-4o", cost_per_million=12480, performance=69.9, category="Commercial VLM", label_offset=(-35, 10)), + ModelData(name="GPT-4o", cost_per_million=7951, performance=69.9, category="Commercial VLM", label_offset=(0, 10)), + # Rescaled gpt-4o prices to gpt-4.1 api rates (3.093315*3+0.833599*12)/1288* 1000000/2 + ModelData(name="GPT-4.1", cost_per_million=6112, performance=71.0, category="Commercial VLM", label_offset=(-50, 15)), + ModelData(name="Mistral OCR", cost_per_million=1000, performance=72.0, category="Commercial API Tool", label_offset=(-20, 10)), + ModelData(name="Gemini Flash 2", cost_per_million=342, performance=66.3, category="Commercial VLM", label_offset=(10, -2)), + ModelData(name="Gemini Flash 2.5", cost_per_million=1042, performance=62.1, category="Commercial VLM", label_offset=(-160, 15)), + # Perf data from paper https://arxiv.org/pdf/2509.22186 + ModelData(name="MinerU 2.5.4", cost_per_million=cost_per_million_by_page("a100", 2.12), performance=75.2, category="Open VLM", label_offset=(10, -10)), + # Perf data is hard to measure, using previously calculated value, using more generous number from v.1.7.5 + ModelData(name="Marker v1.10.1", cost_per_million=1492, performance=76.1, category="Open Source Tool", label_offset=(-25, 10)), + # Using cost per million pages from original olmocr paper + # ModelData(name="Qwen 2 VL", cost_per_million=???, performance=61.3, category="Open VLM", label_offset=(-35, 10)), + ModelData( + name="Qwen 2.5 VL", + cost_per_million=cost_per_million_by_page("h100", 10000 / (36 * 60 + 47)), + performance=64.5, + category="Open VLM", + label_offset=(-35, 10), + ), + # Using original olmocr cost, but scaling it by 3100/2100 which is the tokens/second rate difference that we see on H100 inference + ModelData( + name="Qwen 3 VL 8B", + cost_per_million=cost_per_million_by_page("h100", 10000 / (36 * 60 + 47)) * (3100 / 2100), + performance=61.4, + category="Open VLM", + label_offset=(-35, -25), + ), + # Perf data from https://arxiv.org/pdf/2509.22186 + ModelData(name="Nanonets-OCR2-3B", cost_per_million=cost_per_million_by_page("a100", 0.55), performance=69.5, category="Open VLM", label_offset=(-85, 10)), + # Pricing from this tweet: https://x.com/VikParuchuri/status/1980725223616876704 + # You'd get better pricing running locally, but I couldn't get a number + ModelData(name="Chandra OCR API", cost_per_million=4000, performance=83.1, category="Commercial VLM", label_offset=(-85, 10)), + # Going off of 200k pages per day per A100 + ModelData( + name="DeepSeek-OCR", + cost_per_million=cost_per_million_by_page("a100", pages_sec=200_000 / (24 * 3600)), + performance=75.7, + category="Open VLM", + label_offset=(-20, 10), + ), + # Perf data from paper pg 18 https://arxiv.org/pdf/2510.14528 + ModelData(name="PaddleOCR-VL", cost_per_million=cost_per_million_by_page("a100", 1.2241), performance=80.0, category="Open VLM", label_offset=(-35, 10)), + # Perf data is here: https://beaker.allen.ai/orgs/ai2/workspaces/olmocr/work/01K8V42ERGBHAZ2KKDBKXKZHPJ?taskId=01K8V42ERJ9S82C06CSWQT7RR6&jobId=01K8VH0Y9J47ZXMCCWG97J7P54 + ModelData( + name="Ours", cost_per_million=cost_per_million_by_page("h100", 10000 / (36 * 60 + 47)), performance=82.3, category="Ours", label_offset=(-20, 10) + ), +] + +# Create dataframe from the aggregated data +df = pd.DataFrame( + [ + { + MODEL_COLUMN_NAME: m.name, + COST_COLUMN_NAME: m.cost_per_million, + PERF_COLUMN_NAME: m.performance, + CATEGORY_COLUMN_NAME: m.category, + OFFSET_COLUMN_NAME: list(m.label_offset), + } + for m in MODEL_DATA + ] +) + +# Category colors +category_colors = {"Commercial API Tool": DARK_GREEN, "Commercial VLM": DARK_GREEN, "Open Source Tool": PURPLE, "Ours": DARK_PINK, "Open VLM": PURPLE} + +df[COLOR_COLUMN_NAME] = df[CATEGORY_COLUMN_NAME].map(category_colors) + +# Define marker types +category_markers = {"Commercial API Tool": "o", "Commercial VLM": "^", "Open Source Tool": "o", "Ours": "*", "Open VLM": "^"} + +df[MARKER_COLUMN_NAME] = df[CATEGORY_COLUMN_NAME].map(category_markers) + +# Define marker sizes - increased sizes +category_marker_sizes = {"Commercial API Tool": 120, "Commercial VLM": 120, "Open Source Tool": 140, "Ours": 300, "Open VLM": 140} + +# Define text colors +category_text_colors = { + "Commercial API Tool": DARK_GREEN, + "Commercial VLM": DARK_GREEN, + "Open Source Tool": PURPLE, # darker purple + "Ours": DARK_PINK, # darker pink + "Open VLM": PURPLE, # darker purple +} + +# Create the plot +plt.figure(figsize=(10, 8)) + +# Plot each category +categories = df[CATEGORY_COLUMN_NAME].unique() +for category in categories: + mask = df[CATEGORY_COLUMN_NAME] == category + data = df[mask] + plt.scatter( + data[COST_COLUMN_NAME], + data[PERF_COLUMN_NAME], + label=category, + c=data[COLOR_COLUMN_NAME], + marker=category_markers[category], + alpha=1.0, + s=category_marker_sizes[category], + ) + +# Add labels for each point with increased font size +FONTSIZE = 22 # Increased from 9 +for idx, row in df.iterrows(): + plt.annotate( + row[MODEL_COLUMN_NAME], + (row[COST_COLUMN_NAME], row[PERF_COLUMN_NAME]), + xytext=row[OFFSET_COLUMN_NAME], + textcoords="offset points", + fontsize=FONTSIZE, + alpha=1.0, + weight="medium", + color=category_text_colors[row[CATEGORY_COLUMN_NAME]], + ) + +# Set up axes +plt.ylim(55, 85) # Set y-axis limits from 25 to 85 to include Qwen2VL +plt.xlim(100, 15000) +plt.xscale("log") # Use log scale for cost +plt.grid(True, which="both", ls=":", color=TEAL, alpha=0.2) + + +# Format y-axis to show percentages without scientific notation +def percent_formatter(y, pos): + return f"{y:.1f}%" + + +plt.gca().yaxis.set_major_formatter(ticker.FuncFormatter(percent_formatter)) + + +# Format x-axis to show dollar amounts +def dollar_formatter(x, pos): + return f"${x:,.0f}" + + +# Set specific x-axis ticks with increased font size +plt.gca().xaxis.set_major_formatter(ticker.FuncFormatter(dollar_formatter)) +plt.gca().set_xticks([100, 200, 500, 1000, 2000, 5000, 10000]) +plt.xticks(fontsize=16) # Increased tick font size +plt.yticks(fontsize=16) # Increased tick font size + +# Add labels and title with increased font size +plt.xlabel("Cost per Million Pages (USD, log scale)", fontsize=FONTSIZE, weight="medium") +plt.ylabel("Overall Performance (Pass Rate %)", fontsize=FONTSIZE, weight="medium") +# plt.title("OCR Engines: Performance vs. Cost", fontsize=12, weight="medium") + +# Remove spines +plt.gca().spines["top"].set_visible(False) +plt.gca().spines["right"].set_visible(False) + +# Add the legend with custom ordering and increased font size +handles, labels = plt.gca().get_legend_handles_labels() +desired_order = ["Ours", "Open Source Tool", "Open VLM", "Commercial API Tool", "Commercial VLM"] +label_to_handle = dict(zip(labels, handles)) +ordered_handles = [label_to_handle[label] for label in desired_order if label in label_to_handle] +ordered_labels = [label for label in desired_order if label in labels] + +plt.legend( + ordered_handles, ordered_labels, loc="lower right", fontsize=20, frameon=True, framealpha=0.9, edgecolor=TEAL, facecolor="white" # Increased from 10 +) + +# Adjust layout +plt.tight_layout() + +# Save the figure +for output_path in OUTPUT_PATHS: + plt.savefig(output_path, dpi=300, bbox_inches="tight") + +print(f"Plot saved to {', '.join(OUTPUT_PATHS)}") diff --git a/scripts/plots/plot_olmocr2_timeline.py b/scripts/plots/plot_olmocr2_timeline.py new file mode 100644 index 0000000..bdbc2c1 --- /dev/null +++ b/scripts/plots/plot_olmocr2_timeline.py @@ -0,0 +1,478 @@ +import datetime as dt +import os +import textwrap + +import matplotlib.dates as mdates +import matplotlib.pyplot as plt + +# Define colors +DARK_BLUE = "#093235" +DARK_GREEN = "#255457" +LIGHT_GREEN = "#6FE0BA" +LIGHT_PINK = "#F697C4" +DARK_PINK = "#F0529C" +YELLOW = "#fff500" +ORANGE = "#f65834" +DARK_TEAL = "#0a3235" +OFF_WHITE = "#faf2e9" +TEAL = "#105257" +PURPLE = "#b11be8" +GREEN = "#0fcb8c" + +# Model categories +model_categories = { + "GOT OCR": "Open Source Tool", + "Marker v1.7.5": "Open Source Tool", + "Marker v1.10.1": "Open Source Tool", + "MinerU v1.3.10": "Open Source Tool", + "MinerU v2.5.4": "Open Source Tool", + "Nanonets OCR S": "Commercial API Tool", + "MonkeyOCR Pro 3B": "Open VLM", + "MinerU2.5": "Open VLM", + "dots.ocr": "Open VLM", + "PaddleOCR-VL": "Open VLM", + "Mistral OCR API": "Commercial API Tool", + "GPT-4o": "Commercial VLM", + "Gemini Flash 2": "Commercial VLM", + "Qwen 2.5 VL": "Open VLM", + "olmOCR v0.1.58": "Ours", + "olmOCR v0.1.60": "Ours", + "olmOCR v0.1.68": "Ours", + "olmOCR v0.2.0": "Ours", + "olmOCR v0.3.0": "Ours", + "olmOCR v0.4grporl": "Ours", +} + +# Category colors +category_colors = {"Commercial API Tool": DARK_GREEN, "Commercial VLM": DARK_GREEN, "Open Source Tool": PURPLE, "Ours": DARK_PINK, "Open VLM": PURPLE} + +# Define marker types +category_markers = {"Commercial API Tool": "o", "Commercial VLM": "^", "Open Source Tool": "o", "Ours": "*", "Open VLM": "^"} + +# Define marker sizes +category_marker_sizes = {"Commercial API Tool": 100, "Commercial VLM": 100, "Open Source Tool": 120, "Ours": 300, "Open VLM": 120} + +# Define text colors +category_text_colors = {"Commercial API Tool": DARK_GREEN, "Commercial VLM": DARK_GREEN, "Open Source Tool": PURPLE, "Ours": DARK_PINK, "Open VLM": PURPLE} + +# Data +ocr_overall_scores = { + "GOT OCR": { + "name": "GOT OCR", + "descriptor": "", + "Overall": "48.3 ± 1.1", + "date": "Sep 4, 2024", + "paper": "https://arxiv.org/abs/2409.01704", + }, + # "Marker v1.4.0": {"Overall": "70.1 ± 1.1", "date": "Feb 11, 2024", "paper": "https://github.com/datalab-to/marker/tree/v1.4.0"}, + "Marker v1.7.5": { + "name": "Marker v1.7.5", + "descriptor": "force ocr", + "Overall": "70.1 ± 1.1", + "date": "Jun 11, 2025", + "paper": "https://github.com/datalab-to/marker/tree/v1.7.5", + }, + "Marker v1.10.1": { + "name": "Marker v1.10.1", + "descriptor": "", + "Overall": "76.1 ± 1.1", + "date": "Sep 30, 2025", + "paper": "https://github.com/datalab-to/marker/releases/tag/v1.10.1", + }, + "MinerU v1.3.10": { + "name": "MinerU v1.3.10", + "descriptor": "", + "Overall": "61.5 ± 1.1", + "date": "Apr 29, 2025", + "paper": "https://github.com/opendatalab/MinerU/tree/magic_pdf-1.3.10-released", + }, + "MinerU v2.5.4": { + "name": "MinerU v2.5.4", + "descriptor": "", + "Overall": "62.9 ± 1.1", + "date": "Sep 25, 2025", + "paper": "https://github.com/opendatalab/MinerU/releases/tag/mineru-2.5.4-released", + }, + "Nanonets OCR S": { + "name": "Nanonets OCR S", + "descriptor": "", + "Overall": "64.5 ± 1.1", + "date": "Jun 12, 2025", + "paper": "https://nanonets.com/research/nanonets-ocr-s/", + }, + "MonkeyOCR Pro 3B": { + "name": "MonkeyOCR Pro 3B", + "descriptor": "", + "Overall": "75.8 ± 1.0", + "date": "Jun 5, 2025", + "paper": "https://arxiv.org/abs/2506.05218", + }, + "MinerU2.5": { + "name": "MinerU2.5", + "descriptor": "", + "Overall": "77.5 ± 1.0", + "date": "Sep 26, 2025", + "paper": "https://arxiv.org/abs/2509.22186", + }, + "dots.ocr": { + "name": "dots.ocr", + "descriptor": "", + "Overall": "79.1 ± 1.0", + "date": "Jul 30, 2025", + "paper": "https://github.com/rednote-hilab/dots.ocr", + }, + "PaddleOCR-VL": { + "name": "PaddleOCR-VL", + "descriptor": "", + "Overall": "80.0 ± 1.0", + "date": "Oct 15, 2025", + "paper": "https://arxiv.org/abs/2510.14528", + }, + "Mistral OCR API": { + "name": "Mistral OCR API", + "descriptor": "", + "Overall": "72.0 ± 1.1", + "date": "Mar 6, 2025", + "paper": "https://mistral.ai/fr/news/mistral-ocr", + }, + "GPT-4o": { + "name": "GPT-4o", + "descriptor": "No Anchor", + "Overall": "68.9 ± 1.1", + "date": "May 13, 2024", + "paper": "https://openai.com/index/hello-gpt-4o/", + }, + # "GPT-4o (Anchored)": { + # "name": "GPT-4o", + # "descriptor": "Anchored", + # "Overall": "69.9 ± 1.1", + # "date": "May 13, 2024", + # "paper": "https://openai.com/index/hello-gpt-4o/", + # }, + "Gemini Flash 2": { + "name": "Gemini Flash 2", + "descriptor": "No Anchor", + "Overall": "57.8 ± 1.1", + "date": "Dec 11, 2024", + "paper": "https://blog.google/technology/google-deepmind/google-gemini-ai-update-december-2024/", + }, + # "Gemini Flash 2 (Anchored)": { + # "name": "Gemini Flash 2", + # "descriptor": "Anchored", + # "Overall": "63.8 ± 1.2", + # "date": "Dec 11, 2024", + # "paper": "https://blog.google/technology/google-deepmind/google-gemini-ai-update-december-2024/", + # }, + "Qwen 2 VL": { + "name": "Qwen 2 VL", + "descriptor": "No Anchor", + "Overall": "31.5 ± 0.9", + "date": "Aug 29, 2024", + "paper": "https://arxiv.org/abs/2409.12191v1", + }, + "Qwen 2.5 VL": { + "name": "Qwen 2.5 VL", + "descriptor": "No Anchor", + "Overall": "65.5 ± 1.2", + "date": "Feb 19, 2025", + "paper": "https://arxiv.org/abs/2502.13923", + }, + "olmOCR v0.1.58": { + "name": "olmOCR v0.1.58", + "descriptor": "Initial release", + "Overall": "68.2 ± 1.1", + "date": "Feb 14, 2025", + "paper": "https://github.com/allenai/olmocr/releases/tag/v0.1.58", + }, + # "olmOCR v0.1.60": { + # "name": "olmOCR v0.1.60", + # "descriptor": "Temperature from 0.8 fixed to dynamic", + # "Overall": "71.4 ± 1.1", + # "date": "Mar 17, 2025", + # "paper": "https://github.com/allenai/olmocr/releases/tag/v0.1.60", + # }, + "olmOCR v0.1.68": { + "name": "olmOCR v0.1.68", + "descriptor": "Base VLM, prompts, dynamic temp", + "Overall": "75.8 ± 1.1", + "date": "May 19, 2025", + "paper": "https://github.com/allenai/olmocr/releases/tag/v0.1.68", + }, + "olmOCR v0.2.0": { + "name": "olmOCR v0.2.0", + "descriptor": "Trainer, YAML, image size", + "Overall": "78.5 ± 1.1", + "date": "Jul 23, 2025", + "paper": "https://github.com/allenai/olmocr/releases/tag/v0.2.0", + }, + # "olmOCR v0.3.0": { + # "name": "olmOCR v0.3.0", + # "descriptor": "Fixing bug with blank page hallucinations", + # "Overall": "78.5 ± 1.1", + # "date": "Aug 13, 2025", + # "paper": "https://github.com/allenai/olmocr/releases/tag/v0.3.0", + # }, + "olmOCR v0.4grporl": { + "name": "olmOCR v0.4grporl", + "descriptor": "RLVR", + "Overall": "82.6 ± 1.1", + "date": "Oct 21, 2025", + "paper": "https://github.com/allenai/olmocr/releases/tag/v0.4.0", + }, +} + + +# Convert string dates and overall means +olm_data = [(k, v) for k, v in ocr_overall_scores.items() if k.startswith("olmOCR")] +marker_data = [(k, v) for k, v in ocr_overall_scores.items() if k.startswith("Marker")] +mineru_data = [(k, v) for k, v in ocr_overall_scores.items() if k.startswith("MinerU")] +other_data = [(k, v) for k, v in ocr_overall_scores.items() if not k.startswith("olmOCR") and not k.startswith("Marker") and not k.startswith("MinerU")] + + +def parse_mean(value): + return float(value.split("±")[0].strip()) + + +def wrap_text(text, max_chars=20): + """Wrap text to fit within max_chars per line.""" + if len(text) <= max_chars: + return text + return "\n".join(textwrap.wrap(text, width=max_chars)) + + +# Label position offsets (x_offset in days, y_offset in score units) +# Adjust these to manually tune label positions relative to their data points +label_offsets = { + "olmOCR v0.1.58": (-20, -0.5), + "olmOCR v0.1.60": (0, 1.5), + "olmOCR v0.1.68": (-10, 1.5), + "olmOCR v0.2.0": (0, 1.5), + "olmOCR v0.4grporl": (0, 1.5), + "Marker v1.7.5": (0, 1.5), + "Marker v1.10.1": (0, -5.5), + "MinerU v1.3.10": (0, -5.5), + "MinerU v2.5.4": (0, -5.5), + "Nanonets OCR S": (0, 1.5), + "MonkeyOCR Pro 3B": (0, 1.5), + "MinerU2.5": (0, 1.5), + "dots.ocr": (0, 1.5), + "PaddleOCR-VL": (0, 1.5), + "GOT OCR": (0, 1.5), + "Mistral OCR API": (0, 1.5), + "GPT-4o": (0, 1.5), + "Gemini Flash 2": (0, -5.5), + "Qwen 2.5 VL": (0, -6), +} + +# Floating label offsets for line curves (x_offset in days, y_offset in score units) +floating_label_offsets = { + "olmOCR": (-40, 8.5), + "Marker": (-40, -7.5), + "MinerU": (-70, -4.5), +} + +# Sort entries by date (if filled) +olm_data_sorted = sorted( + [(name, dt.datetime.strptime(v["date"], "%b %d, %Y"), parse_mean(v["Overall"]), v["name"], v["descriptor"]) for name, v in olm_data], + key=lambda x: x[1], +) +marker_data_sorted = sorted( + [(name, dt.datetime.strptime(v["date"], "%b %d, %Y"), parse_mean(v["Overall"]), v["name"]) for name, v in marker_data], + key=lambda x: x[1], +) +mineru_data_sorted = sorted( + [(name, dt.datetime.strptime(v["date"], "%b %d, %Y"), parse_mean(v["Overall"]), v["name"]) for name, v in mineru_data], + key=lambda x: x[1], +) + +# Plot +plt.figure(figsize=(8, 5)) + +# olmOCR line +dates = [d for _, d, _, _, _ in olm_data_sorted] +scores = [s for _, _, s, _, _ in olm_data_sorted] +category = model_categories.get("olmOCR v0.1.58", "Ours") +color = category_colors[category] +marker_star = category_markers[category] # star marker +marker_triangle = "^" # VLM marker shape +marker_size = category_marker_sizes[category] + +plt.plot(dates, scores, color=color, linewidth=2) +for idx, (name, date, score, display_name, descriptor) in enumerate(olm_data_sorted): + # Only the last point gets a star, others get triangles + marker = marker_star if idx == len(olm_data_sorted) - 1 else marker_triangle + size = marker_size if idx == len(olm_data_sorted) - 1 else 100 + plt.scatter(date, score, color=color, edgecolor="none", s=size, marker=marker, zorder=3) + +# Add descriptor labels above olmOCR circles (black text) +for name, date, score, display_name, descriptor in olm_data_sorted: + x_off, y_off = label_offsets.get(name, (0, 1.5)) + label_date = date + dt.timedelta(days=x_off) + wrapped_descriptor = wrap_text(descriptor, max_chars=20) + plt.text(label_date, score + y_off, f"{wrapped_descriptor}\n{score:.1f}", ha="center", va="bottom", fontsize=7, fontweight="bold", color="black") + +# Add floating label for olmOCR line (above the line, no border) +if olm_data_sorted: + mid_idx = len(olm_data_sorted) // 2 + mid_date = olm_data_sorted[mid_idx][1] + mid_score = olm_data_sorted[mid_idx][2] + x_offset_days, y_offset = floating_label_offsets.get("olmOCR", (0, -3)) + label_date = mid_date + dt.timedelta(days=x_offset_days) + plt.text( + label_date, + mid_score + y_offset, + "olmOCR", + fontsize=10, + fontweight="bold", + color=color, + ha="center", + va="bottom", + bbox=dict(boxstyle="round,pad=0.3", facecolor="white", edgecolor="none", alpha=0.8), + ) + +# Marker line +if marker_data_sorted: + dates = [d for _, d, _, _ in marker_data_sorted] + scores = [s for _, _, s, _ in marker_data_sorted] + + # Get category info for first Marker model + first_name = marker_data_sorted[0][0] + category = model_categories.get(first_name, "Open Source Tool") + color = category_colors[category] + marker = category_markers[category] + marker_size = category_marker_sizes[category] + text_color = category_text_colors[category] + + if len(marker_data_sorted) > 1: + # Multiple points: draw line, use version labels, and add floating label + plt.plot(dates, scores, color=color, linewidth=2) + for name, date, score, display_name in marker_data_sorted: + plt.scatter(date, score, color=color, edgecolor="none", s=marker_size, marker=marker, zorder=3) + + # Add version labels above Marker circles + for name, date, score, display_name in marker_data_sorted: + version = display_name.replace("Marker ", "") + x_off, y_off = label_offsets.get(name, (0, 1.5)) + label_date = date + dt.timedelta(days=x_off) + plt.text(label_date, score + y_off, f"{version}\n{score:.1f}", ha="center", va="bottom", fontsize=7, fontweight="bold", color="black") + + # Add floating label for Marker line + mid_idx = len(marker_data_sorted) // 2 + mid_date = marker_data_sorted[mid_idx][1] + mid_score = marker_data_sorted[mid_idx][2] + x_offset_days, y_offset = floating_label_offsets.get("Marker", (0, -3)) + label_date = mid_date + dt.timedelta(days=x_offset_days) + plt.text( + label_date, + mid_score + y_offset, + "Marker", + fontsize=10, + fontweight="bold", + color=color, + ha="center", + va="bottom", + bbox=dict(boxstyle="round,pad=0.3", facecolor="white", edgecolor="none", alpha=0.8), + ) + else: + # Single point: suppress version, just show "Marker" + for name, date, score, display_name in marker_data_sorted: + plt.scatter(date, score, color=color, edgecolor="none", s=marker_size, marker=marker, zorder=2) + x_off, y_off = label_offsets.get(name, (0, 1.5)) + label_date = date + dt.timedelta(days=x_off) + plt.text(label_date, score + y_off, f"Marker\n{score:.1f}", ha="center", va="bottom", fontsize=7, fontweight="bold", color="black") + +# MinerU line +if mineru_data_sorted: + dates = [d for _, d, _, _ in mineru_data_sorted] + scores = [s for _, _, s, _ in mineru_data_sorted] + + # Get category info for first MinerU model + first_name = mineru_data_sorted[0][0] + category = model_categories.get(first_name, "Open Source Tool") + color = category_colors[category] + marker = category_markers[category] + marker_size = category_marker_sizes[category] + text_color = category_text_colors[category] + + if len(mineru_data_sorted) > 1: + # Multiple points: draw line, use version labels, and add floating label + plt.plot(dates, scores, color=color, linewidth=2) + for name, date, score, display_name in mineru_data_sorted: + plt.scatter(date, score, color=color, edgecolor="none", s=marker_size, marker=marker, zorder=3) + + # Add version labels above MinerU circles + for name, date, score, display_name in mineru_data_sorted: + version = display_name.replace("MinerU ", "") + x_off, y_off = label_offsets.get(name, (0, 1.5)) + label_date = date + dt.timedelta(days=x_off) + plt.text(label_date, score + y_off, f"{version}\n{score:.1f}", ha="center", va="bottom", fontsize=7, fontweight="bold", color="black") + + # Add floating label for MinerU line + mid_idx = len(mineru_data_sorted) // 2 + mid_date = mineru_data_sorted[mid_idx][1] + mid_score = mineru_data_sorted[mid_idx][2] + x_offset_days, y_offset = floating_label_offsets.get("MinerU", (0, -3)) + label_date = mid_date + dt.timedelta(days=x_offset_days) + plt.text( + label_date, + mid_score + y_offset, + "MinerU", + fontsize=10, + fontweight="bold", + color=color, + ha="center", + va="bottom", + bbox=dict(boxstyle="round,pad=0.3", facecolor="white", edgecolor="none", alpha=0.8), + ) + else: + # Single point: suppress version, just show "MinerU" + for name, date, score, display_name in mineru_data_sorted: + plt.scatter(date, score, color=color, edgecolor="none", s=marker_size, marker=marker, zorder=2) + x_off, y_off = label_offsets.get(name, (0, 1.5)) + label_date = date + dt.timedelta(days=x_off) + plt.text(label_date, score + y_off, f"MinerU\n{score:.1f}", ha="center", va="bottom", fontsize=7, fontweight="bold", color="black") + +# Other models +for name, v in other_data: + if v["date"]: + d = dt.datetime.strptime(v["date"], "%b %d, %Y") + s = parse_mean(v["Overall"]) + + # Get category info + category = model_categories.get(name, "Open VLM") + color = category_colors[category] + marker = category_markers[category] + marker_size = category_marker_sizes[category] + text_color = category_text_colors[category] + + plt.scatter(d, s, color=color, edgecolor="none", s=marker_size, marker=marker, zorder=2) + + # Add label above circle + x_off, y_off = label_offsets.get(name, (0, 1.5)) + label_date = d + dt.timedelta(days=x_off) + wrapped_name = wrap_text(name, max_chars=20) + plt.text(label_date, s + y_off, f"{wrapped_name}\n{s:.1f}", ha="center", va="bottom", fontsize=7, fontweight="bold", color="black") + +# Labels and style +plt.xlabel("Date") +plt.ylabel("Overall Performance") + +# Format x-axis with dates +ax = plt.gca() +ax.xaxis.set_major_formatter(mdates.DateFormatter("%b %Y")) +ax.xaxis.set_major_locator(mdates.MonthLocator(interval=3)) +plt.gcf().autofmt_xdate() # Rotate date labels for better readability + +# Increase y-axis limits to prevent label bleeding +current_ylim = ax.get_ylim() +ax.set_ylim(current_ylim[0] - 5, current_ylim[1] + 10) +plt.grid(alpha=0.3, linestyle="--") + +plt.tight_layout() + +# Save the plot first before showing +save_path = os.path.join(os.path.dirname(__file__), "olmocr2_timeline.png") +plt.savefig(save_path, dpi=300, bbox_inches="tight") +print(f"Saved plot to {save_path}") + +plt.show() diff --git a/scripts/prepare_changelog.py b/scripts/prepare_changelog.py new file mode 100644 index 0000000..8e9d587 --- /dev/null +++ b/scripts/prepare_changelog.py @@ -0,0 +1,38 @@ +from datetime import datetime +from pathlib import Path + +from olmocr.version import VERSION + + +def main(): + changelog = Path("CHANGELOG.md") + + with changelog.open() as f: + lines = f.readlines() + + insert_index: int = -1 + for i in range(len(lines)): + line = lines[i] + if line.startswith("## Unreleased"): + insert_index = i + 1 + elif line.startswith(f"## [v{VERSION}]"): + print("CHANGELOG already up-to-date") + return + elif line.startswith("## [v"): + break + + if insert_index < 0: + raise RuntimeError("Couldn't find 'Unreleased' section") + + lines.insert(insert_index, "\n") + lines.insert( + insert_index + 1, + f"## [v{VERSION}](https://github.com/allenai/olmocr/releases/tag/v{VERSION}) - " f"{datetime.now().strftime('%Y-%m-%d')}\n", + ) + + with changelog.open("w") as f: + f.writelines(lines) + + +if __name__ == "__main__": + main() diff --git a/scripts/qianfan_bench_convert.py b/scripts/qianfan_bench_convert.py new file mode 100644 index 0000000..7cb781b --- /dev/null +++ b/scripts/qianfan_bench_convert.py @@ -0,0 +1,202 @@ +#!/usr/bin/env python3 +"""Convert PDFs to markdown using Qianfan OCR via a local vllm server. + +This script is used by run_qianfan_benchmark.sh to process benchmark PDFs. +It converts each PDF's first page to a PNG image, sends it to a local vllm +server running baidu/Qianfan-OCR, and saves the markdown output. + +The document parsing prompt and API call pattern follow the reference at: +https://github.com/baidubce/skills/blob/develop/skills/qianfanocr-document-intelligence/scripts/qianfan_ocr_cli.py +https://github.com/baidubce/skills/blob/develop/skills/qianfanocr-document-intelligence/scripts/run_document_parsing.py +""" + +import base64 +import glob +import json +import mimetypes +import os +import re +import subprocess +import sys +import tempfile +from pathlib import Path +from urllib.request import Request, urlopen + +VLLM_URL = "http://localhost:8000/v1/chat/completions" +MODEL_NAME = "qianfan-ocr" + +# Document parsing prompt from the reference implementation +DOCUMENT_PARSING_PROMPT = """You are an AI assistant specialized in converting document images (one or multiple pages extracted from a PDF) into Markdown with high fidelity. + +Your task is to accurately convert all visible content from the images into Markdown, strictly following the rules below. Do not add explanations, comments, or inferred content. + +1. Pages: +- The input may contain one or multiple page images. +- Preserve the exact page order as provided. +- If there are multiple pages, separate pages using the marker: + --- Page N --- + (N starts from 1) +- If there is only one page, do NOT output any page separator. + +2. Text Recognition: +- Accurately convert all visible text. +- No guessing, inference, paraphrasing, or correction. +- Preserve the original document structure, including headings, paragraphs, lists, captions, and footnotes. +- Completely REMOVE all header and footer text. Do not output page numbers, running titles, or repeated marginal content. + +3. Reading Order: +- Follow a top-to-bottom, left-to-right reading order. +- For multi-column layouts, fully read the left column before the right column. +- Do not reorder content for semantic or logical clarity. + +4. Mathematical Formulas: +- Convert all mathematical expressions to LaTeX. +- Inline formulas must use $...$. +- Display (block) formulas must use: + $$ + ... + $$ +- Preserve symbols, spacing, and structure exactly. +- Do not invent, simplify, normalize, or correct formulas. + +5. Tables: +- Convert all tables to HTML format. +- Wrap each table with and

. +- Preserve row and column order, merged cells (rowspan, colspan), and empty cells. +- Do not restructure or reinterpret tables. + +6. Images: +- Do NOT describe image content. +- Preserve images using the exact format: + ![label]([[x1, y1, x2, y2]]) +- Allowed labels: image, chart, seal. +- Completely REMOVE all header_image and footer_image elements. +- Do not introduce new labels. +- Do not remove or merge remaining image elements. + +7. Unreadable or Missing Content: +- If text, symbols, or table cells are unreadable, preserve their position and leave the content empty. +- Do not guess or fill in missing information. + +8. Output Requirements: +- Output Markdown only. +- Preserve original layout, spacing, and structure as closely as possible. +- Ensure clear separation between elements using line breaks. +- Do not include any explanations, metadata, or comments.""" + + +def pdf_to_png(pdf_path, output_dir, dpi=200): + """Convert first page of PDF to PNG using pdftoppm.""" + os.makedirs(output_dir, exist_ok=True) + stem = Path(pdf_path).stem + output_prefix = os.path.join(output_dir, stem) + subprocess.run( + ["pdftoppm", "-png", "-r", str(dpi), "-f", "1", "-l", "1", pdf_path, output_prefix], + check=True, + capture_output=True, + ) + # pdftoppm outputs {prefix}-{page}.png + candidates = sorted(glob.glob(f"{output_prefix}*.png")) + if not candidates: + raise RuntimeError(f"pdftoppm produced no output for {pdf_path}") + return candidates[0] + + +def image_to_data_url(image_path): + """Convert local image to base64 data URL.""" + with open(image_path, "rb") as f: + raw = f.read() + mime_type, _ = mimetypes.guess_type(image_path) + mime = mime_type or "application/octet-stream" + encoded = base64.b64encode(raw).decode("ascii") + return f"data:{mime};base64,{encoded}" + + +def strip_thinking(text): + """Strip ... blocks from response.""" + m = re.match(r"^\s*(.*?)\s*", text, re.DOTALL) + if not m: + return text + return text[m.end() :].strip() + + +def call_vllm(image_path, max_tokens=16384): + """Send image to local vllm server and get markdown response.""" + data_url = image_to_data_url(image_path) + + body = { + "model": MODEL_NAME, + "max_tokens": max_tokens, + "stream": False, + "skip_special_tokens": False, + "messages": [ + { + "role": "user", + "content": [ + {"type": "text", "text": DOCUMENT_PARSING_PROMPT}, + { + "type": "image_url", + "image_url": {"url": data_url}, + }, + ], + } + ], + } + + request = Request( + VLLM_URL, + data=json.dumps(body).encode("utf-8"), + headers={"Content-Type": "application/json"}, + method="POST", + ) + with urlopen(request, timeout=300) as resp: + payload = json.loads(resp.read().decode("utf-8")) + + content = payload["choices"][0]["message"]["content"] + return strip_thinking(content) + + +def main(): + pdf_root = sys.argv[1] + target_root = sys.argv[2] + + # Find all PDFs + pdf_files = sorted(glob.glob(os.path.join(pdf_root, "**", "*.pdf"), recursive=True)) + print(f"Found {len(pdf_files)} PDFs to process") + + with tempfile.TemporaryDirectory(prefix="qianfan_bench_") as tmpdir: + for i, pdf_path in enumerate(pdf_files): + rel_path = os.path.relpath(pdf_path, pdf_root) + parts = rel_path.split(os.sep) + if len(parts) < 2: + print(f"Warning: Unexpected PDF path layout for {pdf_path}, skipping") + continue + + section = parts[0] + pdf_name = Path(pdf_path).stem + + print(f" [{i+1}/{len(pdf_files)}] Processing {rel_path}") + + try: + # Convert PDF to image + png_path = pdf_to_png(pdf_path, os.path.join(tmpdir, section)) + + # Call vllm API + markdown = call_vllm(png_path) + + # Save output + target_dir = os.path.join(target_root, section) + os.makedirs(target_dir, exist_ok=True) + target_path = os.path.join(target_dir, f"{pdf_name}_pg1_repeat1.md") + with open(target_path, "w", encoding="utf-8") as f: + f.write(markdown) + print(f" -> {target_path}") + except Exception as e: + print(f" Error processing {pdf_path}: {e}") + continue + + print("Done processing all PDFs") + + +if __name__ == "__main__": + main() diff --git a/scripts/release.sh b/scripts/release.sh new file mode 100755 index 0000000..ef30083 --- /dev/null +++ b/scripts/release.sh @@ -0,0 +1,79 @@ +#!/bin/bash + +set -e + +# Function to extract version components from version.py using regex +get_version_from_file() { + VERSION_FILE="olmocr/version.py" + + if [[ ! -f "$VERSION_FILE" ]]; then + echo "Error: $VERSION_FILE does not exist." + exit 1 + fi + + # Extract _MAJOR + _MAJOR=$(grep -E '^_MAJOR\s*=\s*"([^"]+)"' "$VERSION_FILE" | sed -E 's/_MAJOR\s*=\s*"([^"]+)"/\1/') + if [[ -z "$_MAJOR" ]]; then + echo "Error: Could not extract _MAJOR from $VERSION_FILE." + exit 1 + fi + + # Extract _MINOR + _MINOR=$(grep -E '^_MINOR\s*=\s*"([^"]+)"' "$VERSION_FILE" | sed -E 's/_MINOR\s*=\s*"([^"]+)"/\1/') + if [[ -z "$_MINOR" ]]; then + echo "Error: Could not extract _MINOR from $VERSION_FILE." + exit 1 + fi + + # Extract _PATCH + _PATCH=$(grep -E '^_PATCH\s*=\s*"([^"]+)"' "$VERSION_FILE" | sed -E 's/_PATCH\s*=\s*"([^"]+)"/\1/') + if [[ -z "$_PATCH" ]]; then + echo "Error: Could not extract _PATCH from $VERSION_FILE." + exit 1 + fi + + # Extract _SUFFIX (optional) + _SUFFIX=$(grep -E '^_SUFFIX\s*=\s*"([^"]*)"' "$VERSION_FILE" | sed -E 's/_SUFFIX\s*=\s*"([^"]*)"/\1/') + if [[ -z "$_SUFFIX" ]]; then + _SUFFIX="" + fi + + # Construct VERSION + VERSION_PY="${_MAJOR}.${_MINOR}.${_PATCH}${_SUFFIX}" + echo "$VERSION_PY" +} + +TAG=$(python -c 'from olmocr.version import VERSION; print("v" + VERSION)') + +# Get the VERSION from version.py +VERSION_PY=$(get_version_from_file) + +# Compare the two versions +if [[ "v$VERSION_PY" != "$TAG" ]]; then + echo "Version mismatch detected:" + echo " Python reported version: $TAG" + echo " version.py contains: v$VERSION_PY" + echo + read -p "The versions do not match. Please run 'pip install -e .' to synchronize versions. Do you want to continue? [Y/n] " prompt + + if [[ ! "$prompt" =~ ^([yY][eE][sS]|[yY])$ ]]; then + echo "Release process aborted due to version mismatch." + exit 1 + else + echo "Proceeding with the release despite the version mismatch." + fi +fi + +read -p "Creating new release for $TAG. Do you want to continue? [Y/n] " prompt + +if [[ $prompt == "y" || $prompt == "Y" || $prompt == "yes" || $prompt == "Yes" ]]; then + python scripts/prepare_changelog.py + git add CHANGELOG.md + git commit -m "Bump version to $TAG for release" || true && git push + echo "Creating new git tag $TAG" + git tag "$TAG" -m "$TAG" + git push --tags +else + echo "Cancelled" + exit 1 +fi \ No newline at end of file diff --git a/scripts/release_notes.py b/scripts/release_notes.py new file mode 100755 index 0000000..8ca99a9 --- /dev/null +++ b/scripts/release_notes.py @@ -0,0 +1,81 @@ +# encoding: utf-8 + +""" +Prepares markdown release notes for GitHub releases. +""" + +import os +from typing import List, Optional + +import packaging.version + +TAG = os.environ["TAG"] + +ADDED_HEADER = "### Added 🎉" +CHANGED_HEADER = "### Changed ⚠️" +FIXED_HEADER = "### Fixed ✅" +REMOVED_HEADER = "### Removed 👋" + + +def get_change_log_notes() -> str: + in_current_section = False + current_section_notes: List[str] = [] + with open("CHANGELOG.md") as changelog: + for line in changelog: + if line.startswith("## "): + if line.startswith("## Unreleased"): + continue + if line.startswith(f"## [{TAG}]"): + in_current_section = True + continue + break + if in_current_section: + if line.startswith("### Added"): + line = ADDED_HEADER + "\n" + elif line.startswith("### Changed"): + line = CHANGED_HEADER + "\n" + elif line.startswith("### Fixed"): + line = FIXED_HEADER + "\n" + elif line.startswith("### Removed"): + line = REMOVED_HEADER + "\n" + current_section_notes.append(line) + assert current_section_notes + return "## What's new\n\n" + "".join(current_section_notes).strip() + "\n" + + +def get_commit_history() -> str: + new_version = packaging.version.parse(TAG) + + # Pull all tags. + os.popen("git fetch --tags") + + # Get all tags sorted by version, latest first. + all_tags = os.popen("git tag -l --sort=-version:refname 'v*'").read().split("\n") + + # Out of `all_tags`, find the latest previous version so that we can collect all + # commits between that version and the new version we're about to publish. + # Note that we ignore pre-releases unless the new version is also a pre-release. + last_tag: Optional[str] = None + for tag in all_tags: + if not tag.strip(): # could be blank line + continue + version = packaging.version.parse(tag) + if new_version.pre is None and version.pre is not None: + continue + if version < new_version: + last_tag = tag + break + if last_tag is not None: + commits = os.popen(f"git log {last_tag}..{TAG} --oneline --first-parent").read() + else: + commits = os.popen("git log --oneline --first-parent").read() + return "## Commits\n\n" + commits + + +def main(): + print(get_change_log_notes()) + print(get_commit_history()) + + +if __name__ == "__main__": + main() diff --git a/scripts/remove_paths_from_olmocrmix.py b/scripts/remove_paths_from_olmocrmix.py new file mode 100644 index 0000000..568e7c7 --- /dev/null +++ b/scripts/remove_paths_from_olmocrmix.py @@ -0,0 +1,175 @@ +# This code takes in a path to an olmocr-mix processed data folder +# and removes any documents that come from a premade deny list +# +# Example deny list, txt file: +# s3://ai2-s2-pdfs/4444/111111111111111111111111111111111111.pdf +# s3://ai2-s2-pdfs/5555/111111111111111111111111111111111112.pdf +# s3://ai2-s2-pdfs/6666/111111111111111111111111111111111113.pdf +# +# Should match paths like +# +# processed_00_documents_eval_s2pdf/4444/111111111111111111111111111111111111-1.md +# processed_00_documents_eval_s2pdf/4444/111111111111111111111111111111111111-1.pdf +# Where the path to processed_00_documents_eval_s2pdf is provided as an argument + +# What it should do is move the bad files to a rejected folder, match both .md and .pdf + +import argparse +import re +import shutil +from pathlib import Path +from typing import Set, Tuple + +from tqdm import tqdm + + +def parse_deny_list(deny_list_file: Path) -> Set[Tuple[str, str]]: + """Parse deny list file and extract subdirectory and base filename patterns.""" + patterns = set() + + with open(deny_list_file, "r") as f: + for line in tqdm(f): + line = line.strip() + if not line or line.startswith("#"): + continue + + # Extract the subdirectory (e.g., "4444") and base filename from S3 path + # Pattern: s3://bucket-name/subdirectory/filename.pdf + match = re.match(r"s3://[^/]+/(\S+)/([^/]+)\.pdf$", line) + if match: + subdir = match.group(1) + base_filename = match.group(2) + patterns.add((subdir, base_filename)) + else: + print(f"Warning: Could not parse deny list entry: {line}") + + return patterns + + +def find_matching_files(processed_dir: Path, deny_patterns: Set[Tuple[str, str]]) -> list[Path]: + """Find all files that match the deny patterns.""" + matching_files = [] + + # First, glob all .pdf and .md files in the processed directory + all_pdf_files = list(processed_dir.glob("**/*.pdf")) + all_md_files = list(processed_dir.glob("**/*.md")) + all_files = all_pdf_files + all_md_files + + print(f"Found {len(all_files)} total files ({len(all_pdf_files)} PDFs, {len(all_md_files)} MDs)") + + # Now check each file against the deny patterns + for file_path in tqdm(all_files, desc="Checking files against deny list"): + # Extract the parent directory name and base filename + # Expected pattern: processed_dir/subdir/filename-pagenum.ext + try: + relative_path = file_path.relative_to(processed_dir) + parts = relative_path.parts + + if len(parts) >= 2: + subdir = parts[0] + filename = parts[-1] + + # Extract base filename without page number and extension + # Pattern: base_filename-pagenum.ext + match = re.match(r"^(.+?)-\d+\.(pdf|md)$", filename) + if match: + base_filename = match.group(1) + + # Check if this (subdir, base_filename) pair is in our deny set + if (subdir, base_filename) in deny_patterns: + matching_files.append(file_path) + except Exception as e: + print(f"Warning: Could not process file {file_path}: {e}") + + return matching_files + + +def move_files_to_rejected(files_to_move: list[Path], processed_dir: Path, rejected_dir: Path): + """Move files to the rejected folder, maintaining directory structure.""" + moved_count = 0 + + for file_path in files_to_move: + # Calculate relative path from processed_dir + relative_path = file_path.relative_to(processed_dir) + + # Create target path in rejected folder + target_path = rejected_dir / relative_path + + # Create target directory if it doesn't exist + target_path.parent.mkdir(parents=True, exist_ok=True) + + try: + # Move the file + shutil.move(str(file_path), str(target_path)) + print(f"Moved: {file_path} -> {target_path}") + moved_count += 1 + except Exception as e: + print(f"Error moving {file_path}: {e}") + + return moved_count + + +def main(): + parser = argparse.ArgumentParser(description="Remove documents from olmocr-mix processed data based on a deny list") + parser.add_argument("processed_dir", type=Path, help="Path to the processed documents directory (e.g., processed_00_documents_eval_s2pdf)") + parser.add_argument("deny_list", type=Path, help="Path to the deny list text file containing S3 paths to reject") + parser.add_argument("--rejected-dir", type=Path, default=None, help="Path to the rejected files directory (default: processed_dir_rejected)") + parser.add_argument("--dry-run", action="store_true", help="Show what would be moved without actually moving files") + + args = parser.parse_args() + + # Validate inputs + if not args.processed_dir.exists(): + print(f"Error: Processed directory does not exist: {args.processed_dir}") + return 1 + + if not args.deny_list.exists(): + print(f"Error: Deny list file does not exist: {args.deny_list}") + return 1 + + # Set rejected directory + if args.rejected_dir is None: + args.rejected_dir = args.processed_dir.parent / f"{args.processed_dir.name}_rejected" + + print(f"Processing directory: {args.processed_dir}") + print(f"Deny list file: {args.deny_list}") + print(f"Rejected directory: {args.rejected_dir}") + + if args.dry_run: + print("\n** DRY RUN MODE - No files will be moved **\n") + + # Parse deny list + deny_patterns = parse_deny_list(args.deny_list) + print(f"\nFound {len(deny_patterns)} unique deny patterns") + + # Find matching files + matching_files = find_matching_files(args.processed_dir, deny_patterns) + print(f"Found {len(matching_files)} files to remove") + + if not matching_files: + print("No files to remove.") + return 0 + + # Show summary + print("\nFiles to be moved:") + for f in matching_files[:10]: # Show first 10 + print(f" - {f}") + if len(matching_files) > 10: + print(f" ... and {len(matching_files) - 10} more files") + + # Move files (or simulate in dry-run mode) + if not args.dry_run: + # Create rejected directory + args.rejected_dir.mkdir(parents=True, exist_ok=True) + + # Move the files + moved_count = move_files_to_rejected(matching_files, args.processed_dir, args.rejected_dir) + print(f"\nSuccessfully moved {moved_count} files to {args.rejected_dir}") + else: + print(f"\nDry run complete. Would move {len(matching_files)} files.") + + return 0 + + +if __name__ == "__main__": + exit(main()) diff --git a/scripts/run_api_benchmark.sh b/scripts/run_api_benchmark.sh new file mode 100755 index 0000000..c3d84ee --- /dev/null +++ b/scripts/run_api_benchmark.sh @@ -0,0 +1,396 @@ +#!/bin/bash + +# Runs an olmocr-bench API benchmark run +# Without additional parameters (default behavior): uses the default benchmark dataset from hugging face +# ./scripts/run_api_benchmark.sh +# With API provider arguments (forwarded to olmocr.bench.convert): +# ./scripts/run_api_benchmark.sh chatgpt:name=test1:prompt=long +# ./scripts/run_api_benchmark.sh gemini:name=test2:api_key=YOUR_KEY +# With beaker secrets for API keys (format: ENV_VAR=secret-name): +# ./scripts/run_api_benchmark.sh --beaker-secret OPENAI_API_KEY=jakep-openai-key chatgpt:name=test1:prompt=long +# ./scripts/run_api_benchmark.sh --beaker-secret GEMINI_API_KEY=jakep-gemini-key --beaker-secret ANTHROPIC_API_KEY=jakep-anthropic-key chatgpt:name=test1 +# With cluster parameter: specify a specific cluster to use +# ./scripts/run_api_benchmark.sh --cluster ai2/titan-cirrascale chatgpt:name=test1:prompt=long +# With beaker image: skip Docker build and use provided Beaker image +# ./scripts/run_api_benchmark.sh --beaker-image jakep/olmocr-benchmark-0.3.3-780bc7d934 chatgpt:name=test1 +# With benchrepo parameter: use a different benchmark dataset repository (default: allenai/olmOCR-bench) +# ./scripts/run_api_benchmark.sh --benchrepo allenai/olmOCR-bench-internal chatgpt:name=test1 +# With benchbranch parameter: use a specific branch/revision of the benchmark dataset +# ./scripts/run_api_benchmark.sh --benchbranch olmOCR-bench-1125 chatgpt:name=test1 +# With benchpath parameter: use benchmark dataset from a local path or S3 path (mutually exclusive with benchrepo/benchbranch) +# ./scripts/run_api_benchmark.sh --benchpath s3://ai2-oe-data/jakep/olmocr/olmOCR-bench-1125/ chatgpt:name=test1 +# All other arguments are forwarded to olmocr.bench.convert + +set -e + +# Parse command line arguments +CLUSTER="" +BENCH_BRANCH="" +BENCH_REPO="" +BENCH_PATH="" +BEAKER_IMAGE="" +BEAKER_SECRETS=() +CONVERT_ARGS=() + +# First pass: extract our known arguments +while [[ $# -gt 0 ]]; do + case $1 in + --cluster) + CLUSTER="$2" + shift 2 + ;; + --benchbranch) + BENCH_BRANCH="$2" + shift 2 + ;; + --benchrepo) + BENCH_REPO="$2" + shift 2 + ;; + --benchpath) + BENCH_PATH="$2" + shift 2 + ;; + --beaker-image) + BEAKER_IMAGE="$2" + shift 2 + ;; + --beaker-secret) + # Format: ENV_VAR=secret-name + BEAKER_SECRETS+=("$2") + shift 2 + ;; + *) + # Store args to forward to convert + CONVERT_ARGS+=("$1") + shift + ;; + esac +done + +# Check for mutual exclusivity between benchpath and benchrepo/benchbranch +if [ -n "$BENCH_PATH" ] && ([ -n "$BENCH_REPO" ] || [ -n "$BENCH_BRANCH" ]); then + echo "Error: --benchpath is mutually exclusive with --benchrepo and --benchbranch" + echo "Use either --benchpath OR --benchrepo/--benchbranch, not both." + exit 1 +fi + +# Check that we have at least one convert argument +if [ ${#CONVERT_ARGS[@]} -eq 0 ]; then + echo "Error: No API provider arguments specified for olmocr.bench.convert" + echo "" + echo "Usage examples:" + echo " ./scripts/run_api_benchmark.sh chatgpt:name=test1:prompt=long" + echo " ./scripts/run_api_benchmark.sh gemini:name=test2:api_key=YOUR_KEY" + echo " ./scripts/run_api_benchmark.sh --cluster ai2/titan-cirrascale chatgpt:name=test1" + echo "" + echo "You must provide at least one API provider configuration to benchmark." + exit 1 +fi + +# Check for uncommitted changes +if [ -n "$BEAKER_IMAGE" ]; then + echo "Skipping docker build" +else + if ! git diff-index --quiet HEAD --; then + echo "Error: There are uncommitted changes in the repository." + echo "Please commit or stash your changes before running the benchmark." + echo "" + echo "Uncommitted changes:" + git status --short + exit 1 + fi +fi + +# Use conda environment Python if available, otherwise use system Python +if [ -n "$CONDA_PREFIX" ]; then + PYTHON="$CONDA_PREFIX/bin/python" + echo "Using conda Python from: $CONDA_PREFIX" +else + PYTHON="python" + echo "Warning: No conda environment detected, using system Python" +fi + +# Get version from version.py +VERSION=$($PYTHON -c 'import olmocr.version; print(olmocr.version.VERSION)') +echo "OlmOCR version: $VERSION" + +# Get first 10 characters of git hash +GIT_HASH=$(git rev-parse HEAD | cut -c1-10) +echo "Git hash: $GIT_HASH" + +# Get current git branch name +GIT_BRANCH=$(git rev-parse --abbrev-ref HEAD) +echo "Git branch: $GIT_BRANCH" + +# Check if a Beaker image was provided +if [ -n "$BEAKER_IMAGE" ]; then + echo "Using provided Beaker image: $BEAKER_IMAGE" + IMAGE_TAG="$BEAKER_IMAGE" +else + # Create full image tag + IMAGE_TAG="olmocr-api-benchmark-${VERSION}-${GIT_HASH}" + echo "Building Docker image with tag: $IMAGE_TAG" + + # Build the Docker image + echo "Building Docker image..." + docker build --platform linux/amd64 -f ./Dockerfile -t $IMAGE_TAG . + + # Push image to beaker + echo "Trying to push image to Beaker..." + if ! beaker image create --workspace ai2/oe-data-pdf --name $IMAGE_TAG $IMAGE_TAG 2>/dev/null; then + echo "Warning: Beaker image with tag $IMAGE_TAG already exists. Using existing image." + fi +fi + +# Get Beaker username +BEAKER_USER=$(beaker account whoami --format json | jq -r '.[0].name') +echo "Beaker user: $BEAKER_USER" + +# Create Python script to run beaker experiment +cat << 'EOF' > /tmp/run_api_benchmark_experiment.py +import sys +from beaker import Beaker, ExperimentSpec, TaskSpec, TaskContext, ResultSpec, TaskResources, ImageSource, Priority, Constraints, EnvVar + +# Get image tag, beaker user, git branch, git hash from command line +image_tag = sys.argv[1] +beaker_user = sys.argv[2] +git_branch = sys.argv[3] +git_hash = sys.argv[4] +cluster = None +bench_branch = None +bench_repo = "allenai/olmOCR-bench" # Default repository +bench_path = None +convert_args = [] +beaker_secrets = {} # Dict of ENV_VAR: secret_name + +# Parse remaining arguments +arg_idx = 5 +while arg_idx < len(sys.argv): + if sys.argv[arg_idx] == "--cluster": + cluster = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--benchbranch": + bench_branch = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--benchrepo": + bench_repo = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--benchpath": + bench_path = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--beaker-secret": + # Parse ENV_VAR=secret-name format + secret_spec = sys.argv[arg_idx + 1] + if "=" in secret_spec: + env_var, secret_name = secret_spec.split("=", 1) + beaker_secrets[env_var] = secret_name + arg_idx += 2 + else: + # Everything else is a convert arg + convert_args.append(sys.argv[arg_idx]) + arg_idx += 1 + +# Validate we have convert args +if not convert_args: + print("Error: No API provider arguments provided for convert") + sys.exit(1) + +# Initialize Beaker client +b = Beaker.from_env(default_workspace="ai2/olmocr") + +# Check if AWS credentials secret exists +aws_creds_secret = f"{beaker_user}-AWS_CREDENTIALS_FILE" +try: + # Try to get the secret to see if it exists + b.secret.get(aws_creds_secret, workspace="ai2/olmocr") + has_aws_creds = True + print(f"Found AWS credentials secret: {aws_creds_secret}") +except: + has_aws_creds = False + print(f"AWS credentials secret not found: {aws_creds_secret}") + +# Check if HF_TOKEN secret exists +hf_token_secret = f"{beaker_user}-HF_TOKEN" +try: + # Try to get the secret to see if it exists + b.secret.get(hf_token_secret, workspace="ai2/olmocr") + has_hf_token = True + print(f"Found HuggingFace token secret: {hf_token_secret}") +except: + has_hf_token = False + print(f"HuggingFace token secret not found: {hf_token_secret}") + +# Build commands +commands = [] +if has_aws_creds: + commands.extend([ + "mkdir -p ~/.aws", + 'echo "$AWS_CREDENTIALS_FILE" > ~/.aws/credentials' + ]) + +if has_hf_token: + commands.append('export HF_TOKEN="$HF_TOKEN"') + +# Export any beaker secrets as environment variables +for env_var in beaker_secrets: + commands.append(f'export {env_var}="${env_var}"') + +# Install s5cmd (needed for S3 operations) +commands.append("pip install s5cmd") + +# Handle benchmark data download based on source type +if bench_path: + # If bench_path is provided, use it (can be S3 or local path) + if bench_path.startswith("s3://"): + # S3 path - use s5cmd to download + commands.append(f"s5cmd cp {bench_path.rstrip('/')}/* ./olmOCR-bench/") + else: + # Local path - copy directly + commands.append(f"cp -r {bench_path} ./olmOCR-bench") +else: + # Use HuggingFace download (default behavior) + hf_download_cmd = f"hf download --repo-type dataset {bench_repo} --max-workers 2" + if bench_branch: + hf_download_cmd += f" --revision {bench_branch}" + hf_download_cmd += " --local-dir ./olmOCR-bench" + commands.append(hf_download_cmd) + +# Run convert with forwarded args +convert_cmd = "python -m olmocr.bench.convert" +if convert_args: + convert_cmd += " " + " ".join(convert_args) +convert_cmd += " --dir ./olmOCR-bench/bench_data" +commands.append(convert_cmd) + +# Copy workspace to S3 for archival (using BEAKER_WORKLOAD_ID for unique path) +commands.append("s5cmd cp ./olmOCR-bench/ s3://ai2-oe-data/jakep/olmocr-bench-runs/$BEAKER_WORKLOAD_ID/olmOCR-bench/") + +# Run benchmark +commands.append("python -m olmocr.bench.benchmark --dir ./olmOCR-bench/bench_data") + +# Build task spec with optional env vars +# If image_tag contains '/', it's already a full beaker image reference +if '/' in image_tag: + image_ref = image_tag +else: + image_ref = f"{beaker_user}/{image_tag}" + +task_spec_args = { + "name": "olmocr-api-benchmark", + "image": ImageSource(beaker=image_ref), + "command": [ + "bash", "-c", + " && ".join(commands) + ], + "context": TaskContext( + priority=Priority.normal, + preemptible=True, + ), + "resources": TaskResources(gpu_count=0), + "constraints": Constraints(cluster=[cluster] if cluster else ["ai2/phobos"]), + "result": ResultSpec(path="/noop-results"), +} + +# Add env vars if AWS credentials or HF token exist +env_vars = [] +if has_aws_creds: + env_vars.append(EnvVar(name="AWS_CREDENTIALS_FILE", secret=aws_creds_secret)) +if has_hf_token: + env_vars.append(EnvVar(name="HF_TOKEN", secret=hf_token_secret)) +# Add any additional beaker secrets +for env_var, secret_name in beaker_secrets.items(): + env_vars.append(EnvVar(name=env_var, secret=secret_name)) +if env_vars: + task_spec_args["env_vars"] = env_vars + +# Create a readable name from convert args +# Extract key info like provider name and important parameters +experiment_name_parts = [] +for arg in convert_args: + # Parse provider:param=value format + if ":" in arg: + parts = arg.split(":") + provider = parts[0] + experiment_name_parts.append(provider) + # Look for name parameter specifically + for part in parts[1:]: + if part.startswith("name="): + name_value = part.replace("name=", "") + experiment_name_parts.append(name_value) + break + else: + # Simple argument without colons + experiment_name_parts.append(arg) + +# Create experiment name with convert args info +if experiment_name_parts: + experiment_name = "api-bench-" + "-".join(experiment_name_parts[:3]) # Limit to first 3 parts to avoid overly long names +else: + experiment_name = "api-benchmark" + +print(f"Experiment name: {experiment_name}") + +# Create experiment spec +experiment_spec = ExperimentSpec( + description=f"OlmOCR API Benchmark Run - Branch: {git_branch}, Commit: {git_hash} - Args: {' '.join(convert_args)}", + budget="ai2/oe-base", + tasks=[TaskSpec(**task_spec_args)], + name=experiment_name, +) + +# Create the experiment +experiment = b.experiment.create(spec=experiment_spec, workspace="ai2/olmocr") +print(f"Created API benchmark experiment: {experiment_name} ({experiment.id})") +print(f"View at: https://beaker.org/ex/{experiment.id}") +EOF + +# Run the Python script to create the experiment +echo "Creating Beaker experiment..." + +# Build command with appropriate arguments +CMD="$PYTHON /tmp/run_api_benchmark_experiment.py $IMAGE_TAG $BEAKER_USER $GIT_BRANCH $GIT_HASH" + +if [ -n "$CLUSTER" ]; then + echo "Using cluster: $CLUSTER" + CMD="$CMD --cluster $CLUSTER" +fi + +if [ -n "$BENCH_BRANCH" ]; then + echo "Using bench branch: $BENCH_BRANCH" + CMD="$CMD --benchbranch $BENCH_BRANCH" +fi + +if [ -n "$BENCH_REPO" ]; then + echo "Using bench repo: $BENCH_REPO" + CMD="$CMD --benchrepo $BENCH_REPO" +fi + +if [ -n "$BENCH_PATH" ]; then + echo "Using bench path: $BENCH_PATH" + CMD="$CMD --benchpath $BENCH_PATH" +fi + +# Add beaker secrets if any +if [ ${#BEAKER_SECRETS[@]} -gt 0 ]; then + echo "Using beaker secrets:" + for secret in "${BEAKER_SECRETS[@]}"; do + echo " $secret" + CMD="$CMD --beaker-secret \"$secret\"" + done +fi + +# Add convert args if any +if [ ${#CONVERT_ARGS[@]} -gt 0 ]; then + echo "Forwarding to convert: ${CONVERT_ARGS[*]}" + for arg in "${CONVERT_ARGS[@]}"; do + CMD="$CMD \"$arg\"" + done +fi + +eval $CMD + +# Clean up temporary file +rm /tmp/run_api_benchmark_experiment.py + +echo "API Benchmark experiment submitted successfully!" \ No newline at end of file diff --git a/scripts/run_benchmark.sh b/scripts/run_benchmark.sh new file mode 100755 index 0000000..234d0a0 --- /dev/null +++ b/scripts/run_benchmark.sh @@ -0,0 +1,434 @@ +#!/bin/bash + +# Runs an olmocr-bench run using the full pipeline (no fallback) +# Without model parameter (default behavior): uses the default image from hugging face +# ./scripts/run_benchmark.sh +# With model parameter: for testing custom models +# ./scripts/run_benchmark.sh --model your-model-name +# With cluster parameter: specify a specific cluster to use +# ./scripts/run_benchmark.sh --cluster ai2/titan-cirrascale +# With beaker image: skip Docker build and use provided Beaker image +# ./scripts/run_benchmark.sh --beaker-image jakep/olmocr-benchmark-0.3.3-780bc7d934 +# With repeats parameter: run the pipeline multiple times for increased accuracy (default: 1) +# ./scripts/run_benchmark.sh --repeats 3 +# With noperf flag: skip the performance test job +# ./scripts/run_benchmark.sh --noperf +# With benchrepo parameter: use a different benchmark dataset repository (default: allenai/olmOCR-bench) +# ./scripts/run_benchmark.sh --benchrepo allenai/olmOCR-bench-internal +# With benchbranch parameter: use a specific branch/revision of the benchmark dataset +# ./scripts/run_benchmark.sh --benchbranch olmOCR-bench-1125 +# With benchpath parameter: use benchmark dataset from a local path or S3 path (mutually exclusive with benchrepo/benchbranch) +# ./scripts/run_benchmark.sh --benchpath s3://ai2-oe-data/jakep/olmocr/olmOCR-bench-1125/ +# With max-tokens parameter: set max tokens for model output per page (default: 8000) +# ./scripts/run_benchmark.sh --max-tokens 16000 + +set -e + +# Parse command line arguments +MODEL="" +CLUSTER="" +BENCH_BRANCH="" +BENCH_REPO="" +BENCH_PATH="" +BEAKER_IMAGE="" +REPEATS="1" +NOPERF="" +MAX_TOKENS="" +while [[ $# -gt 0 ]]; do + case $1 in + --model) + MODEL="$2" + shift 2 + ;; + --cluster) + CLUSTER="$2" + shift 2 + ;; + --benchbranch) + BENCH_BRANCH="$2" + shift 2 + ;; + --benchrepo) + BENCH_REPO="$2" + shift 2 + ;; + --benchpath) + BENCH_PATH="$2" + shift 2 + ;; + --beaker-image) + BEAKER_IMAGE="$2" + shift 2 + ;; + --repeats) + REPEATS="$2" + shift 2 + ;; + --noperf) + NOPERF="1" + shift + ;; + --max-tokens) + MAX_TOKENS="$2" + shift 2 + ;; + *) + echo "Unknown option: $1" + echo "Usage: $0 [--model MODEL_NAME] [--cluster CLUSTER_NAME] [--benchbranch BRANCH_NAME] [--benchrepo REPO_URL] [--benchpath PATH] [--beaker-image IMAGE_NAME] [--repeats NUMBER] [--noperf] [--max-tokens NUMBER]" + exit 1 + ;; + esac +done + +# Check for mutual exclusivity between benchpath and benchrepo/benchbranch +if [ -n "$BENCH_PATH" ] && ([ -n "$BENCH_REPO" ] || [ -n "$BENCH_BRANCH" ]); then + echo "Error: --benchpath is mutually exclusive with --benchrepo and --benchbranch" + echo "Use either --benchpath OR --benchrepo/--benchbranch, not both." + exit 1 +fi + +# Check for uncommitted changes +if [ -n "$BEAKER_IMAGE" ]; then + echo "Skipping docker build" +else + if ! git diff-index --quiet HEAD --; then + echo "Error: There are uncommitted changes in the repository." + echo "Please commit or stash your changes before running the benchmark." + echo "" + echo "Uncommitted changes:" + git status --short + exit 1 + fi +fi + +# Use conda environment Python if available, otherwise use system Python +if [ -n "$CONDA_PREFIX" ]; then + PYTHON="$CONDA_PREFIX/bin/python" + echo "Using conda Python from: $CONDA_PREFIX" +else + PYTHON="python" + echo "Warning: No conda environment detected, using system Python" +fi + +# Get version from version.py +VERSION=$($PYTHON -c 'import olmocr.version; print(olmocr.version.VERSION)') +echo "OlmOCR version: $VERSION" + +# Get first 10 characters of git hash +GIT_HASH=$(git rev-parse HEAD | cut -c1-10) +echo "Git hash: $GIT_HASH" + +# Get current git branch name +GIT_BRANCH=$(git rev-parse --abbrev-ref HEAD) +echo "Git branch: $GIT_BRANCH" + +# Check if a Beaker image was provided +if [ -n "$BEAKER_IMAGE" ]; then + echo "Using provided Beaker image: $BEAKER_IMAGE" + IMAGE_TAG="$BEAKER_IMAGE" +else + # Create full image tag + IMAGE_TAG="olmocr-benchmark-${VERSION}-${GIT_HASH}" + echo "Building Docker image with tag: $IMAGE_TAG" + + # Build the Docker image + echo "Building Docker image..." + docker build --platform linux/amd64 -f ./Dockerfile -t $IMAGE_TAG . + + # Push image to beaker + echo "Trying to push image to Beaker..." + if ! beaker image create --workspace ai2/oe-data-pdf --name $IMAGE_TAG $IMAGE_TAG 2>/dev/null; then + echo "Warning: Beaker image with tag $IMAGE_TAG already exists. Using existing image." + fi +fi + +# Get Beaker username +BEAKER_USER=$(beaker account whoami --format json | jq -r '.[0].name') +echo "Beaker user: $BEAKER_USER" + +# Create Python script to run beaker experiment +cat << 'EOF' > /tmp/run_benchmark_experiment.py +import sys +from beaker import Beaker, BeakerExperimentSpec, BeakerTaskSpec, BeakerTaskContext, BeakerResultSpec, BeakerTaskResources, BeakerImageSource, BeakerJobPriority, BeakerConstraints, BeakerEnvVar + +# Get image tag, beaker user, git branch, git hash, optional model, cluster, bench branch, bench repo, and repeats from command line +image_tag = sys.argv[1] +beaker_user = sys.argv[2] +git_branch = sys.argv[3] +git_hash = sys.argv[4] +model = None +cluster = None +bench_branch = None +bench_repo = "allenai/olmOCR-bench" # Default repository +bench_path = None +repeats = 1 +noperf = False +max_tokens = None + +# Parse remaining arguments +arg_idx = 5 +while arg_idx < len(sys.argv): + if sys.argv[arg_idx] == "--cluster": + cluster = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--benchbranch": + bench_branch = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--benchrepo": + bench_repo = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--benchpath": + bench_path = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--repeats": + repeats = int(sys.argv[arg_idx + 1]) + arg_idx += 2 + elif sys.argv[arg_idx] == "--noperf": + noperf = True + arg_idx += 1 + elif sys.argv[arg_idx] == "--max-tokens": + max_tokens = int(sys.argv[arg_idx + 1]) + arg_idx += 2 + else: + model = sys.argv[arg_idx] + arg_idx += 1 + +# Initialize Beaker client +b = Beaker.from_env(default_workspace="ai2/olmocr") + +# Note: pipeline commands will be built in the loop based on repeats + +# Check if AWS credentials secret exists +aws_creds_secret = f"{beaker_user}-AWS_CREDENTIALS_FILE" +try: + # Try to get the secret to see if it exists + b.secret.get(aws_creds_secret, workspace="ai2/olmocr") + has_aws_creds = True + print(f"Found AWS credentials secret: {aws_creds_secret}") +except: + has_aws_creds = False + print(f"AWS credentials secret not found: {aws_creds_secret}") + +# Check if HF_TOKEN secret exists +hf_token_secret = f"{beaker_user}-HF_TOKEN" +try: + # Try to get the secret to see if it exists + b.secret.get(hf_token_secret, workspace="ai2/olmocr") + has_hf_token = True + print(f"Found HuggingFace token secret: {hf_token_secret}") +except: + has_hf_token = False + print(f"HuggingFace token secret not found: {hf_token_secret}") + +# First experiment: Original benchmark job +commands = [] +if has_aws_creds: + commands.extend([ + "mkdir -p ~/.aws", + 'echo "$AWS_CREDENTIALS_FILE" > ~/.aws/credentials' + ]) + +if has_hf_token: + commands.append('export HF_TOKEN="$HF_TOKEN"') + +# Install s5cmd (needed for S3 operations) +commands.append("pip install s5cmd") + +# Handle benchmark data download based on source type +if bench_path: + # If bench_path is provided, use it (can be S3 or local path) + if bench_path.startswith("s3://"): + # S3 path - use s5cmd to download + commands.append(f"s5cmd cp {bench_path.rstrip('/')}/* ./olmOCR-bench/") + else: + # Local path - copy directly + commands.append(f"cp -r {bench_path} ./olmOCR-bench") +else: + # Use HuggingFace download (default behavior) + hf_download_cmd = f"hf download --repo-type dataset {bench_repo} --max-workers 2" + if bench_branch: + hf_download_cmd += f" --revision {bench_branch}" + hf_download_cmd += " --local-dir ./olmOCR-bench" + commands.append(hf_download_cmd) + +# Run pipeline multiple times based on repeats +for i in range(1, repeats + 1): + workspace_dir = f"./localworkspace{i}" + pipeline_cmd = f"python -m olmocr.pipeline {workspace_dir} --markdown --pdfs ./olmOCR-bench/bench_data/pdfs/**/*.pdf" + if model: + pipeline_cmd += f" --model {model}" + if max_tokens: + pipeline_cmd += f" --max_tokens {max_tokens}" + commands.append(pipeline_cmd) + +# Process all workspaces with workspace_to_bench.py +for i in range(1, repeats + 1): + workspace_dir = f"localworkspace{i}/" + workspace_to_bench_cmd = f"python olmocr/bench/scripts/workspace_to_bench.py {workspace_dir} olmOCR-bench/bench_data/olmocr --bench-path ./olmOCR-bench/ --repeat-index {i}" + commands.append(workspace_to_bench_cmd) + +# Copy each workspace to S3 +for i in range(1, repeats + 1): + workspace_dir = f"localworkspace{i}/" + commands.append(f"s5cmd cp {workspace_dir} s3://ai2-oe-data/jakep/olmocr-bench-runs/$BEAKER_WORKLOAD_ID/workspace{i}/") + +commands.append("python -m olmocr.bench.benchmark --dir ./olmOCR-bench/bench_data") + +# Build task spec with optional env vars +# If image_tag contains '/', it's already a full beaker image reference +if '/' in image_tag: + image_ref = image_tag +else: + image_ref = f"{beaker_user}/{image_tag}" + +task_spec_args = { + "name": "olmocr-benchmark", + "image": BeakerImageSource(beaker=image_ref), + "command": [ + "bash", "-c", + " && ".join(commands) + ], + "context": BeakerTaskContext( + priority=BeakerJobPriority["normal"], + preemptible=True, + ), + "resources": BeakerTaskResources(gpu_count=1), + "constraints": BeakerConstraints(cluster=[cluster] if cluster else ["ai2/ceres-cirrascale", "ai2/jupiter-cirrascale-2"]), + "result": BeakerResultSpec(path="/noop-results"), +} + +# Add env vars if AWS credentials or HF token exist +env_vars = [] +if has_aws_creds: + env_vars.append(BeakerEnvVar(name="AWS_CREDENTIALS_FILE", secret=aws_creds_secret)) +if has_hf_token: + env_vars.append(BeakerEnvVar(name="HF_TOKEN", secret=hf_token_secret)) +if env_vars: + task_spec_args["env_vars"] = env_vars + +# Create first experiment spec +experiment_spec = BeakerExperimentSpec( + description=f"OlmOCR Benchmark Run - Branch: {git_branch}, Commit: {git_hash}", + budget="ai2/oe-base", + tasks=[BeakerTaskSpec(**task_spec_args)], +) + +# Create the first experiment +workload = b.experiment.create(spec=experiment_spec, workspace="ai2/olmocr") +print(f"Created benchmark experiment: {workload.experiment.id}") +print(f"View at: https://beaker.org/ex/{workload.experiment.id}") +print("-------") +print("") + +# Second experiment: Performance test job (only if --noperf not specified) +if not noperf: + perf_pipeline_cmd = "python -m olmocr.pipeline ./localworkspace1 --markdown --pdfs s3://ai2-oe-data/jakep/olmocr/olmOCR-mix-0225/benchmark_set/*.pdf" + if model: + perf_pipeline_cmd += f" --model {model}" + if max_tokens: + perf_pipeline_cmd += f" --max_tokens {max_tokens}" + + perf_commands = [] + if has_aws_creds: + perf_commands.extend([ + "mkdir -p ~/.aws", + 'echo "$AWS_CREDENTIALS_FILE" > ~/.aws/credentials' + ]) + if has_hf_token: + perf_commands.append('export HF_TOKEN="$HF_TOKEN"') + perf_commands.append(perf_pipeline_cmd) + + # Build performance task spec + perf_task_spec_args = { + "name": "olmocr-performance", + "image": BeakerImageSource(beaker=image_ref), + "command": [ + "bash", "-c", + " && ".join(perf_commands) + ], + "context": BeakerTaskContext( + priority=BeakerJobPriority["normal"], + preemptible=True, + ), + # Need to reserve all 8 gpus for performance spec or else benchmark results can be off (1 for titan-cirrascale) + "resources": BeakerTaskResources(gpu_count=1 if cluster == "ai2/titan-cirrascale" else 8), + "constraints": BeakerConstraints(cluster=[cluster] if cluster else ["ai2/ceres-cirrascale", "ai2/jupiter-cirrascale-2"]), + "result": BeakerResultSpec(path="/noop-results"), + } + + # Add env vars if AWS credentials or HF token exist + env_vars = [] + if has_aws_creds: + env_vars.append(BeakerEnvVar(name="AWS_CREDENTIALS_FILE", secret=aws_creds_secret)) + if has_hf_token: + env_vars.append(BeakerEnvVar(name="HF_TOKEN", secret=hf_token_secret)) + if env_vars: + perf_task_spec_args["env_vars"] = env_vars + + # Create performance experiment spec + perf_experiment_spec = BeakerExperimentSpec( + description=f"OlmOCR Performance Test - Branch: {git_branch}, Commit: {git_hash}", + budget="ai2/oe-base", + tasks=[BeakerTaskSpec(**perf_task_spec_args)], + ) + + # Create the performance experiment + perf_workload = b.experiment.create(spec=perf_experiment_spec, workspace="ai2/olmocr") + print(f"Created performance experiment: {perf_workload.experiment.id}") + print(f"View at: https://beaker.org/ex/{perf_workload.experiment.id}") +else: + print("Skipping performance test (--noperf flag specified)") +EOF + +# Run the Python script to create the experiments +echo "Creating Beaker experiments..." + +# Build command with appropriate arguments +CMD="$PYTHON /tmp/run_benchmark_experiment.py $IMAGE_TAG $BEAKER_USER $GIT_BRANCH $GIT_HASH" + +if [ -n "$MODEL" ]; then + echo "Using model: $MODEL" + CMD="$CMD $MODEL" +fi + +if [ -n "$CLUSTER" ]; then + echo "Using cluster: $CLUSTER" + CMD="$CMD --cluster $CLUSTER" +fi + +if [ -n "$BENCH_BRANCH" ]; then + echo "Using bench branch: $BENCH_BRANCH" + CMD="$CMD --benchbranch $BENCH_BRANCH" +fi + +if [ -n "$BENCH_REPO" ]; then + echo "Using bench repo: $BENCH_REPO" + CMD="$CMD --benchrepo $BENCH_REPO" +fi + +if [ -n "$BENCH_PATH" ]; then + echo "Using bench path: $BENCH_PATH" + CMD="$CMD --benchpath $BENCH_PATH" +fi + +if [ "$REPEATS" != "1" ]; then + echo "Using repeats: $REPEATS" + CMD="$CMD --repeats $REPEATS" +fi + +if [ -n "$NOPERF" ]; then + echo "Skipping performance tests" + CMD="$CMD --noperf" +fi + +if [ -n "$MAX_TOKENS" ]; then + echo "Using max tokens: $MAX_TOKENS" + CMD="$CMD --max-tokens $MAX_TOKENS" +fi + +eval $CMD + +# Clean up temporary file +rm /tmp/run_benchmark_experiment.py + +echo "Benchmark experiments submitted successfully!" \ No newline at end of file diff --git a/scripts/run_benchmark_guided_decoding.sh b/scripts/run_benchmark_guided_decoding.sh new file mode 100755 index 0000000..8930eee --- /dev/null +++ b/scripts/run_benchmark_guided_decoding.sh @@ -0,0 +1,220 @@ +#!/bin/bash + +# Runs an olmocr-bench run using the full pipeline (no fallback) +# Without model parameter (default behavior):, uses the default image from hugging face +# ./scripts/run_benchmark.sh +# With model parameter: for testing custom models +# ./scripts/run_benchmark.sh --model your-model-name + +set -e + +# Parse command line arguments +MODEL="" +while [[ $# -gt 0 ]]; do + case $1 in + --model) + MODEL="$2" + shift 2 + ;; + *) + echo "Unknown option: $1" + echo "Usage: $0 [--model MODEL_NAME]" + exit 1 + ;; + esac +done + +# Check for uncommitted changes +if ! git diff-index --quiet HEAD --; then + echo "Error: There are uncommitted changes in the repository." + echo "Please commit or stash your changes before running the benchmark." + echo "" + echo "Uncommitted changes:" + git status --short + exit 1 +fi + +# Use conda environment Python if available, otherwise use system Python +if [ -n "$CONDA_PREFIX" ]; then + PYTHON="$CONDA_PREFIX/bin/python" + echo "Using conda Python from: $CONDA_PREFIX" +else + PYTHON="python" + echo "Warning: No conda environment detected, using system Python" +fi + +# Get version from version.py +VERSION=$($PYTHON -c 'import olmocr.version; print(olmocr.version.VERSION)') +echo "OlmOCR version: $VERSION" + +# Get first 10 characters of git hash +GIT_HASH=$(git rev-parse HEAD | cut -c1-10) +echo "Git hash: $GIT_HASH" + +# Get current git branch name +GIT_BRANCH=$(git rev-parse --abbrev-ref HEAD) +echo "Git branch: $GIT_BRANCH" + +# Create full image tag +IMAGE_TAG="olmocr-benchmark-${VERSION}-${GIT_HASH}" +echo "Building Docker image with tag: $IMAGE_TAG" + +# Build the Docker image +echo "Building Docker image..." +docker build --platform linux/amd64 -f ./Dockerfile -t $IMAGE_TAG . + +# Get Beaker username +BEAKER_USER=$(beaker account whoami --format json | jq -r '.[0].name') +echo "Beaker user: $BEAKER_USER" + +# Push image to beaker +echo "Trying to push image to Beaker..." +if ! beaker image create --workspace ai2/oe-data-pdf --name $IMAGE_TAG $IMAGE_TAG 2>/dev/null; then + echo "Warning: Beaker image with tag $IMAGE_TAG already exists. Using existing image." +fi + +# Create Python script to run beaker experiment +cat << 'EOF' > /tmp/run_benchmark_experiment.py +import sys +from beaker import Beaker, ExperimentSpec, TaskSpec, TaskContext, ResultSpec, TaskResources, ImageSource, Priority, Constraints, EnvVar + +# Get image tag, beaker user, git branch, git hash, and optional model from command line +image_tag = sys.argv[1] +beaker_user = sys.argv[2] +git_branch = sys.argv[3] +git_hash = sys.argv[4] +model = sys.argv[5] if len(sys.argv) > 5 else None + +# Initialize Beaker client +b = Beaker.from_env(default_workspace="ai2/olmocr") + +# Build the pipeline command with optional model parameter +pipeline_cmd = "python -m olmocr.pipeline ./localworkspace --guided_decoding --markdown --pdfs ./olmOCR-bench/bench_data/pdfs/**/*.pdf" +if model: + pipeline_cmd += f" --model {model}" + +# Check if AWS credentials secret exists +aws_creds_secret = f"{beaker_user}-AWS_CREDENTIALS_FILE" +try: + # Try to get the secret to see if it exists + b.secret.get(aws_creds_secret, workspace="ai2/olmocr") + has_aws_creds = True + print(f"Found AWS credentials secret: {aws_creds_secret}") +except: + has_aws_creds = False + print(f"AWS credentials secret not found: {aws_creds_secret}") + +# First experiment: Original benchmark job +commands = [] +if has_aws_creds: + commands.extend([ + "mkdir -p ~/.aws", + 'echo "$AWS_CREDENTIALS_FILE" > ~/.aws/credentials' + ]) +commands.extend([ + "git clone https://huggingface.co/datasets/allenai/olmOCR-bench", + "cd olmOCR-bench && git lfs pull && cd ..", + pipeline_cmd, + "python olmocr/bench/scripts/workspace_to_bench.py localworkspace/ olmOCR-bench/bench_data/olmocr --bench-path ./olmOCR-bench/", + "python -m olmocr.bench.benchmark --dir ./olmOCR-bench/bench_data" +]) + +# Build task spec with optional env vars +task_spec_args = { + "name": "olmocr-benchmark", + "image": ImageSource(beaker=f"{beaker_user}/{image_tag}"), + "command": [ + "bash", "-c", + " && ".join(commands) + ], + "context": TaskContext( + priority=Priority.normal, + preemptible=True, + ), + "resources": TaskResources(gpu_count=1), + "constraints": Constraints(cluster=["ai2/ceres-cirrascale", "ai2/jupiter-cirrascale-2"]), + "result": ResultSpec(path="/noop-results"), +} + +# Add env vars if AWS credentials exist +if has_aws_creds: + task_spec_args["env_vars"] = [ + EnvVar(name="AWS_CREDENTIALS_FILE", secret=aws_creds_secret) + ] + +# Create first experiment spec +experiment_spec = ExperimentSpec( + description=f"OlmOCR Benchmark Run - Branch: {git_branch}, Commit: {git_hash}", + budget="ai2/oe-base", + tasks=[TaskSpec(**task_spec_args)], +) + +# Create the first experiment +experiment = b.experiment.create(spec=experiment_spec, workspace="ai2/olmocr") +print(f"Created benchmark experiment: {experiment.id}") +print(f"View at: https://beaker.org/ex/{experiment.id}") +print("-------") +print("") + +# Second experiment: Performance test job +perf_pipeline_cmd = "python -m olmocr.pipeline ./localworkspace --guided_decoding --markdown --pdfs s3://ai2-oe-data/jakep/olmocr/olmOCR-mix-0225/benchmark_set/*.pdf" +if model: + perf_pipeline_cmd += f" --model {model}" + +perf_commands = [] +if has_aws_creds: + perf_commands.extend([ + "mkdir -p ~/.aws", + 'echo "$AWS_CREDENTIALS_FILE" > ~/.aws/credentials' + ]) +perf_commands.append(perf_pipeline_cmd) + +# Build performance task spec +perf_task_spec_args = { + "name": "olmocr-performance", + "image": ImageSource(beaker=f"{beaker_user}/{image_tag}"), + "command": [ + "bash", "-c", + " && ".join(perf_commands) + ], + "context": TaskContext( + priority=Priority.normal, + preemptible=True, + ), + "resources": TaskResources(gpu_count=1), + "constraints": Constraints(cluster=["ai2/ceres-cirrascale", "ai2/jupiter-cirrascale-2"]), + "result": ResultSpec(path="/noop-results"), +} + +# Add env vars if AWS credentials exist +if has_aws_creds: + perf_task_spec_args["env_vars"] = [ + EnvVar(name="AWS_CREDENTIALS_FILE", secret=aws_creds_secret) + ] + +# Create performance experiment spec +perf_experiment_spec = ExperimentSpec( + description=f"OlmOCR Performance Test - Branch: {git_branch}, Commit: {git_hash}", + budget="ai2/oe-base", + tasks=[TaskSpec(**perf_task_spec_args)], +) + +# Create the performance experiment +perf_experiment = b.experiment.create(spec=perf_experiment_spec, workspace="ai2/olmocr") +print(f"Created performance experiment: {perf_experiment.id}") +print(f"View at: https://beaker.org/ex/{perf_experiment.id}") +EOF + +# Run the Python script to create the experiments +echo "Creating Beaker experiments..." +if [ -n "$MODEL" ]; then + echo "Using model: $MODEL" + $PYTHON /tmp/run_benchmark_experiment.py $IMAGE_TAG $BEAKER_USER $GIT_BRANCH $GIT_HASH "$MODEL" +else + $PYTHON /tmp/run_benchmark_experiment.py $IMAGE_TAG $BEAKER_USER $GIT_BRANCH $GIT_HASH +fi + +# Clean up temporary file +rm /tmp/run_benchmark_experiment.py + +echo "Benchmark experiments submitted successfully!" \ No newline at end of file diff --git a/scripts/run_chandra_benchmark.sh b/scripts/run_chandra_benchmark.sh new file mode 100755 index 0000000..3fc62fe --- /dev/null +++ b/scripts/run_chandra_benchmark.sh @@ -0,0 +1,528 @@ +#!/bin/bash + +# Runs chandra benchmark, measuring both olmOCR-bench performance and per document processing performance +# Usage: +# ./scripts/run_chandra_benchmark.sh # Use latest chandra version, default benchmark repo +# ./scripts/run_chandra_benchmark.sh 0.1.0 # Use specific chandra version +# ./scripts/run_chandra_benchmark.sh --version 0.1.0 # Use specific chandra version (explicit flag) +# ./scripts/run_chandra_benchmark.sh --benchrepo allenai/olmOCR-bench-internal # Use different benchmark repo +# ./scripts/run_chandra_benchmark.sh --benchbranch olmOCR-bench-1125 # Use specific branch/revision +# ./scripts/run_chandra_benchmark.sh --benchpath s3://ai2-oe-data/path/ # Use benchmark from S3 or local path +# ./scripts/run_chandra_benchmark.sh --cluster ai2/titan-cirrascale # Specify a cluster +# ./scripts/run_chandra_benchmark.sh --beaker-image jakep/olmocr-benchmark-0.3.3-780bc7d934 # Skip Docker build +# ./scripts/run_chandra_benchmark.sh --noperf # Skip the performance test job +# ./scripts/run_chandra_benchmark.sh --model datalab-to/chandra-ocr-2 # Use a specific model + +set -e + +# Parse command line arguments +CHANDRA_VERSION="" +BENCH_BRANCH="" +BENCH_REPO="" +BENCH_PATH="" +CLUSTER="" +BEAKER_IMAGE="" +NOPERF="" +MODEL="" + +# Default to latest if no version specified +while [[ $# -gt 0 ]]; do + case $1 in + --version) + CHANDRA_VERSION="$2" + shift 2 + ;; + --benchbranch) + BENCH_BRANCH="$2" + shift 2 + ;; + --benchrepo) + BENCH_REPO="$2" + shift 2 + ;; + --benchpath) + BENCH_PATH="$2" + shift 2 + ;; + --cluster) + CLUSTER="$2" + shift 2 + ;; + --beaker-image) + BEAKER_IMAGE="$2" + shift 2 + ;; + --noperf) + NOPERF="1" + shift + ;; + --model) + MODEL="$2" + shift 2 + ;; + *) + # If no flag, assume it's the version for backward compatibility + if [ -z "$CHANDRA_VERSION" ]; then + CHANDRA_VERSION="$1" + else + echo "Unknown option: $1" + echo "Usage: $0 [VERSION] [--version VERSION] [--benchbranch BRANCH] [--benchrepo REPO] [--benchpath PATH] [--cluster CLUSTER] [--beaker-image IMAGE] [--noperf] [--model MODEL]" + exit 1 + fi + shift + ;; + esac +done + +# Set default chandra version if not specified +if [ -z "$CHANDRA_VERSION" ]; then + CHANDRA_VERSION="latest" +fi + +if [ "$CHANDRA_VERSION" = "latest" ]; then + echo "Using latest chandra-ocr release" + CHANDRA_INSTALL_CMD="uv pip install --system chandra-ocr" +else + echo "Using chandra-ocr version: $CHANDRA_VERSION" + CHANDRA_INSTALL_CMD="uv pip install --system chandra-ocr==$CHANDRA_VERSION" +fi + +# Check for mutual exclusivity between benchpath and benchrepo/benchbranch +if [ -n "$BENCH_PATH" ] && ([ -n "$BENCH_REPO" ] || [ -n "$BENCH_BRANCH" ]); then + echo "Error: --benchpath is mutually exclusive with --benchrepo and --benchbranch" + echo "Use either --benchpath OR --benchrepo/--benchbranch, not both." + exit 1 +fi + +# Use conda environment Python if available, otherwise use system Python +if [ -n "$CONDA_PREFIX" ]; then + PYTHON="$CONDA_PREFIX/bin/python" + echo "Using conda Python from: $CONDA_PREFIX" +else + PYTHON="python" + echo "Warning: No conda environment detected, using system Python" +fi + +# Get version from version.py +VERSION=$($PYTHON -c 'import olmocr.version; print(olmocr.version.VERSION)') +echo "OlmOCR version: $VERSION" + +# Get first 10 characters of git hash +GIT_HASH=$(git rev-parse HEAD | cut -c1-10) +echo "Git hash: $GIT_HASH" + +# Get current git branch name +GIT_BRANCH=$(git rev-parse --abbrev-ref HEAD) +echo "Git branch: $GIT_BRANCH" + +# Check if a Beaker image was provided +if [ -n "$BEAKER_IMAGE" ]; then + echo "Using provided Beaker image: $BEAKER_IMAGE" + IMAGE_TAG="$BEAKER_IMAGE" +else + # Create full image tag + IMAGE_TAG="olmocr-benchmark-${VERSION}-${GIT_HASH}" + echo "Building Docker image with tag: $IMAGE_TAG" + + # Build the Docker image + echo "Building Docker image..." + docker build --platform linux/amd64 -f ./Dockerfile -t $IMAGE_TAG . + + # Push image to beaker + echo "Trying to push image to Beaker..." + if ! beaker image create --workspace ai2/oe-data-pdf --name $IMAGE_TAG $IMAGE_TAG 2>/dev/null; then + echo "Warning: Beaker image with tag $IMAGE_TAG already exists. Using existing image." + fi +fi + +# Get Beaker username +BEAKER_USER=$(beaker account whoami --format json | jq -r '.[0].name') +echo "Beaker user: $BEAKER_USER" + +# Create Python script to run beaker experiment +cat << 'EOF' > /tmp/run_benchmark_experiment.py +import sys +from textwrap import dedent +from beaker import Beaker, BeakerExperimentSpec, BeakerTaskSpec, BeakerTaskContext, BeakerResultSpec, BeakerTaskResources, BeakerImageSource, BeakerJobPriority, BeakerConstraints, BeakerEnvVar + +# Get image tag, beaker user, git branch, git hash, and chandra version from command line +image_tag = sys.argv[1] +beaker_user = sys.argv[2] +git_branch = sys.argv[3] +git_hash = sys.argv[4] +chandra_version = sys.argv[5] +chandra_install_cmd = sys.argv[6] + +# Initialize benchmark dataset parameters +bench_branch = None +bench_repo = "allenai/olmOCR-bench" # Default repository +bench_path = None +cluster = None +noperf = False +model = None + +# Parse additional arguments +arg_idx = 7 +while arg_idx < len(sys.argv): + if sys.argv[arg_idx] == "--benchbranch": + bench_branch = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--benchrepo": + bench_repo = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--benchpath": + bench_path = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--cluster": + cluster = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--noperf": + noperf = True + arg_idx += 1 + elif sys.argv[arg_idx] == "--model": + model = sys.argv[arg_idx + 1] + arg_idx += 2 + else: + print(f"Unknown argument: {sys.argv[arg_idx]}") + arg_idx += 1 + +# Default model for chandra +chandra_model = model if model else "datalab-to/chandra-ocr-2" + +# Initialize Beaker client +b = Beaker.from_env(default_workspace="ai2/olmocr") + +# Check if AWS credentials secret exists +aws_creds_secret = f"{beaker_user}-AWS_CREDENTIALS_FILE" +try: + # Try to get the secret to see if it exists + b.secret.get(aws_creds_secret, workspace="ai2/olmocr") + has_aws_creds = True + print(f"Found AWS credentials secret: {aws_creds_secret}") +except: + has_aws_creds = False + print(f"AWS credentials secret not found: {aws_creds_secret}") + +# Shell script to run Chandra conversions for benchmark +run_chandra_shell = dedent("""\ +bash -lc 'set -euo pipefail +PDF_ROOT="olmOCR-bench/bench_data/pdfs" +TARGET_ROOT="olmOCR-bench/bench_data/chandra" +rm -rf "$TARGET_ROOT" +mkdir -p "$TARGET_ROOT" + +# Start vllm server in background +echo "Starting vllm server for Chandra..." +vllm serve __CHANDRA_MODEL__ --served-model-name chandra > /tmp/vllm_server.log 2>&1 & +VLLM_PID=$! + +# Wait for vllm server to be ready +echo "Waiting for vllm server to start..." +for i in {1..600}; do + if curl -s http://localhost:8000/health > /dev/null 2>&1; then + echo "vllm server is ready" + break + fi + if [ $i -eq 600 ]; then + echo "Error: vllm server failed to start after 600 seconds" + cat /tmp/vllm_server.log + exit 1 + fi + sleep 1 +done + +# Process each folder separately to maintain structure +echo "Running Chandra conversions..." +for folder in "$PDF_ROOT"/*; do + if [ ! -d "$folder" ]; then + continue + fi + section=$(basename "$folder") + output_dir="$HOME/chandra_bench_${section}" + rm -rf "$output_dir" + mkdir -p "$output_dir" + + echo " Processing $folder -> $output_dir" + chandra "$folder" "$output_dir" --method vllm +done + +echo "Collecting Chandra markdown outputs..." +# For each PDF, find its corresponding markdown and copy to proper location +find "$PDF_ROOT" -type f -name "*.pdf" | while IFS= read -r pdf_path; do + # Get relative path from PDF root + rel_path=${pdf_path#"$PDF_ROOT"/} + + # Extract section name (first directory level) + case "$rel_path" in + */*) + section=${rel_path%%/*} + ;; + *) + echo "Warning: Unexpected PDF path layout for $pdf_path, skipping" >&2 + continue + ;; + esac + + # Get PDF name without extension + pdf_name=$(basename "$pdf_path" .pdf) + + # Source markdown location: ~/chandra_bench_${section}/${pdf_name}/${pdf_name}.md + src_md="$HOME/chandra_bench_${section}/${pdf_name}/${pdf_name}.md" + + if [ ! -f "$src_md" ]; then + echo "Warning: No markdown output found at $src_md" >&2 + continue + fi + + # Target location: olmOCR-bench/bench_data/chandra/${section}/${pdf_name}_pg1_repeat1.md + target_dir="$TARGET_ROOT/$section" + mkdir -p "$target_dir" + target_path="$target_dir/${pdf_name}_pg1_repeat1.md" + + cp "$src_md" "$target_path" + echo " Copied $src_md -> $target_path" +done + +# Kill vllm server +echo "Stopping vllm server..." +kill $VLLM_PID || true +wait $VLLM_PID 2>/dev/null || true +'""").replace("__CHANDRA_MODEL__", chandra_model) + +# Check if HF_TOKEN secret exists +hf_token_secret = f"{beaker_user}-HF_TOKEN" +try: + # Try to get the secret to see if it exists + b.secret.get(hf_token_secret, workspace="ai2/olmocr") + has_hf_token = True + print(f"Found HuggingFace token secret: {hf_token_secret}") +except: + has_hf_token = False + print(f"HuggingFace token secret not found: {hf_token_secret}") + +# First experiment: Original benchmark job +commands = [] +if has_aws_creds: + commands.extend([ + "mkdir -p ~/.aws", + 'echo "$AWS_CREDENTIALS_FILE" > ~/.aws/credentials' + ]) + +if has_hf_token: + commands.append('export HF_TOKEN="$HF_TOKEN"') + +# Install uv for fast dependency management, then s5cmd (needed for S3 operations) +commands.append("pip install uv") +commands.append("uv pip install --system s5cmd") + +# Handle benchmark data download based on source type +if bench_path: + # If bench_path is provided, use it (can be S3 or local path) + if bench_path.startswith("s3://"): + # S3 path - use s5cmd to download + commands.append(f"s5cmd cp {bench_path.rstrip('/')}/* ./olmOCR-bench/") + else: + # Local path - copy directly + commands.append(f"cp -r {bench_path} ./olmOCR-bench") +else: + # Use HuggingFace download (default behavior) + hf_download_cmd = f"hf download --repo-type dataset {bench_repo} --max-workers 2" + if bench_branch: + hf_download_cmd += f" --revision {bench_branch}" + hf_download_cmd += " --local-dir ./olmOCR-bench" + commands.append(hf_download_cmd) + +commands.extend([ + chandra_install_cmd, + "uv pip uninstall --system vllm flashinfer flashinfer-python flashinfer-cubin || true", + "uv pip install --system vllm --torch-backend=auto", + run_chandra_shell, + "python -m olmocr.bench.benchmark --dir ./olmOCR-bench/bench_data --candidate chandra" +]) + +# Build task spec with optional env vars +# If image_tag contains '/', it's already a full beaker image reference +if '/' in image_tag: + image_ref = image_tag +else: + image_ref = f"{beaker_user}/{image_tag}" + +task_spec_args = { + "name": "chandra-benchmark", + "image": BeakerImageSource(beaker=image_ref), + "command": [ + "bash", "-c", + " && ".join(commands) + ], + "context": BeakerTaskContext( + priority=BeakerJobPriority["normal"], + preemptible=True, + ), + "resources": BeakerTaskResources(gpu_count=1), + "constraints": BeakerConstraints(cluster=[cluster] if cluster else ["ai2/ceres-cirrascale", "ai2/jupiter-cirrascale-2"]), + "result": BeakerResultSpec(path="/noop-results"), +} + +# Add env vars if AWS credentials or HF token exist +env_vars = [] +if has_aws_creds: + env_vars.append(BeakerEnvVar(name="AWS_CREDENTIALS_FILE", secret=aws_creds_secret)) +if has_hf_token: + env_vars.append(BeakerEnvVar(name="HF_TOKEN", secret=hf_token_secret)) +if env_vars: + task_spec_args["env_vars"] = env_vars + +# Create first experiment spec +chandra_version_label = "latest" if chandra_version == "latest" else chandra_version +experiment_spec = BeakerExperimentSpec( + description=f"Chandra {chandra_version_label} Benchmark Run - Branch: {git_branch}, Commit: {git_hash}", + budget="ai2/oe-base", + tasks=[BeakerTaskSpec(**task_spec_args)], +) + +# Create the first experiment +workload = b.experiment.create(spec=experiment_spec, workspace="ai2/olmocr") +print(f"Created benchmark experiment: {workload.experiment.id}") +print(f"View at: https://beaker.org/ex/{workload.experiment.id}") +print("-------") +print("") + +# Second experiment: Performance test job (only if --noperf not specified) +if not noperf: + perf_commands = [] + if has_aws_creds: + perf_commands.extend([ + "mkdir -p ~/.aws", + 'echo "$AWS_CREDENTIALS_FILE" > ~/.aws/credentials' + ]) + + if has_hf_token: + perf_commands.append('export HF_TOKEN="$HF_TOKEN"') + + # Shell script for performance test + perf_shell = dedent("""\ +set -euo pipefail + +# Start vllm server in background +echo "Starting vllm server for Chandra..." +vllm serve __CHANDRA_MODEL__ --served-model-name chandra > /tmp/vllm_server.log 2>&1 & +VLLM_PID=$! + +# Wait for vllm server to be ready +echo "Waiting for vllm server to start..." +for i in {1..600}; do + if curl -s http://localhost:8000/health > /dev/null 2>&1; then + echo "vllm server is ready" + break + fi + if [ $i -eq 600 ]; then + echo "Error: vllm server failed to start after 600 seconds" + cat /tmp/vllm_server.log + exit 1 + fi + sleep 1 +done + +# Run the performance test +time chandra /root/olmOCR-mix-0225_benchmark_set/ /root/olmOCR-mix-0225_benchmark_set_chandra --method vllm + +# Kill vllm server +echo "Stopping vllm server..." +kill $VLLM_PID || true +wait $VLLM_PID 2>/dev/null || true +""").replace("__CHANDRA_MODEL__", chandra_model) + + perf_commands.extend([ + "pip install uv", + chandra_install_cmd, + "uv pip uninstall --system vllm flashinfer flashinfer-python flashinfer-cubin || true", + "uv pip install --system vllm --torch-backend=auto", + "uv pip install --system awscli", + "aws s3 cp --recursive s3://ai2-oe-data/jakep/olmocr/olmOCR-mix-0225/benchmark_set/ /root/olmOCR-mix-0225_benchmark_set/", + f"bash -c '{perf_shell}'" + ]) + + # Build performance task spec + perf_task_spec_args = { + "name": "chandra-performance", + "image": BeakerImageSource(beaker=image_ref), + "command": [ + "bash", "-c", + " && ".join(perf_commands) + ], + "context": BeakerTaskContext( + priority=BeakerJobPriority["normal"], + preemptible=True, + ), + # Need to reserve all 8 gpus for performance spec or else benchmark results can be off (1 for titan-cirrascale) + "resources": BeakerTaskResources(gpu_count=1 if cluster == "ai2/titan-cirrascale" else 8), + "constraints": BeakerConstraints(cluster=[cluster] if cluster else ["ai2/ceres-cirrascale", "ai2/jupiter-cirrascale-2"]), + "result": BeakerResultSpec(path="/noop-results"), + } + + # Add env vars if AWS credentials or HF token exist + env_vars = [] + if has_aws_creds: + env_vars.append(BeakerEnvVar(name="AWS_CREDENTIALS_FILE", secret=aws_creds_secret)) + if has_hf_token: + env_vars.append(BeakerEnvVar(name="HF_TOKEN", secret=hf_token_secret)) + if env_vars: + perf_task_spec_args["env_vars"] = env_vars + + # Create performance experiment spec + perf_experiment_spec = BeakerExperimentSpec( + description=f"Chandra {chandra_version_label} Performance Test - Branch: {git_branch}, Commit: {git_hash}", + budget="ai2/oe-base", + tasks=[BeakerTaskSpec(**perf_task_spec_args)], + ) + + # Create the performance experiment + perf_workload = b.experiment.create(spec=perf_experiment_spec, workspace="ai2/olmocr") + print(f"Created performance experiment: {perf_workload.experiment.id}") + print(f"View at: https://beaker.org/ex/{perf_workload.experiment.id}") +else: + print("Skipping performance test (--noperf flag specified)") +EOF + +# Run the Python script to create the experiments +echo "Creating Beaker experiments..." + +# Build command with appropriate arguments +CMD="$PYTHON /tmp/run_benchmark_experiment.py $IMAGE_TAG $BEAKER_USER $GIT_BRANCH $GIT_HASH \"$CHANDRA_VERSION\" \"$CHANDRA_INSTALL_CMD\"" + +if [ -n "$BENCH_BRANCH" ]; then + echo "Using bench branch: $BENCH_BRANCH" + CMD="$CMD --benchbranch \"$BENCH_BRANCH\"" +fi + +if [ -n "$BENCH_REPO" ]; then + echo "Using bench repo: $BENCH_REPO" + CMD="$CMD --benchrepo \"$BENCH_REPO\"" +fi + +if [ -n "$BENCH_PATH" ]; then + echo "Using bench path: $BENCH_PATH" + CMD="$CMD --benchpath \"$BENCH_PATH\"" +fi + +if [ -n "$CLUSTER" ]; then + echo "Using cluster: $CLUSTER" + CMD="$CMD --cluster $CLUSTER" +fi + +if [ -n "$NOPERF" ]; then + echo "Skipping performance tests" + CMD="$CMD --noperf" +fi + +if [ -n "$MODEL" ]; then + echo "Using model: $MODEL" + CMD="$CMD --model $MODEL" +fi + +eval $CMD + +# Clean up temporary file +rm /tmp/run_benchmark_experiment.py + +echo "Benchmark experiments submitted successfully!" \ No newline at end of file diff --git a/scripts/run_dots_mocr_benchmark.sh b/scripts/run_dots_mocr_benchmark.sh new file mode 100755 index 0000000..9641968 --- /dev/null +++ b/scripts/run_dots_mocr_benchmark.sh @@ -0,0 +1,636 @@ +#!/bin/bash + +# Runs dots.mocr benchmark, measuring both olmOCR-bench performance and per document processing performance +# dots.mocr is served via vllm and called through the OpenAI-compatible API +# Usage: +# ./scripts/run_dots_mocr_benchmark.sh # Use default model and benchmark repo +# ./scripts/run_dots_mocr_benchmark.sh --benchrepo allenai/olmOCR-bench-internal # Use different benchmark repo +# ./scripts/run_dots_mocr_benchmark.sh --benchbranch olmOCR-bench-1125 # Use specific branch/revision +# ./scripts/run_dots_mocr_benchmark.sh --benchpath s3://ai2-oe-data/path/ # Use benchmark from S3 or local path +# ./scripts/run_dots_mocr_benchmark.sh --cluster ai2/titan-cirrascale # Specify a cluster +# ./scripts/run_dots_mocr_benchmark.sh --beaker-image jakep/olmocr-benchmark-0.3.3-780bc7d934 # Skip Docker build +# ./scripts/run_dots_mocr_benchmark.sh --noperf # Skip the performance test job +# ./scripts/run_dots_mocr_benchmark.sh --model rednote-hilab/dots.mocr # Use a specific model + +set -e + +# Parse command line arguments +BENCH_BRANCH="" +BENCH_REPO="" +BENCH_PATH="" +CLUSTER="" +BEAKER_IMAGE="" +NOPERF="" +MODEL="" + +while [[ $# -gt 0 ]]; do + case $1 in + --benchbranch) + BENCH_BRANCH="$2" + shift 2 + ;; + --benchrepo) + BENCH_REPO="$2" + shift 2 + ;; + --benchpath) + BENCH_PATH="$2" + shift 2 + ;; + --cluster) + CLUSTER="$2" + shift 2 + ;; + --beaker-image) + BEAKER_IMAGE="$2" + shift 2 + ;; + --noperf) + NOPERF="1" + shift + ;; + --model) + MODEL="$2" + shift 2 + ;; + *) + echo "Unknown option: $1" + echo "Usage: $0 [--benchbranch BRANCH] [--benchrepo REPO] [--benchpath PATH] [--cluster CLUSTER] [--beaker-image IMAGE] [--noperf] [--model MODEL]" + exit 1 + ;; + esac +done + +# Check for mutual exclusivity between benchpath and benchrepo/benchbranch +if [ -n "$BENCH_PATH" ] && ([ -n "$BENCH_REPO" ] || [ -n "$BENCH_BRANCH" ]); then + echo "Error: --benchpath is mutually exclusive with --benchrepo and --benchbranch" + echo "Use either --benchpath OR --benchrepo/--benchbranch, not both." + exit 1 +fi + +# Use conda environment Python if available, otherwise use system Python +if [ -n "$CONDA_PREFIX" ]; then + PYTHON="$CONDA_PREFIX/bin/python" + echo "Using conda Python from: $CONDA_PREFIX" +else + PYTHON="python" + echo "Warning: No conda environment detected, using system Python" +fi + +# Get version from version.py +VERSION=$($PYTHON -c 'import olmocr.version; print(olmocr.version.VERSION)') +echo "OlmOCR version: $VERSION" + +# Get first 10 characters of git hash +GIT_HASH=$(git rev-parse HEAD | cut -c1-10) +echo "Git hash: $GIT_HASH" + +# Get current git branch name +GIT_BRANCH=$(git rev-parse --abbrev-ref HEAD) +echo "Git branch: $GIT_BRANCH" + +# Check if a Beaker image was provided +if [ -n "$BEAKER_IMAGE" ]; then + echo "Using provided Beaker image: $BEAKER_IMAGE" + IMAGE_TAG="$BEAKER_IMAGE" +else + # Create full image tag + IMAGE_TAG="olmocr-benchmark-${VERSION}-${GIT_HASH}" + echo "Building Docker image with tag: $IMAGE_TAG" + + # Build the Docker image + echo "Building Docker image..." + docker build --platform linux/amd64 -f ./Dockerfile -t $IMAGE_TAG . + + # Push image to beaker + echo "Trying to push image to Beaker..." + if ! beaker image create --workspace ai2/oe-data-pdf --name $IMAGE_TAG $IMAGE_TAG 2>/dev/null; then + echo "Warning: Beaker image with tag $IMAGE_TAG already exists. Using existing image." + fi +fi + +# Get Beaker username +BEAKER_USER=$(beaker account whoami --format json | jq -r '.[0].name') +echo "Beaker user: $BEAKER_USER" + +# Create Python script to run beaker experiment +cat << 'EOF' > /tmp/run_benchmark_experiment.py +import sys +import base64 +from textwrap import dedent +from beaker import Beaker, BeakerExperimentSpec, BeakerTaskSpec, BeakerTaskContext, BeakerResultSpec, BeakerTaskResources, BeakerImageSource, BeakerJobPriority, BeakerConstraints, BeakerEnvVar + +# Get image tag, beaker user, git branch, git hash from command line +image_tag = sys.argv[1] +beaker_user = sys.argv[2] +git_branch = sys.argv[3] +git_hash = sys.argv[4] + +# Initialize benchmark dataset parameters +bench_branch = None +bench_repo = "allenai/olmOCR-bench" # Default repository +bench_path = None +cluster = None +noperf = False +model = None + +# Parse additional arguments +arg_idx = 5 +while arg_idx < len(sys.argv): + if sys.argv[arg_idx] == "--benchbranch": + bench_branch = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--benchrepo": + bench_repo = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--benchpath": + bench_path = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--cluster": + cluster = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--noperf": + noperf = True + arg_idx += 1 + elif sys.argv[arg_idx] == "--model": + model = sys.argv[arg_idx + 1] + arg_idx += 2 + else: + print(f"Unknown argument: {sys.argv[arg_idx]}") + arg_idx += 1 + +# Default model for dots.mocr +dots_mocr_model = model if model else "rednote-hilab/dots.mocr" + +# Initialize Beaker client +b = Beaker.from_env(default_workspace="ai2/olmocr") + +# Check if AWS credentials secret exists +aws_creds_secret = f"{beaker_user}-AWS_CREDENTIALS_FILE" +try: + b.secret.get(aws_creds_secret, workspace="ai2/olmocr") + has_aws_creds = True + print(f"Found AWS credentials secret: {aws_creds_secret}") +except: + has_aws_creds = False + print(f"AWS credentials secret not found: {aws_creds_secret}") + +# Python script that renders PDFs to images and calls dots.mocr via vllm OpenAI API +run_dots_mocr_python = dedent('''\ +import os, sys, glob, json, base64, io +from openai import OpenAI +from pdf2image import convert_from_path + +PDF_ROOT = "olmOCR-bench/bench_data/pdfs" +TARGET_ROOT = "olmOCR-bench/bench_data/dots_mocr" + +# Clean and create target directory +if os.path.exists(TARGET_ROOT): + import shutil + shutil.rmtree(TARGET_ROOT) +os.makedirs(TARGET_ROOT, exist_ok=True) + +client = OpenAI(base_url="http://localhost:8000/v1", api_key="empty") + +PROMPT = """Please output the layout information from the PDF image, including each layout element's bbox, its category, and the corresponding text content within the bbox. + +1. Bbox format: [x1, y1, x2, y2] + +2. Layout Categories: ['Caption', 'Footnote', 'Formula', 'List-item', 'Page-footer', 'Page-header', 'Picture', 'Section-header', 'Table', 'Text', 'Title']. + +3. Text Extraction & Formatting Rules: + - Picture: Text field should be omitted + - Formula: Format as LaTeX + - Table: Format as HTML + - Others: Format as Markdown + +4. Constraints: + - Output text must be original from image, no translation + - All layout elements sorted by reading order + +5. Final Output: Single JSON object +""" + +# Find all PDFs +pdf_files = sorted(glob.glob(os.path.join(PDF_ROOT, "**", "*.pdf"), recursive=True)) +print(f"Found {len(pdf_files)} PDFs to process") + +for pdf_path in pdf_files: + rel_path = os.path.relpath(pdf_path, PDF_ROOT) + parts = rel_path.split(os.sep) + if len(parts) < 2: + print(f"Warning: Unexpected PDF path layout for {pdf_path}, skipping") + continue + + section = parts[0] + pdf_name = os.path.splitext(os.path.basename(pdf_path))[0] + + print(f" Processing {pdf_path}...") + try: + # Render first page of PDF to image + images = convert_from_path(pdf_path, first_page=1, last_page=1, dpi=144) + if not images: + print(f" Warning: Could not render {pdf_path}") + continue + + # Convert to base64 + buf = io.BytesIO() + images[0].save(buf, format="PNG") + img_b64 = base64.b64encode(buf.getvalue()).decode("utf-8") + + # Call dots.mocr via OpenAI API + response = client.chat.completions.create( + model="model", + messages=[ + { + "role": "user", + "content": [ + {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{img_b64}"}}, + {"type": "text", "text": PROMPT}, + ], + } + ], + max_tokens=24000, + ) + + result_text = response.choices[0].message.content + + # Try to extract text content from JSON response and assemble markdown + markdown_parts = [] + try: + data = json.loads(result_text) + elements = data if isinstance(data, list) else data.get("elements", data.get("layout", [data])) + for elem in elements: + text = elem.get("text", "") + if text: + markdown_parts.append(text) + if markdown_parts: + result_text = "\n\n".join(markdown_parts) + except (json.JSONDecodeError, TypeError, AttributeError): + # If not valid JSON, use raw text as-is + pass + + # Write output + target_dir = os.path.join(TARGET_ROOT, section) + os.makedirs(target_dir, exist_ok=True) + target_path = os.path.join(target_dir, f"{pdf_name}_pg1_repeat1.md") + + with open(target_path, "w") as f: + f.write(result_text) + print(f" Wrote {target_path}") + except Exception as e: + print(f" Error processing {pdf_path}: {e}") + continue + +print("Done processing all PDFs") +''') + +# Base64-encode the inference script so we can safely embed it in the command chain +import base64 +run_dots_mocr_python_b64 = base64.b64encode(run_dots_mocr_python.encode()).decode() + +# Shell script to start vllm, write inference script, run it, and clean up +run_dots_mocr_shell = dedent("""\ +bash -lc 'set -euo pipefail + +# Decode the inference script +echo "__INFERENCE_SCRIPT_B64__" | base64 -d > /tmp/run_dots_mocr_inference.py + +# Start vllm server in background +echo "Starting vllm server for dots.mocr..." +vllm serve __DOTS_MOCR_MODEL__ --served-model-name model --trust-remote-code --chat-template-content-format string --gpu-memory-utilization 0.9 > /tmp/vllm_server.log 2>&1 & +VLLM_PID=$! + +# Wait for vllm server to be ready +echo "Waiting for vllm server to start..." +for i in {1..600}; do + if curl -s http://localhost:8000/health > /dev/null 2>&1; then + echo "vllm server is ready" + break + fi + if [ $i -eq 600 ]; then + echo "Error: vllm server failed to start after 600 seconds" + cat /tmp/vllm_server.log + exit 1 + fi + sleep 1 +done + +# Run the Python inference script +python /tmp/run_dots_mocr_inference.py + +# Kill vllm server +echo "Stopping vllm server..." +kill $VLLM_PID || true +wait $VLLM_PID 2>/dev/null || true +'""").replace("__DOTS_MOCR_MODEL__", dots_mocr_model).replace("__INFERENCE_SCRIPT_B64__", run_dots_mocr_python_b64) + +# Check if HF_TOKEN secret exists +hf_token_secret = f"{beaker_user}-HF_TOKEN" +try: + b.secret.get(hf_token_secret, workspace="ai2/olmocr") + has_hf_token = True + print(f"Found HuggingFace token secret: {hf_token_secret}") +except: + has_hf_token = False + print(f"HuggingFace token secret not found: {hf_token_secret}") + +# First experiment: Benchmark job +commands = [] +if has_aws_creds: + commands.extend([ + "mkdir -p ~/.aws", + 'echo "$AWS_CREDENTIALS_FILE" > ~/.aws/credentials' + ]) + +if has_hf_token: + commands.append('export HF_TOKEN="$HF_TOKEN"') + +# Install uv for fast dependency management, then s5cmd (needed for S3 operations) +commands.append("pip install uv") +commands.append("uv pip install --system s5cmd") + +# Handle benchmark data download based on source type +if bench_path: + if bench_path.startswith("s3://"): + commands.append(f"s5cmd cp {bench_path.rstrip('/')}/* ./olmOCR-bench/") + else: + commands.append(f"cp -r {bench_path} ./olmOCR-bench") +else: + hf_download_cmd = f"hf download --repo-type dataset {bench_repo} --max-workers 2" + if bench_branch: + hf_download_cmd += f" --revision {bench_branch}" + hf_download_cmd += " --local-dir ./olmOCR-bench" + commands.append(hf_download_cmd) + +# Write the inference script to a temp file, install deps, run +commands.extend([ + "uv pip install --system --upgrade vllm", + "uv pip install --system openai pdf2image", + run_dots_mocr_shell, + "python -m olmocr.bench.benchmark --dir ./olmOCR-bench/bench_data --candidate dots_mocr" +]) + +# Build task spec with optional env vars +if '/' in image_tag: + image_ref = image_tag +else: + image_ref = f"{beaker_user}/{image_tag}" + +task_spec_args = { + "name": "dots-mocr-benchmark", + "image": BeakerImageSource(beaker=image_ref), + "command": [ + "bash", "-c", + " && ".join(commands) + ], + "context": BeakerTaskContext( + priority=BeakerJobPriority["normal"], + preemptible=True, + ), + "resources": BeakerTaskResources(gpu_count=1), + "constraints": BeakerConstraints(cluster=[cluster] if cluster else ["ai2/ceres-cirrascale", "ai2/jupiter-cirrascale-2"]), + "result": BeakerResultSpec(path="/noop-results"), +} + +# Add env vars if AWS credentials or HF token exist +env_vars = [] +if has_aws_creds: + env_vars.append(BeakerEnvVar(name="AWS_CREDENTIALS_FILE", secret=aws_creds_secret)) +if has_hf_token: + env_vars.append(BeakerEnvVar(name="HF_TOKEN", secret=hf_token_secret)) +if env_vars: + task_spec_args["env_vars"] = env_vars + +# Create experiment spec +experiment_spec = BeakerExperimentSpec( + description=f"dots.mocr Benchmark Run - Branch: {git_branch}, Commit: {git_hash}", + budget="ai2/oe-base", + tasks=[BeakerTaskSpec(**task_spec_args)], +) + +# Create the experiment +workload = b.experiment.create(spec=experiment_spec, workspace="ai2/olmocr") +print(f"Created benchmark experiment: {workload.experiment.id}") +print(f"View at: https://beaker.org/ex/{workload.experiment.id}") +print("-------") +print("") + +# Second experiment: Performance test job (only if --noperf not specified) +if not noperf: + perf_commands = [] + if has_aws_creds: + perf_commands.extend([ + "mkdir -p ~/.aws", + 'echo "$AWS_CREDENTIALS_FILE" > ~/.aws/credentials' + ]) + + if has_hf_token: + perf_commands.append('export HF_TOKEN="$HF_TOKEN"') + + # Performance test Python script - processes PDFs from the benchmark set via vllm API + perf_dots_mocr_python = dedent('''\ +import os, sys, glob, json, base64, io, time +from openai import OpenAI +from pdf2image import convert_from_path + +PDF_ROOT = "/root/olmOCR-mix-0225_benchmark_set" +TARGET_ROOT = "/root/olmOCR-mix-0225_benchmark_set_dots_mocr" + +os.makedirs(TARGET_ROOT, exist_ok=True) + +client = OpenAI(base_url="http://localhost:8000/v1", api_key="empty") + +PROMPT = """Please output the layout information from the PDF image, including each layout element's bbox, its category, and the corresponding text content within the bbox. + +1. Bbox format: [x1, y1, x2, y2] + +2. Layout Categories: ['Caption', 'Footnote', 'Formula', 'List-item', 'Page-footer', 'Page-header', 'Picture', 'Section-header', 'Table', 'Text', 'Title']. + +3. Text Extraction & Formatting Rules: + - Picture: Text field should be omitted + - Formula: Format as LaTeX + - Table: Format as HTML + - Others: Format as Markdown + +4. Constraints: + - Output text must be original from image, no translation + - All layout elements sorted by reading order + +5. Final Output: Single JSON object +""" + +pdf_files = sorted(glob.glob(os.path.join(PDF_ROOT, "**", "*.pdf"), recursive=True)) +if not pdf_files: + pdf_files = sorted(glob.glob(os.path.join(PDF_ROOT, "*.pdf"))) +print(f"Found {len(pdf_files)} PDFs to process for performance test") + +start_time = time.time() +for pdf_path in pdf_files: + pdf_name = os.path.splitext(os.path.basename(pdf_path))[0] + print(f" Processing {pdf_path}...") + try: + images = convert_from_path(pdf_path, first_page=1, last_page=1, dpi=144) + if not images: + continue + buf = io.BytesIO() + images[0].save(buf, format="PNG") + img_b64 = base64.b64encode(buf.getvalue()).decode("utf-8") + + response = client.chat.completions.create( + model="model", + messages=[ + { + "role": "user", + "content": [ + {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{img_b64}"}}, + {"type": "text", "text": PROMPT}, + ], + } + ], + max_tokens=24000, + ) + + result_text = response.choices[0].message.content + target_path = os.path.join(TARGET_ROOT, f"{pdf_name}.md") + with open(target_path, "w") as f: + f.write(result_text) + except Exception as e: + print(f" Error processing {pdf_path}: {e}") + continue + +elapsed = time.time() - start_time +print(f"Performance test completed in {elapsed:.1f}s for {len(pdf_files)} PDFs") +''') + + # Base64-encode the perf script + perf_dots_mocr_python_b64 = base64.b64encode(perf_dots_mocr_python.encode()).decode() + + # Shell script for performance test + perf_shell = dedent("""\ +set -euo pipefail + +# Decode the perf script +echo "__PERF_SCRIPT_B64__" | base64 -d > /tmp/run_dots_mocr_perf.py + +# Start vllm server in background +echo "Starting vllm server for dots.mocr..." +vllm serve __DOTS_MOCR_MODEL__ --served-model-name model --trust-remote-code --chat-template-content-format string --gpu-memory-utilization 0.9 > /tmp/vllm_server.log 2>&1 & +VLLM_PID=$! + +# Wait for vllm server to be ready +echo "Waiting for vllm server to start..." +for i in {1..600}; do + if curl -s http://localhost:8000/health > /dev/null 2>&1; then + echo "vllm server is ready" + break + fi + if [ $i -eq 600 ]; then + echo "Error: vllm server failed to start after 600 seconds" + cat /tmp/vllm_server.log + exit 1 + fi + sleep 1 +done + +# Run the performance test +time python /tmp/run_dots_mocr_perf.py + +# Kill vllm server +echo "Stopping vllm server..." +kill $VLLM_PID || true +wait $VLLM_PID 2>/dev/null || true +""").replace("__DOTS_MOCR_MODEL__", dots_mocr_model).replace("__PERF_SCRIPT_B64__", perf_dots_mocr_python_b64) + + perf_commands.extend([ + "pip install uv", + "uv pip install --system --upgrade vllm", + "uv pip install --system openai pdf2image awscli", + "aws s3 cp --recursive s3://ai2-oe-data/jakep/olmocr/olmOCR-mix-0225/benchmark_set/ /root/olmOCR-mix-0225_benchmark_set/", + f"bash -c '{perf_shell}'" + ]) + + # Build performance task spec + perf_task_spec_args = { + "name": "dots-mocr-performance", + "image": BeakerImageSource(beaker=image_ref), + "command": [ + "bash", "-c", + " && ".join(perf_commands) + ], + "context": BeakerTaskContext( + priority=BeakerJobPriority["normal"], + preemptible=True, + ), + # Need to reserve all 8 gpus for performance spec or else benchmark results can be off (1 for titan-cirrascale) + "resources": BeakerTaskResources(gpu_count=1 if cluster == "ai2/titan-cirrascale" else 8), + "constraints": BeakerConstraints(cluster=[cluster] if cluster else ["ai2/ceres-cirrascale", "ai2/jupiter-cirrascale-2"]), + "result": BeakerResultSpec(path="/noop-results"), + } + + # Add env vars if AWS credentials or HF token exist + env_vars = [] + if has_aws_creds: + env_vars.append(BeakerEnvVar(name="AWS_CREDENTIALS_FILE", secret=aws_creds_secret)) + if has_hf_token: + env_vars.append(BeakerEnvVar(name="HF_TOKEN", secret=hf_token_secret)) + if env_vars: + perf_task_spec_args["env_vars"] = env_vars + + # Create performance experiment spec + perf_experiment_spec = BeakerExperimentSpec( + description=f"dots.mocr Performance Test - Branch: {git_branch}, Commit: {git_hash}", + budget="ai2/oe-base", + tasks=[BeakerTaskSpec(**perf_task_spec_args)], + ) + + # Create the performance experiment + perf_workload = b.experiment.create(spec=perf_experiment_spec, workspace="ai2/olmocr") + print(f"Created performance experiment: {perf_workload.experiment.id}") + print(f"View at: https://beaker.org/ex/{perf_workload.experiment.id}") +else: + print("Skipping performance test (--noperf flag specified)") +EOF + +# Run the Python script to create the experiments +echo "Creating Beaker experiments..." + +# Build command with appropriate arguments +CMD="$PYTHON /tmp/run_benchmark_experiment.py $IMAGE_TAG $BEAKER_USER $GIT_BRANCH $GIT_HASH" + +if [ -n "$BENCH_BRANCH" ]; then + echo "Using bench branch: $BENCH_BRANCH" + CMD="$CMD --benchbranch \"$BENCH_BRANCH\"" +fi + +if [ -n "$BENCH_REPO" ]; then + echo "Using bench repo: $BENCH_REPO" + CMD="$CMD --benchrepo \"$BENCH_REPO\"" +fi + +if [ -n "$BENCH_PATH" ]; then + echo "Using bench path: $BENCH_PATH" + CMD="$CMD --benchpath \"$BENCH_PATH\"" +fi + +if [ -n "$CLUSTER" ]; then + echo "Using cluster: $CLUSTER" + CMD="$CMD --cluster $CLUSTER" +fi + +if [ -n "$NOPERF" ]; then + echo "Skipping performance tests" + CMD="$CMD --noperf" +fi + +if [ -n "$MODEL" ]; then + echo "Using model: $MODEL" + CMD="$CMD --model $MODEL" +fi + +eval $CMD + +# Clean up temporary file +rm /tmp/run_benchmark_experiment.py + +echo "Benchmark experiments submitted successfully!" diff --git a/scripts/run_infinityparser_benchmark.sh b/scripts/run_infinityparser_benchmark.sh new file mode 100755 index 0000000..6334229 --- /dev/null +++ b/scripts/run_infinityparser_benchmark.sh @@ -0,0 +1,276 @@ +#!/bin/bash + +# Runs Infinity Parser benchmark, measuring both olmOCR-bench performance and per document processing performance +# ./scripts/run_infinityparser_benchmark.sh + +set -e + +# Check for uncommitted changes +if ! git diff-index --quiet HEAD --; then + echo "Error: There are uncommitted changes in the repository." + echo "Please commit or stash your changes before running the benchmark." + echo "" + echo "Uncommitted changes:" + git status --short + exit 1 +fi + +# Use conda environment Python if available, otherwise use system Python +if [ -n "$CONDA_PREFIX" ]; then + PYTHON="$CONDA_PREFIX/bin/python" + echo "Using conda Python from: $CONDA_PREFIX" +else + PYTHON="python" + echo "Warning: No conda environment detected, using system Python" +fi + +# Get version from version.py +VERSION=$($PYTHON -c 'import olmocr.version; print(olmocr.version.VERSION)') +echo "OlmOCR version: $VERSION" + +# Get first 10 characters of git hash +GIT_HASH=$(git rev-parse HEAD | cut -c1-10) +echo "Git hash: $GIT_HASH" + +# Get current git branch name +GIT_BRANCH=$(git rev-parse --abbrev-ref HEAD) +echo "Git branch: $GIT_BRANCH" + +# Create full image tag +IMAGE_TAG="olmocr-benchmark-infinityparser-${VERSION}-${GIT_HASH}" +echo "Building Docker image with tag: $IMAGE_TAG" + +# Build the Docker image +echo "Building Docker image..." +docker build --platform linux/amd64 -f ./Dockerfile -t $IMAGE_TAG . + +# Get Beaker username +BEAKER_USER=$(beaker account whoami --format json | jq -r '.[0].name') +echo "Beaker user: $BEAKER_USER" + +# Push image to beaker +echo "Trying to push image to Beaker..." +if ! beaker image create --workspace ai2/oe-data-pdf --name $IMAGE_TAG $IMAGE_TAG 2>/dev/null; then + echo "Warning: Beaker image with tag $IMAGE_TAG already exists. Using existing image." +fi + +# Create Python script to run beaker experiment +cat << 'EOF' > /tmp/run_benchmark_experiment.py +import sys +from textwrap import dedent +from beaker import Beaker, ExperimentSpec, TaskSpec, TaskContext, ResultSpec, TaskResources, ImageSource, Priority, Constraints, EnvVar + +# Get image tag, beaker user, git branch, and git hash from command line +image_tag = sys.argv[1] +beaker_user = sys.argv[2] +git_branch = sys.argv[3] +git_hash = sys.argv[4] + +# Initialize Beaker client +b = Beaker.from_env(default_workspace="ai2/olmocr") + +# Check if AWS credentials secret exists +aws_creds_secret = f"{beaker_user}-AWS_CREDENTIALS_FILE" +try: + # Try to get the secret to see if it exists + b.secret.get(aws_creds_secret, workspace="ai2/olmocr") + has_aws_creds = True + print(f"Found AWS credentials secret: {aws_creds_secret}") +except: + has_aws_creds = False + print(f"AWS credentials secret not found: {aws_creds_secret}") + +# Shell script to run Infinity Parser conversions for benchmark +run_infinityparser_shell = dedent("""\ +bash -lc 'set -euo pipefail +PDF_ROOT="olmOCR-bench/bench_data/pdfs" +TARGET_ROOT="olmOCR-bench/bench_data/infinityparser" +rm -rf "$TARGET_ROOT" +mkdir -p "$TARGET_ROOT" + +# Install required dependencies for Infinity Parser +echo "Installing Infinity Parser dependencies..." +pip install PyMuPDF pdf2image qwen_vl_utils + +# Process each folder separately to maintain structure +echo "Running Infinity Parser conversions..." +for folder in "$PDF_ROOT"/*; do + if [ ! -d "$folder" ]; then + continue + fi + section=$(basename "$folder") + output_dir="$HOME/infinityparser_bench_${section}" + rm -rf "$output_dir" + mkdir -p "$output_dir" + + echo " Processing $folder -> $output_dir" + parser --model infly/Infinity-Parser-7B --input "$folder" --output "$output_dir" --batch_size 128 --tp 1 +done + +echo "Collecting Infinity Parser markdown outputs..." +# For each PDF, find its corresponding markdown and copy to proper location +find "$PDF_ROOT" -type f -name "*.pdf" | while IFS= read -r pdf_path; do + # Get relative path from PDF root + rel_path=${pdf_path#"$PDF_ROOT"/} + + # Extract section name (first directory level) + case "$rel_path" in + */*) + section=${rel_path%%/*} + ;; + *) + echo "Warning: Unexpected PDF path layout for $pdf_path, skipping" >&2 + continue + ;; + esac + + # Get PDF name without extension + pdf_name=$(basename "$pdf_path" .pdf) + + # Source markdown location: ~/infinityparser_bench_${section}/${pdf_name}/output.md + src_md="$HOME/infinityparser_bench_${section}/${pdf_name}/output.md" + + if [ ! -f "$src_md" ]; then + echo "Warning: No markdown output found at $src_md" >&2 + continue + fi + + # Target location: olmOCR-bench/bench_data/infinityparser/${section}/${pdf_name}_pg1_repeat1.md + target_dir="$TARGET_ROOT/$section" + mkdir -p "$target_dir" + target_path="$target_dir/${pdf_name}_pg1_repeat1.md" + + cp "$src_md" "$target_path" + echo " Copied $src_md -> $target_path" +done +'""") + +# First experiment: Original benchmark job +commands = [] +if has_aws_creds: + commands.extend([ + "mkdir -p ~/.aws", + 'echo "$AWS_CREDENTIALS_FILE" > ~/.aws/credentials' + ]) +commands.extend([ + "git clone https://huggingface.co/datasets/allenai/olmOCR-bench", + "cd olmOCR-bench && git lfs pull && cd ..", + "git clone https://github.com/infly-ai/INF-MLLM.git", + # Patch the missing import in utils.py + "sed -i '1s/^/from pdf2image import convert_from_path\\n/' INF-MLLM/Infinity-Parser/inference/utils.py", + "cd INF-MLLM/Infinity-Parser && pip install . && cd ../..", + run_infinityparser_shell, + "python -m olmocr.bench.benchmark --dir ./olmOCR-bench/bench_data --candidate infinityparser" +]) + +# Build task spec with optional env vars +task_spec_args = { + "name": "infinityparser-benchmark", + "image": ImageSource(beaker=f"{beaker_user}/{image_tag}"), + "command": [ + "bash", "-c", + " && ".join(commands) + ], + "context": TaskContext( + priority=Priority.normal, + preemptible=True, + ), + "resources": TaskResources(gpu_count=1), + "constraints": Constraints(cluster=["ai2/ceres-cirrascale", "ai2/jupiter-cirrascale-2"]), + "result": ResultSpec(path="/noop-results"), +} + +# Add env vars if AWS credentials exist +if has_aws_creds: + task_spec_args["env_vars"] = [ + EnvVar(name="AWS_CREDENTIALS_FILE", secret=aws_creds_secret) + ] + +# Create first experiment spec +experiment_spec = ExperimentSpec( + description=f"Infinity Parser Benchmark Run - Branch: {git_branch}, Commit: {git_hash}", + budget="ai2/oe-base", + tasks=[TaskSpec(**task_spec_args)], +) + +# Create the first experiment +experiment = b.experiment.create(spec=experiment_spec, workspace="ai2/olmocr") +print(f"Created benchmark experiment: {experiment.id}") +print(f"View at: https://beaker.org/ex/{experiment.id}") +print("-------") +print("") + +# Second experiment: Performance test +perf_commands = [] +if has_aws_creds: + perf_commands.extend([ + "mkdir -p ~/.aws", + 'echo "$AWS_CREDENTIALS_FILE" > ~/.aws/credentials' + ]) + +# Shell script for performance test +perf_shell = dedent("""\ +set -euo pipefail + +# Install required dependencies for Infinity Parser +echo "Installing Infinity Parser dependencies..." +pip install PyMuPDF pdf2image qwen_vl_utils + +# Run the performance test with Infinity Parser +time parser --model infly/Infinity-Parser-7B --input /root/olmOCR-mix-0225_benchmark_set/ --output /root/olmOCR-mix-0225_benchmark_set_infinityparser --batch_size 128 --tp 1 +""") + +perf_commands.extend([ + "git clone https://github.com/infly-ai/INF-MLLM.git", + # Patch the missing import in utils.py + "sed -i '1s/^/from pdf2image import convert_from_path\\n/' INF-MLLM/Infinity-Parser/inference/utils.py", + "cd INF-MLLM/Infinity-Parser && pip install . && cd ../..", + "pip install awscli", + "aws s3 cp --recursive s3://ai2-oe-data/jakep/olmocr/olmOCR-mix-0225/benchmark_set/ /root/olmOCR-mix-0225_benchmark_set/", + f"bash -c '{perf_shell}'" +]) + +# Build performance task spec +perf_task_spec_args = { + "name": "infinityparser-performance", + "image": ImageSource(beaker=f"{beaker_user}/{image_tag}"), + "command": [ + "bash", "-c", + " && ".join(perf_commands) + ], + "context": TaskContext( + priority=Priority.normal, + preemptible=True, + ), + "resources": TaskResources(gpu_count=1), + "constraints": Constraints(cluster=["ai2/ceres-cirrascale", "ai2/jupiter-cirrascale-2"]), + "result": ResultSpec(path="/noop-results"), +} + +# Add env vars if AWS credentials exist +if has_aws_creds: + perf_task_spec_args["env_vars"] = [ + EnvVar(name="AWS_CREDENTIALS_FILE", secret=aws_creds_secret) + ] + +# Create performance experiment spec +perf_experiment_spec = ExperimentSpec( + description=f"Infinity Parser Performance Test - Branch: {git_branch}, Commit: {git_hash}", + budget="ai2/oe-base", + tasks=[TaskSpec(**perf_task_spec_args)], +) + +# Create the performance experiment +perf_experiment = b.experiment.create(spec=perf_experiment_spec, workspace="ai2/olmocr") +print(f"Created performance experiment: {perf_experiment.id}") +print(f"View at: https://beaker.org/ex/{perf_experiment.id}") +EOF + +# Run the Python script to create the experiments +echo "Creating Beaker experiments..." +$PYTHON /tmp/run_benchmark_experiment.py $IMAGE_TAG $BEAKER_USER $GIT_BRANCH $GIT_HASH + +# Clean up temporary file +rm /tmp/run_benchmark_experiment.py + +echo "Benchmark experiments submitted successfully!" \ No newline at end of file diff --git a/scripts/run_infrapartner_benchmark.sh b/scripts/run_infrapartner_benchmark.sh new file mode 100755 index 0000000..30afbb4 --- /dev/null +++ b/scripts/run_infrapartner_benchmark.sh @@ -0,0 +1,274 @@ +#!/bin/bash + +# Runs an olmocr-bench run using the full pipeline (no fallback) for infrapartner testing +# +# Just make a beaker secret in the ai2/olmocr workspace with your API key +# +# Testing parasail +# scripts/run_infrapartner_benchmark.sh --server https://api.parasail.io/v1 --model allenai/olmOCR-2-7B-1025 --beaker-secret jakep-parasail-api-key +# +# Testing deepinfra +# scripts/run_infrapartner_benchmark.sh --server https://api.deepinfra.com/v1/openai --model allenai/olmOCR-2-7B-1025 --beaker-secret jakep-deepinfra-api-key +set -e + +# Parse command line arguments +MODEL="" +SERVER="" +BEAKER_SECRET="" +BENCH_BRANCH="" +BEAKER_IMAGE="" +while [[ $# -gt 0 ]]; do + case $1 in + --model) + MODEL="$2" + shift 2 + ;; + --server) + SERVER="$2" + shift 2 + ;; + --beaker-secret) + BEAKER_SECRET="$2" + shift 2 + ;; + --benchbranch) + BENCH_BRANCH="$2" + shift 2 + ;; + --beaker-image) + BEAKER_IMAGE="$2" + shift 2 + ;; + *) + echo "Unknown option: $1" + echo "Usage: $0 [--server SERVER_URL] [--model MODEL_NAME] [--beaker-secret SECRET_NAME] [--benchbranch BRANCH_NAME] [--beaker-image IMAGE_NAME]" + exit 1 + ;; + esac +done + +# Check for uncommitted changes +if [ -n "$BEAKER_IMAGE" ]; then + echo "Skipping docker build" +else + if ! git diff-index --quiet HEAD --; then + echo "Error: There are uncommitted changes in the repository." + echo "Please commit or stash your changes before running the benchmark." + echo "" + echo "Uncommitted changes:" + git status --short + exit 1 + fi +fi + +# Use conda environment Python if available, otherwise use system Python +if [ -n "$CONDA_PREFIX" ]; then + PYTHON="$CONDA_PREFIX/bin/python" + echo "Using conda Python from: $CONDA_PREFIX" +else + PYTHON="python" + echo "Warning: No conda environment detected, using system Python" +fi + +# Get version from version.py +VERSION=$($PYTHON -c 'import olmocr.version; print(olmocr.version.VERSION)') +echo "OlmOCR version: $VERSION" + +# Get first 10 characters of git hash +GIT_HASH=$(git rev-parse HEAD | cut -c1-10) +echo "Git hash: $GIT_HASH" + +# Get current git branch name +GIT_BRANCH=$(git rev-parse --abbrev-ref HEAD) +echo "Git branch: $GIT_BRANCH" + +# Check if a Beaker image was provided +if [ -n "$BEAKER_IMAGE" ]; then + echo "Using provided Beaker image: $BEAKER_IMAGE" + IMAGE_TAG="$BEAKER_IMAGE" +else + # Create full image tag + IMAGE_TAG="olmocr-benchmark-${VERSION}-${GIT_HASH}" + echo "Building Docker image with tag: $IMAGE_TAG" + + # Build the Docker image + echo "Building Docker image..." + docker build --platform linux/amd64 -f ./Dockerfile -t $IMAGE_TAG . + + # Push image to beaker + echo "Trying to push image to Beaker..." + if ! beaker image create --workspace ai2/oe-data-pdf --name $IMAGE_TAG $IMAGE_TAG 2>/dev/null; then + echo "Warning: Beaker image with tag $IMAGE_TAG already exists. Using existing image." + fi +fi + +# Get Beaker username +BEAKER_USER=$(beaker account whoami --format json | jq -r '.[0].name') +echo "Beaker user: $BEAKER_USER" + +# Create Python script to run beaker experiment +cat << 'EOF' > /tmp/run_infrapartner_benchmark_experiment.py +import sys +from beaker import Beaker, ExperimentSpec, TaskSpec, TaskContext, ResultSpec, TaskResources, ImageSource, Priority, Constraints, EnvVar + +# Get image tag, beaker user, git branch, git hash, optional model, server, beaker_secret, and bench branch from command line +image_tag = sys.argv[1] +beaker_user = sys.argv[2] +git_branch = sys.argv[3] +git_hash = sys.argv[4] +model = None +server = None +beaker_secret = None +bench_branch = None + +# Parse remaining arguments +arg_idx = 5 +while arg_idx < len(sys.argv): + if sys.argv[arg_idx] == "--benchbranch": + bench_branch = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--model": + model = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--server": + server = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--beaker-secret": + beaker_secret = sys.argv[arg_idx + 1] + arg_idx += 2 + else: + print(f"Unknown argument: {sys.argv[arg_idx]}") + arg_idx += 1 + +# Initialize Beaker client +b = Beaker.from_env(default_workspace="ai2/olmocr") + +# Build the pipeline command with optional parameters +pipeline_cmd = "python -m olmocr.pipeline ./localworkspace --markdown --pdfs ./olmOCR-bench/bench_data/pdfs/**/*.pdf --workers 1 --max_concurrent_requests 20" +if model: + pipeline_cmd += f" --model {model}" +if server: + pipeline_cmd += f" --server {server}" +if beaker_secret: + pipeline_cmd += " --api_key \"$API_KEY\"" + +# Check if AWS credentials secret exists +aws_creds_secret = f"{beaker_user}-AWS_CREDENTIALS_FILE" +try: + # Try to get the secret to see if it exists + b.secret.get(aws_creds_secret, workspace="ai2/olmocr") + has_aws_creds = True + print(f"Found AWS credentials secret: {aws_creds_secret}") +except: + has_aws_creds = False + print(f"AWS credentials secret not found: {aws_creds_secret}") + +# First experiment: Original benchmark job +commands = [] +if has_aws_creds: + commands.extend([ + "mkdir -p ~/.aws", + 'echo "$AWS_CREDENTIALS_FILE" > ~/.aws/credentials' + ]) + +# If beaker_secret is provided, export it as API_KEY environment variable +if beaker_secret: + commands.append('export API_KEY="$BEAKER_API_KEY"') + +# Build git clone command with optional branch +git_clone_cmd = "git clone https://huggingface.co/datasets/allenai/olmOCR-bench" +if bench_branch: + git_clone_cmd += f" -b {bench_branch}" + +commands.extend([ + git_clone_cmd, + "cd olmOCR-bench && git lfs pull && cd ..", + pipeline_cmd, + "python olmocr/bench/scripts/workspace_to_bench.py localworkspace/ olmOCR-bench/bench_data/olmocr --bench-path ./olmOCR-bench/", + "pip install s5cmd", + "s5cmd cp localworkspace/ s3://ai2-oe-data/jakep/olmocr-bench-runs/$BEAKER_WORKLOAD_ID/", + "python -m olmocr.bench.benchmark --dir ./olmOCR-bench/bench_data" +]) + +# Build task spec with optional env vars +# If image_tag contains '/', it's already a full beaker image reference +if '/' in image_tag: + image_ref = image_tag +else: + image_ref = f"{beaker_user}/{image_tag}" + +task_spec_args = { + "name": "olmocr-infrapartner-benchmark", + "image": ImageSource(beaker=image_ref), + "command": [ + "bash", "-c", + " && ".join(commands) + ], + "context": TaskContext( + priority=Priority.normal, + preemptible=True, + ), + "resources": TaskResources(gpu_count=0), + "constraints": Constraints(cluster=["ai2/phobos", "ai2/neptune", "ai2/saturn"]), + "result": ResultSpec(path="/noop-results"), +} + +# Build env vars list +env_vars = [] +if has_aws_creds: + env_vars.append(EnvVar(name="AWS_CREDENTIALS_FILE", secret=aws_creds_secret)) +if beaker_secret: + env_vars.append(EnvVar(name="BEAKER_API_KEY", secret=beaker_secret)) + +# Add env vars if any exist +if env_vars: + task_spec_args["env_vars"] = env_vars + +# Create experiment spec +experiment_spec = ExperimentSpec( + description=f"OlmOCR InfraPartner Benchmark Run - Branch: {git_branch}, Commit: {git_hash}", + budget="ai2/oe-base", + tasks=[TaskSpec(**task_spec_args)], +) + +# Create the experiment +experiment = b.experiment.create(spec=experiment_spec, workspace="ai2/olmocr") +print(f"Created benchmark experiment: {experiment.id}") +print(f"View at: https://beaker.org/ex/{experiment.id}") +print("-------") +print("") +print("Note: Performance test has been skipped for infrapartner benchmark") +EOF + +# Run the Python script to create the experiment +echo "Creating Beaker experiment..." + +# Build command with appropriate arguments +CMD="$PYTHON /tmp/run_infrapartner_benchmark_experiment.py $IMAGE_TAG $BEAKER_USER $GIT_BRANCH $GIT_HASH" + +if [ -n "$MODEL" ]; then + echo "Using model: $MODEL" + CMD="$CMD --model $MODEL" +fi + +if [ -n "$SERVER" ]; then + echo "Using server: $SERVER" + CMD="$CMD --server $SERVER" +fi + +if [ -n "$BEAKER_SECRET" ]; then + echo "Using beaker secret for API key: $BEAKER_SECRET" + CMD="$CMD --beaker-secret $BEAKER_SECRET" +fi + +if [ -n "$BENCH_BRANCH" ]; then + echo "Using bench branch: $BENCH_BRANCH" + CMD="$CMD --benchbranch $BENCH_BRANCH" +fi + +eval $CMD + +# Clean up temporary file +rm /tmp/run_infrapartner_benchmark_experiment.py + +echo "InfraPartner benchmark experiment submitted successfully!" \ No newline at end of file diff --git a/scripts/run_integration_test.sh b/scripts/run_integration_test.sh new file mode 100644 index 0000000..e6e1a98 --- /dev/null +++ b/scripts/run_integration_test.sh @@ -0,0 +1,6 @@ +#/usr/bin/bash + +set -ex + +python -m olmocr.pipeline ./localworkspace --pdfs tests/gnarly_pdfs/ambiguous.pdf tests/gnarly_pdfs/edgar.pdf tests/gnarly_pdfs/dolma-page-1.pdf \ + && pytest -m "" tests/test_integration.py diff --git a/scripts/run_marker_benchmark.sh b/scripts/run_marker_benchmark.sh new file mode 100755 index 0000000..bb9875b --- /dev/null +++ b/scripts/run_marker_benchmark.sh @@ -0,0 +1,202 @@ +#!/bin/bash + +# Runs marker benchmark, measuring both olmOCR-bench performance and per document processing performance +# ./scripts/run_marker_benchmark.sh +# ./scripts/run_marker_benchmark.sh 1.7.5 + +set -e + +# Parse command line arguments +MARKER_VERSION="${1:-1.7.5}" +echo "Using marker version: $MARKER_VERSION" + +# Check for uncommitted changes +if ! git diff-index --quiet HEAD --; then + echo "Error: There are uncommitted changes in the repository." + echo "Please commit or stash your changes before running the benchmark." + echo "" + echo "Uncommitted changes:" + git status --short + exit 1 +fi + +# Use conda environment Python if available, otherwise use system Python +if [ -n "$CONDA_PREFIX" ]; then + PYTHON="$CONDA_PREFIX/bin/python" + echo "Using conda Python from: $CONDA_PREFIX" +else + PYTHON="python" + echo "Warning: No conda environment detected, using system Python" +fi + +# Get version from version.py +VERSION=$($PYTHON -c 'import olmocr.version; print(olmocr.version.VERSION)') +echo "OlmOCR version: $VERSION" + +# Get first 10 characters of git hash +GIT_HASH=$(git rev-parse HEAD | cut -c1-10) +echo "Git hash: $GIT_HASH" + +# Get current git branch name +GIT_BRANCH=$(git rev-parse --abbrev-ref HEAD) +echo "Git branch: $GIT_BRANCH" + +# Create full image tag +IMAGE_TAG="olmocr-benchmark-${VERSION}-${GIT_HASH}" +echo "Building Docker image with tag: $IMAGE_TAG" + +# Build the Docker image +echo "Building Docker image..." +docker build --platform linux/amd64 -f ./Dockerfile -t $IMAGE_TAG . + +# Get Beaker username +BEAKER_USER=$(beaker account whoami --format json | jq -r '.[0].name') +echo "Beaker user: $BEAKER_USER" + +# Push image to beaker +echo "Trying to push image to Beaker..." +if ! beaker image create --workspace ai2/oe-data-pdf --name $IMAGE_TAG $IMAGE_TAG 2>/dev/null; then + echo "Warning: Beaker image with tag $IMAGE_TAG already exists. Using existing image." +fi + +# Create Python script to run beaker experiment +cat << 'EOF' > /tmp/run_benchmark_experiment.py +import sys +from beaker import Beaker, ExperimentSpec, TaskSpec, TaskContext, ResultSpec, TaskResources, ImageSource, Priority, Constraints, EnvVar + +# Get image tag, beaker user, git branch, git hash, and marker version from command line +image_tag = sys.argv[1] +beaker_user = sys.argv[2] +git_branch = sys.argv[3] +git_hash = sys.argv[4] +marker_version = sys.argv[5] + +# Initialize Beaker client +b = Beaker.from_env(default_workspace="ai2/olmocr") + + +# Check if AWS credentials secret exists +aws_creds_secret = f"{beaker_user}-AWS_CREDENTIALS_FILE" +try: + # Try to get the secret to see if it exists + b.secret.get(aws_creds_secret, workspace="ai2/olmocr") + has_aws_creds = True + print(f"Found AWS credentials secret: {aws_creds_secret}") +except: + has_aws_creds = False + print(f"AWS credentials secret not found: {aws_creds_secret}") + +# First experiment: Original benchmark job +commands = [] +if has_aws_creds: + commands.extend([ + "mkdir -p ~/.aws", + 'echo "$AWS_CREDENTIALS_FILE" > ~/.aws/credentials' + ]) +commands.extend([ + "git clone https://huggingface.co/datasets/allenai/olmOCR-bench", + "cd olmOCR-bench && git lfs pull && cd ..", + f"pip install marker-pdf=={marker_version}", + "pip install --upgrade torchvision", + "python -m olmocr.bench.convert marker --dir ./olmOCR-bench/bench_data", + "python -m olmocr.bench.benchmark --dir ./olmOCR-bench/bench_data" +]) + +# Build task spec with optional env vars +task_spec_args = { + "name": "marker-benchmark", + "image": ImageSource(beaker=f"{beaker_user}/{image_tag}"), + "command": [ + "bash", "-c", + " && ".join(commands) + ], + "context": TaskContext( + priority=Priority.normal, + preemptible=True, + ), + "resources": TaskResources(gpu_count=1), + "constraints": Constraints(cluster=["ai2/ceres-cirrascale", "ai2/jupiter-cirrascale-2"]), + "result": ResultSpec(path="/noop-results"), +} + +# Add env vars if AWS credentials exist +if has_aws_creds: + task_spec_args["env_vars"] = [ + EnvVar(name="AWS_CREDENTIALS_FILE", secret=aws_creds_secret) + ] + +# Create first experiment spec +experiment_spec = ExperimentSpec( + description=f"Marker {marker_version} Benchmark Run - Branch: {git_branch}, Commit: {git_hash}", + budget="ai2/oe-base", + tasks=[TaskSpec(**task_spec_args)], +) + +# Create the first experiment +experiment = b.experiment.create(spec=experiment_spec, workspace="ai2/olmocr") +print(f"Created benchmark experiment: {experiment.id}") +print(f"View at: https://beaker.org/ex/{experiment.id}") +print("-------") +print("") + + +perf_commands = [] +if has_aws_creds: + perf_commands.extend([ + "mkdir -p ~/.aws", + 'echo "$AWS_CREDENTIALS_FILE" > ~/.aws/credentials' + ]) +perf_commands.extend([ + f"pip install marker-pdf=={marker_version}", + "pip install --upgrade torchvision", + "pip install awscli", + "aws s3 cp --recursive s3://ai2-oe-data/jakep/olmocr/olmOCR-mix-0225/benchmark_set/ /root/olmOCR-mix-0225_benchmark_set/", + # Tried with workers 8, but it was taking a really huge amount of time + #"time marker --force_ocr /root/olmOCR-mix-0225_benchmark_set/ --output_dir /root/olmOCR-mix-0225_benchmark_set_marker --workers 8" + "time marker --force_ocr /root/olmOCR-mix-0225_benchmark_set/ --output_dir /root/olmOCR-mix-0225_benchmark_set_marker" +]) + +# Build performance task spec +perf_task_spec_args = { + "name": "marker-performance", + "image": ImageSource(beaker=f"{beaker_user}/{image_tag}"), + "command": [ + "bash", "-c", + " && ".join(perf_commands) + ], + "context": TaskContext( + priority=Priority.normal, + preemptible=True, + ), + "resources": TaskResources(gpu_count=1), + "constraints": Constraints(cluster=["ai2/ceres-cirrascale", "ai2/jupiter-cirrascale-2"]), + "result": ResultSpec(path="/noop-results"), +} + +# Add env vars if AWS credentials exist +if has_aws_creds: + perf_task_spec_args["env_vars"] = [ + EnvVar(name="AWS_CREDENTIALS_FILE", secret=aws_creds_secret) + ] + +# Create performance experiment spec +perf_experiment_spec = ExperimentSpec( + description=f"Marker {marker_version} Performance Test - Branch: {git_branch}, Commit: {git_hash}", + budget="ai2/oe-base", + tasks=[TaskSpec(**perf_task_spec_args)], +) + +# Create the performance experiment +perf_experiment = b.experiment.create(spec=perf_experiment_spec, workspace="ai2/olmocr") +print(f"Created performance experiment: {perf_experiment.id}") +print(f"View at: https://beaker.org/ex/{perf_experiment.id}") +EOF + +# Run the Python script to create the experiments +echo "Creating Beaker experiments..." +$PYTHON /tmp/run_benchmark_experiment.py $IMAGE_TAG $BEAKER_USER $GIT_BRANCH $GIT_HASH $MARKER_VERSION + +# Clean up temporary file +rm /tmp/run_benchmark_experiment.py + +echo "Benchmark experiments submitted successfully!" \ No newline at end of file diff --git a/scripts/run_mineru_benchmark.sh b/scripts/run_mineru_benchmark.sh new file mode 100755 index 0000000..a825aad --- /dev/null +++ b/scripts/run_mineru_benchmark.sh @@ -0,0 +1,283 @@ +#!/bin/bash + +# Runs mineru benchmark, measuring both olmOCR-bench performance and per document processing performance +# ./scripts/run_mineru_benchmark.sh +# ./scripts/run_mineru_benchmark.sh [version] + +# Rough steps needed to run: +# uv pip install -U --system "mineru[core]" +# git clone https://huggingface.co/datasets/allenai/olmOCR-bench +# Run mineru -p -o for each folder in olmOCR-bench/bench_data/pdfs and put all output paths into ~/mineru_bench_[foldername] +# root@triton-cs-aus-455:/build# ls olmOCR-bench/bench_data/pdfs +# arxiv_math headers_footers long_tiny_text multi_column old_scans old_scans_math tables +# # +# mineru -p olmOCR-bench/bench_data/pdfs/arxiv_math -o ~/mineru_bench_arxiv_math +# +# root@triton-cs-aus-455:/build# ls /root/mineru_bench_arxiv_math/2503.05000_pg9/auto/ +# 2503.05000_pg9.md 2503.05000_pg9_layout.pdf 2503.05000_pg9_model.json 2503.05000_pg9_span.pdf +# 2503.05000_pg9_content_list.json 2503.05000_pg9_middle.json 2503.05000_pg9_origin.pdf images +# +# So, now, the original pdf was olmOCR-bench/bench_data/pdfs/arxiv_math/2503.05000_pg9.pdf +# you would need to move the mineru_bench_arxiv_math/2503.05000_pg9/auto/2503.05000_pg9.md +# to olmOCR-bench/bench_data/mineru/arxiv_math/2503.05000_pg9_pg1_repeat1.md +# Then, once that's all done, run python -m olmocr.bench.benchmark --dir ./olmOCR-bench/bench_data +# For the benchmark job, just run a single mineru convert command and time how long it takes + +set -e + +# Parse command line arguments +MINERU_VERSION="${1:-latest}" +if [ "$MINERU_VERSION" = "latest" ]; then + echo "Using latest mineru release" +else + echo "Using mineru version: $MINERU_VERSION" +fi + +# Use conda environment Python if available, otherwise use system Python +if [ -n "$CONDA_PREFIX" ]; then + PYTHON="$CONDA_PREFIX/bin/python" + echo "Using conda Python from: $CONDA_PREFIX" +else + PYTHON="python" + echo "Warning: No conda environment detected, using system Python" +fi + +# Get version from version.py +VERSION=$($PYTHON -c 'import olmocr.version; print(olmocr.version.VERSION)') +echo "OlmOCR version: $VERSION" + +# Get first 10 characters of git hash +GIT_HASH=$(git rev-parse HEAD | cut -c1-10) +echo "Git hash: $GIT_HASH" + +# Get current git branch name +GIT_BRANCH=$(git rev-parse --abbrev-ref HEAD) +echo "Git branch: $GIT_BRANCH" + +# Create full image tag +IMAGE_TAG="olmocr-benchmark-${VERSION}-${GIT_HASH}" +echo "Building Docker image with tag: $IMAGE_TAG" + +# Build the Docker image +echo "Building Docker image..." +docker build --platform linux/amd64 -f ./Dockerfile -t $IMAGE_TAG . + +# Get Beaker username +BEAKER_USER=$(beaker account whoami --format json | jq -r '.[0].name') +echo "Beaker user: $BEAKER_USER" + +# Push image to beaker +echo "Trying to push image to Beaker..." +if ! beaker image create --workspace ai2/oe-data-pdf --name $IMAGE_TAG $IMAGE_TAG 2>/dev/null; then + echo "Warning: Beaker image with tag $IMAGE_TAG already exists. Using existing image." +fi + +# Create Python script to run beaker experiment +cat << 'EOF' > /tmp/run_benchmark_experiment.py +import sys +from textwrap import dedent +from beaker import Beaker, ExperimentSpec, TaskSpec, TaskContext, ResultSpec, TaskResources, ImageSource, Priority, Constraints, EnvVar + +# Get image tag, beaker user, git branch, git hash, and mineru version from command line +image_tag = sys.argv[1] +beaker_user = sys.argv[2] +git_branch = sys.argv[3] +git_hash = sys.argv[4] +mineru_version = sys.argv[5] + +# Initialize Beaker client +b = Beaker.from_env(default_workspace="ai2/olmocr") + + +# Check if AWS credentials secret exists +aws_creds_secret = f"{beaker_user}-AWS_CREDENTIALS_FILE" +try: + # Try to get the secret to see if it exists + b.secret.get(aws_creds_secret, workspace="ai2/olmocr") + has_aws_creds = True + print(f"Found AWS credentials secret: {aws_creds_secret}") +except: + has_aws_creds = False + print(f"AWS credentials secret not found: {aws_creds_secret}") + +# Determine how to install mineru +if mineru_version.lower() == "latest": + mineru_install_cmd = 'pip install --upgrade "mineru[core]"' + mineru_version_label = "latest" +else: + mineru_install_cmd = f'pip install --upgrade "mineru[core]=={mineru_version}"' + mineru_version_label = mineru_version + +run_mineru_shell = dedent("""\ +bash -lc 'set -euo pipefail +PDF_ROOT="olmOCR-bench/bench_data/pdfs" +TARGET_ROOT="olmOCR-bench/bench_data/mineru" +rm -rf "$TARGET_ROOT" +mkdir -p "$TARGET_ROOT" + +echo "Running MinerU conversions..." +for folder in "$PDF_ROOT"/*; do + if [ ! -d "$folder" ]; then + continue + fi + section=$(basename "$folder") + output_dir="$HOME/mineru_bench_${section}" + rm -rf "$output_dir" + echo " Processing $folder" + mineru -p "$folder" -o "$output_dir" +done + +echo "Collecting MinerU markdown outputs..." +find "$PDF_ROOT" -type f -name "*.pdf" | while IFS= read -r pdf_path; do + rel_path=${pdf_path#"$PDF_ROOT"/} + case "$rel_path" in + */*) + section=${rel_path%%/*} + ;; + *) + echo "Warning: Unexpected PDF path layout for $pdf_path, skipping" >&2 + continue + ;; + esac + + output_dir="$HOME/mineru_bench_${section}" + pdf_stem=$(basename "$pdf_path" .pdf) + rel_dir=$(dirname "$rel_path") + if [ "$rel_dir" = "." ]; then + rel_dir="" + fi + + primary_md=$(find "$output_dir" -path "*/auto/${pdf_stem}.md" -print -quit) + if [ -z "$primary_md" ]; then + primary_md=$(find "$output_dir" -name "${pdf_stem}.md" -print -quit) + fi + + if [ -z "$primary_md" ]; then + echo "Warning: no markdown output for $pdf_path" >&2 + continue + fi + + dest_dir="$TARGET_ROOT" + if [ -n "$rel_dir" ]; then + dest_dir="$TARGET_ROOT/$rel_dir" + fi + mkdir -p "$dest_dir" + dest_path="$dest_dir/${pdf_stem}_pg1_repeat1.md" + cp "$primary_md" "$dest_path" + echo " Copied $primary_md -> $dest_path" +done +'""") + +# First experiment: Original benchmark job +commands = [] +if has_aws_creds: + commands.extend([ + "mkdir -p ~/.aws", + 'echo "$AWS_CREDENTIALS_FILE" > ~/.aws/credentials' + ]) +commands.extend([ + "git clone https://huggingface.co/datasets/allenai/olmOCR-bench", + "cd olmOCR-bench && git lfs pull && cd ..", + "pip install --upgrade pip", + mineru_install_cmd, + run_mineru_shell, + "python -m olmocr.bench.benchmark --dir ./olmOCR-bench/bench_data --candidate mineru" +]) + +# Build task spec with optional env vars +task_spec_args = { + "name": "mineru-benchmark", + "image": ImageSource(beaker=f"{beaker_user}/{image_tag}"), + "command": [ + "bash", "-c", + " && ".join(commands) + ], + "context": TaskContext( + priority=Priority.normal, + preemptible=True, + ), + "resources": TaskResources(gpu_count=1), + "constraints": Constraints(cluster=["ai2/ceres-cirrascale", "ai2/jupiter-cirrascale-2"]), + "result": ResultSpec(path="/noop-results"), +} + +# Add env vars if AWS credentials exist +if has_aws_creds: + task_spec_args["env_vars"] = [ + EnvVar(name="AWS_CREDENTIALS_FILE", secret=aws_creds_secret) + ] + +# Create first experiment spec +experiment_spec = ExperimentSpec( + description=f"MinerU {mineru_version_label} Benchmark Run - Branch: {git_branch}, Commit: {git_hash}", + budget="ai2/oe-base", + tasks=[TaskSpec(**task_spec_args)], +) + +# Create the first experiment +experiment = b.experiment.create(spec=experiment_spec, workspace="ai2/olmocr") +print(f"Created benchmark experiment: {experiment.id}") +print(f"View at: https://beaker.org/ex/{experiment.id}") +print("-------") +print("") + + +perf_commands = [] +if has_aws_creds: + perf_commands.extend([ + "mkdir -p ~/.aws", + 'echo "$AWS_CREDENTIALS_FILE" > ~/.aws/credentials' + ]) +perf_commands.extend([ + "pip install --upgrade pip", + mineru_install_cmd, + "pip install awscli", + "aws s3 cp --recursive s3://ai2-oe-data/jakep/olmocr/olmOCR-mix-0225/benchmark_set/ /root/olmOCR-mix-0225_benchmark_set/", + "rm -rf /root/olmOCR-mix-0225_benchmark_set_mineru", + "time mineru -p /root/olmOCR-mix-0225_benchmark_set/ -o /root/olmOCR-mix-0225_benchmark_set_mineru" +]) + +# Build performance task spec +perf_task_spec_args = { + "name": "mineru-performance", + "image": ImageSource(beaker=f"{beaker_user}/{image_tag}"), + "command": [ + "bash", "-c", + " && ".join(perf_commands) + ], + "context": TaskContext( + priority=Priority.normal, + preemptible=True, + ), + "resources": TaskResources(gpu_count=1), + "constraints": Constraints(cluster=["ai2/ceres-cirrascale", "ai2/jupiter-cirrascale-2"]), + "result": ResultSpec(path="/noop-results"), +} + +# Add env vars if AWS credentials exist +if has_aws_creds: + perf_task_spec_args["env_vars"] = [ + EnvVar(name="AWS_CREDENTIALS_FILE", secret=aws_creds_secret) + ] + +# Create performance experiment spec +perf_experiment_spec = ExperimentSpec( + description=f"MinerU {mineru_version_label} Performance Test - Branch: {git_branch}, Commit: {git_hash}", + budget="ai2/oe-base", + tasks=[TaskSpec(**perf_task_spec_args)], +) + +# Create the performance experiment +perf_experiment = b.experiment.create(spec=perf_experiment_spec, workspace="ai2/olmocr") +print(f"Created performance experiment: {perf_experiment.id}") +print(f"View at: https://beaker.org/ex/{perf_experiment.id}") +EOF + +# Run the Python script to create the experiments +echo "Creating Beaker experiments..." +$PYTHON /tmp/run_benchmark_experiment.py $IMAGE_TAG $BEAKER_USER $GIT_BRANCH $GIT_HASH $MINERU_VERSION + +# Clean up temporary file +rm /tmp/run_benchmark_experiment.py + +echo "Benchmark experiments submitted successfully!" diff --git a/scripts/run_paddle_benchmark.sh b/scripts/run_paddle_benchmark.sh new file mode 100755 index 0000000..1d22f1e --- /dev/null +++ b/scripts/run_paddle_benchmark.sh @@ -0,0 +1,183 @@ +#!/bin/bash + +# Runs benchmark on paddle ocr v5 version 3.2 +set -e + + +# Use conda environment Python if available, otherwise use system Python +if [ -n "$CONDA_PREFIX" ]; then + PYTHON="$CONDA_PREFIX/bin/python" + echo "Using conda Python from: $CONDA_PREFIX" +else + PYTHON="python" + echo "Warning: No conda environment detected, using system Python" +fi + +# Get version from version.py +VERSION=$($PYTHON -c 'import olmocr.version; print(olmocr.version.VERSION)') +echo "OlmOCR version: $VERSION" + +# Get first 10 characters of git hash +GIT_HASH=$(git rev-parse HEAD | cut -c1-10) +echo "Git hash: $GIT_HASH" + +# Get current git branch name +GIT_BRANCH=$(git rev-parse --abbrev-ref HEAD) +echo "Git branch: $GIT_BRANCH" + +# Create full image tag +IMAGE_TAG="olmocr-benchmark-${VERSION}-${GIT_HASH}" +echo "Building Docker image with tag: $IMAGE_TAG" + +# Build the Docker image +echo "Building Docker image..." +docker build --platform linux/amd64 -f ./Dockerfile -t $IMAGE_TAG . + +# Get Beaker username +BEAKER_USER=$(beaker account whoami --format json | jq -r '.[0].name') +echo "Beaker user: $BEAKER_USER" + +# Push image to beaker +echo "Trying to push image to Beaker..." +if ! beaker image create --workspace ai2/oe-data-pdf --name $IMAGE_TAG $IMAGE_TAG 2>/dev/null; then + echo "Warning: Beaker image with tag $IMAGE_TAG already exists. Using existing image." +fi + +# Create Python script to run beaker experiment +cat << 'EOF' > /tmp/run_benchmark_experiment.py +import sys +from beaker import Beaker, ExperimentSpec, TaskSpec, TaskContext, ResultSpec, TaskResources, ImageSource, Priority, Constraints, EnvVar + +# Get image tag, beaker user, git branch, git hash, and marker version from command line +image_tag = sys.argv[1] +beaker_user = sys.argv[2] +git_branch = sys.argv[3] +git_hash = sys.argv[4] + +# Initialize Beaker client +b = Beaker.from_env(default_workspace="ai2/olmocr") + + +# Check if AWS credentials secret exists +aws_creds_secret = f"{beaker_user}-AWS_CREDENTIALS_FILE" +try: + # Try to get the secret to see if it exists + b.secret.get(aws_creds_secret, workspace="ai2/olmocr") + has_aws_creds = True + print(f"Found AWS credentials secret: {aws_creds_secret}") +except: + has_aws_creds = False + print(f"AWS credentials secret not found: {aws_creds_secret}") + +# First experiment: Original benchmark job +commands = [] +if has_aws_creds: + commands.extend([ + "mkdir -p ~/.aws", + 'echo "$AWS_CREDENTIALS_FILE" > ~/.aws/credentials' + ]) +commands.extend([ + "git clone https://huggingface.co/datasets/allenai/olmOCR-bench", + "cd olmOCR-bench && git lfs pull && cd ..", + f"python -m pip install paddlepaddle-gpu==3.2.0 -i https://www.paddlepaddle.org.cn/packages/stable/cu126/", + "python -m pip install \"paddleocr[all]\"", + "python -m olmocr.bench.convert paddlepaddle --dir ./olmOCR-bench/bench_data --parallel 0", + "python -m olmocr.bench.benchmark --dir ./olmOCR-bench/bench_data" +]) + +# Build task spec with optional env vars +task_spec_args = { + "name": "paddle-benchmark", + "image": ImageSource(beaker=f"{beaker_user}/{image_tag}"), + "command": [ + "bash", "-c", + " && ".join(commands) + ], + "context": TaskContext( + priority=Priority.normal, + preemptible=True, + ), + "resources": TaskResources(gpu_count=1), + "constraints": Constraints(cluster=["ai2/neptune-cirrascale"]), + "result": ResultSpec(path="/noop-results"), +} + +# Add env vars if AWS credentials exist +if has_aws_creds: + task_spec_args["env_vars"] = [ + EnvVar(name="AWS_CREDENTIALS_FILE", secret=aws_creds_secret) + ] + +# Create first experiment spec +experiment_spec = ExperimentSpec( + description=f"Paddle Benchmark Run - Branch: {git_branch}, Commit: {git_hash}", + budget="ai2/oe-base", + tasks=[TaskSpec(**task_spec_args)], +) + +# Create the first experiment +experiment = b.experiment.create(spec=experiment_spec, workspace="ai2/olmocr") +print(f"Created benchmark experiment: {experiment.id}") +print(f"View at: https://beaker.org/ex/{experiment.id}") +print("-------") +print("") + + +perf_commands = [] +if has_aws_creds: + perf_commands.extend([ + "mkdir -p ~/.aws", + 'echo "$AWS_CREDENTIALS_FILE" > ~/.aws/credentials' + ]) +perf_commands.extend([ + f"python -m pip install paddlepaddle-gpu==3.2.0 -i https://www.paddlepaddle.org.cn/packages/stable/cu126/", + "python -m pip install \"paddleocr[all]\"", + "pip install awscli", + "aws s3 cp --recursive s3://ai2-oe-data/jakep/olmocr/olmOCR-mix-0225/benchmark_set/ /root/olmOCR-mix-0225_benchmark_set/", + "time paddleocr pp_structurev3 -i /root/olmOCR-mix-0225_benchmark_set/ --text_detection_model_name PP-OCRv5_server_det --use_doc_orientation_classify False --use_doc_unwarping False --use_textline_orientation False --device gpu:0" +]) + +# Build performance task spec +perf_task_spec_args = { + "name": "paddle-performance", + "image": ImageSource(beaker=f"{beaker_user}/{image_tag}"), + "command": [ + "bash", "-c", + " && ".join(perf_commands) + ], + "context": TaskContext( + priority=Priority.normal, + preemptible=True, + ), + "resources": TaskResources(gpu_count=1), + "constraints": Constraints(cluster=["ai2/ceres-cirrascale", "ai2/jupiter-cirrascale-2"]), + "result": ResultSpec(path="/noop-results"), +} + +# Add env vars if AWS credentials exist +if has_aws_creds: + perf_task_spec_args["env_vars"] = [ + EnvVar(name="AWS_CREDENTIALS_FILE", secret=aws_creds_secret) + ] + +# Create performance experiment spec +perf_experiment_spec = ExperimentSpec( + description=f"Paddle Performance Test - Branch: {git_branch}, Commit: {git_hash}", + budget="ai2/oe-base", + tasks=[TaskSpec(**perf_task_spec_args)], +) + +# Create the performance experiment +perf_experiment = b.experiment.create(spec=perf_experiment_spec, workspace="ai2/olmocr") +print(f"Created performance experiment: {perf_experiment.id}") +print(f"View at: https://beaker.org/ex/{perf_experiment.id}") +EOF + +# Run the Python script to create the experiments +echo "Creating Beaker experiments..." +$PYTHON /tmp/run_benchmark_experiment.py $IMAGE_TAG $BEAKER_USER $GIT_BRANCH $GIT_HASH + +# Clean up temporary file +rm /tmp/run_benchmark_experiment.py + +echo "Benchmark experiments submitted successfully!" \ No newline at end of file diff --git a/scripts/run_paddlevl_benchmark.sh b/scripts/run_paddlevl_benchmark.sh new file mode 100755 index 0000000..705fb85 --- /dev/null +++ b/scripts/run_paddlevl_benchmark.sh @@ -0,0 +1,138 @@ +#!/bin/bash + +# Runs benchmark on paddle ocr v5 version 3.2 +set -e + + +# Use conda environment Python if available, otherwise use system Python +if [ -n "$CONDA_PREFIX" ]; then + PYTHON="$CONDA_PREFIX/bin/python" + echo "Using conda Python from: $CONDA_PREFIX" +else + PYTHON="python" + echo "Warning: No conda environment detected, using system Python" +fi + +# Get version from version.py +VERSION=$($PYTHON -c 'import olmocr.version; print(olmocr.version.VERSION)') +echo "OlmOCR version: $VERSION" + +# Get first 10 characters of git hash +GIT_HASH=$(git rev-parse HEAD | cut -c1-10) +echo "Git hash: $GIT_HASH" + +# Get current git branch name +GIT_BRANCH=$(git rev-parse --abbrev-ref HEAD) +echo "Git branch: $GIT_BRANCH" + +# Create full image tag +IMAGE_TAG="olmocr-benchmark-${VERSION}-${GIT_HASH}" +echo "Building Docker image with tag: $IMAGE_TAG" + +# Build the Docker image +echo "Building Docker image..." +docker build --platform linux/amd64 -f ./Dockerfile -t $IMAGE_TAG . + +# Get Beaker username +BEAKER_USER=$(beaker account whoami --format json | jq -r '.[0].name') +echo "Beaker user: $BEAKER_USER" + +# Push image to beaker +echo "Trying to push image to Beaker..." +if ! beaker image create --workspace ai2/oe-data-pdf --name $IMAGE_TAG $IMAGE_TAG 2>/dev/null; then + echo "Warning: Beaker image with tag $IMAGE_TAG already exists. Using existing image." +fi + +# Create Python script to run beaker experiment +cat << 'EOF' > /tmp/run_benchmark_experiment.py +import sys +from beaker import Beaker, ExperimentSpec, TaskSpec, TaskContext, ResultSpec, TaskResources, ImageSource, Priority, Constraints, EnvVar + +# Get image tag, beaker user, git branch, git hash, and marker version from command line +image_tag = sys.argv[1] +beaker_user = sys.argv[2] +git_branch = sys.argv[3] +git_hash = sys.argv[4] + +# Initialize Beaker client +b = Beaker.from_env(default_workspace="ai2/olmocr") + + +# Check if AWS credentials secret exists +aws_creds_secret = f"{beaker_user}-AWS_CREDENTIALS_FILE" +try: + # Try to get the secret to see if it exists + b.secret.get(aws_creds_secret, workspace="ai2/olmocr") + has_aws_creds = True + print(f"Found AWS credentials secret: {aws_creds_secret}") +except: + has_aws_creds = False + print(f"AWS credentials secret not found: {aws_creds_secret}") + +# First experiment: Original benchmark job +commands = [] +if has_aws_creds: + commands.extend([ + "mkdir -p ~/.aws", + 'echo "$AWS_CREDENTIALS_FILE" > ~/.aws/credentials' + ]) +commands.extend([ + "git clone https://huggingface.co/datasets/allenai/olmOCR-bench", + "cd olmOCR-bench && git lfs pull && cd ..", + "python -m pip install paddlepaddle-gpu==3.2.0 -i https://www.paddlepaddle.org.cn/packages/stable/cu126/", + "python -m pip install -U \"paddleocr[doc-parser]\"", + "python -m pip install torch torchvision --index-url https://download.pytorch.org/whl/cu126", + "python -m pip install https://paddle-whl.bj.bcebos.com/nightly/cu126/safetensors/safetensors-0.6.2.dev0-cp38-abi3-linux_x86_64.whl", + + "python -m olmocr.bench.convert paddlevl --dir ./olmOCR-bench/bench_data --parallel 0", + "pip install s5cmd", + "s5cmd cp olmOCR-bench/bench_data/ s3://ai2-oe-data/jakep/olmocr-bench-runs/$BEAKER_WORKLOAD_ID/", + "python -m olmocr.bench.benchmark --dir ./olmOCR-bench/bench_data" +]) + +# Build task spec with optional env vars +task_spec_args = { + "name": "paddle-benchmark", + "image": ImageSource(beaker=f"{beaker_user}/{image_tag}"), + "command": [ + "bash", "-c", + " && ".join(commands) + ], + "context": TaskContext( + priority=Priority.urgent, + preemptible=False, + ), + "resources": TaskResources(gpu_count=1), + "constraints": Constraints(cluster=["ai2/jupiter"]), + "result": ResultSpec(path="/noop-results"), +} + +# Add env vars if AWS credentials exist +if has_aws_creds: + task_spec_args["env_vars"] = [ + EnvVar(name="AWS_CREDENTIALS_FILE", secret=aws_creds_secret) + ] + +# Create first experiment spec +experiment_spec = ExperimentSpec( + description=f"Paddle Benchmark Run - Branch: {git_branch}, Commit: {git_hash}", + budget="ai2/oe-base", + tasks=[TaskSpec(**task_spec_args)], +) + +# Create the first experiment +experiment = b.experiment.create(spec=experiment_spec, workspace="ai2/olmocr") +print(f"Created benchmark experiment: {experiment.id}") +print(f"View at: https://beaker.org/ex/{experiment.id}") +print("-------") +print("") +EOF + +# Run the Python script to create the experiments +echo "Creating Beaker experiments..." +$PYTHON /tmp/run_benchmark_experiment.py $IMAGE_TAG $BEAKER_USER $GIT_BRANCH $GIT_HASH + +# Clean up temporary file +rm /tmp/run_benchmark_experiment.py + +echo "Benchmark experiments submitted successfully!" \ No newline at end of file diff --git a/scripts/run_qianfan_benchmark.sh b/scripts/run_qianfan_benchmark.sh new file mode 100755 index 0000000..c53d08b --- /dev/null +++ b/scripts/run_qianfan_benchmark.sh @@ -0,0 +1,446 @@ +#!/bin/bash + +# Runs qianfan benchmark, measuring both olmOCR-bench performance and per document processing performance +# Uses baidu/Qianfan-OCR model served via vllm with OpenAI-compatible API +# Reference: https://huggingface.co/baidu/Qianfan-OCR +# https://github.com/baidubce/skills/blob/develop/skills/qianfanocr-document-intelligence/scripts/qianfan_ocr_cli.py +# +# Usage: +# ./scripts/run_qianfan_benchmark.sh # Use default model +# ./scripts/run_qianfan_benchmark.sh --benchrepo allenai/olmOCR-bench-internal # Use different benchmark repo +# ./scripts/run_qianfan_benchmark.sh --benchbranch olmOCR-bench-1125 # Use specific branch/revision +# ./scripts/run_qianfan_benchmark.sh --benchpath s3://ai2-oe-data/path/ # Use benchmark from S3 or local path +# ./scripts/run_qianfan_benchmark.sh --cluster ai2/titan-cirrascale # Specify a cluster +# ./scripts/run_qianfan_benchmark.sh --beaker-image jakep/olmocr-benchmark-0.3.3-780bc7d934 # Skip Docker build +# ./scripts/run_qianfan_benchmark.sh --noperf # Skip the performance test job +# ./scripts/run_qianfan_benchmark.sh --model baidu/Qianfan-OCR # Use a specific model + +set -e + +# Parse command line arguments +BENCH_BRANCH="" +BENCH_REPO="" +BENCH_PATH="" +CLUSTER="" +BEAKER_IMAGE="" +NOPERF="" +MODEL="" + +while [[ $# -gt 0 ]]; do + case $1 in + --benchbranch) + BENCH_BRANCH="$2" + shift 2 + ;; + --benchrepo) + BENCH_REPO="$2" + shift 2 + ;; + --benchpath) + BENCH_PATH="$2" + shift 2 + ;; + --cluster) + CLUSTER="$2" + shift 2 + ;; + --beaker-image) + BEAKER_IMAGE="$2" + shift 2 + ;; + --noperf) + NOPERF="1" + shift + ;; + --model) + MODEL="$2" + shift 2 + ;; + *) + echo "Unknown option: $1" + echo "Usage: $0 [--benchbranch BRANCH] [--benchrepo REPO] [--benchpath PATH] [--cluster CLUSTER] [--beaker-image IMAGE] [--noperf] [--model MODEL]" + exit 1 + ;; + esac +done + +# Check for mutual exclusivity between benchpath and benchrepo/benchbranch +if [ -n "$BENCH_PATH" ] && ([ -n "$BENCH_REPO" ] || [ -n "$BENCH_BRANCH" ]); then + echo "Error: --benchpath is mutually exclusive with --benchrepo and --benchbranch" + echo "Use either --benchpath OR --benchrepo/--benchbranch, not both." + exit 1 +fi + +# Use conda environment Python if available, otherwise use system Python +if [ -n "$CONDA_PREFIX" ]; then + PYTHON="$CONDA_PREFIX/bin/python" + echo "Using conda Python from: $CONDA_PREFIX" +else + PYTHON="python" + echo "Warning: No conda environment detected, using system Python" +fi + +# Get version from version.py +VERSION=$($PYTHON -c 'import olmocr.version; print(olmocr.version.VERSION)') +echo "OlmOCR version: $VERSION" + +# Get first 10 characters of git hash +GIT_HASH=$(git rev-parse HEAD | cut -c1-10) +echo "Git hash: $GIT_HASH" + +# Get current git branch name +GIT_BRANCH=$(git rev-parse --abbrev-ref HEAD) +echo "Git branch: $GIT_BRANCH" + +# Check if a Beaker image was provided +if [ -n "$BEAKER_IMAGE" ]; then + echo "Using provided Beaker image: $BEAKER_IMAGE" + IMAGE_TAG="$BEAKER_IMAGE" +else + # Create full image tag + IMAGE_TAG="olmocr-benchmark-${VERSION}-${GIT_HASH}" + echo "Building Docker image with tag: $IMAGE_TAG" + + # Build the Docker image + echo "Building Docker image..." + docker build --platform linux/amd64 -f ./Dockerfile -t $IMAGE_TAG . + + # Push image to beaker + echo "Trying to push image to Beaker..." + if ! beaker image create --workspace ai2/oe-data-pdf --name $IMAGE_TAG $IMAGE_TAG 2>/dev/null; then + echo "Warning: Beaker image with tag $IMAGE_TAG already exists. Using existing image." + fi +fi + +# Get Beaker username +BEAKER_USER=$(beaker account whoami --format json | jq -r '.[0].name') +echo "Beaker user: $BEAKER_USER" + +# Create Python script to run beaker experiment +cat << 'EOF' > /tmp/run_qianfan_benchmark_experiment.py +import sys +from textwrap import dedent +from beaker import Beaker, BeakerExperimentSpec, BeakerTaskSpec, BeakerTaskContext, BeakerResultSpec, BeakerTaskResources, BeakerImageSource, BeakerJobPriority, BeakerConstraints, BeakerEnvVar + +# Get image tag, beaker user, git branch, git hash from command line +image_tag = sys.argv[1] +beaker_user = sys.argv[2] +git_branch = sys.argv[3] +git_hash = sys.argv[4] + +# Initialize benchmark dataset parameters +bench_branch = None +bench_repo = "allenai/olmOCR-bench" # Default repository +bench_path = None +cluster = None +noperf = False +model = None + +# Parse additional arguments +arg_idx = 5 +while arg_idx < len(sys.argv): + if sys.argv[arg_idx] == "--benchbranch": + bench_branch = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--benchrepo": + bench_repo = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--benchpath": + bench_path = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--cluster": + cluster = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--noperf": + noperf = True + arg_idx += 1 + elif sys.argv[arg_idx] == "--model": + model = sys.argv[arg_idx + 1] + arg_idx += 2 + else: + print(f"Unknown argument: {sys.argv[arg_idx]}") + arg_idx += 1 + +# Default model for Qianfan OCR +qianfan_model = model if model else "baidu/Qianfan-OCR" + +# Initialize Beaker client +b = Beaker.from_env(default_workspace="ai2/olmocr") + +# Check if AWS credentials secret exists +aws_creds_secret = f"{beaker_user}-AWS_CREDENTIALS_FILE" +try: + b.secret.get(aws_creds_secret, workspace="ai2/olmocr") + has_aws_creds = True + print(f"Found AWS credentials secret: {aws_creds_secret}") +except: + has_aws_creds = False + print(f"AWS credentials secret not found: {aws_creds_secret}") + +# Check if HF_TOKEN secret exists +hf_token_secret = f"{beaker_user}-HF_TOKEN" +try: + b.secret.get(hf_token_secret, workspace="ai2/olmocr") + has_hf_token = True + print(f"Found HuggingFace token secret: {hf_token_secret}") +except: + has_hf_token = False + print(f"HuggingFace token secret not found: {hf_token_secret}") + +# Shell script to run Qianfan OCR conversions for benchmark +# Uses scripts/qianfan_bench_convert.py which is baked into the Docker image +run_qianfan_shell = dedent("""\ +bash -lc 'set -euo pipefail +PDF_ROOT="olmOCR-bench/bench_data/pdfs" +TARGET_ROOT="olmOCR-bench/bench_data/qianfan" +rm -rf "$TARGET_ROOT" +mkdir -p "$TARGET_ROOT" + +# Start vllm server in background +echo "Starting vllm server for Qianfan OCR..." +vllm serve __QIANFAN_MODEL__ --trust-remote-code --served-model-name qianfan-ocr --max-model-len 16384 > /tmp/vllm_server.log 2>&1 & +VLLM_PID=$! + +# Wait for vllm server to be ready +echo "Waiting for vllm server to start..." +for i in {1..600}; do + if curl -s http://localhost:8000/health > /dev/null 2>&1; then + echo "vllm server is ready" + break + fi + if [ $i -eq 600 ]; then + echo "Error: vllm server failed to start after 600 seconds" + cat /tmp/vllm_server.log + exit 1 + fi + sleep 1 +done + +python scripts/qianfan_bench_convert.py "$PDF_ROOT" "$TARGET_ROOT" + +# Kill vllm server +echo "Stopping vllm server..." +kill $VLLM_PID || true +wait $VLLM_PID 2>/dev/null || true +'""").replace("__QIANFAN_MODEL__", qianfan_model) + +# First experiment: Original benchmark job +commands = [] +if has_aws_creds: + commands.extend([ + "mkdir -p ~/.aws", + 'echo "$AWS_CREDENTIALS_FILE" > ~/.aws/credentials' + ]) + +if has_hf_token: + commands.append('export HF_TOKEN="$HF_TOKEN"') + +# Install uv for fast dependency management, then s5cmd (needed for S3 operations) +commands.append("pip install uv") +commands.append("uv pip install s5cmd") + +# Handle benchmark data download based on source type +if bench_path: + if bench_path.startswith("s3://"): + commands.append(f"s5cmd cp {bench_path.rstrip('/')}/* ./olmOCR-bench/") + else: + commands.append(f"cp -r {bench_path} ./olmOCR-bench") +else: + hf_download_cmd = f"hf download --repo-type dataset {bench_repo} --max-workers 2" + if bench_branch: + hf_download_cmd += f" --revision {bench_branch}" + hf_download_cmd += " --local-dir ./olmOCR-bench" + commands.append(hf_download_cmd) + +# Install poppler-utils for pdftoppm (PDF to image conversion), upgrade vllm, then run +commands.extend([ + "apt-get update && apt-get install -y poppler-utils", + "uv pip install --upgrade vllm", + run_qianfan_shell, + "python -m olmocr.bench.benchmark --dir ./olmOCR-bench/bench_data --candidate qianfan" +]) + +# Build task spec with optional env vars +# If image_tag contains '/', it's already a full beaker image reference +if '/' in image_tag: + image_ref = image_tag +else: + image_ref = f"{beaker_user}/{image_tag}" + +task_spec_args = { + "name": "qianfan-benchmark", + "image": BeakerImageSource(beaker=image_ref), + "command": [ + "bash", "-c", + " && ".join(commands) + ], + "context": BeakerTaskContext( + priority=BeakerJobPriority["normal"], + preemptible=True, + ), + "resources": BeakerTaskResources(gpu_count=1), + "constraints": BeakerConstraints(cluster=[cluster] if cluster else ["ai2/ceres-cirrascale", "ai2/jupiter-cirrascale-2"]), + "result": BeakerResultSpec(path="/noop-results"), +} + +# Add env vars if AWS credentials or HF token exist +env_vars = [] +if has_aws_creds: + env_vars.append(BeakerEnvVar(name="AWS_CREDENTIALS_FILE", secret=aws_creds_secret)) +if has_hf_token: + env_vars.append(BeakerEnvVar(name="HF_TOKEN", secret=hf_token_secret)) +if env_vars: + task_spec_args["env_vars"] = env_vars + +# Create first experiment spec +experiment_spec = BeakerExperimentSpec( + description=f"Qianfan OCR Benchmark Run - Branch: {git_branch}, Commit: {git_hash}, Model: {qianfan_model}", + budget="ai2/oe-base", + tasks=[BeakerTaskSpec(**task_spec_args)], +) + +# Create the first experiment +workload = b.experiment.create(spec=experiment_spec, workspace="ai2/olmocr") +print(f"Created benchmark experiment: {workload.experiment.id}") +print(f"View at: https://beaker.org/ex/{workload.experiment.id}") +print("-------") +print("") + +# Second experiment: Performance test job (only if --noperf not specified) +if not noperf: + perf_commands = [] + if has_aws_creds: + perf_commands.extend([ + "mkdir -p ~/.aws", + 'echo "$AWS_CREDENTIALS_FILE" > ~/.aws/credentials' + ]) + + if has_hf_token: + perf_commands.append('export HF_TOKEN="$HF_TOKEN"') + + # Shell script for performance test + perf_shell = dedent("""\ +set -euo pipefail + +# Start vllm server in background +echo "Starting vllm server for Qianfan OCR..." +vllm serve __QIANFAN_MODEL__ --trust-remote-code --served-model-name qianfan-ocr --max-model-len 16384 > /tmp/vllm_server.log 2>&1 & +VLLM_PID=$! + +# Wait for vllm server to be ready +echo "Waiting for vllm server to start..." +for i in {1..600}; do + if curl -s http://localhost:8000/health > /dev/null 2>&1; then + echo "vllm server is ready" + break + fi + if [ $i -eq 600 ]; then + echo "Error: vllm server failed to start after 600 seconds" + cat /tmp/vllm_server.log + exit 1 + fi + sleep 1 +done + +# Run the performance test +time python scripts/qianfan_bench_convert.py /root/olmOCR-mix-0225_benchmark_set/ /root/olmOCR-mix-0225_benchmark_set_qianfan + +# Kill vllm server +echo "Stopping vllm server..." +kill $VLLM_PID || true +wait $VLLM_PID 2>/dev/null || true +""").replace("__QIANFAN_MODEL__", qianfan_model) + + perf_commands.extend([ + "pip install uv", + "apt-get update && apt-get install -y poppler-utils", + "uv pip install --upgrade vllm", + "uv pip install awscli", + "aws s3 cp --recursive s3://ai2-oe-data/jakep/olmocr/olmOCR-mix-0225/benchmark_set/ /root/olmOCR-mix-0225_benchmark_set/", + f"bash -c '{perf_shell}'" + ]) + + # Build performance task spec + perf_task_spec_args = { + "name": "qianfan-performance", + "image": BeakerImageSource(beaker=image_ref), + "command": [ + "bash", "-c", + " && ".join(perf_commands) + ], + "context": BeakerTaskContext( + priority=BeakerJobPriority["normal"], + preemptible=True, + ), + # Need to reserve all 8 gpus for performance spec or else benchmark results can be off (1 for titan-cirrascale) + "resources": BeakerTaskResources(gpu_count=1 if cluster == "ai2/titan-cirrascale" else 8), + "constraints": BeakerConstraints(cluster=[cluster] if cluster else ["ai2/ceres-cirrascale", "ai2/jupiter-cirrascale-2"]), + "result": BeakerResultSpec(path="/noop-results"), + } + + # Add env vars if AWS credentials or HF token exist + env_vars = [] + if has_aws_creds: + env_vars.append(BeakerEnvVar(name="AWS_CREDENTIALS_FILE", secret=aws_creds_secret)) + if has_hf_token: + env_vars.append(BeakerEnvVar(name="HF_TOKEN", secret=hf_token_secret)) + if env_vars: + perf_task_spec_args["env_vars"] = env_vars + + # Create performance experiment spec + perf_experiment_spec = BeakerExperimentSpec( + description=f"Qianfan OCR Performance Test - Branch: {git_branch}, Commit: {git_hash}, Model: {qianfan_model}", + budget="ai2/oe-base", + tasks=[BeakerTaskSpec(**perf_task_spec_args)], + ) + + # Create the performance experiment + perf_workload = b.experiment.create(spec=perf_experiment_spec, workspace="ai2/olmocr") + print(f"Created performance experiment: {perf_workload.experiment.id}") + print(f"View at: https://beaker.org/ex/{perf_workload.experiment.id}") +else: + print("Skipping performance test (--noperf flag specified)") +EOF + +# Run the Python script to create the experiments +echo "Creating Beaker experiments..." + +# Build command with appropriate arguments +CMD="$PYTHON /tmp/run_qianfan_benchmark_experiment.py $IMAGE_TAG $BEAKER_USER $GIT_BRANCH $GIT_HASH" + +if [ -n "$BENCH_BRANCH" ]; then + echo "Using bench branch: $BENCH_BRANCH" + CMD="$CMD --benchbranch \"$BENCH_BRANCH\"" +fi + +if [ -n "$BENCH_REPO" ]; then + echo "Using bench repo: $BENCH_REPO" + CMD="$CMD --benchrepo \"$BENCH_REPO\"" +fi + +if [ -n "$BENCH_PATH" ]; then + echo "Using bench path: $BENCH_PATH" + CMD="$CMD --benchpath \"$BENCH_PATH\"" +fi + +if [ -n "$CLUSTER" ]; then + echo "Using cluster: $CLUSTER" + CMD="$CMD --cluster $CLUSTER" +fi + +if [ -n "$NOPERF" ]; then + echo "Skipping performance tests" + CMD="$CMD --noperf" +fi + +if [ -n "$MODEL" ]; then + echo "Using model: $MODEL" + CMD="$CMD --model $MODEL" +fi + +eval $CMD + +# Clean up temporary file +rm /tmp/run_qianfan_benchmark_experiment.py + +echo "Benchmark experiments submitted successfully!" diff --git a/scripts/run_server_benchmark.sh b/scripts/run_server_benchmark.sh new file mode 100755 index 0000000..3141b22 --- /dev/null +++ b/scripts/run_server_benchmark.sh @@ -0,0 +1,424 @@ +#!/bin/bash + +# Runs an olmocr-bench server benchmark run with vLLM model serving +# Basic usage with default model: +# ./scripts/run_server_benchmark.sh +# With custom vLLM model and served name: +# ./scripts/run_server_benchmark.sh --model facebook/opt-125m --served-model-name opt-125m +# With custom benchmark dataset: +# ./scripts/run_server_benchmark.sh --benchrepo allenai/olmOCR-bench-internal --model gpt2 +# ./scripts/run_server_benchmark.sh --benchbranch olmOCR-bench-1125 --model gpt2 +# ./scripts/run_server_benchmark.sh --benchpath s3://ai2-oe-data/jakep/olmocr/olmOCR-bench-1125/ --model gpt2 +# With beaker secrets for API keys (format: ENV_VAR=secret-name): +# ./scripts/run_server_benchmark.sh --beaker-secret OPENAI_API_KEY=jakep-openai-key --model gpt2 +# With cluster parameter: specify a specific cluster to use +# ./scripts/run_server_benchmark.sh --cluster ai2/titan-cirrascale --model gpt2 +# With beaker image: skip Docker build and use provided Beaker image +# ./scripts/run_server_benchmark.sh --beaker-image jakep/olmocr-benchmark-0.3.3-780bc7d934 --model gpt2 +# With additional server convert arguments: +# ./scripts/run_server_benchmark.sh --model gpt2 server:name=test1 + +set -e + +# Parse command line arguments +CLUSTER="" +BENCH_BRANCH="" +BENCH_REPO="" +BENCH_PATH="" +BEAKER_IMAGE="" +VLLM_MODEL="" +SERVED_MODEL_NAME="" +BEAKER_SECRETS=() +CONVERT_ARGS=() + +# First pass: extract our known arguments +while [[ $# -gt 0 ]]; do + case $1 in + --cluster) + CLUSTER="$2" + shift 2 + ;; + --benchbranch) + BENCH_BRANCH="$2" + shift 2 + ;; + --benchrepo) + BENCH_REPO="$2" + shift 2 + ;; + --benchpath) + BENCH_PATH="$2" + shift 2 + ;; + --beaker-image) + BEAKER_IMAGE="$2" + shift 2 + ;; + --model) + VLLM_MODEL="$2" + shift 2 + ;; + --served-model-name) + SERVED_MODEL_NAME="$2" + shift 2 + ;; + --beaker-secret) + # Format: ENV_VAR=secret-name + BEAKER_SECRETS+=("$2") + shift 2 + ;; + *) + # Store args to forward to convert + CONVERT_ARGS+=("$1") + shift + ;; + esac +done + +# Set default values if not provided +if [ -z "$VLLM_MODEL" ]; then + echo "Error: --model argument is required to specify the vLLM model to serve" + echo "" + echo "Usage examples:" + echo " ./scripts/run_server_benchmark.sh --model facebook/opt-125m" + echo " ./scripts/run_server_benchmark.sh --model gpt2 --served-model-name my-gpt2" + echo " ./scripts/run_server_benchmark.sh --model meta-llama/Llama-2-7b-hf server:name=llama2" + exit 1 +fi + +# If served-model-name not specified, use the model name +if [ -z "$SERVED_MODEL_NAME" ]; then + SERVED_MODEL_NAME="$VLLM_MODEL" + echo "Using served-model-name: $SERVED_MODEL_NAME" +fi + +# Check for mutual exclusivity between benchpath and benchrepo/benchbranch +if [ -n "$BENCH_PATH" ] && ([ -n "$BENCH_REPO" ] || [ -n "$BENCH_BRANCH" ]); then + echo "Error: --benchpath is mutually exclusive with --benchrepo and --benchbranch" + echo "Use either --benchpath OR --benchrepo/--benchbranch, not both." + exit 1 +fi + +# Check for uncommitted changes +if [ -n "$BEAKER_IMAGE" ]; then + echo "Skipping docker build" +else + if ! git diff-index --quiet HEAD --; then + echo "Error: There are uncommitted changes in the repository." + echo "Please commit or stash your changes before running the benchmark." + echo "" + echo "Uncommitted changes:" + git status --short + exit 1 + fi +fi + +# Use conda environment Python if available, otherwise use system Python +if [ -n "$CONDA_PREFIX" ]; then + PYTHON="$CONDA_PREFIX/bin/python" + echo "Using conda Python from: $CONDA_PREFIX" +else + PYTHON="python" + echo "Warning: No conda environment detected, using system Python" +fi + +# Get version from version.py +VERSION=$($PYTHON -c 'import olmocr.version; print(olmocr.version.VERSION)') +echo "OlmOCR version: $VERSION" + +# Get first 10 characters of git hash +GIT_HASH=$(git rev-parse HEAD | cut -c1-10) +echo "Git hash: $GIT_HASH" + +# Get current git branch name +GIT_BRANCH=$(git rev-parse --abbrev-ref HEAD) +echo "Git branch: $GIT_BRANCH" + +# Check if a Beaker image was provided +if [ -n "$BEAKER_IMAGE" ]; then + echo "Using provided Beaker image: $BEAKER_IMAGE" + IMAGE_TAG="$BEAKER_IMAGE" +else + # Create full image tag + IMAGE_TAG="olmocr-server-benchmark-${VERSION}-${GIT_HASH}" + echo "Building Docker image with tag: $IMAGE_TAG" + + # Build the Docker image + echo "Building Docker image..." + docker build --platform linux/amd64 -f ./Dockerfile -t $IMAGE_TAG . + + # Push image to beaker + echo "Trying to push image to Beaker..." + if ! beaker image create --workspace ai2/oe-data-pdf --name $IMAGE_TAG $IMAGE_TAG 2>/dev/null; then + echo "Warning: Beaker image with tag $IMAGE_TAG already exists. Using existing image." + fi +fi + +# Get Beaker username +BEAKER_USER=$(beaker account whoami --format json | jq -r '.[0].name') +echo "Beaker user: $BEAKER_USER" + +# Create Python script to run beaker experiment +cat << 'EOF' > /tmp/run_server_benchmark_experiment.py +import sys +from textwrap import dedent +from beaker import Beaker, ExperimentSpec, TaskSpec, TaskContext, ResultSpec, TaskResources, ImageSource, Priority, Constraints, EnvVar + +# Get image tag, beaker user, git branch, git hash from command line +image_tag = sys.argv[1] +beaker_user = sys.argv[2] +git_branch = sys.argv[3] +git_hash = sys.argv[4] +vllm_model = sys.argv[5] +served_model_name = sys.argv[6] +cluster = None +bench_branch = None +bench_repo = "allenai/olmOCR-bench" # Default repository +bench_path = None +convert_args = [] +beaker_secrets = {} # Dict of ENV_VAR: secret_name + +# Parse remaining arguments +arg_idx = 7 +while arg_idx < len(sys.argv): + if sys.argv[arg_idx] == "--cluster": + cluster = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--benchbranch": + bench_branch = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--benchrepo": + bench_repo = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--benchpath": + bench_path = sys.argv[arg_idx + 1] + arg_idx += 2 + elif sys.argv[arg_idx] == "--beaker-secret": + # Parse ENV_VAR=secret-name format + secret_spec = sys.argv[arg_idx + 1] + if "=" in secret_spec: + env_var, secret_name = secret_spec.split("=", 1) + beaker_secrets[env_var] = secret_name + arg_idx += 2 + else: + # Everything else is a convert arg + convert_args.append(sys.argv[arg_idx]) + arg_idx += 1 + +# Initialize Beaker client +b = Beaker.from_env(default_workspace="ai2/olmocr") + +# Check if AWS credentials secret exists +aws_creds_secret = f"{beaker_user}-AWS_CREDENTIALS_FILE" +try: + # Try to get the secret to see if it exists + b.secret.get(aws_creds_secret, workspace="ai2/olmocr") + has_aws_creds = True + print(f"Found AWS credentials secret: {aws_creds_secret}") +except: + has_aws_creds = False + print(f"AWS credentials secret not found: {aws_creds_secret}") + +# Check if HF_TOKEN secret exists +hf_token_secret = f"{beaker_user}-HF_TOKEN" +try: + # Try to get the secret to see if it exists + b.secret.get(hf_token_secret, workspace="ai2/olmocr") + has_hf_token = True + print(f"Found HuggingFace token secret: {hf_token_secret}") +except: + has_hf_token = False + print(f"HuggingFace token secret not found: {hf_token_secret}") + +# Shell script to run server benchmark with vLLM +run_server_shell = dedent(f"""\ +bash -lc 'set -euo pipefail + +# Start vllm server in background +echo "Starting vllm server for model: {vllm_model}..." +vllm serve {vllm_model} --served-model-name {served_model_name} > /tmp/vllm_server.log 2>&1 & +VLLM_PID=$! + +# Wait for vllm server to be ready +echo "Waiting for vllm server to start..." +for i in {{1..600}}; do + if curl -s http://localhost:8000/health > /dev/null 2>&1; then + echo "vllm server is ready" + break + fi + if [ $i -eq 600 ]; then + echo "Error: vllm server failed to start after 600 seconds" + echo "Last 100 lines of server log:" + tail -100 /tmp/vllm_server.log + exit 1 + fi + sleep 1 +done + +# Show server info +echo "vLLM server started successfully" +curl -s http://localhost:8000/v1/models | python -m json.tool || true + +# Run the convert command +echo "Running convert with server endpoint..." +python -m olmocr.bench.convert server:model={served_model_name} {"" + " ".join(convert_args) if convert_args else ""} --dir ./olmOCR-bench/bench_data + +# Kill vllm server +echo "Stopping vllm server..." +kill $VLLM_PID || true +wait $VLLM_PID 2>/dev/null || true +'""") + +# Build commands +commands = [] +if has_aws_creds: + commands.extend([ + "mkdir -p ~/.aws", + 'echo "$AWS_CREDENTIALS_FILE" > ~/.aws/credentials' + ]) + +if has_hf_token: + commands.append('export HF_TOKEN="$HF_TOKEN"') + +# Export any beaker secrets as environment variables +for env_var in beaker_secrets: + commands.append(f'export {env_var}="${env_var}"') + +# Install dependencies +commands.extend([ + "pip install s5cmd", + "pip install --upgrade vllm" # Ensure vllm is installed +]) + +# Handle benchmark data download based on source type +if bench_path: + # If bench_path is provided, use it (can be S3 or local path) + if bench_path.startswith("s3://"): + # S3 path - use s5cmd to download + commands.append(f"s5cmd cp {bench_path.rstrip('/')}/* ./olmOCR-bench/") + else: + # Local path - copy directly + commands.append(f"cp -r {bench_path} ./olmOCR-bench") +else: + # Use HuggingFace download (default behavior) + hf_download_cmd = f"hf download --repo-type dataset {bench_repo} --max-workers 2" + if bench_branch: + hf_download_cmd += f" --revision {bench_branch}" + hf_download_cmd += " --local-dir ./olmOCR-bench" + commands.append(hf_download_cmd) + +# Run the server and convert +commands.append(run_server_shell) + +# Copy workspace to S3 for archival (using BEAKER_WORKLOAD_ID for unique path) +commands.append("s5cmd cp ./olmOCR-bench/ s3://ai2-oe-data/jakep/olmocr-bench-runs/$BEAKER_WORKLOAD_ID/olmOCR-bench/") + +# Run benchmark +commands.append("python -m olmocr.bench.benchmark --dir ./olmOCR-bench/bench_data") + +# Build task spec with optional env vars +# If image_tag contains '/', it's already a full beaker image reference +if '/' in image_tag: + image_ref = image_tag +else: + image_ref = f"{beaker_user}/{image_tag}" + +task_spec_args = { + "name": "olmocr-server-benchmark", + "image": ImageSource(beaker=image_ref), + "command": [ + "bash", "-c", + " && ".join(commands) + ], + "context": TaskContext( + priority=Priority.normal, + preemptible=True, + ), + "resources": TaskResources(gpu_count=1), # Need GPU for vLLM + "constraints": Constraints(cluster=[cluster] if cluster else ["ai2/ceres-cirrascale", "ai2/jupiter-cirrascale-2"]), + "result": ResultSpec(path="/noop-results"), +} + +# Add env vars if AWS credentials or HF token exist +env_vars = [] +if has_aws_creds: + env_vars.append(EnvVar(name="AWS_CREDENTIALS_FILE", secret=aws_creds_secret)) +if has_hf_token: + env_vars.append(EnvVar(name="HF_TOKEN", secret=hf_token_secret)) +# Add any additional beaker secrets +for env_var, secret_name in beaker_secrets.items(): + env_vars.append(EnvVar(name=env_var, secret=secret_name)) +if env_vars: + task_spec_args["env_vars"] = env_vars + +# Create a readable experiment name +experiment_name = f"server-bench-{vllm_model.replace('/', '-')}" +if len(experiment_name) > 50: + # Truncate long model names + experiment_name = f"server-bench-{vllm_model.split('/')[-1]}" + +print(f"Experiment name: {experiment_name}") + +# Create experiment spec +experiment_spec = ExperimentSpec( + description=f"OlmOCR Server Benchmark - Model: {vllm_model}, Branch: {git_branch}, Commit: {git_hash}", + budget="ai2/oe-base", + tasks=[TaskSpec(**task_spec_args)], + name=experiment_name, +) + +# Create the experiment +experiment = b.experiment.create(spec=experiment_spec, workspace="ai2/olmocr") +print(f"Created server benchmark experiment: {experiment_name} ({experiment.id})") +print(f"View at: https://beaker.org/ex/{experiment.id}") +EOF + +# Run the Python script to create the experiment +echo "Creating Beaker experiment..." + +# Build command with appropriate arguments +CMD="$PYTHON /tmp/run_server_benchmark_experiment.py '$IMAGE_TAG' '$BEAKER_USER' '$GIT_BRANCH' '$GIT_HASH' '$VLLM_MODEL' '$SERVED_MODEL_NAME'" + +if [ -n "$CLUSTER" ]; then + echo "Using cluster: $CLUSTER" + CMD="$CMD --cluster '$CLUSTER'" +fi + +if [ -n "$BENCH_BRANCH" ]; then + echo "Using bench branch: $BENCH_BRANCH" + CMD="$CMD --benchbranch '$BENCH_BRANCH'" +fi + +if [ -n "$BENCH_REPO" ]; then + echo "Using bench repo: $BENCH_REPO" + CMD="$CMD --benchrepo '$BENCH_REPO'" +fi + +if [ -n "$BENCH_PATH" ]; then + echo "Using bench path: $BENCH_PATH" + CMD="$CMD --benchpath '$BENCH_PATH'" +fi + +# Add beaker secrets if any +if [ ${#BEAKER_SECRETS[@]} -gt 0 ]; then + echo "Using beaker secrets:" + for secret in "${BEAKER_SECRETS[@]}"; do + echo " $secret" + CMD="$CMD --beaker-secret '$secret'" + done +fi + +# Add convert args if any +if [ ${#CONVERT_ARGS[@]} -gt 0 ]; then + echo "Forwarding to convert: ${CONVERT_ARGS[*]}" + for arg in "${CONVERT_ARGS[@]}"; do + CMD="$CMD '$arg'" + done +fi + +eval $CMD + +# Clean up temporary file +rm /tmp/run_server_benchmark_experiment.py + +echo "Server Benchmark experiment submitted successfully!" \ No newline at end of file diff --git a/scripts/run_transformers_benchmark.sh b/scripts/run_transformers_benchmark.sh new file mode 100755 index 0000000..ce2260c --- /dev/null +++ b/scripts/run_transformers_benchmark.sh @@ -0,0 +1,146 @@ +#!/bin/bash + +# Runs transformers benchmark in olmocr-bench +set -e + + +# Check for uncommitted changes +# if ! git diff-index --quiet HEAD --; then +# echo "Error: There are uncommitted changes in the repository." +# echo "Please commit or stash your changes before running the benchmark." +# echo "" +# echo "Uncommitted changes:" +# git status --short +# exit 1 +# fi + +# Use conda environment Python if available, otherwise use system Python +if [ -n "$CONDA_PREFIX" ]; then + PYTHON="$CONDA_PREFIX/bin/python" + echo "Using conda Python from: $CONDA_PREFIX" +else + PYTHON="python" + echo "Warning: No conda environment detected, using system Python" +fi + +# Get version from version.py +VERSION=$($PYTHON -c 'import olmocr.version; print(olmocr.version.VERSION)') +echo "OlmOCR version: $VERSION" + +# Get first 10 characters of git hash +GIT_HASH=$(git rev-parse HEAD | cut -c1-10) +echo "Git hash: $GIT_HASH" + +# Get current git branch name +GIT_BRANCH=$(git rev-parse --abbrev-ref HEAD) +echo "Git branch: $GIT_BRANCH" + +# Create full image tag +IMAGE_TAG="olmocr-benchmark-${VERSION}-${GIT_HASH}" +echo "Building Docker image with tag: $IMAGE_TAG" + +# Build the Docker image +echo "Building Docker image..." +docker build --platform linux/amd64 -f ./Dockerfile -t $IMAGE_TAG . + +# Get Beaker username +BEAKER_USER=$(beaker account whoami --format json | jq -r '.[0].name') +echo "Beaker user: $BEAKER_USER" + +# Push image to beaker +echo "Trying to push image to Beaker..." +if ! beaker image create --workspace ai2/oe-data-pdf --name $IMAGE_TAG $IMAGE_TAG 2>/dev/null; then + echo "Warning: Beaker image with tag $IMAGE_TAG already exists. Using existing image." +fi + +# Create Python script to run beaker experiment +cat << 'EOF' > /tmp/run_benchmark_experiment.py +import sys +from beaker import Beaker, ExperimentSpec, TaskSpec, TaskContext, ResultSpec, TaskResources, ImageSource, Priority, Constraints, EnvVar + +# Get image tag, beaker user, git branch, git hash, version from command line +image_tag = sys.argv[1] +beaker_user = sys.argv[2] +git_branch = sys.argv[3] +git_hash = sys.argv[4] + + +# Initialize Beaker client +b = Beaker.from_env(default_workspace="ai2/olmocr") + + +# Check if AWS credentials secret exists +aws_creds_secret = f"{beaker_user}-AWS_CREDENTIALS_FILE" +try: + # Try to get the secret to see if it exists + b.secret.get(aws_creds_secret, workspace="ai2/olmocr") + has_aws_creds = True + print(f"Found AWS credentials secret: {aws_creds_secret}") +except: + has_aws_creds = False + print(f"AWS credentials secret not found: {aws_creds_secret}") + +# First experiment: Original benchmark job +commands = [] +if has_aws_creds: + commands.extend([ + "mkdir -p ~/.aws", + 'echo "$AWS_CREDENTIALS_FILE" > ~/.aws/credentials' + ]) +commands.extend([ + "git clone https://huggingface.co/datasets/allenai/olmOCR-bench", + "cd olmOCR-bench && git lfs pull && cd ..", + "pip install accelerate", + "pip install flash-attn==2.8.0.post2 --no-build-isolation", + "PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True python -m olmocr.bench.convert transformers:target_longest_image_dim=1288:prompt_template=yaml:response_template=yaml --dir ./olmOCR-bench/bench_data", + "python -m olmocr.bench.benchmark --dir ./olmOCR-bench/bench_data" +]) + +# Build task spec with optional env vars +task_spec_args = { + "name": "transformers-benchmark", + "image": ImageSource(beaker=f"{beaker_user}/{image_tag}"), + "command": [ + "bash", "-c", + " && ".join(commands) + ], + "context": TaskContext( + priority=Priority.normal, + preemptible=True, + ), + "resources": TaskResources(gpu_count=1), + "constraints": Constraints(cluster=["ai2/titan-cirrascale"]), + "result": ResultSpec(path="/noop-results"), +} + +# Add env vars if AWS credentials exist +if has_aws_creds: + task_spec_args["env_vars"] = [ + EnvVar(name="AWS_CREDENTIALS_FILE", secret=aws_creds_secret) + ] + +# Create first experiment spec +experiment_spec = ExperimentSpec( + description=f"Transformers Benchmark Run - Branch: {git_branch}, Commit: {git_hash}", + budget="ai2/oe-data", + tasks=[TaskSpec(**task_spec_args)], +) + +# Create the first experiment +experiment = b.experiment.create(spec=experiment_spec, workspace="ai2/olmocr") +print(f"Created benchmark experiment: {experiment.id}") +print(f"View at: https://beaker.org/ex/{experiment.id}") +print("-------") +print("") + + +EOF + +# Run the Python script to create the experiments +echo "Creating Beaker experiments..." +$PYTHON /tmp/run_benchmark_experiment.py $IMAGE_TAG $BEAKER_USER $GIT_BRANCH $GIT_HASH + +# Clean up temporary file +rm /tmp/run_benchmark_experiment.py + +echo "Benchmark experiments submitted successfully!" \ No newline at end of file diff --git a/scripts/s2orc_extractor.sh b/scripts/s2orc_extractor.sh new file mode 100644 index 0000000..9c1e82b --- /dev/null +++ b/scripts/s2orc_extractor.sh @@ -0,0 +1,36 @@ +#!/bin/bash + +# Define the output file for the metadata.sha1 fields +OUTPUT_FILE="s2orc_pdfs_v2.txt" + +# Clear the output file if it already exists +> "$OUTPUT_FILE" + +# Create a temporary directory for partial outputs +temp_output_dir=$(mktemp -d) + +# Ensure the temporary directory is cleaned up on exit or error +trap 'rm -rf "$temp_output_dir"' EXIT + +# Export the temporary output directory variable for use in xargs +export temp_output_dir + +echo "temp dir $temp_output_dir" + +# Find all .gz files recursively from the current directory +find 'split=train' -type f -name "*.gz" | \ + xargs -P 30 -I{} bash -c ' + gz_file="$1" + partial_output="$temp_output_dir/$(basename "$gz_file").txt" + + # Stream uncompressed data directly into jq and format the output + gunzip -c "$gz_file" | jq -r '"'"' + select(.metadata.sha1 != null) | + "s3://ai2-s2-pdfs/" + (.metadata.sha1[:4]) + "/" + (.metadata.sha1[4:]) + ".pdf" + '"'"' >> "$partial_output" + ' _ {} + +# Concatenate all partial outputs into the final output file +cat "$temp_output_dir"/*.txt >> "$OUTPUT_FILE" + +echo "All metadata.sha1 fields have been extracted to $OUTPUT_FILE." diff --git a/scripts/sample_pdf_pages.py b/scripts/sample_pdf_pages.py new file mode 100644 index 0000000..6f09f65 --- /dev/null +++ b/scripts/sample_pdf_pages.py @@ -0,0 +1,110 @@ +#!/usr/bin/env python3 +""" +Sample random pages from PDFs in an input directory and save them as individual PDFs. + +This script takes PDFs from an input directory, randomly selects N pages from each, +and writes each selected page as a separate PDF file in the output directory. +Uses a fixed random seed for reproducibility. +""" + +import argparse +import random +from pathlib import Path + +from pypdf import PdfReader, PdfWriter + + +def sample_pdf_pages(input_dir: Path, output_dir: Path, num_pages: int, seed: int = 42): + """ + Sample random pages from PDFs and save them as individual files. + + Args: + input_dir: Directory containing input PDF files + output_dir: Directory to save sampled pages + num_pages: Number of pages to sample from each PDF + seed: Random seed for reproducibility + """ + # Set random seed for consistency + random.seed(seed) + + # Create output directory if it doesn't exist + output_dir.mkdir(parents=True, exist_ok=True) + + # Get all PDF files from input directory + pdf_files = sorted(input_dir.glob("*.pdf")) + + if not pdf_files: + print(f"No PDF files found in {input_dir}") + return + + print(f"Found {len(pdf_files)} PDF file(s)") + + for pdf_path in pdf_files: + print(f"\nProcessing: {pdf_path.name}") + + try: + reader = PdfReader(pdf_path) + total_pages = len(reader.pages) + + print(f" Total pages: {total_pages}") + + # Determine how many pages to sample (min of requested and available) + pages_to_sample = min(num_pages, total_pages) + + if pages_to_sample < num_pages: + print(f" Warning: Only {total_pages} pages available, sampling all") + + # Randomly select page indices + selected_indices = sorted(random.sample(range(total_pages), pages_to_sample)) + print(f" Selected page indices: {selected_indices}") + + # Extract and save each selected page + for page_idx in selected_indices: + # Create writer for single page + writer = PdfWriter() + writer.add_page(reader.pages[page_idx]) + + # Create output filename: original_name_page_N.pdf + output_filename = f"{pdf_path.stem}_page_{page_idx}.pdf" + output_path = output_dir / output_filename + + # Write the single page PDF + with open(output_path, "wb") as output_file: + writer.write(output_file) + + print(f" Saved: {output_filename}") + + except Exception as e: + print(f" Error processing {pdf_path.name}: {e}") + continue + + print(f"\nDone! Output saved to {output_dir}") + + +def main(): + parser = argparse.ArgumentParser(description="Sample random pages from PDFs and save as individual files") + parser.add_argument("--input-dir", type=Path, required=True, help="Directory containing input PDF files") + parser.add_argument("--output-dir", type=Path, required=True, help="Directory to save sampled pages") + parser.add_argument("--pages", type=int, required=True, help="Number of pages to sample from each PDF") + parser.add_argument("--seed", type=int, default=42, help="Random seed for reproducibility (default: 42)") + + args = parser.parse_args() + + # Validate arguments + if not args.input_dir.exists(): + print(f"Error: Input directory does not exist: {args.input_dir}") + return + + if not args.input_dir.is_dir(): + print(f"Error: Input path is not a directory: {args.input_dir}") + return + + if args.pages < 1: + print(f"Error: Number of pages must be at least 1") + return + + sample_pdf_pages(args.input_dir, args.output_dir, args.pages, args.seed) + + +if __name__ == "__main__": + main() diff --git a/scripts/scan_dolmadocs.py b/scripts/scan_dolmadocs.py new file mode 100644 index 0000000..bd9abb2 --- /dev/null +++ b/scripts/scan_dolmadocs.py @@ -0,0 +1,2030 @@ +import argparse +import base64 +import csv +import datetime +import json +import os +import random +import re +import sqlite3 +import tempfile +from concurrent.futures import ThreadPoolExecutor +from pathlib import Path +from typing import Any, Dict, List, Optional, Tuple + +import boto3 +import requests +import tinyhost +from tqdm import tqdm + +from olmocr.data.renderpdf import render_pdf_to_base64webp +from olmocr.s3_utils import get_s3_bytes, parse_s3_path + + +def parse_args(): + parser = argparse.ArgumentParser(description="Scan OLMO OCR workspace results and create visual samples") + parser.add_argument("workspace", help="OLMO OCR workspace path (s3://bucket/workspace)") + parser.add_argument("--pages_per_output", type=int, default=30, help="Number of pages per output file") + parser.add_argument("--repeats", type=int, default=1, help="Number of output files to generate") + parser.add_argument("--pdf_profile", help="AWS profile for accessing PDFs") + parser.add_argument("--output_dir", default="dolma_samples", help="Directory to save output HTML files") + parser.add_argument("--max_workers", type=int, default=4, help="Maximum number of worker threads") + parser.add_argument( + "--db_path", + default="~/s2pdf_url_data/d65142df-6588-4b68-a12c-d468b3761189.csv.db", + help="Path to the SQLite database containing PDF hash to URL mapping", + ) + parser.add_argument( + "--prolific_code", + required=True, + help="Fixed completion code to use for all outputs", + ) + parser.add_argument( + "--prolific_csv", + default="prolific_codes.csv", + help="Path to save the file with tinyhost links (one URL per line)", + ) + parser.add_argument( + "--read_results", + help="Path to a CSV file containing previously generated tinyhost links to extract annotations", + ) + return parser.parse_args() + + +# Fixed prolific code is now passed in as a command line argument + + +def obfuscate_code(code): + """Gently obfuscate the Prolific code so it's not immediately visible in source.""" + # Convert to base64 and reverse + encoded = base64.b64encode(code.encode()).decode() + return encoded[::-1] + + +def deobfuscate_code(obfuscated_code): + """Deobfuscate the code - this will be done in JavaScript.""" + # Reverse and decode from base64 + reversed_encoded = obfuscated_code[::-1] + try: + return base64.b64decode(reversed_encoded).decode() + except: + return "ERROR_DECODING" + + +def parse_pdf_hash(pretty_pdf_path: str) -> Optional[str]: + pattern = r"s3://ai2-s2-pdfs/([a-f0-9]{4})/([a-f0-9]+)\.pdf" + match = re.match(pattern, pretty_pdf_path) + if match: + return match.group(1) + match.group(2) + return None + + +def get_original_url(pdf_hash: str, db_path: str) -> Optional[str]: + """Look up the original URL for a PDF hash in the SQLite database.""" + if not pdf_hash: + return None + + try: + sqlite_db_path = os.path.expanduser(db_path) + if not os.path.exists(sqlite_db_path): + print(f"SQLite database not found at {sqlite_db_path}") + return None + + conn = sqlite3.connect(sqlite_db_path) + cursor = conn.cursor() + + cursor.execute("SELECT uri FROM pdf_mapping WHERE pdf_hash = ?", (pdf_hash,)) + result = cursor.fetchone() + + conn.close() + + if result: + return result[0] + return None + except Exception as e: + print(f"Error looking up URL for PDF hash {pdf_hash}: {e}") + return None + + +def list_result_files(s3_client, workspace_path): + """List all JSON result files in the workspace results directory.""" + bucket, prefix = parse_s3_path(workspace_path) + results_prefix = os.path.join(prefix, "results").rstrip("/") + "/" + + all_files = [] + paginator = s3_client.get_paginator("list_objects_v2") + for page in paginator.paginate(Bucket=bucket, Prefix=results_prefix): + if "Contents" in page: + all_files.extend([f"s3://{bucket}/{obj['Key']}" for obj in page["Contents"] if obj["Key"].endswith(".jsonl") or obj["Key"].endswith(".json")]) + + # if len(all_files) > 1000: + # break + + return all_files + + +def get_random_pages(s3_client, result_files, count=30): + """Get random pages from the result files.""" + random_pages = [] + + # Try to collect the requested number of pages + attempts = 0 + max_attempts = count * 3 # Allow extra attempts to handle potential failures + + while len(random_pages) < count and attempts < max_attempts: + attempts += 1 + + # Pick a random result file + if not result_files: + print("No result files found!") + break + + result_file = random.choice(result_files) + + try: + # Get the content of the file + content = get_s3_bytes(s3_client, result_file) + lines = content.decode("utf-8").strip().split("\n") + + if not lines: + continue + + # Pick a random line (which contains a complete document) + line = random.choice(lines) + doc = json.loads(line) + + # A Dolma document has "text", "metadata", and "attributes" fields + if "text" not in doc or "metadata" not in doc or "attributes" not in doc: + print(f"Document in {result_file} is not a valid Dolma document") + continue + + # Get the original PDF path from metadata + pdf_path = doc["metadata"].get("Source-File") + if not pdf_path: + continue + + # Get page spans from attributes + page_spans = doc["attributes"].get("pdf_page_numbers", []) + if not page_spans: + continue + + # Pick a random page span + page_span = random.choice(page_spans) + if len(page_span) >= 3: + # Page spans are [start_pos, end_pos, page_num] + page_num = page_span[2] + + # Extract text for this page + start_pos, end_pos = page_span[0], page_span[1] + page_text = doc["text"][start_pos:end_pos].strip() + + # Include the text snippet with the page info + random_pages.append((pdf_path, page_num, page_text, result_file)) + + if len(random_pages) >= count: + break + + except Exception as e: + print(f"Error processing {result_file}: {e}") + continue + + print(f"Found {len(random_pages)} random pages from Dolma documents") + return random_pages + + +def create_presigned_url(s3_client, pdf_path, expiration=3600 * 24 * 7): + """Create a presigned URL for the given S3 path.""" + try: + bucket, key = parse_s3_path(pdf_path) + url = s3_client.generate_presigned_url("get_object", Params={"Bucket": bucket, "Key": key}, ExpiresIn=expiration) + return url + except Exception as e: + print(f"Error creating presigned URL for {pdf_path}: {e}") + return None + + +def create_html_output(random_pages, pdf_s3_client, output_path, workspace_path, db_path, prolific_code, resolution=2048): + """Create an HTML file with rendered PDF pages.""" + # Obfuscate the provided Prolific code + obfuscated_code = obfuscate_code(prolific_code) + + # Get current date and time for the report + current_time = datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S") + + html_content = f""" + + + + + + OLMO OCR Samples + + + + +

Task Overview

In this task, you will review {len(random_pages)} document pages and determine whether they contain any Personally Identifiable Information (PII). For each page, please follow the decision flow outlined in the "How to Annotate" section below.

Carefully but efficiently inspect each page and select the appropriate response. You do not need to read every word. Instead, focus on ascertaining the document's intended use and spotting information that would qualify as PII.

The entire task should take about 20-25 minutes.

+ + + +

How to Annotate

The current annotation will be highlighted with a blue outline and a set of response buttons will be displayed directly below the page preview. If you are having trouble viewing the displayed page, click the “View Cached PDF” link for a better look. However, DO NOT examine the entire document; ONLY review the single page being previewed (also indicated in the parentheses after “Viewed Cached PDF”).

For each page, complete the following steps:

+ +

+
Determine if the document is intended for public release.
+
Inspect the page and answer: "Is this document intended for public release or dissemination?"
+
- Yes - If the document appears to be a publication, research paper, public information, etc.
- No - If the document appears to be private, personal, or not intended for public release
- Cannot Read - If you are unable to read the page (e.g., foreign language, no text, etc.)
- Report Content - If the content is inappropriate or disturbing
+
If you selected "Yes," "Cannot Read," or "Report Content," you will automatically move to the next document. If you selected "No," proceed to Step 2.
+
+
Identify the kind of PII found in the private document (if any).
+
You will be shown a checklist with a set of PII options.
+
- Refer to the "How to Identify PII" section below and mark all options that apply.
- If you select "Other," describe the kind of other PII in the expanded text box.
+
+
Press the blue Continue button to complete your annotation.
+
You will automatically be moved to the next annotation.
+

+ +

Note: If you cannot confidently tell that a page is private, treat it as public and do not mark any PII you are unsure about. We anticipate very few private pages or instances of PII in these documents, so erring towards public and no PII minimizes false positives and keeps the review process consistent.

+ +

You may review and edit your previous annotations at any time. To do so, press the green Edit button directly above the page preview for the annotation you want to edit.

After completing all {len(random_pages)} document pages, you will receive a Prolific completion code.

+ +

How to Identify PII

+ +

Identifiers for PII

Some personal information needs to be accompanied by an identifier to be considered PII. Identifiers that trigger PII include:

Names (full names, first/last names, maiden names, nicknames, aliases)
Email Addresses
Phone Numbers

Note that the reverse is also true - an identifier must be accompanied by additional personal information or another identifier (e.g., name + email address) to be considered PII.

+
+ +

PII that must co-occur with an Identifier

The following types of information should only be marked as PII if they occur alongside an identifier (commonly, a person's name):

Addresses (street address, postal code, etc.)
Biographical Information (date of birth, place of birth, gender, sexual orientation, race, ethnicity, citizenship/immigration status, religion)
Location Information (geolocations, specific coordinates)
Employment Information (job titles, workplace names, employment history)
Education Information (school names, degrees, transcripts)
Medical Information (health records, diagnoses, genetic or neural data)

For example, a street address might be personal information, but is not PII by itself; however, a street address associated with a name is regulated PII.

+ +
+

PII that occurs even without an Identifier

Certain types of sensitive information should always be classified as PII because the information is inherently self-identifying. The following should always be marked as PII even if they do not occur alongside an identifier:

Government IDs (SSNs, passport numbers, driver's license numbers, tax IDs)
Financial Information (credit card numbers, bank account/routing numbers)
Biometric Data (fingerprints, retina scans, facial recognition data, voice signatures)
Login information (only mark as PII when a username, password, and login location are present together)

+ + + + + """ + ) + + with open(output_path, "w") as f: + f.write(html_content) + + print(f"Created HTML output at {output_path}") + + +def generate_sample_set(args, i, s3_client, pdf_s3_client, result_files): + """Generate a single sample set.""" + output_filename = Path(args.output_dir) / f"dolma_samples_{i+1}.html" + + print(f"\nGenerating sample set {i+1} of {args.repeats}") + + # Get random pages + random_pages = get_random_pages(s3_client, result_files, args.pages_per_output) + + # Use the fixed prolific code from command line arguments + prolific_code = args.prolific_code + + # Create HTML output with the Prolific code + create_html_output(random_pages, pdf_s3_client, output_filename, args.workspace, args.db_path, prolific_code) + + return output_filename + + +def extract_datastore_url(html_content: str) -> Optional[str]: + """Extract the presigned datastore URL from HTML content.""" + match = re.search(r'const\s+presignedGetUrl\s*=\s*"([^"]+)"', html_content) + if match: + return match.group(1) + return None + + +def extract_page_number_from_html(html_content: str, page_id: str) -> Optional[int]: + """Extract PDF page number from HTML content for a specific page_id. + + This is a fallback mechanism for older versions of the annotation page + that didn't store the page number in a data attribute. + """ + # Try to find the page number in the "View Cached PDF (page X)" text + # Look for section with this page_id + page_section_pattern = '

]*data-index="([^"]*)"[^>]*>.*?

.*?]*>View Cached PDF \$page ([0-9]+)\$' + matches = re.finditer(page_section_pattern, html_content, re.DOTALL) + + for match in matches: + container_index = match.group(1) + pdf_page_from_url = match.group(2) + pdf_page_from_text = match.group(3) + + # Check if this container index matches our page_id (page-X) + if f"page-{container_index}" == page_id: + # Both numbers should be the same, but prefer the one from the URL fragment + try: + return int(pdf_page_from_url) + except (ValueError, TypeError): + try: + return int(pdf_page_from_text) + except (ValueError, TypeError): + pass + + return None + + +def fetch_annotations(tinyhost_link: str) -> Tuple[Dict[str, Any], str, str]: + """Fetch and parse annotations from a tinyhost link.""" + # Request the HTML content + print(f"Fetching annotations from {tinyhost_link}") + response = requests.get(tinyhost_link) + response.raise_for_status() + html_content = response.text + + # Extract the datastore URL + datastore_url = extract_datastore_url(html_content) + if not datastore_url: + print(f"Could not find datastore URL in {tinyhost_link}") + return {}, tinyhost_link, html_content + + # Fetch the datastore content + print(f"Found datastore URL: {datastore_url}") + try: + datastore_response = requests.get(datastore_url) + datastore_response.raise_for_status() + annotations = datastore_response.json() + return annotations, tinyhost_link, html_content + except Exception as e: + print(f"Error fetching datastore from {datastore_url}: {e}") + return {}, tinyhost_link, html_content + + +def process_annotations(annotations_by_link: List[Tuple[Dict[str, Any], str, str]]) -> Dict[str, List[Dict[str, Any]]]: + """Process and categorize annotations by feedback type.""" + results = { + "public_document": [], + "private_document": [], + "cannot_read": [], + "report_content": [], + "no_annotation": [], + } + + # Process each annotation + for annotations, link, html_content in annotations_by_link: + # Extract Prolific PID from datastore if available + prolific_pid = annotations.get("prolific_pid", None) + + for page_id, annotation in annotations.items(): + # Skip non-page entries like prolific_pid + if page_id == "prolific_pid": + continue + + # Handle case where annotation might be a boolean or non-dict value + if not isinstance(annotation, dict) or "primaryOption" not in annotation: + continue + + primary_option = annotation["primaryOption"] + pdf_path = annotation.get("pdfPath", "Unknown") + + # Get PDF page number from annotation data + # This is the actual page number in the PDF that was annotated + pdf_page = None + + # First try to get it from the annotation data (for new format) + if annotation.get("pdfPage"): + try: + pdf_page = int(annotation.get("pdfPage")) + except (ValueError, TypeError): + pass + + # Fallback: try to extract page number from HTML content (for older format) + if pdf_page is None: + pdf_page = extract_page_number_from_html(html_content, page_id) + + # Build a result item based on the new annotation structure + if primary_option == "yes-public": + # Public document - no PII info collected with new flow + results["public_document"].append( + { + "page_id": page_id, + "link": link, + "pdf_path": pdf_path, + "pdf_page": pdf_page, + "pii_types": [], + "has_pii": False, + "description": "", + "prolific_pid": prolific_pid, + } + ) + + elif primary_option == "no-public": + # Private document with potential PII + private_pii_options = annotation.get("privatePiiOptions", []) + other_desc = annotation.get("otherPrivateDesc", "") + + if not private_pii_options: + # No PII selected in a private document + results["private_document"].append( + { + "page_id": page_id, + "link": link, + "pdf_path": pdf_path, + "pdf_page": pdf_page, + "pii_types": [], + "has_pii": False, + "description": "", + "prolific_pid": prolific_pid, + } + ) + else: + # PII found in a private document + results["private_document"].append( + { + "page_id": page_id, + "link": link, + "pdf_path": pdf_path, + "pdf_page": pdf_page, + "pii_types": private_pii_options, + "has_pii": True, + "description": other_desc if "other" in private_pii_options else "", + "prolific_pid": prolific_pid, + } + ) + + elif primary_option == "cannot-read": + results["cannot_read"].append({"page_id": page_id, "link": link, "pdf_path": pdf_path, "pdf_page": pdf_page, "prolific_pid": prolific_pid}) + + elif primary_option == "report-content": + results["report_content"].append({"page_id": page_id, "link": link, "pdf_path": pdf_path, "pdf_page": pdf_page, "prolific_pid": prolific_pid}) + + else: + results["no_annotation"].append({"page_id": page_id, "link": link, "pdf_path": pdf_path, "pdf_page": pdf_page, "prolific_pid": prolific_pid}) + + return results + + +def print_annotation_report(annotation_results: Dict[str, List[Dict[str, Any]]], pdf_s3_client=None): + """Print a summary report of annotations.""" + total_pages = sum(len(items) for items in annotation_results.values()) + + print("\n" + "=" * 80) + print(f"ANNOTATION REPORT - Total Pages: {total_pages}") + print("=" * 80) + + # Count pages with PII in public documents + public_with_pii = [page for page in annotation_results["public_document"] if page.get("has_pii", False)] + public_without_pii = [page for page in annotation_results["public_document"] if not page.get("has_pii", False)] + + # Count pages with PII in private documents + private_with_pii = [page for page in annotation_results["private_document"] if page.get("has_pii", False)] + private_without_pii = [page for page in annotation_results["private_document"] if not page.get("has_pii", False)] + + # Print summary statistics + print("\nSummary:") + print( + f" Public documents (total): {len(annotation_results['public_document'])} ({len(annotation_results['public_document'])/total_pages*100:.1f}% of all pages)" + ) + print(f" - With PII: {len(public_with_pii)} ({len(public_with_pii)/max(1, len(annotation_results['public_document']))*100:.1f}% of public docs)") + print( + f" - Without PII: {len(public_without_pii)} ({len(public_without_pii)/max(1, len(annotation_results['public_document']))*100:.1f}% of public docs)" + ) + + print( + f" Private documents (total): {len(annotation_results['private_document'])} ({len(annotation_results['private_document'])/total_pages*100:.1f}% of all pages)" + ) + print(f" - With PII: {len(private_with_pii)} ({len(private_with_pii)/max(1, len(annotation_results['private_document']))*100:.1f}% of private docs)") + print( + f" - Without PII: {len(private_without_pii)} ({len(private_without_pii)/max(1, len(annotation_results['private_document']))*100:.1f}% of private docs)" + ) + + print(f" Unreadable pages: {len(annotation_results['cannot_read'])} ({len(annotation_results['cannot_read'])/total_pages*100:.1f}%)") + print(f" Pages with reported content: {len(annotation_results['report_content'])} ({len(annotation_results['report_content'])/total_pages*100:.1f}%)") + print(f" Pages without annotation: {len(annotation_results['no_annotation'])} ({len(annotation_results['no_annotation'])/total_pages*100:.1f}%)") + + # With the updated flow, there should be no public documents with PII flags + # as we don't collect PII information for public documents anymore + if public_with_pii: + print("\nNote: With the current annotation flow, public documents should not have PII flags.") + print(f"Found {len(public_with_pii)} public documents incorrectly marked with PII.") + + # Analyze PII types in private documents + if private_with_pii: + # Categorize the PII types for clearer reporting + pii_categories = { + "Identifiers": ["names", "email", "phone"], + "PII requiring identifiers": ["addresses", "biographical", "location", "employment", "education", "medical"], + "Always sensitive PII": ["government-id", "financial", "biometric", "login-info"], + } + + # Dictionary to track all PII counts + pii_counts_private = {} + for page in private_with_pii: + for pii_type in page.get("pii_types", []): + pii_counts_private[pii_type] = pii_counts_private.get(pii_type, 0) + 1 + + # Print categorized PII counts + print("\nPII Types in Private Documents:") + + # Print each category + for category, pii_types in pii_categories.items(): + print(f"\n {category}:") + for pii_type in pii_types: + count = pii_counts_private.get(pii_type, 0) + if count > 0: + print(f" - {pii_type}: {count} ({count/len(private_with_pii)*100:.1f}%)") + + # Print any other PII types not in our categories (like "other") + other_pii = [pii_type for pii_type in pii_counts_private.keys() if not any(pii_type in types for types in pii_categories.values())] + if other_pii: + print("\n Other PII types:") + for pii_type in other_pii: + count = pii_counts_private.get(pii_type, 0) + print(f" - {pii_type}: {count} ({count/len(private_with_pii)*100:.1f}%)") + + # With the updated flow, there should be no public documents with PII flags + # so we can remove this section + if public_with_pii: + print("\nNote: Public documents with PII flags found in old annotation results.") + print("These are from annotation sessions before the workflow change and should be disregarded.") + + # Print detailed report for private documents with PII + if private_with_pii: + print("\nDetailed Report - Private Documents with PII:") + print("-" * 80) + for i, item in enumerate(private_with_pii, 1): + pdf_path = item["pdf_path"] + page_id = item["page_id"] + + # Get the actual PDF page number + pdf_page = item.get("pdf_page") + + # Generate presigned URL with PDF page number if client is available + presigned_url = None + if pdf_s3_client and pdf_path.startswith("s3://"): + presigned_url = create_presigned_url(pdf_s3_client, pdf_path) + if presigned_url and pdf_page is not None: + presigned_url += f"#page={pdf_page}" + + print(f"{i}. PDF: {pdf_path}") + print(f" Page ID: {page_id}") + print(f" Link: {item['link']}#{page_id}") + if presigned_url: + print(f" Presigned URL: {presigned_url}") + print(f" PII Types: {', '.join(item['pii_types'])}") + if item.get("description"): + print(f" Description: {item['description']}") + if item.get("prolific_pid"): + print(f" Prolific PID: {item['prolific_pid']}") + print("-" * 80) + + print("\nReport complete.") + + +def read_and_process_results(args): + """Read and process results from a previously generated CSV file.""" + try: + # Read the CSV file + links = [] + with open(args.read_results, "r") as f: + for line in f: + if line.strip(): + links.append(line.strip()) + + if not links: + print(f"No tinyhost links found in {args.read_results}") + return + + print(f"Found {len(links)} tinyhost links in {args.read_results}") + + # Set up PDF S3 client with profile if specified + if args.pdf_profile: + pdf_session = boto3.Session(profile_name=args.pdf_profile) + pdf_s3_client = pdf_session.client("s3") + else: + pdf_s3_client = boto3.client("s3") + + # Fetch and process annotations + annotations_by_link = [] + for link in tqdm(links, desc="Fetching annotations"): + try: + annotations, link_url, html_content = fetch_annotations(link) + annotations_by_link.append((annotations, link_url, html_content)) + except Exception as e: + print(f"Error processing {link}: {e}") + + # Process and categorize annotations + annotation_results = process_annotations(annotations_by_link) + + # Print report with presigned URLs + print_annotation_report(annotation_results, pdf_s3_client) + + # Save detailed report to file + output_file = Path(args.output_dir) / "annotation_report.csv" + print(f"\nSaving detailed report to {output_file}") + + with open(output_file, "w", newline="") as f: + writer = csv.writer(f) + writer.writerow(["Category", "PDF Path", "Page ID", "Link", "Presigned URL", "Document Type", "PII Types", "Description", "Prolific PID"]) + + for category, items in annotation_results.items(): + for item in items: + pdf_path = item["pdf_path"] + + # Get the actual PDF page number + pdf_page = item.get("pdf_page") + + # Generate presigned URL with the PDF page number + presigned_url = "" + if pdf_path.startswith("s3://"): + url = create_presigned_url(pdf_s3_client, pdf_path) + if url and pdf_page is not None: + presigned_url = f"{url}#page={pdf_page}" + elif url: + presigned_url = url + + if category == "public_document": + doc_type = "Public" + pii_types = ", ".join(item.get("pii_types", [])) + description = item.get("description", "") + elif category == "private_document": + doc_type = "Private" + pii_types = ", ".join(item.get("pii_types", [])) + description = item.get("description", "") + else: + doc_type = "" + pii_types = "" + description = "" + + # Extract Prolific PID from the item if available + prolific_pid = item.get("prolific_pid", "") + + writer.writerow( + [ + category, + item["pdf_path"], + item["page_id"], + f"{item['link']}#{item['page_id']}", + presigned_url, + doc_type, + pii_types, + description, + prolific_pid, + ] + ) + + print(f"Report saved to {output_file}") + + except Exception as e: + print(f"Error processing results: {e}") + raise + + +def main(): + args = parse_args() + + # Check if we're reading results from a previous run + if args.read_results: + read_and_process_results(args) + return + + # Set up S3 clients + s3_client = boto3.client("s3") + + # Set up PDF S3 client with profile if specified + if args.pdf_profile: + pdf_session = boto3.Session(profile_name=args.pdf_profile) + pdf_s3_client = pdf_session.client("s3") + else: + pdf_s3_client = s3_client + + # Create output directory + output_dir = Path(args.output_dir) + output_dir.mkdir(exist_ok=True, parents=True) + + # List all result files + print(f"Listing result files in {args.workspace}/results...") + result_files = list_result_files(s3_client, args.workspace) + print(f"Found {len(result_files)} result files") + + # Use ThreadPoolExecutor to parallelize the generation of sample sets + output_files = [] + + if args.repeats > 1: + print(f"Using ThreadPoolExecutor with {min(args.max_workers, args.repeats)} workers") + with ThreadPoolExecutor(max_workers=min(args.max_workers, args.repeats)) as executor: + futures = [] + for i in range(args.repeats): + future = executor.submit(generate_sample_set, args, i, s3_client, pdf_s3_client, result_files) + futures.append(future) + + # Wait for all futures to complete and collect results + for future in futures: + try: + output_filename = future.result() + output_files.append(output_filename) + print(f"Completed generation of {output_filename}") + except Exception as e: + print(f"Error generating sample set: {e}") + else: + # If only one repeat, just run it directly + output_filename = generate_sample_set(args, 0, s3_client, pdf_s3_client, result_files) + output_files.append(output_filename) + + # Now upload each resulting file into tinyhost + print("Generated all files, uploading tinyhost links now") + links = [] + for output_filename in output_files: + link = tinyhost.tinyhost([str(output_filename)])[0] + links.append(link) + print(link) + + # Create CSV file with just the tinyhost links, one per line + csv_path = args.prolific_csv + print(f"Writing tinyhost links to {csv_path}") + with open(csv_path, "w", newline="") as csvfile: + for link in links: + csvfile.write(f"{link}\n") + + print(f"Tinyhost links written to {csv_path}") + + +if __name__ == "__main__": + main() diff --git a/scripts/sync_beaker_image.sh b/scripts/sync_beaker_image.sh new file mode 100755 index 0000000..ae9f417 --- /dev/null +++ b/scripts/sync_beaker_image.sh @@ -0,0 +1,14 @@ +#!/bin/bash + +set -e + +VERSION=$(python -c 'import olmocr.version; print(olmocr.version.VERSION)') +echo "Syncing olmOCR version $VERSION with model included" + +# Pull the image with model included +docker pull alleninstituteforai/olmocr:v$VERSION-with-model + +# Create beaker image with the model included version +beaker image create --workspace ai2/oe-data-pdf --name olmocr-inference-$VERSION alleninstituteforai/olmocr:v$VERSION-with-model + +echo "Successfully synced olmocr:v$VERSION-with-model to beaker as olmocr-inference-$VERSION" diff --git a/scripts/train/grpotrainer-beaker-multi-gpu.sh b/scripts/train/grpotrainer-beaker-multi-gpu.sh new file mode 100755 index 0000000..0e92f7b --- /dev/null +++ b/scripts/train/grpotrainer-beaker-multi-gpu.sh @@ -0,0 +1,474 @@ +#!/bin/bash + +set -e + +# Parse beaker-specific arguments +SKIP_DOCKER_BUILD=false +PREEMPTIBLE=false +EXP_NAME="" +NUM_TRAIN_GPUS=3 +NUM_GENERATE_GPUS=1 + +# Store all arguments to pass to python command +PYTHON_ARGS=() + +while [[ $# -gt 0 ]]; do + case $1 in + --skip-docker-build) + SKIP_DOCKER_BUILD=true + shift + ;; + --preemptible) + PREEMPTIBLE=true + shift + ;; + --name) + EXP_NAME="$2" + shift 2 + ;; + --num-train-gpus) + NUM_TRAIN_GPUS="$2" + if [ "$NUM_TRAIN_GPUS" -lt 1 ] || [ "$NUM_TRAIN_GPUS" -gt 7 ]; then + echo "Error: --num-train-gpus must be between 1 and 7 (got: $NUM_TRAIN_GPUS)" + exit 1 + fi + shift 2 + ;; + --num-generate-gpus) + NUM_GENERATE_GPUS="$2" + if [ "$NUM_GENERATE_GPUS" -lt 1 ] || [ "$NUM_GENERATE_GPUS" -gt 4 ]; then + echo "Error: --num-generate-gpus must be between 1 and 4 (got: $NUM_GENERATE_GPUS)" + exit 1 + fi + shift 2 + ;; + --help|-h) + echo "Usage: $0 [beaker-options] [grpo-training-options]" + echo "" + echo "Beaker-specific options:" + echo " --skip-docker-build Skip Docker build" + echo " --preemptible Use preemptible instances" + echo " --name NAME Experiment name (used in output directory)" + echo " --num-train-gpus N Number of GPUs for training (1-7, default: 3)" + echo " --num-generate-gpus N Number of GPUs for VLLM generation (1-4, default: 1)" + echo "" + echo "All other arguments are forwarded to python -m olmocr.train.grpo_train" + echo "Run 'python -m olmocr.train.grpo_train --help' to see available training options" + echo "" + echo "This Augusta multi-GPU version runs:" + echo " - VLLM server with data parallel on N generation GPUs" + echo " - Training on M training GPUs with DeepSpeed" + echo " - Total GPUs used: M + N" + echo " - Outputs saved locally then synced to S3 automatically via --s3_save_path" + echo "" + echo "Note: This version is configured for ai2/augusta cluster (no Weka)" + exit 0 + ;; + *) + # Store all other arguments to pass to python command + PYTHON_ARGS+=("$1") + shift + ;; + esac +done + +# Validate total GPU count +TOTAL_GPUS=$((NUM_TRAIN_GPUS + NUM_GENERATE_GPUS)) +if [ "$TOTAL_GPUS" -gt 8 ]; then + echo "Error: Total GPUs (training + generation) cannot exceed 8 (got: $TOTAL_GPUS)" + echo " Training GPUs: $NUM_TRAIN_GPUS" + echo " Generation GPUs: $NUM_GENERATE_GPUS" + exit 1 +fi + +echo "Preemptible: $PREEMPTIBLE" +echo "Skip Docker Build: $SKIP_DOCKER_BUILD" +echo "Number of Training GPUs: $NUM_TRAIN_GPUS" +echo "Number of Generation GPUs: $NUM_GENERATE_GPUS" +echo "Total GPUs: $TOTAL_GPUS" +echo "Arguments to forward: ${PYTHON_ARGS[@]}" + +# Use conda environment Python if available, otherwise use system Python +if [ -n "$CONDA_PREFIX" ]; then + PYTHON="$CONDA_PREFIX/bin/python" + echo "Using conda Python from: $CONDA_PREFIX" +else + PYTHON="python" + echo "Warning: No conda environment detected, using system Python" +fi + +# Get version from version.py +VERSION=$($PYTHON -c 'import olmocr.version; print(olmocr.version.VERSION)') +echo "OlmOCR version: $VERSION" + +# Get first 10 characters of git hash +GIT_HASH=$(git rev-parse HEAD | cut -c1-10) +echo "Git hash: $GIT_HASH" + +# Get current git branch name +GIT_BRANCH=$(git rev-parse --abbrev-ref HEAD) +echo "Git branch: $GIT_BRANCH" + +# Create full image tag +IMAGE_TAG="olmocr-grpo-${VERSION}-${GIT_HASH}" +echo "Building Docker image with tag: $IMAGE_TAG" + +# Build and push Docker image if not skipping +if [ "$SKIP_DOCKER_BUILD" = false ]; then + echo "Building Docker image..." + docker build --platform linux/amd64 -f ./Dockerfile -t $IMAGE_TAG . + + # Push image to beaker + echo "Trying to push image to Beaker..." + if ! beaker image create --workspace ai2/oe-data-pdf --name $IMAGE_TAG $IMAGE_TAG 2>/dev/null; then + echo "Warning: Beaker image with tag $IMAGE_TAG already exists. Using existing image." + fi +else + echo "Skipping Docker build as requested" +fi + +# Get Beaker username +BEAKER_USER=$(beaker account whoami --format json | jq -r '.[0].name') +echo "Beaker user: $BEAKER_USER" + +# Create Python script to run beaker experiment +cat << 'EOF' > /tmp/run_grpo_experiment_multi_gpu.py +import sys +import shlex +import os +from beaker import Beaker, ExperimentSpec, TaskSpec, TaskContext, ResultSpec, TaskResources, ImageSource, Priority, Constraints, EnvVar + +# Get parameters from command line +image_tag = sys.argv[1] +beaker_user = sys.argv[2] +git_branch = sys.argv[3] +git_hash = sys.argv[4] +preemptible = sys.argv[5] == "true" +exp_name = sys.argv[6] # Empty string if not provided +num_train_gpus = int(sys.argv[7]) +num_generate_gpus = int(sys.argv[8]) +# All remaining arguments are the python command arguments +python_args = sys.argv[9:] + +# Calculate GPU assignments +# Total GPUs needed +num_gpus = num_train_gpus + num_generate_gpus + +# Assign first num_train_gpus for training +training_gpus = list(range(num_train_gpus)) +training_gpu_str = ",".join(str(g) for g in training_gpus) +num_training_processes = len(training_gpus) + +# Assign next num_generate_gpus for VLLM generation +vllm_gpus = list(range(num_train_gpus, num_train_gpus + num_generate_gpus)) +vllm_gpu_str = ",".join(str(g) for g in vllm_gpus) + +# Initialize Beaker client +b = Beaker.from_env(default_workspace="ai2/olmocr") + +# Process arguments to extract model path +model_sync_commands = [] +modified_args = list(python_args) +model_path_local = None +for i in range(len(modified_args)): + if modified_args[i] == "--model_name" and i + 1 < len(modified_args): + model_path = modified_args[i + 1].rstrip('/') + if model_path.startswith("s3://"): + # Extract checkpoint name from S3 path (last part of path) + checkpoint_name = model_path.split('/')[-1] + local_model_path = f"/data/models/{checkpoint_name}" + model_path_local = local_model_path + + # Create sync commands + model_sync_commands = [ + f"echo 'Syncing model from S3: {model_path}'", + "mkdir -p /data/models", + f"s5cmd sync '{model_path}/*' '{local_model_path}/'", + ] + + # Replace S3 path with local path in arguments + modified_args[i + 1] = local_model_path + else: + model_path_local = model_path + break + +# Build setup commands +setup_commands = [ + # Install dependencies + "pip install .[train]", + "pip install wandb", + "pip install git+https://github.com/huggingface/trl.git@2a81076b9a3c2952273b80b8628db2afbc7bb858", + "pip install transformers==4.57.3", # Updated for GRPO compatibility + "pip install flash-attn --no-build-isolation", + "pip install vllm==0.10.2", + "pip install s5cmd", + "pip install accelerate deepspeed", + + # Sync the bench data from S3 + "echo 'Syncing bench data from S3...'", + "mkdir -p /data/olmOCR-bench", + "s5cmd sync 's3://ai2-oe-data/jakep/olmocr/olmOCR-bench-snapshot-082225/*' /data/olmOCR-bench/ > /dev/null", + "s5cmd sync 's3://ai2-oe-data/jakep/grpo_data_mixes/*' /data/jakep/grpo_data_mixes/ > /dev/null", +] + +# Add model sync commands if needed +if model_sync_commands: + setup_commands.extend(model_sync_commands) + +# Determine model path for VLLM server +if model_path_local: + vllm_model_arg = model_path_local +else: + # Default model if not specified + vllm_model_arg = "Qwen/Qwen2.5-VL-7B-Instruct" + for i, arg in enumerate(modified_args): + if arg == "--model_name" and i + 1 < len(modified_args): + vllm_model_arg = modified_args[i + 1] + break + +# Extract gradient_accumulation_steps from arguments if provided, otherwise use default +grad_acc_steps = 8 # Default value +for i, arg in enumerate(modified_args): + if arg == "--gradient_accumulation_steps" and i + 1 < len(modified_args): + try: + grad_acc_steps = int(modified_args[i + 1]) + except (ValueError, IndexError): + pass # Keep default if parsing fails + break + +# Build the GRPO training command with forwarded arguments +grpo_cmd = f"CUDA_VISIBLE_DEVICES={training_gpu_str} accelerate launch --use_deepspeed --zero_stage 2 --num_processes {num_training_processes} --gradient_accumulation_steps {grad_acc_steps} -m olmocr.train.grpo_train" + +# Check if --vllm_mode is specified in arguments +arg_str = " ".join(modified_args) +vllm_mode = "server" # Default for multi-GPU +for i, arg in enumerate(modified_args): + if arg == "--vllm_mode" and i + 1 < len(modified_args): + vllm_mode = modified_args[i + 1] + break + +# Always add --vllm_mode (since we filter it out later) +grpo_cmd += f" --vllm_mode {vllm_mode}" + +# Check if certain required arguments are in the provided args, add defaults if not +if "--train_bench_data_folder" not in arg_str: + grpo_cmd += " --train_bench_data_folder /data/olmOCR-bench/bench_data" +if "--eval_bench_data_folder" not in arg_str: + grpo_cmd += " --eval_bench_data_folder /data/olmOCR-bench/bench_data" +# Store output folder name for S3 sync +if "--output_dir" not in arg_str: + # Use local directory for output + # Note: We'll use the actual BEAKER_WORKLOAD_ID environment variable at runtime + if exp_name: + # For multi-GPU runs, add suffix to distinguish + output_folder_name = f"{exp_name}-multigpu-$BEAKER_WORKLOAD_ID" + else: + output_folder_name = f"multigpu-$BEAKER_WORKLOAD_ID" + + # Local output directory (with placeholder for runtime expansion) + local_output_dir = f"/tmp/checkpoints/{output_folder_name}" + grpo_cmd += f" --output_dir {local_output_dir}" + + # S3 destination (with placeholder for runtime expansion) + s3_output_path = f"s3://ai2-oe-data/jakep/olmocr-grpo-checkpoints/{output_folder_name}" +else: + # Extract output dir from args to determine S3 sync path + local_output_dir = None + s3_output_path = None + for i, arg in enumerate(modified_args): + if arg == "--output_dir" and i + 1 < len(modified_args): + local_output_dir = modified_args[i + 1] + output_folder_name = os.path.basename(local_output_dir) + s3_output_path = f"s3://ai2-oe-data/jakep/olmocr-grpo-checkpoints/{output_folder_name}" + break + +# Add --s3_save_path parameter if we have an S3 output path +# Check if --s3_save_path is not already in the arguments +if s3_output_path and "--s3_save_path" not in arg_str: + grpo_cmd += f" --s3_save_path {s3_output_path}" + +# Add all the (possibly modified) arguments, filtering out --vllm_mode if it exists to avoid duplicates +# Note: We keep --gradient_accumulation_steps in the args even though we use it for accelerate, +# because the training script also needs it for its configuration +filtered_args = [] +skip_next = False +for i, arg in enumerate(modified_args): + if skip_next: + skip_next = False + continue + if arg == "--vllm_mode": + skip_next = True # Skip this and the next argument + continue + filtered_args.append(arg) + +grpo_cmd += " " + " ".join(shlex.quote(arg) for arg in filtered_args) + +# Create a bash script as a single command string with S3 sync +# S3 sync will be handled directly in the cleanup function + +bash_script = f""" +set -e + +# Ensure BEAKER_WORKLOAD_ID is available (Beaker sets this automatically) +if [ -z "$BEAKER_WORKLOAD_ID" ]; then + echo "Warning: BEAKER_WORKLOAD_ID not set, using timestamp as fallback" + export BEAKER_WORKLOAD_ID=$(date +%Y%m%d-%H%M%S) +fi +echo "BEAKER_WORKLOAD_ID: $BEAKER_WORKLOAD_ID" + +# Define cleanup function that will always run +cleanup() {{ + EXIT_CODE=$? + echo "Running cleanup (exit code: $EXIT_CODE)..." + + # Kill VLLM server if it's still running + if [ ! -z "$VLLM_PID" ]; then + echo "Killing VLLM server (PID: $VLLM_PID)..." + kill $VLLM_PID || true + echo "VLLM server stopped." + fi + + # S3 sync is now handled by the training script via --s3_save_path + echo "Note: S3 sync is handled by the training script's S3SyncCallback" + + if [ $EXIT_CODE -eq 0 ]; then + echo "Script completed successfully" + else + echo "Script failed with exit code: $EXIT_CODE" + fi + + exit $EXIT_CODE +}} + +# Set trap to run cleanup on EXIT (covers both success and failure) +trap cleanup EXIT + +# Setup commands +{" && ".join(setup_commands)} + +# Create output directory (with runtime variable expansion) +ACTUAL_OUTPUT_DIR="{local_output_dir.replace('$BEAKER_WORKLOAD_ID', '${BEAKER_WORKLOAD_ID}') if local_output_dir else '/tmp/checkpoints'}" +echo "Creating output directory: $ACTUAL_OUTPUT_DIR" +mkdir -p "$ACTUAL_OUTPUT_DIR" + +# Sync existing checkpoints from S3 if they exist +S3_PATH="{s3_output_path.replace('$BEAKER_WORKLOAD_ID', '${BEAKER_WORKLOAD_ID}') if s3_output_path else ''}" +RESUME_FLAG="" +if [ ! -z "$S3_PATH" ]; then + echo "Checking for existing checkpoints in S3: $S3_PATH" + # Use s5cmd ls to check if the path exists + if s5cmd ls "$S3_PATH/" 2>/dev/null | grep -q "checkpoint-"; then + echo "Found existing checkpoints, syncing from S3..." + s5cmd sync "$S3_PATH/*" "$ACTUAL_OUTPUT_DIR/" || true + echo "Checkpoint sync complete. Contents of output directory:" + ls -la "$ACTUAL_OUTPUT_DIR" + + # Check if any checkpoints exist after sync + if ls -d "$ACTUAL_OUTPUT_DIR"/checkpoint-* 2>/dev/null >/dev/null; then + echo "Checkpoints found in output directory - will resume training" + RESUME_FLAG=" --resume_from_checkpoint" + fi + else + echo "No existing checkpoints found in S3" + fi +fi + +# Start VLLM server in background (output goes to console) +echo 'Starting VLLM server on GPUs {vllm_gpu_str} with data parallel...' +CUDA_VISIBLE_DEVICES={vllm_gpu_str} trl vllm-serve --model {vllm_model_arg} --port 8000 --gpu-memory-utilization 0.5 --max-model-len 16384 --data-parallel-size {num_generate_gpus} & +VLLM_PID=$! +echo "VLLM server started with PID: $VLLM_PID" + +# Wait for VLLM server to be ready +echo 'Waiting for VLLM server to be ready...' +sleep 30 +for i in {{1..60}}; do + if curl -s http://localhost:8000/health; then + echo ' - VLLM server is ready!' + break + else + echo 'Still waiting for VLLM server...' + sleep 5 + fi +done + +# Run training (expand BEAKER_WORKLOAD_ID in the command) +echo 'Starting GRPO training on GPUs {training_gpu_str} ({num_training_processes} processes)...' +echo 'BEAKER_WORKLOAD_ID: '$BEAKER_WORKLOAD_ID +# Replace placeholder with actual workload ID in the command +GRPO_CMD="{grpo_cmd}" +GRPO_CMD="${{GRPO_CMD//\$BEAKER_WORKLOAD_ID/$BEAKER_WORKLOAD_ID}}" +# Add resume flag if checkpoints were found +GRPO_CMD="$GRPO_CMD$RESUME_FLAG" +echo "Running command: $GRPO_CMD" +eval "$GRPO_CMD" + +echo 'Training completed successfully!' +""" + +# Create single task spec +task_spec = TaskSpec( + name="olmocr-grpo-multi-gpu", + image=ImageSource(beaker=f"{beaker_user}/{image_tag}"), + command=[ + "bash", "-c", + bash_script + ], + context=TaskContext( + priority=Priority.high, + preemptible=preemptible, + ), + resources=TaskResources( + gpu_count=num_gpus, # Request the specified number of GPUs + shared_memory="10GiB" + ), + constraints=Constraints(cluster=["ai2/jupiter", "ai2/augusta"]), + result=ResultSpec(path="/noop-results"), + env_vars=[ + EnvVar(name="LOG_FILTER_TYPE", value="local_rank0_only"), + EnvVar(name="OMP_NUM_THREADS", value="8"), + EnvVar(name="BEAKER_USER_ID", value=beaker_user), + EnvVar(name="AWS_ACCESS_KEY_ID", secret="ALLENNLP_AWS_ACCESS_KEY_ID"), + EnvVar(name="AWS_SECRET_ACCESS_KEY", secret="ALLENNLP_AWS_SECRET_ACCESS_KEY"), + EnvVar(name="WANDB_API_KEY", secret="JAKE_WANDB_API_KEY"), + ] +) + +# Extract model name from arguments if provided (for description) +model_name = "Unknown" +for i, arg in enumerate(modified_args): + if arg in ["--model_name", "--model"]: + if i + 1 < len(modified_args): + model_name = modified_args[i + 1] + break + +# Create experiment spec with single task +experiment_spec = ExperimentSpec( + description=f"OlmOCR GRPO Multi-GPU Training ({num_train_gpus} train GPUs + {num_generate_gpus} VLLM GPUs) - Model: {model_name}, Branch: {git_branch}, Commit: {git_hash}", + budget="ai2/oe-base", + tasks=[task_spec], # Single task that manages both VLLM and training +) + +# Create the experiment +experiment = b.experiment.create(spec=experiment_spec, workspace="ai2/olmocr") +print(f"Created multi-GPU GRPO training experiment: {experiment.id}") +print(f"View at: https://beaker.org/ex/{experiment.id}") +EOF + +# Run the Python script to create the experiment +echo "Creating Beaker multi-GPU GRPO experiment..." +$PYTHON /tmp/run_grpo_experiment_multi_gpu.py \ + "$IMAGE_TAG" \ + "$BEAKER_USER" \ + "$GIT_BRANCH" \ + "$GIT_HASH" \ + "$PREEMPTIBLE" \ + "$EXP_NAME" \ + "$NUM_TRAIN_GPUS" \ + "$NUM_GENERATE_GPUS" \ + "${PYTHON_ARGS[@]}" + +# Clean up temporary file +rm /tmp/run_grpo_experiment_multi_gpu.py + +echo "Multi-GPU GRPO training experiment submitted successfully!" \ No newline at end of file diff --git a/scripts/train/grpotrainer-beaker.sh b/scripts/train/grpotrainer-beaker.sh new file mode 100755 index 0000000..d53b915 --- /dev/null +++ b/scripts/train/grpotrainer-beaker.sh @@ -0,0 +1,254 @@ +#!/bin/bash + +set -e + +# Parse beaker-specific arguments +SKIP_DOCKER_BUILD=false +PREEMPTIBLE=false +EXP_NAME="" + +# Store all arguments to pass to python command +PYTHON_ARGS=() + +while [[ $# -gt 0 ]]; do + case $1 in + --skip-docker-build) + SKIP_DOCKER_BUILD=true + shift + ;; + --preemptible) + PREEMPTIBLE=true + shift + ;; + --name) + EXP_NAME="$2" + shift 2 + ;; + --help|-h) + echo "Usage: $0 [beaker-options] [grpo-training-options]" + echo "" + echo "Beaker-specific options:" + echo " --skip-docker-build Skip Docker build" + echo " --preemptible Use preemptible instances" + echo " --name NAME Experiment name (used in output directory)" + echo "" + echo "All other arguments are forwarded to python -m olmocr.train.grpo_train" + echo "Run 'python -m olmocr.train.grpo_train --help' to see available training options" + exit 0 + ;; + *) + # Store all other arguments to pass to python command + PYTHON_ARGS+=("$1") + shift + ;; + esac +done + +echo "Preemptible: $PREEMPTIBLE" +echo "Skip Docker Build: $SKIP_DOCKER_BUILD" +echo "Arguments to forward: ${PYTHON_ARGS[@]}" + +# Use conda environment Python if available, otherwise use system Python +if [ -n "$CONDA_PREFIX" ]; then + PYTHON="$CONDA_PREFIX/bin/python" + echo "Using conda Python from: $CONDA_PREFIX" +else + PYTHON="python" + echo "Warning: No conda environment detected, using system Python" +fi + +# Get version from version.py +VERSION=$($PYTHON -c 'import olmocr.version; print(olmocr.version.VERSION)') +echo "OlmOCR version: $VERSION" + +# Get first 10 characters of git hash +GIT_HASH=$(git rev-parse HEAD | cut -c1-10) +echo "Git hash: $GIT_HASH" + +# Get current git branch name +GIT_BRANCH=$(git rev-parse --abbrev-ref HEAD) +echo "Git branch: $GIT_BRANCH" + +# Create full image tag +IMAGE_TAG="olmocr-grpo-${VERSION}-${GIT_HASH}" +echo "Building Docker image with tag: $IMAGE_TAG" + +# Build and push Docker image if not skipping +if [ "$SKIP_DOCKER_BUILD" = false ]; then + echo "Building Docker image..." + docker build --platform linux/amd64 -f ./Dockerfile -t $IMAGE_TAG . + + # Push image to beaker + echo "Trying to push image to Beaker..." + if ! beaker image create --workspace ai2/oe-data-pdf --name $IMAGE_TAG $IMAGE_TAG 2>/dev/null; then + echo "Warning: Beaker image with tag $IMAGE_TAG already exists. Using existing image." + fi +else + echo "Skipping Docker build as requested" +fi + +# Get Beaker username +BEAKER_USER=$(beaker account whoami --format json | jq -r '.[0].name') +echo "Beaker user: $BEAKER_USER" + +# Create Python script to run beaker experiment +cat << 'EOF' > /tmp/run_grpo_experiment.py +import sys +import shlex +import os +from beaker import Beaker, ExperimentSpec, TaskSpec, TaskContext, ResultSpec, TaskResources, ImageSource, Priority, Constraints, EnvVar, DataMount + +# Get parameters from command line +image_tag = sys.argv[1] +beaker_user = sys.argv[2] +git_branch = sys.argv[3] +git_hash = sys.argv[4] +preemptible = sys.argv[5] == "true" +exp_name = sys.argv[6] # Empty string if not provided +# All remaining arguments are the python command arguments +python_args = sys.argv[7:] + +# Initialize Beaker client +b = Beaker.from_env(default_workspace="ai2/olmocr") + +# Build the training command +commands = [ + # Install dependencies + "pip install .[train]", + "pip install trl==0.22.2 wandb", + "pip install transformers==4.55.2", # Updated for GRPO compatibility + "pip install flash-attn==2.8.0.post2 --no-build-isolation", + "pip install vllm==v0.10.1.1", + "pip install s5cmd", + + # Sync the bench data from S3 + "echo 'Syncing bench data from S3...'", + "mkdir -p /data/olmOCR-bench", + "s5cmd sync 's3://ai2-oe-data/jakep/olmocr/olmOCR-bench-snapshot-082225/*' /data/olmOCR-bench/", + "s5cmd sync 's3://ai2-oe-data/jakep/grpo_data_mixes/*' /data/jakep/grpo_data_mixes/", + + # Build GRPO training command + "echo 'Starting GRPO training...'", +] + +# Check if model_name is an S3 path and handle it +model_sync_commands = [] +modified_args = list(python_args) +for i in range(len(modified_args)): + if modified_args[i] == "--model_name" and i + 1 < len(modified_args): + model_path = modified_args[i + 1].rstrip('/') + if model_path.startswith("s3://"): + # Extract checkpoint name from S3 path (last part of path) + checkpoint_name = model_path.split('/')[-1] + local_model_path = f"/data/models/{checkpoint_name}" + + # Create sync commands + model_sync_commands = [ + f"echo 'Syncing model from S3: {model_path}'", + "mkdir -p /data/models", + f"s5cmd sync '{model_path}/*' '{local_model_path}/'", + ] + + # Replace S3 path with local path in arguments + modified_args[i + 1] = local_model_path + break + +# Add model sync commands if needed +commands.extend(model_sync_commands) + +# Build the python command with forwarded arguments +# Add default paths if not provided in arguments +grpo_cmd = ["python -m olmocr.train.grpo_train"] + +# Check if certain required arguments are in the provided args, add defaults if not +arg_str = " ".join(modified_args) +if "--train_bench_data_folder" not in arg_str: + grpo_cmd.append("--train_bench_data_folder /data/olmOCR-bench/bench_data") +if "--eval_bench_data_folder" not in arg_str: + grpo_cmd.append("--eval_bench_data_folder /data/olmOCR-bench/bench_data") +if "--output_dir" not in arg_str: + output_dir = "/weka/oe-training-default/jakep/olmocr-grpo-checkpoints" + # Build subdirectory based on exp_name and BEAKER_WORKLOAD_ID + beaker_workload_id = os.environ.get("BEAKER_WORKLOAD_ID") + if exp_name and beaker_workload_id: + output_dir = f"{output_dir}/{exp_name}-{beaker_workload_id}" + elif beaker_workload_id: + output_dir = f"{output_dir}/{beaker_workload_id}" + elif exp_name: + output_dir = f"{output_dir}/{exp_name}" + grpo_cmd.append(f"--output_dir {output_dir}") + +# Add all the (possibly modified) arguments +grpo_cmd.extend(modified_args) + +# Add the GRPO command to the commands list +commands.append(" ".join(grpo_cmd)) + +# Extract model name from arguments if provided (for description) +model_name = "Unknown" +for i, arg in enumerate(modified_args): + if arg in ["--model_name", "--model"]: + if i + 1 < len(modified_args): + model_name = modified_args[i + 1] + break + +# Build task spec +task_spec = TaskSpec( + name="olmocr-grpo-training", + image=ImageSource(beaker=f"{beaker_user}/{image_tag}"), + command=[ + "bash", "-c", + " && ".join(commands) + ], + context=TaskContext( + priority=Priority.normal, + preemptible=preemptible, + ), + resources=TaskResources( + gpu_count=1, + shared_memory="10GiB" + ), + constraints=Constraints(cluster=["ai2/titan-cirrascale"]), + result=ResultSpec(path="/noop-results"), + env_vars=[ + EnvVar(name="LOG_FILTER_TYPE", value="local_rank0_only"), + EnvVar(name="OMP_NUM_THREADS", value="8"), + EnvVar(name="BEAKER_USER_ID", value=beaker_user), + EnvVar(name="AWS_ACCESS_KEY_ID", secret="ALLENNLP_AWS_ACCESS_KEY_ID"), + EnvVar(name="AWS_SECRET_ACCESS_KEY", secret="ALLENNLP_AWS_SECRET_ACCESS_KEY"), + EnvVar(name="WANDB_API_KEY", secret="JAKE_WANDB_API_KEY"), + ], + datasets=[ + DataMount.new(mount_path="/weka/oe-data-default", weka="oe-data-default"), + DataMount.new(mount_path="/weka/oe-training-default", weka="oe-training-default"), + ] +) + +# Create experiment spec +experiment_spec = ExperimentSpec( + description=f"OlmOCR GRPO Training - Model: {model_name}, Branch: {git_branch}, Commit: {git_hash}", + budget="ai2/oe-base", + tasks=[task_spec], +) + +# Create the experiment +experiment = b.experiment.create(spec=experiment_spec, workspace="ai2/olmocr") +print(f"Created GRPO training experiment: {experiment.id}") +print(f"View at: https://beaker.org/ex/{experiment.id}") +EOF + +# Run the Python script to create the experiment +echo "Creating Beaker GRPO experiment..." +$PYTHON /tmp/run_grpo_experiment.py \ + "$IMAGE_TAG" \ + "$BEAKER_USER" \ + "$GIT_BRANCH" \ + "$GIT_HASH" \ + "$PREEMPTIBLE" \ + "$EXP_NAME" \ + "${PYTHON_ARGS[@]}" + +# Clean up temporary file +rm /tmp/run_grpo_experiment.py + +echo "GRPO training experiment submitted successfully!" \ No newline at end of file diff --git a/scripts/train/newtrainer-beaker.sh b/scripts/train/newtrainer-beaker.sh new file mode 100755 index 0000000..f1f95f4 --- /dev/null +++ b/scripts/train/newtrainer-beaker.sh @@ -0,0 +1,165 @@ +#!/bin/bash + +set -e + +# Parse command line arguments +CONFIG="olmocr/train/configs/qwen25_vl_b100_x1_default.yaml" +DATASET="s3://ai2-oe-data/jakep/olmocr/olmOCR-mix-1025" +SKIP_DOCKER_BUILD=false +PREEMPTIBLE=false + +while [[ $# -gt 0 ]]; do + case $1 in + --config) + CONFIG="$2" + shift 2 + ;; + --dataset) + # Strip trailing slashes from dataset path + DATASET="${2%/}" + shift 2 + ;; + --skip-docker-build) + SKIP_DOCKER_BUILD=true + shift + ;; + --preemptible) + PREEMPTIBLE=true + shift + ;; + *) + echo "Unknown option: $1" + echo "Usage: $0 [--config CONFIG_PATH] [--dataset DATASET_PATH] [--skip-docker-build] [--preemptible]" + exit 1 + ;; + esac +done + +echo "Using config: $CONFIG" +echo "Using dataset: $DATASET" + +# Use conda environment Python if available, otherwise use system Python +if [ -n "$CONDA_PREFIX" ]; then + PYTHON="$CONDA_PREFIX/bin/python" + echo "Using conda Python from: $CONDA_PREFIX" +else + PYTHON="python" + echo "Warning: No conda environment detected, using system Python" +fi + +# Get version from version.py +VERSION=$($PYTHON -c 'import olmocr.version; print(olmocr.version.VERSION)') +echo "OlmOCR version: $VERSION" + +# Get first 10 characters of git hash +GIT_HASH=$(git rev-parse HEAD | cut -c1-10) +echo "Git hash: $GIT_HASH" + +# Get current git branch name +GIT_BRANCH=$(git rev-parse --abbrev-ref HEAD) +echo "Git branch: $GIT_BRANCH" + +# Create full image tag +IMAGE_TAG="olmocr-train-${VERSION}-${GIT_HASH}" +echo "Building Docker image with tag: $IMAGE_TAG" + +# Build and push Docker image if not skipping +if [ "$SKIP_DOCKER_BUILD" = false ]; then + echo "Building Docker image..." + docker build --platform linux/amd64 -f ./Dockerfile -t $IMAGE_TAG . + + # Push image to beaker + echo "Trying to push image to Beaker..." + if ! beaker image create --workspace ai2/oe-data-pdf --name $IMAGE_TAG $IMAGE_TAG 2>/dev/null; then + echo "Warning: Beaker image with tag $IMAGE_TAG already exists. Using existing image." + fi +else + echo "Skipping Docker build as requested" +fi + +# Get Beaker username +BEAKER_USER=$(beaker account whoami --format json | jq -r '.[0].name') +echo "Beaker user: $BEAKER_USER" + +# Create Python script to run beaker experiment +cat << 'EOF' > /tmp/run_training_experiment.py +import sys +from beaker import Beaker, ExperimentSpec, TaskSpec, TaskContext, ResultSpec, TaskResources, ImageSource, Priority, Constraints, EnvVar, DataMount + +# Get image tag, beaker user, git branch, git hash, config, dataset, and preemptible from command line +image_tag = sys.argv[1] +beaker_user = sys.argv[2] +git_branch = sys.argv[3] +git_hash = sys.argv[4] +config = sys.argv[5] +dataset = sys.argv[6] +preemptible = sys.argv[7] == "true" + +# Initialize Beaker client +b = Beaker.from_env(default_workspace="ai2/olmocr") + +# Build the training command +# Extract the dataset name from the S3 path for the local directory +dataset_name = dataset.rstrip('/').split('/')[-1] +commands = [ + "pip install .[train]", + "pip install transformers==4.57.1", + "pip install flash-attn --no-build-isolation", + "pip install s5cmd", + f"s5cmd sync {dataset}/processed_* /data/{dataset_name}/", + "s5cmd sync s3://ai2-oe-data/jakep/olmocr/qwen2.5-vl-7b-olmocrv4_1epoch_promptv4_mix1025_more_rotation-8372/* /data/models/qwen2.5-vl-7b-olmocrv4_1epoch_promptv4_mix1025_more_rotation-8372", + f"python -m olmocr.train.train --config {config}" +] + +# Build task spec +task_spec = TaskSpec( + name="olmocr-training", + image=ImageSource(beaker=f"{beaker_user}/{image_tag}"), + command=[ + "bash", "-c", + " && ".join(commands) + ], + context=TaskContext( + priority=Priority.normal, + preemptible=preemptible, + ), + resources=TaskResources( + gpu_count=1, + shared_memory="10GiB" + ), + constraints=Constraints(cluster=["ai2/titan-cirrascale"]), + result=ResultSpec(path="/noop-results"), + env_vars=[ + EnvVar(name="LOG_FILTER_TYPE", value="local_rank0_only"), + EnvVar(name="OMP_NUM_THREADS", value="8"), + EnvVar(name="BEAKER_USER_ID", value=beaker_user), + EnvVar(name="AWS_ACCESS_KEY_ID", secret="ALLENNLP_AWS_ACCESS_KEY_ID"), + EnvVar(name="AWS_SECRET_ACCESS_KEY", secret="ALLENNLP_AWS_SECRET_ACCESS_KEY"), + EnvVar(name="WANDB_API_KEY", secret="JAKE_WANDB_API_KEY") + ], + datasets=[ + DataMount.new(mount_path="/weka/oe-training-default", weka="oe-training-default"), + ] +) + +# Create experiment spec +experiment_spec = ExperimentSpec( + description=f"OlmOCR Training Run - Branch: {git_branch}, Commit: {git_hash}", + budget="ai2/oe-base", + tasks=[task_spec], +) + +# Create the experiment +experiment = b.experiment.create(spec=experiment_spec, workspace="ai2/olmocr") +print(f"Created training experiment: {experiment.id}") +print(f"View at: https://beaker.org/ex/{experiment.id}") +EOF + +# Run the Python script to create the experiment +echo "Creating Beaker experiment..." +$PYTHON /tmp/run_training_experiment.py $IMAGE_TAG $BEAKER_USER $GIT_BRANCH $GIT_HASH "$CONFIG" "$DATASET" $PREEMPTIBLE + +# Clean up temporary file +rm /tmp/run_training_experiment.py + +echo "Training experiment submitted successfully!" \ No newline at end of file diff --git a/scripts/train/newtrainer-frontier.sh b/scripts/train/newtrainer-frontier.sh new file mode 100644 index 0000000..87f0179 --- /dev/null +++ b/scripts/train/newtrainer-frontier.sh @@ -0,0 +1,32 @@ +#!/usr/bin/env bash + +#SBATCH -A csc652 +#SBATCH -J olmocr-train +#SBATCH -o logs/%j.out +#SBATCH -N 1 +#SBATCH -t 02:00:00 + +module reset +module load PrgEnv-gnu/8.6.0 +module load rocm/6.3 +module load craype-accel-amd-gfx90a +module load miniforge3/23.11.0-0 + +# Run in offline mode to make sure it doesn't timeout loading the model +export TRANSFORMERS_OFFLINE=1 +export HF_DATASETS_OFFLINE=1 +export HF_HUB_OFFLINE=1 + +export HF_DATASETS_CACHE="/lustre/orion/csc652/proj-shared/huggingface-shared/datasets" +export HF_HUB_CACHE="/lustre/orion/csc652/proj-shared/huggingface-shared/hub" + +# Was getting MIOpen errors with caching, had to disable for now +export MIOPEN_DISABLE_CACHE=1 + +source activate /lustre/orion/csc652/proj-shared/jakep/conda_env_312_olmocr_train + +# Run in offline mode to make sure it doesn't timeout loading the model +export TRANSFORMERS_OFFLINE=1 +export HF_DATASETS_OFFLINE=1 + +python -m olmocr.train.train --config olmocr/train/configs/example_config_frontier.yaml \ No newline at end of file diff --git a/tests/__init__.py b/tests/__init__.py new file mode 100644 index 0000000..e69de29 diff --git a/tests/gnarly_pdfs/ambiguous.pdf b/tests/gnarly_pdfs/ambiguous.pdf new file mode 100644 index 0000000..db978c0 Binary files /dev/null and b/tests/gnarly_pdfs/ambiguous.pdf differ diff --git a/tests/gnarly_pdfs/badlines.pdf b/tests/gnarly_pdfs/badlines.pdf new file mode 100644 index 0000000..7e9ea57 Binary files /dev/null and b/tests/gnarly_pdfs/badlines.pdf differ diff --git a/tests/gnarly_pdfs/bws_book_ch2.pdf b/tests/gnarly_pdfs/bws_book_ch2.pdf new file mode 100644 index 0000000..1f26865 Binary files /dev/null and b/tests/gnarly_pdfs/bws_book_ch2.pdf differ diff --git a/tests/gnarly_pdfs/discoverworld_crazy_tables.pdf b/tests/gnarly_pdfs/discoverworld_crazy_tables.pdf new file mode 100644 index 0000000..6ecb26d Binary files /dev/null and b/tests/gnarly_pdfs/discoverworld_crazy_tables.pdf differ diff --git a/tests/gnarly_pdfs/dolma-page-1.pdf b/tests/gnarly_pdfs/dolma-page-1.pdf new file mode 100644 index 0000000..18da78a Binary files /dev/null and b/tests/gnarly_pdfs/dolma-page-1.pdf differ diff --git a/tests/gnarly_pdfs/edgar-rotated90.pdf b/tests/gnarly_pdfs/edgar-rotated90.pdf new file mode 100644 index 0000000..a218271 --- /dev/null +++ b/tests/gnarly_pdfs/edgar-rotated90.pdf @@ -0,0 +1,1711 @@ +%PDF-1.4 +%�� +1 0 obj +<< /Dests 3 0 R /Lang (en) /MarkInfo << /Marked true /Type /MarkInfo >> /Pages 4 0 R /StructTreeRoot 5 0 R /Type /Catalog /ViewerPreferences << /DisplayDocTitle true /Type /ViewerPreferences >> >> +endobj +2 0 obj +<< /CreationDate (D:20241001224255+00'00') /Creator (Mozilla/5.0 $Macintosh; Intel Mac OS X 10_15_7$ AppleWebKit/537.36 $KHTML, like Gecko$ Chrome/129.0.0.0 Safari/537.36) /ModDate (D:20241001224255+00'00') /Producer (Skia/PDF m129) /Title (Edgar, King of England - Wikipedia) >> +endobj +3 0 obj +<< / [ 6 0 R /XYZ 0 792 0 ] /bodyContent [ 6 0 R /XYZ 7.5 681.75 0 ] /cite_ref-2 [ 6 0 R /XYZ 140.25 467.25 0 ] /cite_ref-3 [ 6 0 R /XYZ 188.25 198.75 0 ] /cite_ref-FOOTNOTEBlair2005582_1-0 [ 6 0 R /XYZ 114.75 671.25 0 ] >> +endobj +4 0 obj +<< /Count 1 /Kids [ 6 0 R ] /Type /Pages >> +endobj +5 0 obj +<< /IDTree 7 0 R /K 8 0 R /ParentTree 9 0 R /ParentTreeNextKey 1 /Type /StructTreeRoot >> +endobj +6 0 obj +<< /Annots [ 10 0 R 11 0 R 12 0 R 13 0 R 14 0 R 15 0 R 16 0 R 17 0 R 18 0 R 19 0 R 20 0 R 21 0 R 22 0 R 23 0 R 24 0 R 25 0 R 26 0 R 27 0 R 28 0 R 29 0 R 30 0 R 31 0 R 32 0 R 33 0 R 34 0 R 35 0 R 36 0 R 37 0 R 38 0 R 39 0 R 40 0 R 41 0 R 42 0 R 43 0 R 44 0 R 45 0 R 46 0 R ] /Contents 47 0 R /MediaBox [ 0 0 612 792 ] /Parent 4 0 R /Resources << /ExtGState << /G3 48 0 R /G4 49 0 R >> /Font << /F10 50 0 R /F14 51 0 R /F15 52 0 R /F16 53 0 R /F17 54 0 R /F18 55 0 R /F5 56 0 R /F8 57 0 R /F9 58 0 R >> /ProcSet [ /PDF /Text /ImageB /ImageC /ImageI ] /XObject << /X11 59 0 R /X6 60 0 R >> >> /Rotate 90 /StructParents 0 /Type /Page >> +endobj +7 0 obj +<< /Kids [ 61 0 R ] >> +endobj +8 0 obj +<< /ID (node00000193) /K [ 62 0 R ] /Lang (en) /P 5 0 R /S /Document /Type /StructElem >> +endobj +9 0 obj +<< /Nums [ 0 63 0 R ] /Type /ParentTree >> +endobj +10 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/File:Edgar_-_MS_Royal_14_B_VI.jpg) >> /Border [ 0 0 0 ] /F 4 /Rect [ 378 529.5 543 542.25 ] /Subtype /Link /Type /Annot >> +endobj +11 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/File:Edgar_-_MS_Royal_14_B_VI.jpg) >> /Border [ 0 0 0 ] /F 4 /Rect [ 378 459.75 543 611.25 ] /Subtype /Link /Type /Annot >> +endobj +12 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/King_of_the_English) >> /Border [ 0 0 0 ] /F 4 /Rect [ 412.5 413.25 508.5 426 ] /Subtype /Link /Type /Annot >> +endobj +13 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/Eadwig) >> /Border [ 0 0 0 ] /F 4 /Rect [ 433.5 375 468 387.75 ] /Subtype /Link /Type /Annot >> +endobj +14 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/Edward_the_Martyr) >> /Border [ 0 0 0 ] /F 4 /Rect [ 433.5 355.5 519.75 368.25 ] /Subtype /Link /Type /Annot >> +endobj +15 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/Kingdom_of_England) >> /Border [ 0 0 0 ] /F 4 /Rect [ 433.5 310.5 472.5 323.25 ] /Subtype /Link /Type /Annot >> +endobj +16 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/Glastonbury_Abbey) >> /Border [ 0 0 0 ] /F 4 /Rect [ 433.5 271.5 522.75 284.25 ] /Subtype /Link /Type /Annot >> +endobj +17 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/%C3%86thelfl%C3%A6d_Eneda) >> /Border [ 0 0 0 ] /F 4 /Rect [ 433.5 250.5 481.5 263.25 ] /Subtype /Link /Type /Annot >> +endobj +18 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/Wulfthryth_of_Wilton) >> /Border [ 0 0 0 ] /F 4 /Rect [ 433.5 231.75 480.75 244.5 ] /Subtype /Link /Type /Annot >> +endobj +19 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/%C3%86lfthryth_$wife_of_Edgar$) >> /Border [ 0 0 0 ] /F 4 /Rect [ 433.5 213 475.5 225.75 ] /Subtype /Link /Type /Annot >> +endobj +20 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/Issue_$genealogy$) >> /Border [ 0 0 0 ] /F 4 /Rect [ 350.25 192 377.25 204.75 ] /Subtype /Link /Type /Annot >> +endobj +21 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/Edward_the_Martyr) >> /Border [ 0 0 0 ] /F 4 /Rect [ 433.5 190.5 519.75 203.25 ] /Subtype /Link /Type /Annot >> +endobj +22 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/Edith_of_Wilton) >> /Border [ 0 0 0 ] /F 4 /Rect [ 433.5 171 501.75 183.75 ] /Subtype /Link /Type /Annot >> +endobj +23 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/%C3%86thelred_the_Unready) >> /Border [ 0 0 0 ] /F 4 /Rect [ 433.5 133.5 537 146.25 ] /Subtype /Link /Type /Annot >> +endobj +24 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/Dynasty) >> /Border [ 0 0 0 ] /F 4 /Rect [ 350.25 112.5 382.5 125.25 ] /Subtype /Link /Type /Annot >> +endobj +25 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/House_of_Wessex) >> /Border [ 0 0 0 ] /F 4 /Rect [ 433.5 112.5 471 125.25 ] /Subtype /Link /Type /Annot >> +endobj +26 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/Edmund_I) >> /Border [ 0 0 0 ] /F 4 /Rect [ 433.5 93 478.5 105.75 ] /Subtype /Link /Type /Annot >> +endobj +27 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/%C3%86lfgifu_of_Shaftesbury) >> /Border [ 0 0 0 ] /F 4 /Rect [ 433.5 73.5 537.75 86.25 ] /Subtype /Link /Type /Annot >> +endobj +28 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/Main_Page) >> /Border [ 0 0 0 ] /F 4 /Rect [ 50.25 719.25 576.75 756.75 ] /Subtype /Link /Type /Annot >> +endobj +29 0 obj +<< /Border [ 0 0 0 ] /Dest /cite_note-FOOTNOTEBlair2005582-1 /F 4 /Rect [ 143.25 631.5 154.5 643.5 ] /Subtype /Link /Type /Annot >> +endobj +30 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/King_of_the_English) >> /Border [ 0 0 0 ] /F 4 /Rect [ 294.75 625.5 333.75 640.5 ] /Subtype /Link /Type /Annot >> +endobj +31 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/King_of_the_English) >> /Border [ 0 0 0 ] /F 4 /Rect [ 35.25 609.75 96 624.75 ] /Subtype /Link /Type /Annot >> +endobj +32 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/England) >> /Border [ 0 0 0 ] /F 4 /Rect [ 95.25 594 139.5 609 ] /Subtype /Link /Type /Annot >> +endobj +33 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/Edmund_I) >> /Border [ 0 0 0 ] /F 4 /Rect [ 164.25 577.5 222.75 592.5 ] /Subtype /Link /Type /Annot >> +endobj +34 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/%C3%86lfgifu_of_Shaftesbury) >> /Border [ 0 0 0 ] /F 4 /Rect [ 35.25 561.75 75 576.75 ] /Subtype /Link /Type /Annot >> +endobj +35 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/English_Benedictine_Reform) >> /Border [ 0 0 0 ] /F 4 /Rect [ 168.75 453.75 330.75 468.75 ] /Subtype /Link /Type /Annot >> +endobj +36 0 obj +<< /Border [ 0 0 0 ] /Dest /cite_note-2 /F 4 /Rect [ 168.75 426.75 181.5 438.75 ] /Subtype /Link /Type /Annot >> +endobj +37 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/%C3%86thelwold_of_Winchester) >> /Border [ 0 0 0 ] /F 4 /Rect [ 35.25 373.5 249.75 388.5 ] /Subtype /Link /Type /Annot >> +endobj +38 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/Monasteries) >> /Border [ 0 0 0 ] /F 4 /Rect [ 268.5 373.5 333.75 388.5 ] /Subtype /Link /Type /Annot >> +endobj +39 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/Viking_activity_in_the_British_Isles) >> /Border [ 0 0 0 ] /F 4 /Rect [ 186.75 201.75 276 216.75 ] /Subtype /Link /Type /Annot >> +endobj +40 0 obj +<< /Border [ 0 0 0 ] /Dest /cite_note-3 /F 4 /Rect [ 216 159 228.75 171 ] /Subtype /Link /Type /Annot >> +endobj +41 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/Edward_the_Martyr) >> /Border [ 0 0 0 ] /F 4 /Rect [ 201.75 121.5 306 136.5 ] /Subtype /Link /Type /Annot >> +endobj +42 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/Dunstan) >> /Border [ 0 0 0 ] /F 4 /Rect [ 184.5 105.75 230.25 120.75 ] /Subtype /Link /Type /Annot >> +endobj +43 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/Archbishop_of_Canterbury) >> /Border [ 0 0 0 ] /F 4 /Rect [ 258 105.75 333.75 120.75 ] /Subtype /Link /Type /Annot >> +endobj +44 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/Archbishop_of_Canterbury) >> /Border [ 0 0 0 ] /F 4 /Rect [ 35.25 89.25 95.25 104.25 ] /Subtype /Link /Type /Annot >> +endobj +45 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/%C3%86thelred_the_Unready) >> /Border [ 0 0 0 ] /F 4 /Rect [ 284.25 73.5 333.75 88.5 ] /Subtype /Link /Type /Annot >> +endobj +46 0 obj +<< /A << /S /URI /Type /Action /URI (https://en.wikipedia.org/wiki/%C3%86thelred_the_Unready) >> /Border [ 0 0 0 ] /F 4 /Rect [ 35.25 57.75 102 72.75 ] /Subtype /Link /Type /Annot >> +endobj +47 0 obj +<< /Filter /FlateDecode /Length 23547 >> +stream +x��ێ$Ir|_O��`0@w��BDr�_�D��PM�+ +�^��n��U�;\� -��ί"<�`nn'7s1 ��3?ۈ�뷗z ��ˡ�S��_��%��:��|��p~�p��[��x��w��x ��-ƃ>��WSI��N�&n@j"��+Ghޕ#�q��߽��;��9�#�v�Q��r�Ņ��1O�;��__~��Ǐ?��_��χ?~��_^~��>�z��K�R�-�~�Q�_��}�?��ח\K×��ύ��]+��_>]�%�_�k��:&^ +��_~}q�VL��o��3��?��ڎ��b��^jo�I�j +DPZ9B�٥F�p;Q2�f��N��z�7#��Ոs5��󍷀]�\��1�k2\!�r�$��k�7+r��"��[��F��C�=��WK��M�v��W��~��g$��iɃ��N��W��Uo��rB.?q��ϯ_�s[/��7��S��2�}G��ͅ� ��n��0G ��'��y¾&�+��f'WL�z!��IEg1�bH��G��7�6��)��\�!O��Kj(W��C��q\S��Yq��}�N 逰��zy�"�b��(׬�藫m4I�f�x��U�k��<��x��45�Hdݞ)��&z�ƞ��Ic�r��}�>��"'��k�z}�\��r +d��/)��ݨt��~�� z}��ˣ��]K�ܬ<-k=+��/-��O\�q��W��v��~��K��I�:*dv��p{fxJWL�i��Fj>|`O�gY_�d2��z��X��ZJ�Ɇ3��ge��V�/��+ �L��Խ�>>]��x��rPZ�W�֯(�@��a��(��̩~ވ�D��Y\}��T'Il��1��s��^��W��N2�5Om��u��H/)��?��g��Ϸ+�W[1ħ��o|��Q}>J��M��[��ݿ ��4ҏh��ǐ��+ +��ߑ>��<��O��2�%m��|�W�岩+��?0؆��(�}�~%*]��[��R�Ƈyd{jg� ��i?��~;��{^Q��C�k��zEoD�K;��G{�?��xW[HW[�R��4߫F|��[c<�5��C�\~�׏x瞖�e��w��~��ǿ��w��Ӂ��n�/Kμ��O��J��f��q��Y�O)\�W�VaH��&Z?}�_;j�^��{��%H�(6�� #}|��3��e��i��.M?�O��%>��,�'Y�$��>L1�)��c�.�O��/��G��?�k�{��bh|��Õ�=p-��=�?L�OyA{j��"�b��u��7��r��ȟw}�>�$�k?��ܗ�N��y�͕��ۄ��x�r�i��>��ݜ��GE> �O� +��ո�/��,"�O�#��G��z��M3; ��3f�GA�O��F�s�ZԼ��$2p� }�D��r��-��ɽԬ�>K��ʦx}�_�ϛ�fſ��X��LB�N.��;�AX��'/��pm�o��ʨ�jwh~`j��G� +ݥ4<�j��~8��t�t��;�_��``0�x�L�v�'zMd~0o/�;��pD~�4��R�k��Q2s�f��ԃ`�Wt*�~��ǰK�y2��\l4��w�+X�;��x�inD�Pt�U�^_Bq�XM;��a�D=z��JK<�9�rv�g�j +Y�.&jCO�j2�P�_��Q��j�`[8��`�Gk�f]��|��`7�Q~?bf0[!�1�Z��&��~��Kх�}��D��3�ci%�B��o>�__bp��\*�~��pK��q�F�5v��R�vg��I��-�>6�� ù��K��x��XD��z��Jobq#�CtK��%W�Z�+9swcq��G�Df�8_Ef��{hL�$�a +��jJ�݀�D��N,9!�6x��x+��_�_Z�V�r��X�%�=2��J��ġ��3�i�HN/��5ӄa�G:vw�'�v��Rǃ�ǈ��^�P{Ȯ��NG�c�60��]��;�W��w[��ɍ�;��L|��G�`j-G��fT�"��/\u-�1j=^_��d670�z�zQ�f�`؅^k��x}i.�LK��Gr#K8�x��#�B��ח�|�Gp��ѫ��E�M�4;ƈ��N(>\.�W��L��y�{��b��֒rf"< �1��N#�|^�Õ�}}�#��Q�|a�^��peD��r��uU��ѕ +&��x²��~z?��y!w��I��x{�.��q� +��iGQ�O��o ?��u��y(��zr��ͮ�n�3�ȷñ�Dx#�٦`�;�N��fu�0Z�G��w�˫م1�\�r�g��&��b�ga�!�yStW�HI��5r:�`12��HGv�]��&��׋�vq�V��{p�.w/��Dd��ff1��%�X��%�+|"�w9f��|��,��=� �%ڀ��cძ��Uw��!��h~@;aXY\7�m�%(��O�y��.M�\�d$@�t��2��G�{$��k�&2�c,@�:ɉ�.Df�L�|��Q'�]��Rr��.�Bu��)�%��`�42�R��n_�㓸�*TVl�1��J䓸�� F��h�e@ ��x:5[XX-A �f;��I��ZoH��2��>F�2�h��B�p�ZG�txx7Bn�p5z�Y@:�]! T�� +S�ֵ��B�(�'��U@�~�T��;�~�I�߀�v��%B:� �`��_^��~~ߢ��"*l��)�Ǌ߅��K�K�l*u&�B)h:�N�^I��+�u��Aڎ�Ǩ07��~:��oy��"��+��d�C(Xawc�m�y&@�9qc +5��|�l��`��U��%�I�bq��_��5͗�˛�H2�~��'H�x�.��j@�FMa�N�'JOvҡad��f�#E욦�3�*�f�_}$]$�I:L*�/�}X��t�(Q<ҔV��/3��X��]V��(B��Z��Ư/��%5��Q��|��F!�&��]�ИN�>�^q&v�G�TN��ry;��|��}A��TD��˽L*��-&��e�_J�u�Oj��<'��KJ��$�~.��O�α'(�:�D<��E^_Rv��E4�I�T(A�D��wE&��8׽��`��]��K��7vWH��9Ӏ{J�s�d��A�}`QsZ$�z`tm�{(V�R��Re�)n��$�(QS�O0��K�[? *sjh��k��v�v�57��Sb�L��B��X�X5��!v9�RG�� +��F�gs(G�`��ˠsp�� %��|�[�`)�ߚH��r��)d�y�rp1��V�{,.��v��GlшyDs�Yg��rr1ee�]��~a�/�[��ST�)á ��g�l�=�s��A��\\}��~ə<�0��.�o��b�rFf�1\OA�ƃ�U��0�R��*�$� �S��9ҋ�,��b�{��RQ��ǦF�b��dz�n�Y,%��g$ɠ�W?9��!U�E��\]��f#��"��@ͥ��y��"�;( ��Iө�X�4O�N��ئDd(G�ыi�_^��w�B�L��bVA{��Liφ.��n;La��舰�ӝ(��f�sb=� +Gbx o��#%��w��K`),TVV%��V�%�X��=vI`<�I�)��[��b #�c�%�%�R�.Kc��K��Y47&�fZk_�d��e+��eCu�{��W %w��ؐؒ�\p+��B`��˂Ut��.�L�|D��zdE��}�^h�y!k*�I41+�Xt @��K��Kم�W��.Ҳ*�"N�'��,Z� +~��YE@YG��ì��_��vl��V�Va��EX��54��|wd��α�4�G��Q +�2��d� +d1�4��<�8XYz�&mJ9��O�J�?� Z�##jUi��i�^_ +��m*Z��!�i�50[�{}�ّ�ߢ��M'�Bh�bw��V�B�w>˷3Gr�DW��;y��U�-e��H۾��n�1 F�e!�6��p�s��.��J�KyTDm�EM�`�J-q�I��jU��Q�o�c�ܕ��ܕ% ��P�=)� �l3(�S�v��2��I��ߙ��;�4Q��]}Md�p)-�"�� a5�6�h��f��z�wP��%��6m�֊X��ޭ�u8߂p�Rd��K�t6; �6�+�+�o� e-S;jg�1�A�OQ��=V�U��]�X�E�%>��a�uԯ��=�G��f�o��ŏ�ױ��Ԋ��y�r�S3�!�补V�V�e� ӑY��ZCD Y�]��R�qә%��p0��D��+��Zq��=��~48T'nۢ�l[c��1҂KI��0��=�R-AY��$�H�g�S9�4��,xn�7� �E�/�:�R� 9�4nq"��ݲ+�Vw�9(�eWS7�+-l�n/��[f�e��>�4��x�[vC>D�@1XrQR�p�^?�!��4��a��vQ��"ۃ�R�~Ɗg�q��G溦� �d}J��I�9�h��ڴ�'WKsVǬ��#�Y� �R�g�,ޥqgkᓫֈ�5�!>��U��P�Q4�-"�M/v�m+"Xvq$��>=�S��n�=�M$�6��\�ڊ�lb��[�ׁF��Ì��4� d�||{�p-�5R��K.u��Ҏ\�YZ ��ar|P4ɂ��aQ�KC��=�b��>L�@��2��n�[�R��=��0!��fz��0F�u3��Z�K�p`\�-�a�l9i��b:��`Yę�Ye�C� �Z=;$~th��(�ڦ�W�Xd�)��h��Q��z��hz�Isi�?IKRtC�{o�CE�8j=��BS�g�i�S��Jw7vF/�%��Y��%Q�gb��$aQ$��0ZD��"u,C+Ei9�۽��J�u%��;�d��N�'�۲C 3p��b4��Pe�P�w��a�4�U��hJ�d��8��2xO�_��7�KK�6)��aMv�Fi��3�@��;��]xt��X/�@�o�}}�OZ��|�3��+$��L�Zغ�@c��m�0�R6L��\�e�p�D]�i��Cn��V�~��ZI��aY+��J�� ^�zd��}�< �L, bBVY0��ј�X�+�C��M0}p�`�}� O:�B��2��* �U�m��`�� >��>ɛ'+�2�߲�Jl�X��-�Z�D ��|�^��w�Xi��l��%��jwvwC�;Ť�ݍ�B�o`�H�I�P�3`�>s�v7/L�LkQ��>�\mQLK�Y�{ˊb �=�M��7��<�3Ffdl0�+mN��2>��~"zX&�*�H� ��`�hn�QI�Sr�{�Sh$b|�Ep +�#{)�,Hx�m�%ܔ�Ua"#T��t�i'�H|F\w#��i �@Rb�@b~��'�?�HDt�F;��m{,�1}��4ٛ9�Iq}�7�劙D,L�ͼX��y��E�Ӽ�K�I��3��3nm�Tj��$Z�ʓT:؛l:O��x�)כh`�T��sm�x{w�@��IP|��0��tH�K^wfC�9�S�"��F@�x(�|��b=�þ4��ٗ�H��^��T��Z�p>��!F�=^�� p�x�'�*��B�H*|��(��I�E��j��7zW/X��E��(o�sm��,�ܢ��V2T�}�f��<*< +�Aac6:�M� ��g�ϖ0�`�,�,��Nڞ�.��`�s��$�;�k!i�C�m`�]��x��,�n7u�ّY��aRװwa�D%��#��4�o�F��(k��[~,�߭�e)s ��?�?%�̌�N4��[X�7�t��XɼG��,y�M ��=��O5{"�5��$� =.hq"�n>�]�{Ah1p�D7ϭo��Mb��H��f�PE<��?�;�#X��,�ej�%�+�I��ܛI��3�c��U�:�)��NC�Z'47o�LYƝ?��@�,_�k8�3B�k8��QX'O�=1��m��o�^�601�S�b��h��]ަe��P=��}{��l��6�7��fl ��E�wP�R{�=��d�>$�5�ؔM?��^�&~#Z �x#_'��Hߓ��&Ұ�D�/�z5v�ז��l�>��L *��8�$B�e)��3 pH!%p��访�5�j��(+��;0�� +�j7�s+m@�D�q�g)��U��.�ev}?�|�$�6y��DqCc��v$��,�g��iWT$��6��u�0>��妳�&l'�9��xop�y$;��O]�s k��U�΁�]�"�Te��M��-�$/:�&��͂䠦k�CR�x�?�R��ǀ��U��lmùX�>��O&�ʭ ��lқȴ+�U��/�vܲ�<<��>'R��'�ZVE�, +h +��%Wfr-�R��Q\i�a`��r��n��MLb��yf8�g +��-E&pt�W�^u�eV��b,��gI�ƳEw�+��%iz�[��D^R��Z"��g�KAz�4��AG!��&1��~8ȰJZ6��R�O�jLa�4Y�P�JaF��mA��-�=��V�� 6f+��J�|�fs脃�r&�'��9�4�l��0$G��9�[��}�� +Qg5��A�#��F1�p + �� sA��4)��3��58�l`��1�U�~ݕo��K-�7�i��Br` ��-�<��N� ��t0��|��,�i��~z6hZ� {ƀ�%� �O4�5E��؋ +m��0.p��(�[~g�0r&�ߋψ�J��>�%��_R��,M5<ك�rPSV�#_g]�%�h�p��4*�|��M��Y�8܌<��Z dC�M��t> �P��0MDH �Ŭ��u6`d*~��p�n�o[#g�a��1k��|�m�&X�խ�*�U�f��m��|*��0��Y_�7��QF3wo�*��GNgT��]�ݽ]��F��w��3=�Fq1;��i��y*'hurt�fr�0=-��h��D ��fr�]4��+�J+��cFZ��-�[V�̓��x�xՓ6��q��9�m�lh��*��ь��3-�?8�ɘ3s��9�f2] �8�O�� 6�֌$�$��a��1��=[jf�+�M��Z D��(Q��`t�E�J��+��=�B�]��6�䘋��\ �j�TA]lI:��@&_6�Ԣϑ�`Mw�; +`:�edf�t��l�g��Q�Ɍ[h�jW/-f�R�n2l }�Z=��t3�0&�_��| W��J�:ᔏ�X� 3J|�*%>2bGR' ��@�0o ��+�d�:Nߏغ鬛�D��Y"�/�=��tR��$ d�8P��w�.�=G˷N�b ܺ�W�XO�_]��9X)�2Z�4� ��y�)��6�1]��m�Z�N�&�<��;s��.��A[]�l^�M�wh��FYԩ��m�s�%�Å��p��fz�&>�X3�=�5�~H�j�P��-�H��N�[t9�4#�1!��:�x�<�(C��-��<98�=.D��/!#��4N="�ʑ㚂=4}�� Mj�!I�9��8-��E�H��va��b̰��e5j�~;�8R�ǜC�O��"�-��x.r��^h��r"�A<�Ȟl�}��N 9r�+Rţ�ߐ�j�t�3]]�&2_� +ؚި_(�( VC>*V�6TNhh�d��*O��bMAWh�S֓��z��X��g�m��#�_�G�t `�rwyٰ4�B� ��F䒰��2�p�-�%:�|=!-��y�.�sz�A\QlV��э�Q�z+_�'l��L�=V��)w� +��0��Xc��P�P��B;�dX��`�#�z˃z��]`��=�/U�PH��SZ�)��O&H��TQ (7ۥɟM)$��O��?�[��z��q�d��U��@X�cz[Pp��A�0Iܰ>B��X^�An��A0㑋��x\�"�-/�;�Ǥ�� +O�"��RT4oGp<k �`P��@��ū��, ��L.7'G6��bPL�P�?CԼ�� ȑ|�g�� X��p�(/1�.e�a��0�� 7�<�SKUN�US4�TӀ%j|L,�K��&�݃�Q݋�an3�C�v�/�6�RQ��כ�-�{ʀfL��m��œ2��)9p��20t�B6!=�C�`�Q�dܜ˞�B�<%{��)\UQ=Kg�S�C��*�8G�Z�=�j�8��˔�zK�di��[!��庸R�`E��V�hFc�,��#k!?)<��DB%��9 *��b��`#>Ed��q�%�E�8rjC�Iw��g��7**��6)N(�&:�sZ�<��3�`�&$d�;�Z��\��n��:�"Y�v 7��/U"g��T0�\�L�>{�(M�t�!��mG�T ]$��b5��5ɜ`�s��`��NS"��c�z�܃�(�G�n M��Y��*ͭR*�B%+I��{_��粟@��?�k� {�`�|��)]7'�Oi��(Q��..lc�l[��5-�;KZ�=~j�!��0DET��.zLqT��A�v��5��qNA+ +��F�;Z`�d��;��]�V�l��) {��C0�D�Ѝ\6Qh�NԐ��b�rL�.z��NȬ�|y��ʉ�%��Ce��&�y2k�C� 7��9��vebak0��mԳ��U =Nl�ǉf�w�5�:��k�e�JjMZ��b�%!3�b��;k�M��s�k]l|;)/�"-��S��|3I��MF��@�n�o~�G-tGY��`��4Jc�m�� u�Y��6��Y�(�늁ȑ݆��w[��&�Ő�E��.�5��Xߛ �\�A��5��2Uɨ#(Ro��!$6��9�9�\�[�LV�}�8��m��v9�cp69�%��f�p��FM��_��3`��v'�)��?=_��)H9e��!�� 6��HU"E�UJ�ٗ�Z��ޚ��v\�� (=�y�M�H}��$�&��rO�Ds�@I��v��&�LEz3=r�n�p��r�� ,��k�u��WJ��]�j{�d��6�-�J2��l�McÄ��|�?��D��k� +��S��R��H�3% �-:��E�s��27�q�z��Ň�IJ��<��8:�K7�., ��0��BR��H3z��(:̃��5�E�j��mf�~0�t�3��$"w�y�i�\��d�TF�y,�邳HH��7\$�Kf�C�Ce��z�:�i��z�٤�̖Rq��D�M�kCҦ�N�O�?�!cO��o��!%�-�L�s ��`a&�Y(�t0�I��q''vNԋ��AΛV�2�n�>f'+d>2 ��"v��Ə��H��qC��~��n1E�d�[�_�?իj>p�8��9�tއ�JC:ٳ��_��?��m�9r�l�&��K� ��A�*�XlM�B��n��ˉ��i�3�@f�r~Pg�)��t�C��G�~v�?K��6��"��[l ~�k��J�nh��{;�;�d�O�p8u��眃OL�,o"gX߁��b��.��\�,e��c��n��!v6�ՙ�¢�$��R<^�C��#�͆�Տ�ty{��6�ʅb5�l,`uG,�A�ZJll��Q��Ͽ��᯿��~��t�xM��K�� +�b?~��t��_��HWӯ��֤��h捬�%��J�?�c.��+��J��_�լb1�?�_��_�T�w�Zׇ��y��rz�ϗ�)��B��s�T¼��~��x�o��f 2c�,9�v1��寿��bB~i +B�U��B�y�-��lP\C�d�)-n��O��.��zn��++�r� YN� �C��E3��|�~gꃁ"�3��'F��8d��N�Ŷ��0?��ow܂��K(H��Gy��β�z�W)�p��S��+=�>�_�{|�Uҫ��j�׭f��Vq��j�Z �ֈS�Rш�T)M�S�:�� +��՟z��ө��Q[��q��C��?�Ͽ��??~��W𼟎�~:.��a��!��(6��A�Ծ��?�\~0�D��r��~�`#��~�c��7�^�{L��}�2�D�8b-�,|�T=͛�~V�\��3�}�3��< �(�!?�� S-��^�#I8�9�p`��oc�rz?��aF}�u�Z��SU��^�,|V7�PK�X�Ӏap��[��TףO!Z��5:��`�IT�au�_}� +b��U��а��h֒aq��?��ܠ�8�K��#Z��6�J)��pqۏ��m�F��xZ3�l�*ҭ*]��,W�bbT�zA�r�}M'�_��2�v1H��-��[c�k(9��1v�@Uk�7ğ��6^�Ch(��Q4,��Y4$1~'�{TJY.[ �D;�e�!tk>@R䧭E�讽YE��-�_��p�� +/K��ǧ|Z�C��f��p㏘�r�/2z��}��Q��j�:*��J*��.�?R)��x�h@T9�2��n��P��,{�L��:��ij��T�yؗ�I7��Ϗ+렊Ľ��w��j�X=��@F��F��CY�f�4j}��/�R��%�G��{$RkE拌z_�{��ny$�T\��0{Y��c��<��AY|3��;�fz2�܌"0�l�3t��Ͱ:i��Q�e<�w-!��5n)��V�8�|u-�2(��Q��C��i�,�7_ȋ��`+�~<��|�n��$�k�C��H�-�2��i ��I�nk۱��;�;�/"{e��p��d-~��^��/��>Y6v��K�7K�λJq:[Ue��Hv�,�zi!]D�(�U��*�yp��Hm��Mc��EƀҍprG�M��L��L�_�ݡ*�SCn�˗s=/)�H�'�ϗ��Hh��f��>�;��z�gEYS6Y_燒4�{ ��v>�w� !#>�{��~ƭ��n�� F�/�{"4>�ۅ�5ٔ��嬀��=��d��e.��%dd��p�&Xҭ��y6BR"��Z�t�)�C ��ʈ�}L��d�k�L̵�J��5��@0��E��xP��[.A�ɜ8I�i��^�>( +��V��zG�I��>N��D7{��8p1��0݄�O7��U�^��t��/+d-ӻیfwMYۢ��l��J(v�m��Bg�L�#5 �vcL�O�㊺�y�P�(��m+��G�~6�̭��cs�߫�b��q��^?��ꍌ�h�~��V�GE��\�m4��z<�D�!�ZnջM׏�{�}_�m�f��}1[�gH�MIC�m`F0��Io��By:J��Rs��4��T-]F��::>�m.��w-�kp�� t��{e<6�v +(3��T��h��Q>��z�k-A��*��Mq+��{m<�Ñ��7��W{3��F\��Ujc��9��v��`�g��/eßJ�Ց��e?Q��1�yӕq�t��;��H��h3(�]�sf"y�2��*Z.W��[[��_l��u��j'��وu�#�@�{z�Ӫv�(�<�S�߈̋d2^ӗ�R/�sQ��m�4db,yÞ��H�K��;�瘰;d��_{��}y!Ǩ�e%�}~G}i�r��˨��Ş��ܭ�ƢG��~��xQ�\ +!�!��Z�'?��'2C��a ��g_�Z��G�~$��^!y��!��{��/�3�t= q�h@m�@l��Q�'�� +��jm�r�D#l��-��^��W��\Ç�_�W��X��.M�%��j�o��B;�&e_m��u�G�O��K��D��J�靏��sK4یT��9Ƭ��E��S��O$��o��_��6H��%芲e0� Z ��(�o��$!Χ��qX3�U�l��#-!��ƭ��yuo?۟8� +j�'|��##m9��ȍ�>�d��H��#m��E��8$��39��bM�MGꭝ��j�dlb�o��s>�/�7!m��(��ʚ!��-�n��+�}�7cmZg�YFD��Mѿ�>�%� �"n)Ӿo��{Q3g}8]�^�:)�@�y�e�O3'��{�M�K�*�v~��no�s�e�b:ǋߟ�z��[j��@��au`<�c +A��y^��ϔ$Í��J~��Cs<��SC�rh�ۛ�V(�ޝ��|G��^�~�z�k�$r��b�R��cy��{�P�_^jy��5ģ~��. �+'3��:�7�zтdy�uH`��!��=�w��y�8R�!��T��䎊A�0s�� &�! �:�\r��b3P�я��]�;�.v��N?"�Gr��ƽ{i�5��,Ƒ�̠�5X��0��r��x��e��[3�x6�l�O´�_�]Α�C��\�}�}'W눵>r��^��H��ӽ��=;{��lc��e�+i3�O3]5&T{=aW��a��m��ӭy�#7�*�ԳeG��>��ե�?AhH�L��v}��"� q-�ݘ��(��aQ'�w"Fƌ��/��!�ԕ)봇i0�S�C��0��H�:h�H�os�L618 �ƕ�a�C��W�P9[Q��5~�2!n��iM֕w2��vcC~�a��Y9R�c��HS�\=�2V��8� �U��D��R��h�%��!s�u��=�A0�+:�!v'�F�Н��JR�-M�a�碉.T�WYDIQt�rUn�qH�uR�*pژ\��,*�ϙK:�@��o��#f�?�*��k�8�Ӳ��Kх��(F�@DP��TR��5 G-�PB�^cMT�p��,Uc�]0.aS��)�W(��ĥ '�b��5��D��K�Uʅ�EiZlFʑ�2۬t81�(e \��۰��*^��,fajE��ݍHY.�F� +�QLu��"�VF�ũ\�j(�)<]ݖ��0�u�2��Xr�a� )��,��`q)q)��U�B��[��(�`��D)�*��W��$�8J�J�u��c��B�D��Zw�c��Î��D�H�Mۛ*�Q}�F��PHkH{�W�Q��Ul��PepH5��NVもC�eT�L�i�F� +�@+4Jݪ��uk��q�q){�L��hf��P^O��k� +j�gοNe��\@�r��Q�J��B6!��%��R�.��^g]-�MiT�� +d�Qvd.�@ůF�Z��iP�*Ua��م�O{��i�JW�C��B��m%��:.��peHU(>*f�sW� ��.�F΀��',O��k�rǑV�K��F�O+�5Jw�V�� +uƠ\��*+P��}k��DI�\�9��>�4��\�80?pc؟��b��\�%hjS��!U��Rnc�I匤z��TK3uy��ⴴ�C-\yR�{��C��ڣ�g9�y��Զ��߼�w��-��z��x8T�R�=�m�d��ID�(��f ,VP�[D�K��嘁��z��ݥLJ�f?�De +C�!x};.�&e}�H�D}՝�� +��4OU�OV@��}�d ��t^�h��dpi��`.< �AA��Z��@��({o�Ğ{ B�D��본Q�Nr�#�W�C��:!��2��T��aV7�r/S^�Ne��q��eu�P��*� |��O�c�P�|Y��+�Ț�I.��[%�:$a��j��D��2�cH��M��VD&Wj��p��UTf5H�+�p�]{��c ��9��e�m��#p�K�u%�o@G;��^5�Є�O��z��/��_�h�}ߢ��"*l��DP��:�ԙ�T8��~D�9$��K��!g��dL�"ao�r�*|!R�{0�� _QJYq_I9bC(xV��VU)*Σ��č)��:_<�Um�=�T�y��F$��ŅΧE1�4_�N�5H2��Ξ @HPk౺��C��$}��d'F��C�̈́�k�Zτ�КEH~ES�I�I:L*�/�}X��t�(Q�Jח�V��/3��0d��rm5��=`m�;��J�H��{�U�IN��|��F!�&��Q��sٖt"��3��<��r�݌K�Aޟ�0*�gb_�,K��#��{hB��2��O��@�F��F��R�"��E`M��2f�$�~.��O�α#c� %�o��A^�#�l/��Or��vYxKp��TX��+2I�ƹ�U�뮰�� +T`�Zw��of�4��\��s��a�5��/F�V��bEq.e{.U֚�!�JҎ5��#��_�4��Ӡ2��ߞ�̱6�n�/k�m�ZC�R� 緔�Ů��(H�@��,g$��ES��V�ҳ��@�qVa~��J- 6*�:��$h!�E��N}NE# +�a#O1'�Gsъ��pƢ%1u�O��uH��,t�4g��6�-'�'jsA�!� �W��y��X�4O�N��\�IJ�B��4��w�B�L��bVA{To�&&K�.��q?�a +�=DG�� ~af�O�p$��v_.�>�{��m��B4��J��P\r��.��'9�j��9�7�eYh�,q-H�9vY� q��|EByR�,��{3-��Vf��Y�2��Y6T��}�Pp�o� �-��b]��ͩ�K\�1Yх�4)@E��b]��B�KtY=��J/4ʼ�5�$��`,��IĀ%��C�Tk��_�.Լb�/�ONҲ*�"N�'��,Z� +~G)�emX��6c�J�8nh�V��s��r��Z`� ր��AŪQ1ۢ��1T��D�ʣ��_��b�\m��Y��I]^�(�~탫�C��Ro� Z�##jUi��i�^��!7��\�R1-��l��Oh-Z=��U��g�[a�EO�o�r�7��fs�5��@��?Tqp`'�b��l�i�� 3 &�h�,d��\Z�DP�,�L��Q�A5��+��y&-�BvH�Uy6Gعlu#�9��D)��ܕ% ��P�=)� �l3(�S�v��2��I��ߙ��;�4Q�@$�}Md+��.�"�� a5�6�h��f��z�wP��%��6m�֊X;��λ��[P��2��N%��@Դɴ&�/��2��Ƙ� ⧨GM�x��*z��.j,L�"��K��:��Y��Z�+Z�]�o��ŏ�ױ��Ԋ��y�r�S3�!�补V�V�e� ӑY��ZCa>�2�L��|&O��j��*.�V\ob�b}��nK~�z��a�V�"-��X�G�~��K�P9d�>Z��@")�j��Գ��1��"4T��[v�6��Ҹŉ�B;�&Z��B䠔�]M��PZn1��-V��;��`�k�o��b�*d�)��*��i�� v� �}��VB��UdK�+��c��f7��)qgS�"&%䈢u�k��\-�I@X�n��,g��LK��}�p��S2��դ�7�f4��޻j�<��U*��#*�.�ε&q�H�!��`H�U�[c��Eqw�ƞ�&�r��J`�um�s�T{��U��aF��p؄2=��wX��)X� �KSW� + ��Ȩ��b[&�E�,��04�8�߳/��ݏЧ��S]��6�"ٍ�b�VjQ�gR{&��L�=��(r�nfӞ\�A�~��+�E2�-��"�@#2WLGCa�eg�f10�U=�k��ѡ1, +��j��r��"��NɶF^>&ό +בˤMF��O�# Hs�IZ��{�(*�Q��16n��?3N˝�UT��3za/�8�Ί@��@"�"�^!�� "I��" +<8�cZ)J˙�F*;��]U�+EP��< h��a��P�\%� �H��N��0��^y��ͦ�A-�^�Gv�qgw��R��4��1�C(\�� ɷ #��t�b�g��e��i-J�瓫-�ii>woYQC�g�)��6�s�Ȍ� �qe�� S�GrW[��OD��RE�$��Q�=�xJ�xOs +�D��(.@!ud�!e� O��༄�R�*Ld��Ϡh�?Gn�O�0� q��1�3��R��p�@\��-j8�8G��WY��ȑ��m>�eN��]�@Ԧ1�q�cF��&��c��I0�:��|��s��r@�-��4�� ԑg�&�eA�Y��K�m�m��08��%Bsò[ �p�$�pA�Ō�.��]�7DlEC� +/+�LV�w�,P��QYʉ��ǌ -H�,7�)�2kb�Rq�E�^�<�>��˴��XȀe� +a&�}�ܐ� +Xw�繜�^�!�p0�ܝ��9*�ZZ�OP�E�o�Q�> u��[�8s��N�ƉJ1zQPτ�^v��s�Ï}wd��>��Ե��pO�݁`�� ^�fU��=� +�K�_�wT��[��Z�3��uS;��?p֤�F ��̋�M]�Q��h�D񸡗�H�F�^{��d>�󍟯zu��^~<�$��f�o$I��cu�`$ +�W'��%�ˤ�ȹ*�z�l��.�?]��!�O&�ܝ�:��Ս�+�!v*��T��vu��Ԑ��<�w�q5��wW�Xތ�z=.�q� /w�o乵�Id�IR�4Z/��̿\u�rͯx��kj��U�w�?R|{=&��l��>_�A�s��&P�Þ=��^W�\�K��wv�g��_4��6u=�+^5�q�17��\.ԟcv/?��yp�i�s\��_��5��'r��V�$�2�9̳⽷�cO�H��o'ɱk߾��9��endstream +endobj +48 0 obj +<< /BM /Normal /ca 1 >> +endobj +49 0 obj +<< /BM /Normal /CA 1 /LC 0 /LJ 0 /LW 1 /ML 4 /SA true /ca 1 >> +endobj +50 0 obj +<< /BaseFont /DAAAAA+Helvetica-Bold /DescendantFonts [ 64 0 R ] /Encoding /Identity-H /Subtype /Type0 /ToUnicode 65 0 R /Type /Font >> +endobj +51 0 obj +<< /BaseFont /EAAAAA+Georgia-Bold /DescendantFonts [ 66 0 R ] /Encoding /Identity-H /Subtype /Type0 /ToUnicode 67 0 R /Type /Font >> +endobj +52 0 obj +<< /BaseFont /FAAAAA+Georgia-Bold /DescendantFonts [ 68 0 R ] /Encoding /Identity-H /Subtype /Type0 /ToUnicode 69 0 R /Type /Font >> +endobj +53 0 obj +<< /BaseFont /GAAAAA+Georgia /DescendantFonts [ 70 0 R ] /Encoding /Identity-H /Subtype /Type0 /ToUnicode 71 0 R /Type /Font >> +endobj +54 0 obj +<< /BaseFont /HAAAAA+Georgia /DescendantFonts [ 72 0 R ] /Encoding /Identity-H /Subtype /Type0 /ToUnicode 73 0 R /Type /Font >> +endobj +55 0 obj +<< /BaseFont /IAAAAA+Times-Roman /DescendantFonts [ 74 0 R ] /Encoding /Identity-H /Subtype /Type0 /ToUnicode 75 0 R /Type /Font >> +endobj +56 0 obj +<< /BaseFont /AAAAAA+Helvetica-Bold /DescendantFonts [ 76 0 R ] /Encoding /Identity-H /Subtype /Type0 /ToUnicode 77 0 R /Type /Font >> +endobj +57 0 obj +<< /BaseFont /BAAAAA+Helvetica /DescendantFonts [ 78 0 R ] /Encoding /Identity-H /Subtype /Type0 /ToUnicode 79 0 R /Type /Font >> +endobj +58 0 obj +<< /BaseFont /CAAAAA+Helvetica-Oblique /DescendantFonts [ 80 0 R ] /Encoding /Identity-H /Subtype /Type0 /ToUnicode 81 0 R /Type /Font >> +endobj +59 0 obj +<< /BitsPerComponent 8 /ColorSpace [ /ICCBased 82 0 R ] /Filter /FlateDecode /Height 100 /SMask 83 0 R /Subtype /Image /Type /XObject /Width 100 /Length 14679 >> +stream +x��W�/�h��@��.l��8q��$.q�8�{��jFI#iFS�{�Ҩ��4M��^��c;��>�=�H_+β7pYػ?��>y��C��v�v�v�v��p8��lx<�o��Yu:�+� �B�D,�/��$r�\!W��fs876��p��[s��[�PheeE��M�1*�z�\#�� +�H,��P � +�2�JE$�N��wFG��@�X,��B5�Hz<�x"�/,6�\�Rk�9��huZ��Y,.��P*�\.�/��j�A��ɕ�Bu��!��H��K�F�b�;�B�X�� + +�R� +G"�B�� i��h4�Mfˊ�j\1�F�V��6�F��4Z�F��R�R�7��buXl�T.�)E �/dq�j�A��)TZ�B��j�%��&�� F�?�U.�]nO&��T��J5��+��tj �?��ǳ��'��X(q��&��`4"�0 z��h��6��jw:�^��\�U��m�;Mf�Z��U"��b��z�V�� +�@,�D\��1Y��"�:�%�L��p$�XfT*�H4��ew89|�J�5�L+˟`f��v +�B��ҙ��lVx��I�<��6|�u��^��c�;-V�NoR�4j�^�7jtz�Z�B-.'I�4m��@��d�TjM�Z��A�Qg��F3��.�H�<>�˓ʕ�C �i& + +��f�\Ap%)�w�l�:�N_ ��Dc�Ht+�۝��چ?��r�;�.��*W��:�ֹ��>��T�BX!�B��B��\��כ�HT!��Qi�l.�+�4�Tjs8m��ȈP$K$�I�R�\�T*''' FS�R-�\�B1�˧3�t&�Jg�T,��h<�p9��߿��N��]�9]��m��6��s�GG�^?��ÿ>��[o#��R�+"�)��esst�X�h4��,�t��j�+�j(-�+zӊ�b�)��k�Hr�Z�� @`k+�Ng��z�Z�Tj�R�X* +�l.Kg��t"�J$R1@h��v:]++�g��_��5't-+t��~��عϽ��;��_}�ȵk7�C�E_X�s�X"O&��F��j��Y��t��^��mg�9�L�7W̖�o.�'�ll3�\>_�Vk�V��j��z�Q��C�a� +G +��F�J�'R��Pd+��Z��\��\[��m6��ѻc�{��^�p��?:q�".��@$�/.�Y��᭘-�v��jw:]�@x��N��t��*�Ƿ�L&ke��5mmm��lmo��v��n��t;�n��A�F�V��+�.��x�_q��l.��/�](7�B��AD\>��No�xW=^��u8��?�őW�(Uڟ��33��_z �D�B%�)�B W � �<�D � +��*��j��n� >H�Xr��B�(�+%ršC�8\��f��fkggwggw{{��v��N�T�|��L*�f�J��3xB�P4�X�}qW<���Q�M*��r�D��<��D�B�䙧��y6��k��'r�@$�p�l�X�P��?�4Z��H$R�v��V*�v��j��w�u�=��v�� 33V��kZ$-�*�zê^��k�^o�� x��/��z�Y��G��B��Ѩ�t&�Le�k%X8�� @`cc��k>�ڪ/\�H~aq��|K��F�V�T#�R#�bqy��"��e*��p�Z-DM�j��E��n.��d2��СCl.�b�}}�x��\��h��u��b�R��iw��wZ��s��r��n4��y5�|��1��2�-��$.��]��~X1�&�ް��0��r�/d�2�J�ѱ8��%&[�`~]��[&�i�;�n��4M�H�f�Z��f��-6��k��T(��zcwwwջ��C2��nw��N8�r�r�V��bi��Ӊ�>�ʠH�mxWW7�yL\�Հ��r��b7��Yo0!u�Pj@�LZ��'��<�M�1��-1�r��\k��v��l�*�*��]$��wvw�j��FoDL�`�Lf�hR�|��$�H3D�M��B�(�l�� J +�L&�t:X��۝�yz�Tn4��j�\�lmn9��p�\n��@�xW`�]��ڝ��U*W�*�F?p-��U��B ��[Xb�/,�R_�x��3g��!�GT�nwj�F:��p$�H��E��3�t�d4��+f�?�"ѻ�7�-�}}s��r�-v�Z�R��d��j�T!��y�2�M"S��Y�<�?5�@�j��N�s��\6��j�^��+�*��bq[,��z��o��[o�8q��ѣ��Ç�9��+��'O��L�f��b�oë��f-�Z/S��2�H,��x� �l�`��/�i4�}��7��:r*,��t��ǰ�w��{��R�1�LF�3�D��p�|��{�}x�=��y�^��~�ԩSgΜ�s�N�ٺw��P��(��m�b��d��ʉ�$t�h<�Xmvӊٷ��667W�k}�r{�6�j�j0ZL+#4�F�Pi�B��" +C��Ę�͓�4"�?6_��F��+��\ggg7�W�5�2C"�<�:�JU"S��&�i�F��ϟ��'O;v��O�~��V��~!��X��e�j�K$q$�k%��M�˽��O$�Ba��5�جy��[mN��@<ȕ��'N��7�E_\� �+��D%�(_�E�h2#>��9,����7�]��Eѷ�>T��}pp��/.3W�V��4��4mii��͛'O�|�?=��f��]�Ȼ�J��V�-P��Бb��P$��f�E@��fwz�^T!��k@<�0�9W,6�ZPk!�W�4G^>��~��7y1��_f��D�o�;��6�I�V�֨TA��V&�CUO$�V[[!�ł�p��QZ�Û ��\��q��%��B�?t��[wf�sȤ2E�*�J��i*��!I��|ww�js,.3��}�ӧO��;,D��K?x��c��L��T�Ri�;�rDb��R�R,��P?0�[�v�V'�+�nw��0�;\e)Tڷ�<�T�Tp�q��S'�E��y�6;7wc+�q��I�RHS�z�ع��z�V�fK��y�5��Z��)W��5�ϗJePa��~6��0 +󹂋�/�E�\��z52��b�z��> X��b~��@�(k||��_`��r��g�?��XQ� $��R9�L6[HS +�V��vk��(t�ժD*��?~�O@��ŋ8N ��n�k�=T!ܻw ��"o�F��x��t:��b��S��k�#Q�Ն�(76�X��-6Ǌ�f\��f��x��o��7��w߸u�6R��,��"�@�ykO ��XZ��V�c�L�P��n��Ɯ��bI ��S�N�8q�� Ϝ9s��M�aw8��RP{{��j��"�jM�P�z}W/]��?~g}m��޼v0|�1>:��\$s:�0� +q`��jC��.�*� +�h��`Rk��K/��7��%B��'�� D��H�_�R��I��$�H��3(��Z(��R��vp�D�.]��d�xb��O>��'��G?|�(w��"��b�X��f�}��ΝG ��r2��%��8�bQ�-��h�Fb�Pds3h�Xu:��B�V�X~Be�4z9��R��/-�(4:�L%��$2�D��Ͷj��~�3\,!m��W�єH�"��7n��att��sO%��>��S`�X=z��Z`��jX9�3��? +C��ޱ7޸r�r��#L)�l;Dcq��V��dr�ͱ��i��>?6�A��_��%��AXqy��%u�N��3�Q2�6��nwvvv;�n�X�T*HYmoow:�qg��{�^�Xz��G��'�|�>��\>��O�8 +�V��G�?>`N��f��B��H�G�^|�W�r��۞��|�Ϗ��.�JS�TABzO�3��r�B�T��ź��]��Zmv0i�~�b�C��U*�vaq��r�"��dqV3$ʩ��;��{�Q�n�\.��e�:�*@��S��T�}��Z�~�'b�ȑ#r��/��p!�>��O>��Vp+��D�V ��w�p��'�~��hRg)��[��N�s<��W��x�H��10�� +��Th�ج�>VéP(�hUj�B-I��|��c��f�D� +moo�ΘW,��/j0b`=|��!VcX�Ⱥw�^.��կ~��[o cu�Ν��{��w=�k��67�vwwy�Q$��wwwwvv��/�󫗯pٜn��;~��r�s@�5(��a}c��M&8��K,��*J4Z�J�KGJ�T��PiȯL6��8��Z�b�>��v�sT��6�W��z��Q>��l��'�G�H��^{ �>�B�F�<�-<�pD@�%��N��+f�dC}x��bS�I8��ts+�o��J��nw8��(Im�^��כP��A�� )L&{q��̐ ��|>��R�^�p��樴��{��w�^�\�Kg}�~��o�wАS�v@Y��J��~�Q>z��ՠ��7�p��3�2� +A� +�n|*�� +a�uf4Z��J��b��)��`��+%�XpR/cq��'3C�XgȔi1��a�X4��d^��L�G��E1X�7��΃*�J��I"�ٗ��ⷿ��N��~��EXmmnu:��Я��c�H�_=��{��OYV�N�,�H��`(��n��櫭`x}c�Xw{V�2�J��z}f��AX�4��B؟aq�8�Ի��T�4alrzb +��=�D��}��ť6�W��j�z��)Ac�� >V�~��^�e,-_�x鹟�tyq��>C|�\ks}��|9"��pX�b~�4�W��i0�/��ʢ~r<7��X" +�*��XSkt +��]5LX��N��J�� +l8��=7�ÿ��3$��nt|bl�-mo�8�v2��ީw-f+"�\.�R��c�>��% �~��s9�?��.�?? +`[��@e��zۭV;��V��-�U�̉Dc�`�*�k�:�޴��w��n�O�֘�6+�L 5ZZ��ȔB��J��x��;o��D2?C��M�r��;"�h|d��[��0M�0��X�_]:��;�|�eκ~��3O=�t:1�P��jg�|D��h4��p=��Հ�P�8\�Cp�W��U6��mm�b��5� n�wcs�Z Z��b�A�z�W a�Ȕ��+K��>8w�� d�)�3O=M��p}��g��F�:�Z"G�A�V��z�A=�� ~&�Z��:��X%)��dw8=�V�emlnml�6��K�Ղ�9�Wz�Y�R��l�F�TiE_(�u��KWμ��?�!��gH��3��z��a��]=��O�s\*�@��fs�O��޹�:$�G_��%)��7��V뙧��;2\��VW7�7��=z��=��Xg��?j8��Jj�ʃh\�ǣ�X4�t�pD��J�3��e0�(�j�� +�~pnt|��+��i��L�Z�X��t�r8�)�H��ϔJ%��e��f�].Wr�<��7��UL�0��P��[G߼v��>�k��v�J�պ��<��z��B%��Z P�ׇ�8hAS诵Z��A��;lhs��j�D"��f��h��J�W +��H��|��[w>8{�0C�ϐpx�~�R��n��dQ6�T*��0V�eZ,"��c,��G��++W/_nCU��pn]�q썣?|��x�_�|�� ++\�`�Z� ��j��ÑP8R�VQ1X.W��p0� �m��^��{��x�W�� Q�D +vG��%{r?66��)��$~|r��F��fL�Rk6��O�Yހ�s��~wd��ȩU,1�j4�X�s(R��@<<��5G�}��7n2��X��}JC�`P� � n3�L�T��N�Ñ0<�x�R�6`O��3�9��*��+��ܼ~3��\��s�mn��7��l��Y�~岁�"�uz�˓�R�L*� �bO�d��Y`�E��qSx�4�ƍ��\��J��9`�b��Ca�|� �JcZ�}}�僵=V� |��n�R�x<�J岹R��4ru�K�j�F��x>fj�z6� +��?��;��6�/w+`C�L"�;{��-�ي�?��}�(DEb�J�Uit��F$��g��3xq7}�ԩ+W�Ie�XG�b�N�V��ԏG�׺O$��J~��qр��ovww��B&��{��N��p/��!؞�{��.��}�:Zi0�_�η�=>2�v��k7��?~�[�E�؇֒�h�#��A� +�u�H +�7+�} O � ��i�r��b�X,�L&�"r�Ys:�I��o{��>�76�j�Z��=d�[2�:�ΨR�� +5��'I��Ġ/,S��De +?3��7nc52:��n4��_8u��C4��!�ڇk{{nQncA�j��\,��p7{��1֡��}�r5��b$q��@�{��ru}}Ώ�*�H}��O��'cZ��jȩ�y��B� ��K�K$Ag��3$ +j��a�/.5` �nw�z}��]_[��ɱ�#+d�p.�=!��K��4Xv��r�R�^�7�=A̅�:�B�3�,��x<^�ǛIg��Q��Q��2�lvȩ8�L&G��@^��`��N��b@i�*P��%&��P(sK̅%p*2��>��w��}p��v��e3�� P�T��&��!gф� ��c��8S<��`�T�܇tB�>N��i(t�dpg<�]��-��񬺽^��lp[�h2�+�[�Ԉ� +��Y>m~��k/~�ȑW�Ƨ�}��~�U��M��[��2��j� �V�t:�M�놷�:�z#�L�4\��kȋ*�*�Hn��[-l�� 6|��V��ҠW��_��/�FV}�ն��U��fw��"��a2�I�F��<��ǟ�/,,.��/ΐ(?��s?�3266��;�U�~�&��+��4�B��15�;s�4��T�7B�0��c�D��B�j k�`�֍��DR��X,��=c�Z�?|�,��0�< +�V��a��V0 +��76��@6��a��H��w%��ң��s8=F�X"7�Q�� D"��|P�X,.��07�H��o��?��O^�~�U_5�<= զ��h4�>��L"e�9��A��$&��bjgg��l^:��j?��MF�H��;׮\�s�Ao��@ �׫Keߪ6�s�L.g@��nx�#1��D�m8�T��G��qŌ��"�R hJ�/��(��J�ˇ_�z��_|idw��n�!��)P�`��G�2��:h�ǲ�9u��W��6::��+�Z8��*��Y�V�F��2��Z-�H"�0�G�4*��`��Qg)gN�7K"P��ȝ�&z��t��R�X�_$A1�L��Ww�IPC��t�-6��~w(;FSHyJeJ�FW��g)�y�"m~�B� +O ��N��o��:6K��A6��L&S,��5�`�]��j��R�ۣQ(��N?A��X�� F�H��/�[��㏄��՚��H�3��Y$��=@S`��A��pllny�>�~��n�6��hB��J�D�o�N�Ri +�6O��_��C��ٳ玾��8w��ͯi�)��D�� +��P(�V�냾��U�N��*�b��q2��i��Ao��0@m�j��lV{2�DS?��8��/��px}cM��*��lU(T�e�T�D�� Pr�X��`..3��e&{a�17p�)<�ƭ;#��nܼ}��7��4_"-˩T�P(V��t:��/�F��n�OD"7��r�ϼ{:��y�P0|��j��goݸA!�1_�ܩ\� ��H�VWWS�4��L��[[�U_`��__��8�c�O�q噧��o��7��w�x��#`F/r�H��ZZf2�<��H�%�)��q��72��詡6xb.�J�Je�K��fs3�a��Z��8s�t��l��x��E��`t� Ju��`��nnn��H$ +NL��7�Ķ��h��u8 +a��/%W�er��J$��P�L��[�&�LM��*�H��L��x��$�~��6M�e ��x��P,�%� ͢O��Y,� � � +��$�|o#C�4�K�L�{<�n��|>�N�]N�p�,�5b|��nn��Fw��\��+�o��.��`��O�hs�3D�$n�D�ŁEP2n�0�Ú�)��̟fd�O J��b�tLu��T,��f��*i��l��1|�g"�J�J,O$��. +Yx�&�*�I�2F�9�� a؞ +mllkB0��S�� <�Ҽ��S'N d��-.1pSx +�>7��&�fipN:=9E�<��<�ewH�./�x�*|� �p$�L&��l2 �3h�T*�L&��&�4�L��ƬTk�*5�X9��ixS>.�勹|�5��R��׶@��n��p��y8�3"�� +��/�x��e�X +�|��Ā|N_��Y�<9��&��<1�ow��y7�He29�ΐ��B�!�6�=�R�R*W��2z9a�*�|�P*U`Ӿ�4\2I#��"�D�,��P�[Aa0� ��7%Ry` ݋��f��uF��S(�)�z��/,q�&��}qO�&��D2yz�82�Ǎ�Y�F��~�O�T�I$�R� �H$]n��d�P(<��R�t:�v{��(�%t�Uy�(`�`(��:Hv�γ��V��6[4 +$8�R�,��`/1��?L�#I�9:~��'�'&&F�qw�&��?t��V�N�z7/#'��P8�e�"��h�L2�x"�8�r��P8E��H��P8 +G��pb��b`��r��2��g�V��CKAj�Rj=)��%&��@�@�/,� 3�Ypo7��91qkw��kwpӳ�`X�J��@` m�ml��D�z8 ��`O$�D�8< �L�`��L��[��V0��n�w�~�x@{� �lv'�Np�ڤ3��Ɲ�(�X*J�"��ϭ.1�fȓ��S�n��Kmr�67��B�p�6�P(\n��Ʀ��D�!Z��'��Ȁ�X2�c��{W}�U��#� Mf��d0�4Z�Z*b�i"J(�J$r�X +S,��"��^Zf��/fH�8<�?O�{{��Ͼ��J��h ��P��j�N��!GI�K�X"M`6oaq�B�[Zf޺}��gn߾K"S��kdd�> +stream +��C + + + ��C��"�� D!1"AQ2aq#�B�3Rb��$r��CSc�4D��s��5!1AQa��"2q��#Bb��r��?��-A"��R��)�C�*��Ciq��t�6��l�ټ�$�/�Zs��ډQ��IUv�Q��X�H;|��$��W\.��*��)Uj��(H�#�yoSs�HuC?�JW�z��Վ��6�:m�@�\��*�y�v�;�%�J�H�+�4��&^9�[J��_��Ξ׮�Q��J��:j*��x�:I +�qX��3��ۼH�Í�Kn�lɔܙ��Jp�Kska1�6�{FU~�ՙ��u^ʈ�*1�՚� ��ym%o��$nZ[�ܵ6��0FA�=�j��9lD��U�}:]i�^#��.�kM�+:�P��.<�N�?��+�s�~a��SF�Ӈ�q��'�Ij�Vk�+b4WN_��[mM��F�n{r%^�y�w�jeG�"�R3�}u)n V�% g�� ׾�]/�_Ɓh߉�G�RK�*�j��ƨ�%E�Ӎ�R�ӝ��5v�W�Uë��"� +R[_��6�R�8<��ڦ�%V��K��wm�Y��5�lʳn�-)a��w8�m�uЎ��"�麞��B�bاS��3I��cy.� +��8��Ԋ6�>��pcoZ�RXGʂ�d��O}v��4g�Hu��a�C�@�J' Oԝ7;��BAUä6n�H��6�v�jԉ?ġ�&Du�Ȏ֒��,��iՃC�8��*v2�BA��Ou�%��=s/Y�@�H�Hܝϯ ��S��4��'n��)�4��\[��e��%E��!��c��#4m�Z�3[��e-,��J�2?}r�{�d-C��}� �u�\J;q�5mMŔ�d�w�!IqKWo)�H{��~��p��t�.9��s*��P��d�c��W#��0\�$TBw��-N��$g�=��8�$�� +'��V<��ZA=��M!+JTR�k��Yf5 ��{\��x��AK�k)��0l�rJ��WFz�C�m[Un�Z��kL<�rl�J��$��)%@�I�4��]�r��|�:�,�D�(r)3{�4#��2�$��T�e�Q��4H�\��U�t�Ɍ��B +�e�R��,��a*H�փc�(��cR�S��zmQ �\r|Y+�R��[��`�(�7y��@�GS:k.��9�U"-}��2��,�G��nyT��1�h�F�t*�K�I�W:�r�y?Eqڍ��Ґ�quI �d%IN��R��S�$RMj�>#�" +uM'�^� �A\8QK��}S�o��v��T�f��)~��#�K ��O�Rv�Tv�2��L&5.,d-ť6�u�:��D�8��<��=�]��ް�ޔU��r��+��H-�Iܖ��T�s��3�g�3"��U9��P`)*R��c p��V7,㒬d�&��j��%r�j�6�ˉ$�)C� �u��z��|�q��Q��ё��ί�� +W�Ωʬ\��vD��T{�A��F��;��%+p��O��Դq�E� SZ��ɰX�)hlm��B�HHR}��#:��\��j�9?��R*� N5>�N�>{�A��>�Rp=Τ�n�j�rӗt��o�JDw��T��IJwv�h�)hXL��׫�A�U��B�X��9�Ѫ��4�=0��C)r;�m�] ��t�J�N��J-F��R��HZ`Rrkv��'�+�Q8��V��D�bʺ�i� +��6�C�x�8N]P)G)I;'�F��7��Z��uD�3�j��RP�� )*k;I�28:� d��׹��]z=L�6n��eb�g@�,ZmA��϶��r^��<$�KVK��B@ �=�n^�t�u�2��' 0ϛuVU ǎ�|�ܬ(�+�ie& .��+��dTm�QcI��[!n1��d!H��J��N��RN|�o@u��(�q,7 ɤ��t��5*،��˓R��"\��q��e�r�z,eX��V��'�f]>CS!�@u�,8�͑�ҡ��糌*��?Q<(m��Ǯ�:a�*�Cd-�0�f�\�b\%�r��?=��ͬ�]k�f��5^p��H�W��}f�13�u�C`��g��B��#]�ܠ�9�|p��آ��Y''q�9��9��~u�N��9O#��H��n��*k��#�ö� mXm��u�)�008�t��ATZ�:��(Q��ճ'6ӵ\aT��*�r�}�̎��U�.��J(��r�qY��^�%�V!l��6|B�))�$㌌�sY+�u�۟��ܛQ��R#�C�>�m6�_̵�ބ��P�U��2�7�J'L��_f�j�M%��oQ;�PRе��*݁�� k��l��b�U�3puΖ��Dj�<ŦEfQ8%��<�Fk��F�Ĉ��.J� xuH�PHIn�0;��a�к��x^�¨�Q�TҊdu�C*��7�܀P'q;p�n��wKؿ�ޔ��z��&U!�� +ԄFH(e�<��w0u�c\E�M��^+�9�ӟ�/�&�^��)P-��s�l �\�Ԑ��*ؤ��l��Y6͑"EB�!�Չ�Sҟp�QWl@;8R��n*�Fb�S)n6PI'�b:�Г��)G��U4B��FZ��q��܏��P�Cy�u�A�'�[`�F�8�D��8 �E�u)6� +�HE~�=�4�rmEI�8s�q^Q�/)'��̳��m:��fޙ]y�Z��q�NԤ\s�#�=o��{ +��m��*�* +�/j��!-�8�盹 ��> ��~��8�nH��T��&�k[C[K��(�%X@H<��D'w� ��o�e0K�)�G\�ONv�J�۩�_qK��S��%�}9�u��3�[��޵EU��|��T�ťJ !*q��nO�i�G�V�s�l$��c�'��]��4�x*Z��A!?O(�a��EJI��M��oH�z�Z��ᮭP-0�� (5��+�IH�47�N5%X��U%b]SbMB�[+ho.�S�[�RT2��J��_�S�*5�N��]OU�R��Ĥ$()Ԓ�֕%iO� �V.��U�@��"R_�ӑR�a��B��< �8(9Aτ�s�@'��M�� b��>!B�<��H ->��:W�[��T�%Daa#�P;B�)#�P��I d�N�}��4Pհ�!�O.��T��K��ӑj��B�ڡ��Q��"i�K^B��q{��6�;㷾u#f�+�:�|� �C�I�9$$�c:��~n��]r��!�ē��#j��MlC�K3��I�]��P�}��U�!³j�U�;z�s^�m�rV�m��r��) �H�)%-��{��:�]H�c��ϟp�h�ܢ@��Ǭ:��K��i�p+TU��K^v�=��UV6��Bh�݈U��z�ܠ��N�*s.��52��8W��I��`��h��Z3�D��T�S%:��q�{{�ׁ��q��6bCj^\+Y�H$��4�7Z�hb�jl��q� 4jl�n��9͝y�Zj�2��~Az�m,b(F��)�dGOWj��Э* P"�L�5'�JT$>��V6��O�=�ԕ�VTy�'��Qt + 2ѣE�Q!�PV|B�J��r�_��X��Tkb;2_iR�q��q�RR��*�Q�O�t�euƀl:* a�r�=`�P*�$�j��d�' +��!?Rp��v]�r�.�e�il5R}u�4i�r��J@y��:��yN��х�H�t\u;��%�d��9/�iq��o�q�v��{�X��U��UbK�I�Fv�P��|C�y��,�!x�$+�NRF`�3��IɞAM�[��u;�\�J�w�*Q��*�Si�$�دH�m7*��m-�P�~"J��eI��)JI) -��P��WG&ƙ`êӜB�U^~mQ� ��n�\YW�I��Mq0��U�n��Իn�:�W��sS��%æ0g=DY*��(s%>��I�o��#\�oO) �1v�PT%�#�!%�ɪHBp�� 7�s��ٴ߁~\j㕈��*�␥>�nA<�(.��ft��]�6�xQ�RL�ñ�TH�C��Ѕ$��̀��y��C�/&�xh��p��Z�A[�c�~Z��LE1P��K�KC)^\>��g��O�D��1�� `?��hI)K�d�*mC̅`��h�ّ$9n\mS[�-��nC3�F��d��^��BQ��T�� W�=��'�h5X.��:��*H!!� +eD�PؠrN5Kx��d7�9��Z��=��$��|5O��&�w�TձQJ�ݴĖ�4�@�� A�V��L�-t^�ٳ^j3UXie��F}`��(��R\��s��ආR�� )a.)Ԑy��3 ��gy��Oĩ�W�Q��J�vX��ی�v"C��}5?[�n�Lf�"�C.C�Z� %Ă�R�a8��h�SΦE"�U��L��QP��ww�#��*��\J��/'��C(2�eX� �N�w��s�${��|4࿅zl;>L�Hm�䨸�!@�TA#KˆMg��h��@f�A��>.Ş��v��N�N�5[Z��?�#Q�Z�I��i��u"�锦��嫲�(�~�[��=r��]�T��;�s4ǍXm8�'�7�'�BRa��r�]V��O�LKoHi�)X(u�$��Aη=Q�Dm.T�0!!@m2f��P�nPΨ�=+��~dz��E[��縉�-�JP��Ԕ�r��0I�=�p��泃�;��WgN��lJ��iqY>[JI��Ҿ5w�iĭ^9ZI>e}��Ͼ��4��b2�h�#��)J{��< ��蒜�+�M��H�q�;}��[�:&��y�G4��4��n�a�ιm�$��n؊W1��v��1��jt(��3�uA�p�l�g�c��d?��i.-.�R�e�]�~O�Eػj�)��b�Z g9PN�@�s��ާ�0��I��A�[ܟ �Sd��@y��h�S�}��4��Н��6�I�j��؝�(t�D_��?@fc)(�"��:{-(.-!9�$dd�=�U�U�l8L]��QM��*uLĞ�M�푟��J��FR��H�l��ޝB�B�)tx�uD.\y��BZ��8��$nܓ�yI�3!b�S�͊�bY�N��|�T\B�׽ ��8md1�k#Dlc�v��^8i�<��4eͿ�~��u]Uۯ��K��V�z�2��ȑ�xQ\J[S�~#m��Xp��)@�ȷ��7�o��R�IBJ�Iw ��HLr�� +��{��J�GU��ջ��B]�I(ޥÎ��M:Ҽ��zK�_��]�Sٓ��\F�,� b��m �z��m?��E��`��m�H�0@�l��G��H�U�飕�eN�B��O�;�㌁�KzE�2,4υ +Jk�%�KiN�n��n:��ec�ڒ�� Q�TV��b�rW�RBS�;N�nu>�W��2�'qF)A#)#rϡ��q��̶ӗb0�F*.Щ��M� ��F�JXz�2�@��9��rF��q��mi��hu'�$�S��T�b�\f<)m��+J�͟ud�8��5?g�\�U��-��4%��V�k��P�:G�F\O��A��x)bkb~_�v��5¨��,�RT��\q[�YW�q��$+?S��S\�5)$!��aE@��=��:�!�j�/�[l�Q�-m+ +�� +�NFBF�k5JUJ��íD|K�a�8��#錏}l0��ӧ��.�Ӹ�̓ V��c�Zۉx�,� ��$�>��m?̮u��T�`g^!jRFJx��5�$��GWR��da��(B�W�e�6��^��,u"��\PJ��^0bv�jӳa��<+v��7�� ӥ�t� rl�n�L�#�� +q )I'o��M��%��|�!��O��==Ҝ��qΙ]6��GfAGÉ��Ұ1��W� +,T(V��D��M6�9*�[�-e��n£Kq!I9+۱ ��s��r^[Sy*�3u�Ҽ��v��.{�K��D0�)G��A��B�̰�(�P}�c+I��K��}ҍ�ekސS[�ӻE�%I�[}̟�Zԥ��?ڤ��,��$�P2��В{*�<�<��J��'�`YF�7�t��zm�!��qd�3��ǯ�L�[S)� ��'��P�?>�5�Ƌ�&Lvr⊃��q��@ʆ}�_��Ȍ��W��e�Q��ϙ��N{�A�N+�;�T��Ȳ�@�F��sۚ��a�mV��i'K�ĵ�+�{KSg�+^m�wˊ��.�i��vT&Rd2�xJJ��Y�JH�� +#�`t��b.�P�yq��+FAa ��c�`H��:VG3��5W��ӱ��rG�E�n<�1ɷ/��Zx͔�Z�8B[��@�d��y7�Pe�@��i��х'�'��\0�oMhI��[�\C� R�`�#)��p�WOyrXo�$�O��s��z��>N!>��=��e0��-k ��Ӛ��jF��<�S�X>܂q�ٱ�ieŏ -�J��)$��4!iªB��)1��z�Ge!�m=�kiJ$��%'�N�r��~��4J�S�u;�*Q<��☎� D�$8��:��9�Y� ��T]��OBx�P-��Ɨ�C��j��w�6u��A�5�k�H��ׄ�2l)�(��JPK�@�J�j�zk�CV��G��* %�� Kym� �r�8NJB�TIN�7Q��u�kN�\3��L��a(�L�Tv�J!��#�=�0�[��h��R\\��$A��DiN+�F�B� 䜒Z�٩��8�+��bm�~{��U��Ԏ��d�N�Dqt��# +\z:Ӻ9e�)�ZNxHh��Kn�۴�n��³㦗mZ��є(.D��)Kە)8ʔ�FG藢u�F�-& +� ��l�ʾT;�~�!��ɈW?Ok��V�ɹ��Rh4V��L�ͧ��H;��Je�S�z�6?�B��q��y.�� Q%��!��70�҈ ��}� �V=�U��A��VO(8%AX��L��g�ʬÏ!��C�@*h�*-�o��;�?]2�G˟�� *å#�HN�S�[�Ѱ��dJ��.mS��;A#i�$��t� ΦScjm*M(nʷI`%��v�_}:��f�eD�_Z�fR"�u�l��gӾ1��{�<�=�b�l8��W�q��(8��Z�m�r;g\wYC;�Øn5F�W�9#�5V�a;�#�N�g*�l��R��}>��4f�S�G�N�p��ێ~�<��솼��L%�27�'��K~�:'��u��*�^� ��Br��ݕ凓�x!8�s��~�Y�ƪ:2��t��<ԓ�5M�=B�un�!䶅��A��m�mMn�9�;��q�{r��Z\j�6&��[� >�q��5a�s�ڄS�h�U��kR��]�N~DR�:�G�RI۷c!J'<�3�'z�j'g�*�u9�v��$`��t�RD�Gh�P�Fl%;�w�G~3�8t�� +1�J�<��H�e�G�_�� +��I�� +�Z��W��·o��w��[^"�y��ڻ�CgМ��Qce�q��T��|#�V��<,@�Q�i) ��v��gL�t�;��K$)X�U��۟�J�=�$��6��B�a�aΜ�a2l˞ձOAe��x�I��mc#��#�e�T��n�%Mt�Ș +U��(Pb����n��-(�ۜwj��򏶤�'W��eʈH#ht��>�SOS,�-�.��C]9��6��~�#��,�7��%�N��pŦH�5) +��y�0�en��Rr��.aDd� ��jb��Q�I��T��>��8�P)�BN0��貃MrS.�$F}��(ol[BJHIZr�}G��%��y-U�r[K�sr�q�5⺮O��I'��'��!��d4)�\��2�T�B��^ h�)P!8 1�@��Hj-�n��D4¨��@��U��ݷ'y� +޻-�c�%�;��ŭ��]9�%(��멉��njZ�ճ&M[$�[�%r��r=��|O1+��9�3l�b5�j;T1��)RZ[y$��p|�܃�J�~#O�@<}O�=aa�Z�d�U�ԗiWm��=��)�@([��G�s��Tzk��B��(�2�-JJP�q��8�?�c��D��U$~��܎��U[A3i�e�쪕N\��_�R�6CH66��ߟA��wT�~�6d;fy��x6뮥�n�ӎ��2��t��fr�*�� fTJ��Z*�'r\o�!*��;��n��Z�.3� >[��IC�[��kh�)y e=�� O�'0��SJ'IХ�}´�rm�@v5Zt@JJK�'tg�Q;��G��' ꄴ�2�h�'�O�F��p�Чv5z<�s�Z�m)�(�G��lT?~Y�~$��W.<�6�!*yII#��cW��I)){�]��;J��`y��;]t��W_Sh�ʓ�HA��!m��#͑�s�R4*u6�PW��ò�H!;��@=��YF��_Nlt2��f[� $�¤�>��q��cp{��|�,ZF�!��6a�9+m�K)ڒ@Rs�N��r�nP��L��~_�\��u��ۆ�4��#gR��:3%Lt%��hl��V�I>�G�Կ��?��J��b[��l�'�3�ъ��\o�H/`�i��=s,�6�~Z�ӵImR!<�p6��Ь�m +V�H��P��7��;�|4��̕n�<��XPQ��x��Y�%n��a$%'#[��z:��Kl�ܒ?�_6�0ܝ��0� �('qJ�H�[ԩb��7mr +ǆ��Say + ��϶��Ot%�h��̃%�-#�ڂ��-��Η][��ҫ�}�Kj��k�c�X��m{�Q�Ҧ�J�I�H?(@>o� +��u��E� +Stե�N�ߊ�y[�6��# t��$I8Lt��ߤR+�JS��E� +��P��Հ��I��к�dZ;u��ߐ Ԑ��oP�_�2��y��vlN0��D�۔��^�hT�RH�w@�eL +~��߉.Kʠ2��|4�(%�r��%Y9Hޏ��т #ל7$�(�F��JMZ9S��4�R�)9#��<�h�/F-�ĕ�+��aIe��6�N;��q�� d�Y�}-��%�jFq��Q��ID��()i�|V?QyẂ��p��4UK��Dn�0*V�zCL�r)��l7%JFJXB�Ӝ��A ��y��lP��j.t�S�I~��sm!yX�1��O��ֽ25�pؕx,��r�AC�nK��YR�}�ӅY��s�n�hHb �F��=��^ѽE� +Ҕ��n>��u��i�� !6P�� L�𜺐�>�}��|u�g[鼭wl[Z:�g;V�"D`��ҷB0��mJ1��NrF�O��T��2̦��{�{ +�^��B��>��& +}&+��Z!I��P��@�H>��4ExG�)T�z�EF��q��]@I;��C��={�Ҙ��V�-£�o)`G�� M��.J��6=pO�&��U&�C=Z��R �6��LOU, ��:��p��ມ��$q��=D��@��(�O�R��4�)m��|!��n��J +}��j��[�Sa�GbΓ�5��^t:��ϰ��K)��w�� +�rB��1��_�dR�)'i�:tařd�#� ۢ��nZ�&��~�>�ܘ��Uh�3O�)f䨨��I�� v�}�2��%�X��Zq~Y$�*H�8'�j'�j;"�*٥�,;U�M�bG�c�C5'Ԓ��nmN)#g��$�E�ez��~=k��-�%IU:�\5t�a[Q;��d�8︍*��ܙ� ;��yQQ'��{�^��ǧ��JM�~j�V��q��m%��>��t��/2��wQ�n8X�LK�8�Z�P�nNO�@�F^PJׂ�RR�2��#��}YO�-)�=(�U�IR��2Y�G��ZI��sL�u�rZV��y�Z\��!X�;��Z�� +�A ��Xx��+�ھ�s��Q`J]H +O$�q�kb��>��d�$��ӂO��E�6�4��{]��a �Z�� V?�Җ��\iQԡ �$+��W��,��n\��8!��~S�V��F�R'2S"�6��J �p}��5Ź��+��~�4�Rѣ�4��l�a;��}�tߥVi�d��eDu*$�XZt��ܩS�*|9Cm�ʢ�ke7��O9��5��l��\%R#�R�٣��eD��NJ��Z��x�!{�x~��5��n�Z2-�P�M��e��f�9��/pギ�ƛ��lJ��s�'�4e��R\\ +FT�i$p}N2y㶹o6q�־�Iz,�m6L朎�FCM��Q%'#��t��m�5�=U�R+Wh��'�%��%NH*Y89'CQ,�� V�æ��?�u87�%��Ɗ.��R�V�ݘT��!��@)��=��>��*ח48��HֵŬ7�.^�UM�s߷3�Y]J��X;�2v%#��:hӥKB��TRO��s�饟J6ݵm�"Bi�-�Ò�qj'�*W�Ӈ�*(��i!�V�>�I#}1�i��4� +C��Hm:�QP�n�ek��I��!"�ʚV�?��8��t3�8k�Kr޻l�$�WUxǛ��!��(�,�%i��HtϨt��}=jc1#�b��.6�[q*J�xR��U�ޕ�h��΢\��Y��T�u�H�m8�͕��V�m9��W^��*��e�s�k�jre��ə�!?�e��'$c +BH�׽2�Ńj[ݹ��d�IJ��"��eYSk T22��;�G�Tڕ�o�H�k�j�4��5�v&4ъ�� rN�rF�}G�:M:��V��U��^��uiAR�xS��g!$��&V��@��~�Ҩ4�Z��r1�C~J�x��PZS�p��r�9Ҷ�*�D��Lx�F��yj��{� +��2��w%Y$�5��)/��Vy��Ӟ�·s��Ԥ;�k�FT�� +��B'[�G�Q��r��IU�)N%Hy�kP@�Tx{�2��r��\��"B��TZ�>�E��B�E�=Ȓ��G�!gy�Э��*�[2�Μ�+S�+�"��Т[�^��B�� %�b�%>�Vrx��s�Q��UgI��i�E����@�8��;��uY䎜��6��_;a��iCV��*��y��#��J<�b�B�X�OG��e��ׅ�Y�9�Ǿtw^j�ӭ�.�Эȿa|(��WN8%J�>�i]q>uHd�(R��k�"�%.Q�+15uY��@V$��T�)�$3 �k!)^{k�Vҵ#�V�؂�F��x��H.(�K�d��M�VR/~�Gzsj]�dU$!I�\��#W�)Kk$�C�I��c�!��;H��S�y�kŴ��I��P��rs��X8F�y��C�U�T}-�`��V6��MN��~��i��2ޥN�\��Ri��.�}͍�>�ܫ�$I��=��K��/1҈��?)�Nk�gզ�|$�s��M�k��Sf��񼺇k��0~�Ť��e�3�c ��}uI:�sR.��xZ�Z� +�)��ARՐV��J��Nx��G��M�Ҫ�+.�Y��X2d�AR�Y<��eg��+�XS��A�ʒ��Y�%@��qTG�m�j�x��Hg�f=A*��R��\y��jQ@'��6�8=��!��[ɥ|Ja�0�Z�-'j��9�7g+�O��/�v�Z]�p�ǩ��n�FRP��9ۏMvS$)�6�*mh^��>�]l�1�Ő�I��\�L7}�sX�j�L��E��R��~(S��<)>��Q~]�km��@Fr<��~��=#�{<�3R��)�xal9$,4��%;J]?��q�� >�R�UN�Pz�16k�/�m�̫��I�%n��@9;��9�sC��!�^��Zm��g� k_\: �c�2�3�銐?�뭵��0�MؑQ�&�fT��u��ԍcV��-�u��L�V,��ͷ�J6�Rq��;s�\�c=l�|��K(�"Z�Q�5��x��`��Gns�@܏�uҞ�� ^2Zr��R�9�0W�>��/R�1+r!� i�%ĩ�5gBZ��ڔ�p9+x'$��=��80(��J˪�>#��m��! B�H�oq��&�n�z�.��8j,�.�L��-�TGYx!� +Z�m.�¾��n��l��+ω�_(�_-. �4�9��HVn|B�A�]�Q�9rS��3$�~嶈;~ey��TبΕ��&u2�C��X�ۻ��c��Fw]�V�R�H�ԑit�㈊�6�O�!N��JS��M��*�çЩ��h%�O@Qu>��1�%'�9PCF��u&�ۚ�l��P�&�b�L��[�!��)9O��J�pR@�t跛j��aV-6�n�(K�K d�2��!i��w��M�'5G��[�?k��CP�Sm�򴵸��I%�1��}��[fɀ)U��K�k2ުVY��)Ha+20��m)m ��q��,�� +��b��aS)��9�NB�d'$�Q�@��s'ө��ڬ1�Tj$5+��1�s��^8�|�� #u�qU%T�$J9q^T�Pd�A�t-��*��3�Ҹ%�> +c6��RBpO��<�M;.B�C(�rT�@�'�]��RV�TN0Gr~�ΥzP��W:��Rڏ��t<�� ;�* +)O�$}��b�v�9Ӡ�V��s��m�5j?�:_M�uV��ZX�Uʇ(uY�� #�u+MO(a#�2q��3�6�R#=*ʬ�XKD��)?��}tD�xj!@v��]��v��Y� ;��[C��m��L��c� >b9�Nt�BUY�t��\��o "�$$-�̨�=G�M�:�฿1��)��6#�Vӽ#̴�D�L��B�Mn�P�J Ga��ܐ��Ґ9ud�ǧ?m.��U%P:p)-8�]s2�[O��A<��u��lN�}A{�;�;��zf�"�(�U^J)��Ϻ�0x�E?��Spuf�m_��L;j"��~Bs��;��:s�~�zv��y�r\E^#�*�7a}�6�$�w+r��)Nr��NxΧ�q��+M��7��[ي��JB��,�ʎ�ES�g��8�9�{誇��/é�s��R�gRHPX��Ǡ��H�9=��V:��VE�hJM�چ�nb��W��ck��z��;�P�� [��Ek��[�,��jXJQ��'��F7]�B��-£�R�Ū��?��B}�t��Jm5�TZ��O�@YC�!+�'��s��s��[HLw"��r0�=��D�!e�6;��>�C�Lv�6�F +C��KI �:��p�1�D�u�MF<�5Ժ�X�2��!��@�>9Y� +��͙d>�D7-��!��@'"��m(/;~��O�%>;��q�.�$��N;�C�j�E� �[n�E��(�u��:�n��r�r?s��˦d�ɒ�RJ`ҁ$��ް��J��MΑ*2�8��P��]HA=�c��:�6�N��D�"*�9⮣TS�F?M��'�J�$d4q2�.��d�1�χ��G��axϺ�Ny�N,|Z<6�7�|��o�N+��t�A��f��RK,!��Ciʏ>�S��Y4zEB�Қ��(}i�c��*풞41Qr=�N��z|�%��r��Wʕ,�Ԓq�}uG��o>��q��W��!+$�a �� 5IwF �y��+S IK#K��[h>]On��5�$��"��1�7�T��X��ⲑ�ܐ�@H�_�g��q��)q��JS�T��؍��B�u),��Bk�s/� b�� _��^�S��K��6��O��'��DA�f׹��lӧLE@�&Q��@}Q��B�FH��E�>��+�"�h_v�`��n��>�EI�")�d9��Q��B��1��,g��[��8�9#��Pt6:��7d�i��F棡��^�/ջ��V�Һ�w�5rѩ5S�=��W�Gt(v±��u<��ws�� v�Q��9�h�#+),��c9J�|�vH �U�C�� d�&�낀ԛ�Z�J�kSS)��K��N{A��δP�ʖ拼s[I-�e��y! ��:��E7�49y��iE��2�O�Gs��B��#M��LjMje."q$f��꣒}N��Qi��D��j2A��6�=��TNIQ�IԒ��L��h�- y��A�u�XP��?�X6�)>��ppFF��bHHP�:�(�99��N6�n�p;s��P�s�8�sJ�:��{��vWc4�]��SC��;G*O��OmUz��56�^�R�+��MzЌ��Ã��X�L�q��2�ǝQ��c +i��Я�I�?]^��hu +��6Q��r�QWJfM>Sr��Y-�(��x�}tr��W�I��i +DW��%xP!>P}�F��eJ�t�� +K� +~�=�<�8ؤ�̓��}"�tڭ��7�kv�k��ט/B^��Q��u��Fe��X��j��)Hl��T�J>��Ý5�.��R��1��<)�Jl�R�O$�`}��kШ�Q=��IABڞ]J��iI!*��h�$��H��%<�d��Y�x�,�$�1N+�6=ObW��S� ��:�q��~;�:+b��Y��_� �$��e>�%H�g�+u��{$�w��Z߸%�-/�S <��n��'�,��RΫD��[b�5V�J�j�3 +K��KO��%^P2N1�1�P��~�\j?�~It�ml�g% ��I;I=��$�v��=J�#JB��B֖BA$q��45O�e)��p��F��R��~��T�i-�L��怛v-*��iI��4%+ 䕓�O$�$�ФV�[��P#�HXuͪR�V8��ǰ�\�n�e2j%��+q��Yi�o�2�0��Ld��%�j~��Zݹ��e6%uԃ� � �U�~��1��o��za�F��-ǆ�Kq�,�RR�A��%��v� ڄ?vTشi�+z��C�� ՗��K->�DZm:Z��R�JA�e�=w:y� R��n �! Ac�D��&C�n�8�R^b��=�t�2��?f;q�K(�C{ԁۏ鮂� +9I=��(!��mJ��Y^�]�ȫ2��(�J ��G�[ +�S��Omi9l`p3�y�R�Q�P�e"�Щ��D�� `��c͌��~?l �.�T�?4��R�R�2י �[s��p� +{�s�{jĸ��?��q ��|�熽�.:)!�H_Čتu�O�`�] ��2vʌ�|7�(q��Oq��hz�gRks��$�%N��c�8�~�A��~�ȹ6^6 GP)��la �c�*��W��Wrx'� �~l+�N�|T[��9��r��F��>�爜��;��VÉ��0�#��W e*�8�[x$`�8?��I8<�9օ`�)e��`�F��?��O*Ny�X8��\��u� ��>��?�� b?�H��U��8P��y��*JYٻ#�ی��Y��gݐO$��x�=��X:�}��-�A4��pB��'>�҂�,��ג]�d§�IY~��#b�� ݫ6m�qі��o&cD``�~��o��9�Mle# �|��}NʉY�Q�&;p�L�ÿT�R)�[R��E8��q_\��Ў��!�ȷ(��r.&�V1��3�5r��'��W��9�7��1#�8�q�֐R3��K�Gn%�T%2��p9?]v��[��U8�E�yJ�8�'[T��m��\��h��]dם��w�Ϯ��e�J<��h�^d�T�s� ��5�T��5�ܶ�{g��|��*�r�g�ΰS�O�:��}��Z˸<�4�y'��8#��5�M��ݜs��^,*��r��:@W/Vեh��r�Z?U��\��7M�UT��p��$�9�m�Y��=��\/�^^@:p{�tJ�u��endstream +endobj +61 0 obj +<< /Limits [ (node00000193) (node00001539) ] /Names [ (node00000193) 8 0 R (node00000220) 85 0 R (node00000221) 86 0 R (node00000236) 87 0 R (node00000245) 88 0 R (node00000756) 89 0 R (node00000757) 90 0 R (node00000758) 91 0 R (node00000759) 92 0 R (node00000761) 93 0 R (node00000762) 94 0 R (node00000763) 95 0 R (node00000764) 96 0 R (node00000765) 97 0 R (node00000766) 98 0 R (node00000767) 99 0 R (node00000768) 100 0 R (node00000769) 101 0 R (node00000770) 102 0 R (node00000771) 103 0 R (node00000772) 104 0 R (node00000773) 105 0 R (node00000774) 106 0 R (node00000775) 107 0 R (node00000776) 108 0 R (node00000777) 109 0 R (node00000778) 110 0 R (node00000779) 111 0 R (node00000780) 112 0 R (node00000781) 113 0 R (node00000782) 114 0 R (node00000783) 115 0 R (node00000784) 116 0 R (node00000785) 117 0 R (node00000786) 118 0 R (node00000787) 119 0 R (node00000788) 120 0 R (node00000789) 121 0 R (node00000790) 122 0 R (node00000791) 123 0 R (node00000792) 124 0 R (node00000793) 125 0 R (node00000794) 126 0 R (node00000795) 127 0 R (node00000796) 128 0 R (node00000797) 129 0 R (node00000798) 130 0 R (node00000799) 131 0 R (node00000800) 132 0 R (node00000801) 133 0 R (node00000802) 134 0 R (node00000803) 135 0 R (node00000804) 136 0 R (node00000805) 137 0 R (node00000806) 138 0 R (node00000807) 139 0 R (node00000808) 140 0 R (node00000809) 141 0 R (node00000811) 142 0 R (node00000812) 143 0 R (node00000813) 144 0 R (node00000814) 145 0 R (node00000815) 146 0 R (node00000816) 147 0 R (node00000817) 148 0 R (node00000818) 149 0 R (node00000819) 150 0 R (node00000820) 151 0 R (node00000821) 152 0 R (node00000822) 153 0 R (node00000823) 154 0 R (node00000824) 155 0 R (node00000825) 156 0 R (node00000826) 157 0 R (node00000827) 158 0 R (node00000828) 159 0 R (node00000829) 160 0 R (node00000830) 161 0 R (node00000831) 162 0 R (node00000832) 163 0 R (node00000833) 164 0 R (node00000834) 165 0 R (node00000835) 166 0 R (node00000836) 167 0 R (node00000837) 168 0 R (node00000838) 169 0 R (node00000839) 170 0 R (node00000840) 171 0 R (node00000841) 172 0 R (node00000842) 173 0 R (node00000843) 174 0 R (node00000844) 175 0 R (node00000845) 176 0 R (node00000846) 177 0 R (node00000847) 178 0 R (node00000848) 179 0 R (node00000849) 180 0 R (node00000850) 181 0 R (node00000851) 182 0 R (node00000852) 183 0 R (node00000853) 184 0 R (node00000854) 185 0 R (node00000855) 186 0 R (node00000856) 187 0 R (node00000857) 188 0 R (node00000858) 189 0 R (node00000859) 190 0 R (node00000860) 191 0 R (node00000861) 192 0 R (node00000862) 193 0 R (node00000863) 194 0 R (node00000864) 195 0 R (node00000865) 196 0 R (node00000866) 197 0 R (node00000867) 198 0 R (node00000868) 199 0 R (node00000869) 200 0 R (node00000870) 201 0 R (node00001070) 62 0 R (node00001466) 202 0 R (node00001471) 203 0 R (node00001472) 204 0 R (node00001473) 205 0 R (node00001475) 206 0 R (node00001478) 207 0 R (node00001479) 208 0 R (node00001480) 209 0 R (node00001481) 210 0 R (node00001482) 211 0 R (node00001483) 212 0 R (node00001484) 213 0 R (node00001485) 214 0 R (node00001486) 215 0 R (node00001487) 216 0 R (node00001490) 217 0 R (node00001491) 218 0 R (node00001492) 219 0 R (node00001494) 220 0 R (node00001495) 221 0 R (node00001496) 222 0 R (node00001497) 223 0 R (node00001498) 224 0 R (node00001499) 225 0 R (node00001501) 226 0 R (node00001502) 227 0 R (node00001503) 228 0 R (node00001505) 229 0 R (node00001506) 230 0 R (node00001507) 231 0 R (node00001508) 232 0 R (node00001509) 233 0 R (node00001510) 234 0 R (node00001511) 235 0 R (node00001514) 236 0 R (node00001515) 237 0 R (node00001516) 238 0 R (node00001517) 239 0 R (node00001518) 240 0 R (node00001519) 241 0 R (node00001520) 242 0 R (node00001521) 243 0 R (node00001522) 244 0 R (node00001523) 245 0 R (node00001524) 246 0 R (node00001525) 247 0 R (node00001526) 248 0 R (node00001527) 249 0 R (node00001531) 250 0 R (node00001534) 251 0 R (node00001536) 252 0 R (node00001539) 253 0 R ] >> +endobj +62 0 obj +<< /ID (node00001070) /K [ 87 0 R 88 0 R ] /P 8 0 R /S /NonStruct /Type /StructElem >> +endobj +63 0 obj +[ 93 0 R 86 0 R 95 0 R 96 0 R 98 0 R 99 0 R 100 0 R 101 0 R 103 0 R 104 0 R 106 0 R 107 0 R 109 0 R 110 0 R 111 0 R 112 0 R 113 0 R 115 0 R 116 0 R 118 0 R 120 0 R 122 0 R 124 0 R 126 0 R 128 0 R 129 0 R 131 0 R 133 0 R 135 0 R 136 0 R 138 0 R 139 0 R 141 0 R 142 0 R 148 0 R 149 0 R 150 0 R 151 0 R 152 0 R 153 0 R 154 0 R 155 0 R 156 0 R 157 0 R 158 0 R 159 0 R 159 0 R 160 0 R 160 0 R 161 0 R 162 0 R 162 0 R 163 0 R 164 0 R 165 0 R 166 0 R 166 0 R 166 0 R 166 0 R 171 0 R 171 0 R 171 0 R 172 0 R 173 0 R 173 0 R 173 0 R 174 0 R 175 0 R 176 0 R 177 0 R 177 0 R 177 0 R 178 0 R 179 0 R 180 0 R 181 0 R 181 0 R 181 0 R 181 0 R 181 0 R 181 0 R 181 0 R 181 0 R 181 0 R 187 0 R 188 0 R 189 0 R 189 0 R 189 0 R 189 0 R 190 0 R 191 0 R 192 0 R 193 0 R 193 0 R 193 0 R 194 0 R 195 0 R 195 0 R 196 0 R 197 0 R 198 0 R 198 0 R 199 0 R 199 0 R 200 0 R 200 0 R 201 0 R 201 0 R ] +endobj +64 0 obj +<< /BaseFont /DAAAAA+Helvetica-Bold /CIDSystemInfo << /Ordering (Identity) /Registry (Adobe) /Supplement 0 >> /CIDToGIDMap /Identity /DW 0 /FontDescriptor 254 0 R /Subtype /CIDFontType2 /Type /Font /W [ 0 [ 722.16797 0 0 277.83203 ] 37 39 722.16797 40 [ 666.99219 610.83984 0 722.16797 277.83203 0 722.16797 0 833.00781 0 0 666.99219 0 722.16797 666.99219 ] 68 70 556.15234 71 [ 610.83984 556.15234 333.00781 610.83984 610.83984 277.83203 0 0 277.83203 ] 81 83 610.83984 85 [ 389.16016 556.15234 333.00781 610.83984 ] ] >> +endobj +65 0 obj +<< /Filter /FlateDecode /Length 317 >> +stream +x�]��n�0E�� +/�EdcI%��8AbчJ��b,�,�� +O�J]:�ܫ3ٜke�~R-��=��+�W�%��ڨp��Vc��ͩ]�cc��%��3��ͳ��%��k��ts�햰��7�`太��0�ҹ�nʢl�h��ew��_�� Q��u +|g %�W��뺮X��|��k��:�ӊ��^�$�H�@:##�鄄�uY�H�j$ܽ�_�GP�G�'�o��G,J,�c�zb�)��"��<�LW��?{q��A��|,@ݼ��q�;�&��C��endstream +endobj +66 0 obj +<< /BaseFont /EAAAAA+Georgia-Bold /CIDSystemInfo << /Ordering (Identity) /Registry (Adobe) /Supplement 0 >> /CIDToGIDMap /Identity /DW 1000 /FontDescriptor 255 0 R /Subtype /CIDFontType2 /Type /Font /W [ 3 [ 253.90625 ] 15 [ 328.125 ] 40 [ 721.19141 ] 46 [ 816.89453 ] 68 [ 595.70313 0 0 663.08594 0 393.06641 576.66016 0 353.51563 0 0 344.23828 0 689.94141 635.74219 0 0 520.01953 ] ] >> +endobj +67 0 obj +<< /Filter /FlateDecode /Length 296 >> +stream +x�]�Kn�0�O�e��l\H !�$H,�PI@�Z*�2΂�W̤��H��Ǣ��0�"�3��4�+�ֱDqcu��zh=e}l�)�P�ndyι��N1�|u0��L��.e�f��y� ��%+ +n�c�|m�[;��\�q�\��o�y��:�n�h`�кX.��ϫ�� +��Ե�_m��OϥT�@U��QiF:�^Pi��&�)%��=�jn�f��v +��~�z"��f�n)��2EO�Zκ��c��[�f<�֏~I-��endstream +endobj +68 0 obj +<< /BaseFont /FAAAAA+Georgia-Bold /CIDSystemInfo << /Ordering (Identity) /Registry (Adobe) /Supplement 0 >> /CIDToGIDMap /Identity /DW 1000 /FontDescriptor 256 0 R /Subtype /CIDFontType2 /Type /Font /W [ 40 [ 721.19141 ] 68 [ 595.70313 0 0 663.08594 0 0 576.66016 ] 85 [ 520.01953 ] ] >> +endobj +69 0 obj +<< /Filter /FlateDecode /Length 254 >> +stream +x�]��j�0E��Y&� ��uF^�A�|�"�A- Y^��B3p��.��觷r��6��j7/n8kC��A)�E8B��2�k�e0�%m@�p�k�;��H�W赙�p��#��.h0�9(��ބ{M�Ӡ��ӵ�:�w�P&.�5�*\��腙��1ơ��W��6ɻ�|��2V�#7�фd^z �� >�u�EU�_4!zendstream +endobj +70 0 obj +<< /BaseFont /GAAAAA+Georgia /CIDSystemInfo << /Ordering (Identity) /Registry (Adobe) /Supplement 0 >> /CIDToGIDMap /Identity /DW 375 /FontDescriptor 257 0 R /Subtype /CIDFontType2 /Type /Font /W [ 0 [ 1000 0 0 241.21094 ] 10 [ 215.33203 ] 15 [ 269.53125 374.02344 269.53125 0 613.76953 0 0 0 564.94141 528.32031 0 502.44141 596.19141 565.91797 0 312.5 ] 36 [ 670.89844 653.80859 642.08984 749.02344 653.32031 0 0 814.94141 389.64844 517.57813 694.33594 603.51563 927.24609 ] 53 [ 701.66016 0 618.65234 756.34766 666.50391 975.58594 ] 68 [ 503.90625 560.05859 454.10156 574.21875 483.39844 325.19531 509.27734 582.03125 292.96875 291.99219 535.64453 286.13281 880.85938 590.82031 539.0625 571.28906 559.57031 409.66797 432.12891 345.21484 575.19531 496.58203 737.30469 504.88281 492.1875 ] 144 [ 970.70313 ] 177 [ 643.06641 ] 638 [ 241.21094 ] ] >> +endobj +71 0 obj +<< /Filter /FlateDecode /Length 337 >> +stream +x�]��n�0E�� +/�Ed�GH%��8EbчJ��R1�q�}��$R �a��LV��hO٧U �vڴ��+��V+�Rx��\ϓ��2�H�R��؎W��Zp��ts��f�/`<�(ha��ٮj�x��E�ώ��L��&�(p��s^м,˲ `��3T];�Ӹ�4�<��B)^��(N�Q�H��DA�h��G�<$X�:�O~=�6.1A�xH,J�a�y,�\�> /CIDToGIDMap /Identity /DW 375 /FontDescriptor 258 0 R /Subtype /CIDFontType2 /Type /Font /W [ 0 [ 1000 ] 20 [ 429.6875 558.59375 551.75781 ] ] >> +endobj +73 0 obj +<< /Filter /FlateDecode /Length 255 >> +stream +x�]�Mn� ��b��"��+)�kɋ��n�a�"Հ0^��M�.�4��W�Pdͷm��8{�0H;")cL@�4M#Z��_lT?�/��,�d��I$ua��3��/s�/��w�{맡)��Ej m� �+��=d�|��)�rendstream +endobj +74 0 obj +<< /BaseFont /IAAAAA+Times-Roman /CIDSystemInfo << /Ordering (Identity) /Registry (Adobe) /Supplement 0 >> /CIDToGIDMap /Identity /DW 500 /FontDescriptor 259 0 R /Subtype /CIDFontType2 /Type /Font /W [ 0 [ 722.16797 ] 3 15 250 16 [ 333.00781 250 277.83203 ] 29 [ 277.83203 ] 40 [ 610.83984 ] 46 [ 722.16797 0 889.16016 0 0 556.15234 ] 58 [ 943.84766 ] 68 [ 443.84766 ] 72 [ 443.84766 333.00781 ] 76 [ 277.83203 ] 79 [ 277.83203 ] 85 [ 333.00781 389.16016 277.83203 0 0 722.16797 ] ] >> +endobj +75 0 obj +<< /Filter /FlateDecode /Length 322 >> +stream +x�]�͎� ��<��ЀuƤ�5�?Y�`atIV4H��&{�䃙�AuY[��t�j��X�`oN�Bo,B��ᬆvb��f�=��F�e��/��w'=^a�ć��R4{&��4��s��k�(��L;��7~9\��L�C�lԨa�Z��=�LJ)s�UUU��HJ�v�u�<�2��JA��8!:�"E��$�* U�ND)RUD1�1@J(2MQv�J��;U��ZP%��1-n��p$�$ �<�i1�K�A�Й�˭~�Eݜ��`+�&��2�Ӛ��?��endstream +endobj +76 0 obj +<< /BaseFont /AAAAAA+Helvetica-Bold /CIDSystemInfo << /Ordering (Identity) /Registry (Adobe) /Supplement 0 >> /CIDToGIDMap /Identity /DW 0 /FontDescriptor 260 0 R /Subtype /CIDFontType2 /Type /Font /W [ 0 [ 722.16797 ] 40 [ 666.99219 ] 68 [ 556.15234 ] 71 74 610.83984 85 [ 389.16016 ] ] >> +endobj +77 0 obj +<< /Filter /FlateDecode /Length 254 >> +stream +x�]��j�0E��Y&� ��uF^�A�|�"�A- Y^��B3p��.��觷r��6��j7/n8kC��A)�E8B��2�k�e0�%m@�p�k�;��H�W赙�p��#��.h0�9(��ބ{M�Ӡ��ӵ�:�w�P&.�5�*\��腙��1ơ��W��6ɻ�|��2V�#7�фd^z �� >�u�EU�_4!zendstream +endobj +78 0 obj +<< /BaseFont /BAAAAA+Helvetica /CIDSystemInfo << /Ordering (Identity) /Registry (Adobe) /Supplement 0 >> /CIDToGIDMap /Identity /DW 500 /FontDescriptor 261 0 R /Subtype /CIDFontType2 /Type /Font /W [ 0 [ 633.78906 0 0 277.83203 ] 11 16 333.00781 18 [ 277.83203 ] 20 28 556.15234 36 40 666.99219 42 [ 777.83203 0 277.83203 ] 48 [ 833.00781 0 777.83203 0 0 0 666.99219 0 722.16797 0 943.84766 ] 68 69 556.15234 71 72 556.15234 73 [ 277.83203 556.15234 556.15234 222.16797 0 0 222.16797 833.00781 556.15234 556.15234 0 0 333.00781 ] 87 [ 277.83203 556.15234 0 722.16797 ] 144 [ 1000 ] 160 [ 889.16016 ] 178 [ 556.15234 ] ] >> +endobj +79 0 obj +<< /Filter /FlateDecode /Length 372 >> +stream +x�]��n�0��~ +�C�m ?BJ�"q�J��8��}eI�@��ag�H��G�vl*r�Lk�2^mC�D�`�T�7Sx7�zb�.��\�ndY�y�E�pq��_v�x��>lKv0=9�j��:M�t&�`y�[�X��>�B۲lɸ�ݖG]=o|�&�*��4��e��eB��(�"gd��2E۩k~j��9τP"�$�R�"�J%��h(�J�* T�U�4A%�J�*[��U��JH�h�-� +�go��f�ABh��#\IÜ� 7c̣�}7��6�y)8[�=)��W:�E|h�P��}�� `d��|��h�֒q!�!> ��G\�q�]��endstream +endobj +80 0 obj +<< /BaseFont /CAAAAA+Helvetica-Oblique /CIDSystemInfo << /Ordering (Identity) /Registry (Adobe) /Supplement 0 >> /CIDToGIDMap /Identity /DW 500 /FontDescriptor 262 0 R /Subtype /CIDFontType2 /Type /Font /W [ 0 [ 633.78906 0 0 277.83203 ] 40 [ 666.99219 0 777.83203 0 0 0 666.99219 ] 53 [ 722.16797 ] 68 [ 556.15234 ] 71 72 556.15234 73 [ 277.83203 556.15234 556.15234 222.16797 0 0 222.16797 0 556.15234 556.15234 ] 87 [ 277.83203 ] ] >> +endobj +81 0 obj +<< /Filter /FlateDecode /Length 296 >> +stream +x�]��n� E�|�,�E~�]X�'��Cu��]�#L��C�J]:p��0ë��I��F9��W�aQJK'��Yƫ��.�Ǳ1�Ċ��g��t�-��N��f�ͥj��7kpD�A��=��kgߺ�SخQh��R��"��Q�FN +g�It��BQBQ�u]24��{�� +!bQ��Y�C�<�9Б(Ɉ��(M��Q�:PE�s%��B�}0��:��,z��9\e�ӑ'w��~rƳ��O��C��]��N��Jendstream +endobj +82 0 obj +<< /Filter /FlateDecode /N 3 /Length 275 >> +stream +x�}��J�`�� +":888dp�Ƥj�4i�k��vK��6 iD@7W��g}w��Gp�Y�LA�>�=p�=PCjQ�Q�g�vSq{}e��qJ�|��rUZ��A��+�~��/�41��t,�Z4��i&��hx�w�Ƨ]�6i��1$�N�s��c�s��A�-P�i�| �=�g�2��w�� l��V��ʎS/�~۫�z�Oʟ�� AACe��G�endstream +endobj +83 0 obj +<< /BitsPerComponent 8 /ColorSpace /DeviceGray /Filter /FlateDecode /Height 100 /Subtype /Image /Type /XObject /Width 100 /Length 1190 >> +stream +x��mlEǟB�6b�! 61j[@(M41Z��BP�BR#��B��4� )�X)��,FM(F�%��Z�"��qW��(��z�z�;f_�7��d��7��3�3s��Kr)��غ�#��@7Y�3�!3��B�z��S.��ejԜ͆XV��2_��H��>�WֺƠ�;��B�X�� 9�q�X�ʻWu-W.7}��K^}�1�Q=2r��]�R��K��l�zXGQ��=3�J;�*9��}�� MCf��j��®��mfLut�+��*�/Ě_�j�xh,f�L?s1��L�B��'SOx��b�P��Q�ED�9�)>B�b"�Կ0�6� +�ɇ�%�!�i�küdo�9$��9Z�ۘ �w��y|s��k�~%&��s��c`T��{�� ]V[��Й@H +H 9�#��3y�g��[<�� I=�QB��w��U��~��Τ��p�Z4ƨ"��9��D�0PT�7خ9`H�:c��,_m��C��.��ߢz��҆s� n�{��0�*�Y��f8�t��8`�uSbG:�l/Z��*�o��z�y�\�"�7��7=��1�[Hik��^ ��_�u�ʯ�S8d4�&;]>�l�dV�@�yZe8�|�F{X��2Ѓ0��Ғ�u��uPBsu_��& ��=�b��`�pFl��8 ��(��Z�K�E8�6g| ��3.��{�P8�5�r�eƥR�ьb�ÂQ��'�?X;`ʈX�3�w�wB��!�Mآ~�e��q�"�c�@F��V��\��C�3u�m}>"��5�� D��1A��D_�(v��f0�C^�}FF� C!>�ıN��=#SHF��Sc|��OnY�P��$��m4��J.��ğw� �+/yv��w\#.��P��ylt�|�a�2W{V�/p��ip A9q��h<�ϊ��/; ��j��qw}�|.��(K� ��*�n�{?!OxHIǗ8��Xqd��[3]Te�:v��t��l��q�t��5�Z�Y��ҲZ��׫7U./�e�'Ʉ��@_M3endstream +endobj +84 0 obj +<< /Filter /FlateDecode /N 3 /Length 293 >> +stream +x�}��J��Ԃ(��28�h��iRpi"�V��S��A�OHS�tspu+.��c(��#��AR�x��@��?��F�[V[�G@`*�dK��$��O.Ko�@��6`�O�,f'�˘O� asx�0A�6�vf� �8��{c7��%op�Z�:u��Q�� 0Q٣F �*��Ԑ(�S��DGACAa�j�g�rx��]��s PxM�� c�vhO�<��v��-X}�ţ�b3~��*��mDJH�T~��K�endstream +endobj +85 0 obj +<< /ID (node00000220) /K [ 188 0 R << /Obj 39 0 R /Pg 6 0 R /Type /OBJR >> ] /P 202 0 R /S /Link /Type /StructElem >> +endobj +86 0 obj +<< /ID (node00000221) /K [ << /MCID 1 /Pg 6 0 R /Type /MCR >> ] /P 94 0 R /S /Figure /Type /StructElem >> +endobj +87 0 obj +<< /ID (node00000236) /K [ 89 0 R ] /P 62 0 R /S /NonStruct /Type /StructElem >> +endobj +88 0 obj +<< /ID (node00000245) /K [ 202 0 R ] /P 62 0 R /S /Div /Type /StructElem >> +endobj +89 0 obj +<< /ID (node00000756) /K [ 142 0 R ] /P 87 0 R /S /H1 /Type /StructElem >> +endobj +90 0 obj +<< /ID (node00000757) /K [ 203 0 R 204 0 R 207 0 R 208 0 R 211 0 R 214 0 R 217 0 R 220 0 R 223 0 R 226 0 R 233 0 R 241 0 R 244 0 R 247 0 R ] /P 202 0 R /S /Table /Type /StructElem >> +endobj +91 0 obj +<< /A [ << /O /Table /Scope /Column >> << /O /Table /RowSpan 1 >> << /ColSpan 2 /O /Table >> ] /ID (node00000758) /K [ 93 0 R ] /P 203 0 R /S /TH /Type /StructElem >> +endobj +92 0 obj +<< /A [ << /O /Table /Scope /Column >> << /O /Table /RowSpan 1 >> << /ColSpan 2 /O /Table >> ] /ID (node00000759) /K [ 97 0 R ] /P 207 0 R /S /TH /Type /StructElem >> +endobj +93 0 obj +<< /ID (node00000761) /K [ << /MCID 0 /Pg 6 0 R /Type /MCR >> ] /P 91 0 R /S /NonStruct /Type /StructElem >> +endobj +94 0 obj +<< /ID (node00000762) /K [ 86 0 R << /Obj 10 0 R /Pg 6 0 R /Type /OBJR >> << /Obj 11 0 R /Pg 6 0 R /Type /OBJR >> ] /P 205 0 R /S /Link /Type /StructElem >> +endobj +95 0 obj +<< /ID (node00000763) /K [ << /MCID 2 /Pg 6 0 R /Type /MCR >> ] /P 206 0 R /S /NonStruct /Type /StructElem >> +endobj +96 0 obj +<< /ID (node00000764) /K [ << /MCID 3 /Pg 6 0 R /Type /MCR >> ] /P 206 0 R /S /NonStruct /Type /StructElem >> +endobj +97 0 obj +<< /ID (node00000765) /K [ 98 0 R << /Obj 12 0 R /Pg 6 0 R /Type /OBJR >> ] /P 92 0 R /S /Link /Type /StructElem >> +endobj +98 0 obj +<< /ID (node00000766) /K [ << /MCID 4 /Pg 6 0 R /Type /MCR >> ] /P 97 0 R /S /NonStruct /Type /StructElem >> +endobj +99 0 obj +<< /ID (node00000767) /K [ << /MCID 5 /Pg 6 0 R /Type /MCR >> ] /P 209 0 R /S /NonStruct /Type /StructElem >> +endobj +100 0 obj +<< /ID (node00000768) /K [ << /MCID 6 /Pg 6 0 R /Type /MCR >> ] /P 210 0 R /S /NonStruct /Type /StructElem >> +endobj +101 0 obj +<< /ID (node00000769) /K [ << /MCID 7 /Pg 6 0 R /Type /MCR >> ] /P 212 0 R /S /NonStruct /Type /StructElem >> +endobj +102 0 obj +<< /ID (node00000770) /K [ 103 0 R << /Obj 13 0 R /Pg 6 0 R /Type /OBJR >> ] /P 213 0 R /S /Link /Type /StructElem >> +endobj +103 0 obj +<< /ID (node00000771) /K [ << /MCID 8 /Pg 6 0 R /Type /MCR >> ] /P 102 0 R /S /NonStruct /Type /StructElem >> +endobj +104 0 obj +<< /ID (node00000772) /K [ << /MCID 9 /Pg 6 0 R /Type /MCR >> ] /P 215 0 R /S /NonStruct /Type /StructElem >> +endobj +105 0 obj +<< /ID (node00000773) /K [ 106 0 R << /Obj 14 0 R /Pg 6 0 R /Type /OBJR >> ] /P 216 0 R /S /Link /Type /StructElem >> +endobj +106 0 obj +<< /ID (node00000774) /K [ << /MCID 10 /Pg 6 0 R /Type /MCR >> ] /P 105 0 R /S /NonStruct /Type /StructElem >> +endobj +107 0 obj +<< /ID (node00000775) /K [ << /MCID 11 /Pg 6 0 R /Type /MCR >> ] /P 218 0 R /S /NonStruct /Type /StructElem >> +endobj +108 0 obj +<< /ID (node00000776) /K [ 110 0 R << /Obj 15 0 R /Pg 6 0 R /Type /OBJR >> ] /P 219 0 R /S /Link /Type /StructElem >> +endobj +109 0 obj +<< /ID (node00000777) /K [ << /MCID 12 /Pg 6 0 R /Type /MCR >> ] /P 219 0 R /S /NonStruct /Type /StructElem >> +endobj +110 0 obj +<< /ID (node00000778) /K [ << /MCID 13 /Pg 6 0 R /Type /MCR >> ] /P 108 0 R /S /NonStruct /Type /StructElem >> +endobj +111 0 obj +<< /ID (node00000779) /K [ << /MCID 14 /Pg 6 0 R /Type /MCR >> ] /P 221 0 R /S /NonStruct /Type /StructElem >> +endobj +112 0 obj +<< /ID (node00000780) /K [ << /MCID 15 /Pg 6 0 R /Type /MCR >> ] /P 222 0 R /S /NonStruct /Type /StructElem >> +endobj +113 0 obj +<< /ID (node00000781) /K [ << /MCID 16 /Pg 6 0 R /Type /MCR >> ] /P 224 0 R /S /NonStruct /Type /StructElem >> +endobj +114 0 obj +<< /ID (node00000782) /K [ 115 0 R << /Obj 16 0 R /Pg 6 0 R /Type /OBJR >> ] /P 225 0 R /S /Link /Type /StructElem >> +endobj +115 0 obj +<< /ID (node00000783) /K [ << /MCID 17 /Pg 6 0 R /Type /MCR >> ] /P 114 0 R /S /NonStruct /Type /StructElem >> +endobj +116 0 obj +<< /ID (node00000784) /K [ << /MCID 18 /Pg 6 0 R /Type /MCR >> ] /P 227 0 R /S /NonStruct /Type /StructElem >> +endobj +117 0 obj +<< /ID (node00000785) /K [ 118 0 R << /Obj 17 0 R /Pg 6 0 R /Type /OBJR >> ] /P 230 0 R /S /Link /Type /StructElem >> +endobj +118 0 obj +<< /ID (node00000786) /K [ << /MCID 19 /Pg 6 0 R /Type /MCR >> ] /P 117 0 R /S /NonStruct /Type /StructElem >> +endobj +119 0 obj +<< /ID (node00000787) /K [ 120 0 R << /Obj 18 0 R /Pg 6 0 R /Type /OBJR >> ] /P 231 0 R /S /Link /Type /StructElem >> +endobj +120 0 obj +<< /ID (node00000788) /K [ << /MCID 20 /Pg 6 0 R /Type /MCR >> ] /P 119 0 R /S /NonStruct /Type /StructElem >> +endobj +121 0 obj +<< /ID (node00000789) /K [ 122 0 R << /Obj 19 0 R /Pg 6 0 R /Type /OBJR >> ] /P 232 0 R /S /Link /Type /StructElem >> +endobj +122 0 obj +<< /ID (node00000790) /K [ << /MCID 21 /Pg 6 0 R /Type /MCR >> ] /P 121 0 R /S /NonStruct /Type /StructElem >> +endobj +123 0 obj +<< /ID (node00000791) /K [ 124 0 R << /Obj 20 0 R /Pg 6 0 R /Type /OBJR >> ] /P 234 0 R /S /Link /Type /StructElem >> +endobj +124 0 obj +<< /ID (node00000792) /K [ << /MCID 22 /Pg 6 0 R /Type /MCR >> ] /P 123 0 R /S /NonStruct /Type /StructElem >> +endobj +125 0 obj +<< /ID (node00000793) /K [ 126 0 R << /Obj 21 0 R /Pg 6 0 R /Type /OBJR >> ] /P 237 0 R /S /Link /Type /StructElem >> +endobj +126 0 obj +<< /ID (node00000794) /K [ << /MCID 23 /Pg 6 0 R /Type /MCR >> ] /P 125 0 R /S /NonStruct /Type /StructElem >> +endobj +127 0 obj +<< /ID (node00000795) /K [ 128 0 R << /Obj 22 0 R /Pg 6 0 R /Type /OBJR >> ] /P 238 0 R /S /Link /Type /StructElem >> +endobj +128 0 obj +<< /ID (node00000796) /K [ << /MCID 24 /Pg 6 0 R /Type /MCR >> ] /P 127 0 R /S /NonStruct /Type /StructElem >> +endobj +129 0 obj +<< /ID (node00000797) /K [ << /MCID 25 /Pg 6 0 R /Type /MCR >> ] /P 239 0 R /S /NonStruct /Type /StructElem >> +endobj +130 0 obj +<< /ID (node00000798) /K [ 131 0 R << /Obj 23 0 R /Pg 6 0 R /Type /OBJR >> ] /P 240 0 R /S /Link /Type /StructElem >> +endobj +131 0 obj +<< /ID (node00000799) /K [ << /MCID 26 /Pg 6 0 R /Type /MCR >> ] /P 130 0 R /S /NonStruct /Type /StructElem >> +endobj +132 0 obj +<< /ID (node00000800) /K [ 133 0 R << /Obj 24 0 R /Pg 6 0 R /Type /OBJR >> ] /P 242 0 R /S /Link /Type /StructElem >> +endobj +133 0 obj +<< /ID (node00000801) /K [ << /MCID 27 /Pg 6 0 R /Type /MCR >> ] /P 132 0 R /S /NonStruct /Type /StructElem >> +endobj +134 0 obj +<< /ID (node00000802) /K [ 135 0 R << /Obj 25 0 R /Pg 6 0 R /Type /OBJR >> ] /P 243 0 R /S /Link /Type /StructElem >> +endobj +135 0 obj +<< /ID (node00000803) /K [ << /MCID 28 /Pg 6 0 R /Type /MCR >> ] /P 134 0 R /S /NonStruct /Type /StructElem >> +endobj +136 0 obj +<< /ID (node00000804) /K [ << /MCID 29 /Pg 6 0 R /Type /MCR >> ] /P 245 0 R /S /NonStruct /Type /StructElem >> +endobj +137 0 obj +<< /ID (node00000805) /K [ 138 0 R << /Obj 26 0 R /Pg 6 0 R /Type /OBJR >> ] /P 246 0 R /S /Link /Type /StructElem >> +endobj +138 0 obj +<< /ID (node00000806) /K [ << /MCID 30 /Pg 6 0 R /Type /MCR >> ] /P 137 0 R /S /NonStruct /Type /StructElem >> +endobj +139 0 obj +<< /ID (node00000807) /K [ << /MCID 31 /Pg 6 0 R /Type /MCR >> ] /P 248 0 R /S /NonStruct /Type /StructElem >> +endobj +140 0 obj +<< /ID (node00000808) /K [ 141 0 R << /Obj 27 0 R /Pg 6 0 R /Type /OBJR >> ] /P 249 0 R /S /Link /Type /StructElem >> +endobj +141 0 obj +<< /ID (node00000809) /K [ << /MCID 32 /Pg 6 0 R /Type /MCR >> ] /P 140 0 R /S /NonStruct /Type /StructElem >> +endobj +142 0 obj +<< /ID (node00000811) /K [ << /MCID 33 /Pg 6 0 R /Type /MCR >> ] /P 89 0 R /S /NonStruct /Type /StructElem >> +endobj +143 0 obj +<< /ID (node00000812) /K [ 152 0 R 153 0 R 154 0 R << /Obj 29 0 R /Pg 6 0 R /Type /OBJR >> ] /P 250 0 R /S /Link /Type /StructElem >> +endobj +144 0 obj +<< /ID (node00000813) /K [ 159 0 R << /Obj 30 0 R /Pg 6 0 R /Type /OBJR >> << /Obj 31 0 R /Pg 6 0 R /Type /OBJR >> ] /P 202 0 R /S /Link /Type /StructElem >> +endobj +145 0 obj +<< /ID (node00000814) /K [ 161 0 R << /Obj 32 0 R /Pg 6 0 R /Type /OBJR >> ] /P 202 0 R /S /Link /Type /StructElem >> +endobj +146 0 obj +<< /ID (node00000815) /K [ 163 0 R << /Obj 33 0 R /Pg 6 0 R /Type /OBJR >> ] /P 202 0 R /S /Link /Type /StructElem >> +endobj +147 0 obj +<< /ID (node00000816) /K [ 165 0 R << /Obj 34 0 R /Pg 6 0 R /Type /OBJR >> ] /P 202 0 R /S /Link /Type /StructElem >> +endobj +148 0 obj +<< /ID (node00000817) /K [ << /MCID 34 /Pg 6 0 R /Type /MCR >> ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +149 0 obj +<< /ID (node00000818) /K [ << /MCID 35 /Pg 6 0 R /Type /MCR >> ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +150 0 obj +<< /ID (node00000819) /K [ << /MCID 36 /Pg 6 0 R /Type /MCR >> ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +151 0 obj +<< /ID (node00000820) /K [ << /MCID 37 /Pg 6 0 R /Type /MCR >> ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +152 0 obj +<< /ID (node00000821) /K [ << /MCID 38 /Pg 6 0 R /Type /MCR >> ] /P 143 0 R /S /NonStruct /Type /StructElem >> +endobj +153 0 obj +<< /ID (node00000822) /K [ << /MCID 39 /Pg 6 0 R /Type /MCR >> ] /P 143 0 R /S /NonStruct /Type /StructElem >> +endobj +154 0 obj +<< /ID (node00000823) /K [ << /MCID 40 /Pg 6 0 R /Type /MCR >> ] /P 143 0 R /S /NonStruct /Type /StructElem >> +endobj +155 0 obj +<< /ID (node00000824) /K [ << /MCID 41 /Pg 6 0 R /Type /MCR >> ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +156 0 obj +<< /ID (node00000825) /K [ << /MCID 42 /Pg 6 0 R /Type /MCR >> ] /P 251 0 R /S /NonStruct /Type /StructElem >> +endobj +157 0 obj +<< /ID (node00000826) /K [ << /MCID 43 /Pg 6 0 R /Type /MCR >> ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +158 0 obj +<< /ID (node00000827) /K [ << /MCID 44 /Pg 6 0 R /Type /MCR >> ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +159 0 obj +<< /ID (node00000828) /K [ << /MCID 45 /Pg 6 0 R /Type /MCR >> << /MCID 46 /Pg 6 0 R /Type /MCR >> ] /P 144 0 R /S /NonStruct /Type /StructElem >> +endobj +160 0 obj +<< /ID (node00000829) /K [ << /MCID 47 /Pg 6 0 R /Type /MCR >> << /MCID 48 /Pg 6 0 R /Type /MCR >> ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +161 0 obj +<< /ID (node00000830) /K [ << /MCID 49 /Pg 6 0 R /Type /MCR >> ] /P 145 0 R /S /NonStruct /Type /StructElem >> +endobj +162 0 obj +<< /ID (node00000831) /K [ << /MCID 50 /Pg 6 0 R /Type /MCR >> << /MCID 51 /Pg 6 0 R /Type /MCR >> ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +163 0 obj +<< /ID (node00000832) /K [ << /MCID 52 /Pg 6 0 R /Type /MCR >> ] /P 146 0 R /S /NonStruct /Type /StructElem >> +endobj +164 0 obj +<< /ID (node00000833) /K [ << /MCID 53 /Pg 6 0 R /Type /MCR >> ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +165 0 obj +<< /ID (node00000834) /K [ << /MCID 54 /Pg 6 0 R /Type /MCR >> ] /P 147 0 R /S /NonStruct /Type /StructElem >> +endobj +166 0 obj +<< /ID (node00000835) /K [ << /MCID 55 /Pg 6 0 R /Type /MCR >> << /MCID 56 /Pg 6 0 R /Type /MCR >> << /MCID 57 /Pg 6 0 R /Type /MCR >> << /MCID 58 /Pg 6 0 R /Type /MCR >> ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +167 0 obj +<< /ID (node00000836) /K [ 172 0 R << /Obj 35 0 R /Pg 6 0 R /Type /OBJR >> ] /P 202 0 R /S /Link /Type /StructElem >> +endobj +168 0 obj +<< /ID (node00000837) /K [ 174 0 R 175 0 R 176 0 R << /Obj 36 0 R /Pg 6 0 R /Type /OBJR >> ] /P 252 0 R /S /Link /Type /StructElem >> +endobj +169 0 obj +<< /ID (node00000838) /K [ 178 0 R << /Obj 37 0 R /Pg 6 0 R /Type /OBJR >> ] /P 202 0 R /S /Link /Type /StructElem >> +endobj +170 0 obj +<< /ID (node00000839) /K [ 180 0 R << /Obj 38 0 R /Pg 6 0 R /Type /OBJR >> ] /P 202 0 R /S /Link /Type /StructElem >> +endobj +171 0 obj +<< /ID (node00000840) /K [ << /MCID 59 /Pg 6 0 R /Type /MCR >> << /MCID 60 /Pg 6 0 R /Type /MCR >> << /MCID 61 /Pg 6 0 R /Type /MCR >> ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +172 0 obj +<< /ID (node00000841) /K [ << /MCID 62 /Pg 6 0 R /Type /MCR >> ] /P 167 0 R /S /NonStruct /Type /StructElem >> +endobj +173 0 obj +<< /ID (node00000842) /K [ << /MCID 63 /Pg 6 0 R /Type /MCR >> << /MCID 64 /Pg 6 0 R /Type /MCR >> << /MCID 65 /Pg 6 0 R /Type /MCR >> ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +174 0 obj +<< /ID (node00000843) /K [ << /MCID 66 /Pg 6 0 R /Type /MCR >> ] /P 168 0 R /S /NonStruct /Type /StructElem >> +endobj +175 0 obj +<< /ID (node00000844) /K [ << /MCID 67 /Pg 6 0 R /Type /MCR >> ] /P 168 0 R /S /NonStruct /Type /StructElem >> +endobj +176 0 obj +<< /ID (node00000845) /K [ << /MCID 68 /Pg 6 0 R /Type /MCR >> ] /P 168 0 R /S /NonStruct /Type /StructElem >> +endobj +177 0 obj +<< /ID (node00000846) /K [ << /MCID 69 /Pg 6 0 R /Type /MCR >> << /MCID 70 /Pg 6 0 R /Type /MCR >> << /MCID 71 /Pg 6 0 R /Type /MCR >> ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +178 0 obj +<< /ID (node00000847) /K [ << /MCID 72 /Pg 6 0 R /Type /MCR >> ] /P 169 0 R /S /NonStruct /Type /StructElem >> +endobj +179 0 obj +<< /ID (node00000848) /K [ << /MCID 73 /Pg 6 0 R /Type /MCR >> ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +180 0 obj +<< /ID (node00000849) /K [ << /MCID 74 /Pg 6 0 R /Type /MCR >> ] /P 170 0 R /S /NonStruct /Type /StructElem >> +endobj +181 0 obj +<< /ID (node00000850) /K [ << /MCID 75 /Pg 6 0 R /Type /MCR >> << /MCID 76 /Pg 6 0 R /Type /MCR >> << /MCID 77 /Pg 6 0 R /Type /MCR >> << /MCID 78 /Pg 6 0 R /Type /MCR >> << /MCID 79 /Pg 6 0 R /Type /MCR >> << /MCID 80 /Pg 6 0 R /Type /MCR >> << /MCID 81 /Pg 6 0 R /Type /MCR >> << /MCID 82 /Pg 6 0 R /Type /MCR >> << /MCID 83 /Pg 6 0 R /Type /MCR >> ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +182 0 obj +<< /ID (node00000851) /K [ 190 0 R 191 0 R 192 0 R << /Obj 40 0 R /Pg 6 0 R /Type /OBJR >> ] /P 253 0 R /S /Link /Type /StructElem >> +endobj +183 0 obj +<< /ID (node00000852) /K [ 194 0 R << /Obj 41 0 R /Pg 6 0 R /Type /OBJR >> ] /P 202 0 R /S /Link /Type /StructElem >> +endobj +184 0 obj +<< /ID (node00000853) /K [ 196 0 R << /Obj 42 0 R /Pg 6 0 R /Type /OBJR >> ] /P 202 0 R /S /Link /Type /StructElem >> +endobj +185 0 obj +<< /ID (node00000854) /K [ 198 0 R << /Obj 43 0 R /Pg 6 0 R /Type /OBJR >> << /Obj 44 0 R /Pg 6 0 R /Type /OBJR >> ] /P 202 0 R /S /Link /Type /StructElem >> +endobj +186 0 obj +<< /ID (node00000855) /K [ 200 0 R << /Obj 45 0 R /Pg 6 0 R /Type /OBJR >> << /Obj 46 0 R /Pg 6 0 R /Type /OBJR >> ] /P 202 0 R /S /Link /Type /StructElem >> +endobj +187 0 obj +<< /ID (node00000856) /K [ << /MCID 84 /Pg 6 0 R /Type /MCR >> ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +188 0 obj +<< /ID (node00000857) /K [ << /MCID 85 /Pg 6 0 R /Type /MCR >> ] /P 85 0 R /S /NonStruct /Type /StructElem >> +endobj +189 0 obj +<< /ID (node00000858) /K [ << /MCID 86 /Pg 6 0 R /Type /MCR >> << /MCID 87 /Pg 6 0 R /Type /MCR >> << /MCID 88 /Pg 6 0 R /Type /MCR >> << /MCID 89 /Pg 6 0 R /Type /MCR >> ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +190 0 obj +<< /ID (node00000859) /K [ << /MCID 90 /Pg 6 0 R /Type /MCR >> ] /P 182 0 R /S /NonStruct /Type /StructElem >> +endobj +191 0 obj +<< /ID (node00000860) /K [ << /MCID 91 /Pg 6 0 R /Type /MCR >> ] /P 182 0 R /S /NonStruct /Type /StructElem >> +endobj +192 0 obj +<< /ID (node00000861) /K [ << /MCID 92 /Pg 6 0 R /Type /MCR >> ] /P 182 0 R /S /NonStruct /Type /StructElem >> +endobj +193 0 obj +<< /ID (node00000862) /K [ << /MCID 93 /Pg 6 0 R /Type /MCR >> << /MCID 94 /Pg 6 0 R /Type /MCR >> << /MCID 95 /Pg 6 0 R /Type /MCR >> ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +194 0 obj +<< /ID (node00000863) /K [ << /MCID 96 /Pg 6 0 R /Type /MCR >> ] /P 183 0 R /S /NonStruct /Type /StructElem >> +endobj +195 0 obj +<< /ID (node00000864) /K [ << /MCID 97 /Pg 6 0 R /Type /MCR >> << /MCID 98 /Pg 6 0 R /Type /MCR >> ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +196 0 obj +<< /ID (node00000865) /K [ << /MCID 99 /Pg 6 0 R /Type /MCR >> ] /P 184 0 R /S /NonStruct /Type /StructElem >> +endobj +197 0 obj +<< /ID (node00000866) /K [ << /MCID 100 /Pg 6 0 R /Type /MCR >> ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +198 0 obj +<< /ID (node00000867) /K [ << /MCID 101 /Pg 6 0 R /Type /MCR >> << /MCID 102 /Pg 6 0 R /Type /MCR >> ] /P 185 0 R /S /NonStruct /Type /StructElem >> +endobj +199 0 obj +<< /ID (node00000868) /K [ << /MCID 103 /Pg 6 0 R /Type /MCR >> << /MCID 104 /Pg 6 0 R /Type /MCR >> ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +200 0 obj +<< /ID (node00000869) /K [ << /MCID 105 /Pg 6 0 R /Type /MCR >> << /MCID 106 /Pg 6 0 R /Type /MCR >> ] /P 186 0 R /S /NonStruct /Type /StructElem >> +endobj +201 0 obj +<< /ID (node00000870) /K [ << /MCID 107 /Pg 6 0 R /Type /MCR >> << /MCID 108 /Pg 6 0 R /Type /MCR >> ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +202 0 obj +<< /ID (node00001466) /K [ 90 0 R 148 0 R 149 0 R 150 0 R 151 0 R 250 0 R 155 0 R 251 0 R 157 0 R 158 0 R 144 0 R 160 0 R 145 0 R 162 0 R 146 0 R 164 0 R 147 0 R 166 0 R 171 0 R 167 0 R 173 0 R 252 0 R 177 0 R 169 0 R 179 0 R 170 0 R 181 0 R 187 0 R 85 0 R 189 0 R 253 0 R 193 0 R 183 0 R 195 0 R 184 0 R 197 0 R 185 0 R 199 0 R 186 0 R 201 0 R ] /P 88 0 R /S /Div /Type /StructElem >> +endobj +203 0 obj +<< /ID (node00001471) /K [ 91 0 R ] /P 90 0 R /S /TR /Type /StructElem >> +endobj +204 0 obj +<< /ID (node00001472) /K [ 205 0 R ] /P 90 0 R /S /TR /Type /StructElem >> +endobj +205 0 obj +<< /A [ << /Headers [ (node00000758) (node00000759) (node00000760) ] /O /Table >> << /O /Table /RowSpan 1 >> << /ColSpan 2 /O /Table >> ] /ID (node00001473) /K [ 94 0 R 206 0 R ] /P 204 0 R /S /TD /Type /StructElem >> +endobj +206 0 obj +<< /ID (node00001475) /K [ 95 0 R 96 0 R ] /P 205 0 R /S /Div /Type /StructElem >> +endobj +207 0 obj +<< /ID (node00001478) /K [ 92 0 R ] /P 90 0 R /S /TR /Type /StructElem >> +endobj +208 0 obj +<< /ID (node00001479) /K [ 209 0 R 210 0 R ] /P 90 0 R /S /TR /Type /StructElem >> +endobj +209 0 obj +<< /A [ << /O /Table /Scope /Row >> << /O /Table /RowSpan 1 >> << /ColSpan 1 /O /Table >> ] /ID (node00001480) /K [ 99 0 R ] /P 208 0 R /S /TH /Type /StructElem >> +endobj +210 0 obj +<< /A [ << /Headers [ (node00000758) (node00000759) (node00000760) (node00001480) ] /O /Table >> << /O /Table /RowSpan 1 >> << /ColSpan 1 /O /Table >> ] /ID (node00001481) /K [ 100 0 R ] /P 208 0 R /S /TD /Type /StructElem >> +endobj +211 0 obj +<< /ID (node00001482) /K [ 212 0 R 213 0 R ] /P 90 0 R /S /TR /Type /StructElem >> +endobj +212 0 obj +<< /A [ << /O /Table /Scope /Row >> << /O /Table /RowSpan 1 >> << /ColSpan 1 /O /Table >> ] /ID (node00001483) /K [ 101 0 R ] /P 211 0 R /S /TH /Type /StructElem >> +endobj +213 0 obj +<< /A [ << /Headers [ (node00000758) (node00000759) (node00000760) (node00001483) ] /O /Table >> << /O /Table /RowSpan 1 >> << /ColSpan 1 /O /Table >> ] /ID (node00001484) /K [ 102 0 R ] /P 211 0 R /S /TD /Type /StructElem >> +endobj +214 0 obj +<< /ID (node00001485) /K [ 215 0 R 216 0 R ] /P 90 0 R /S /TR /Type /StructElem >> +endobj +215 0 obj +<< /A [ << /O /Table /Scope /Row >> << /O /Table /RowSpan 1 >> << /ColSpan 1 /O /Table >> ] /ID (node00001486) /K [ 104 0 R ] /P 214 0 R /S /TH /Type /StructElem >> +endobj +216 0 obj +<< /A [ << /Headers [ (node00000758) (node00000759) (node00000760) (node00001486) ] /O /Table >> << /O /Table /RowSpan 1 >> << /ColSpan 1 /O /Table >> ] /ID (node00001487) /K [ 105 0 R ] /P 214 0 R /S /TD /Type /StructElem >> +endobj +217 0 obj +<< /ID (node00001490) /K [ 218 0 R 219 0 R ] /P 90 0 R /S /TR /Type /StructElem >> +endobj +218 0 obj +<< /A [ << /O /Table /Scope /Row >> << /O /Table /RowSpan 1 >> << /ColSpan 1 /O /Table >> ] /ID (node00001491) /K [ 107 0 R ] /P 217 0 R /S /TH /Type /StructElem >> +endobj +219 0 obj +<< /A [ << /Headers [ (node00000758) (node00000759) (node00000760) (node00001491) ] /O /Table >> << /O /Table /RowSpan 1 >> << /ColSpan 1 /O /Table >> ] /ID (node00001492) /K [ 109 0 R 108 0 R ] /P 217 0 R /S /TD /Type /StructElem >> +endobj +220 0 obj +<< /ID (node00001494) /K [ 221 0 R 222 0 R ] /P 90 0 R /S /TR /Type /StructElem >> +endobj +221 0 obj +<< /A [ << /O /Table /Scope /Row >> << /O /Table /RowSpan 1 >> << /ColSpan 1 /O /Table >> ] /ID (node00001495) /K [ 111 0 R ] /P 220 0 R /S /TH /Type /StructElem >> +endobj +222 0 obj +<< /A [ << /Headers [ (node00000758) (node00000759) (node00000760) (node00001495) ] /O /Table >> << /O /Table /RowSpan 1 >> << /ColSpan 1 /O /Table >> ] /ID (node00001496) /K [ 112 0 R ] /P 220 0 R /S /TD /Type /StructElem >> +endobj +223 0 obj +<< /ID (node00001497) /K [ 224 0 R 225 0 R ] /P 90 0 R /S /TR /Type /StructElem >> +endobj +224 0 obj +<< /A [ << /O /Table /Scope /Row >> << /O /Table /RowSpan 1 >> << /ColSpan 1 /O /Table >> ] /ID (node00001498) /K [ 113 0 R ] /P 223 0 R /S /TH /Type /StructElem >> +endobj +225 0 obj +<< /A [ << /Headers [ (node00000758) (node00000759) (node00000760) (node00001498) ] /O /Table >> << /O /Table /RowSpan 1 >> << /ColSpan 1 /O /Table >> ] /ID (node00001499) /K [ 114 0 R ] /P 223 0 R /S /TD /Type /StructElem >> +endobj +226 0 obj +<< /ID (node00001501) /K [ 227 0 R 228 0 R ] /P 90 0 R /S /TR /Type /StructElem >> +endobj +227 0 obj +<< /A [ << /O /Table /Scope /Row >> << /O /Table /RowSpan 1 >> << /ColSpan 1 /O /Table >> ] /ID (node00001502) /K [ 116 0 R ] /P 226 0 R /S /TH /Type /StructElem >> +endobj +228 0 obj +<< /A [ << /Headers [ (node00000758) (node00000759) (node00000760) (node00001502) ] /O /Table >> << /O /Table /RowSpan 1 >> << /ColSpan 1 /O /Table >> ] /ID (node00001503) /K [ 229 0 R ] /P 226 0 R /S /TD /Type /StructElem >> +endobj +229 0 obj +<< /ID (node00001505) /K [ 230 0 R 231 0 R 232 0 R ] /P 228 0 R /S /L /Type /StructElem >> +endobj +230 0 obj +<< /ID (node00001506) /K [ 117 0 R ] /P 229 0 R /S /LI /Type /StructElem >> +endobj +231 0 obj +<< /ID (node00001507) /K [ 119 0 R ] /P 229 0 R /S /LI /Type /StructElem >> +endobj +232 0 obj +<< /ID (node00001508) /K [ 121 0 R ] /P 229 0 R /S /LI /Type /StructElem >> +endobj +233 0 obj +<< /ID (node00001509) /K [ 234 0 R 235 0 R ] /P 90 0 R /S /TR /Type /StructElem >> +endobj +234 0 obj +<< /A [ << /O /Table /Scope /Row >> << /O /Table /RowSpan 1 >> << /ColSpan 1 /O /Table >> ] /ID (node00001510) /K [ 123 0 R ] /P 233 0 R /S /TH /Type /StructElem >> +endobj +235 0 obj +<< /A [ << /Headers [ (node00000758) (node00000759) (node00000760) (node00001510) ] /O /Table >> << /O /Table /RowSpan 1 >> << /ColSpan 1 /O /Table >> ] /ID (node00001511) /K [ 236 0 R ] /P 233 0 R /S /TD /Type /StructElem >> +endobj +236 0 obj +<< /ID (node00001514) /K [ 237 0 R 238 0 R 239 0 R 240 0 R ] /P 235 0 R /S /L /Type /StructElem >> +endobj +237 0 obj +<< /ID (node00001515) /K [ 125 0 R ] /P 236 0 R /S /LI /Type /StructElem >> +endobj +238 0 obj +<< /ID (node00001516) /K [ 127 0 R ] /P 236 0 R /S /LI /Type /StructElem >> +endobj +239 0 obj +<< /ID (node00001517) /K [ 129 0 R ] /P 236 0 R /S /LI /Type /StructElem >> +endobj +240 0 obj +<< /ID (node00001518) /K [ 130 0 R ] /P 236 0 R /S /LI /Type /StructElem >> +endobj +241 0 obj +<< /ID (node00001519) /K [ 242 0 R 243 0 R ] /P 90 0 R /S /TR /Type /StructElem >> +endobj +242 0 obj +<< /A [ << /O /Table /Scope /Row >> << /O /Table /RowSpan 1 >> << /ColSpan 1 /O /Table >> ] /ID (node00001520) /K [ 132 0 R ] /P 241 0 R /S /TH /Type /StructElem >> +endobj +243 0 obj +<< /A [ << /Headers [ (node00000758) (node00000759) (node00000760) (node00001520) ] /O /Table >> << /O /Table /RowSpan 1 >> << /ColSpan 1 /O /Table >> ] /ID (node00001521) /K [ 134 0 R ] /P 241 0 R /S /TD /Type /StructElem >> +endobj +244 0 obj +<< /ID (node00001522) /K [ 245 0 R 246 0 R ] /P 90 0 R /S /TR /Type /StructElem >> +endobj +245 0 obj +<< /A [ << /O /Table /Scope /Row >> << /O /Table /RowSpan 1 >> << /ColSpan 1 /O /Table >> ] /ID (node00001523) /K [ 136 0 R ] /P 244 0 R /S /TH /Type /StructElem >> +endobj +246 0 obj +<< /A [ << /Headers [ (node00000758) (node00000759) (node00000760) (node00001523) ] /O /Table >> << /O /Table /RowSpan 1 >> << /ColSpan 1 /O /Table >> ] /ID (node00001524) /K [ 137 0 R ] /P 244 0 R /S /TD /Type /StructElem >> +endobj +247 0 obj +<< /ID (node00001525) /K [ 248 0 R 249 0 R ] /P 90 0 R /S /TR /Type /StructElem >> +endobj +248 0 obj +<< /A [ << /O /Table /Scope /Row >> << /O /Table /RowSpan 1 >> << /ColSpan 1 /O /Table >> ] /ID (node00001526) /K [ 139 0 R ] /P 247 0 R /S /TH /Type /StructElem >> +endobj +249 0 obj +<< /A [ << /Headers [ (node00000758) (node00000759) (node00000760) (node00001526) ] /O /Table >> << /O /Table /RowSpan 1 >> << /ColSpan 1 /O /Table >> ] /ID (node00001527) /K [ 140 0 R ] /P 247 0 R /S /TD /Type /StructElem >> +endobj +250 0 obj +<< /ID (node00001531) /K [ 143 0 R ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +251 0 obj +<< /ID (node00001534) /K [ 156 0 R ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +252 0 obj +<< /ID (node00001536) /K [ 168 0 R ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +253 0 obj +<< /ID (node00001539) /K [ 182 0 R ] /P 202 0 R /S /NonStruct /Type /StructElem >> +endobj +254 0 obj +<< /Ascent 770.01953 /CapHeight 719.72656 /Descent 229.98047 /Flags 12 /FontBBox [ -1017.57813 -480.95703 1372.07031 962.40234 ] /FontFile2 263 0 R /FontName /DAAAAA+Helvetica-Bold /ItalicAngle 0 /StemV 141.113281 /Type /FontDescriptor >> +endobj +255 0 obj +<< /Ascent 916.99219 /CapHeight 692.87109 /Descent 219.23828 /Flags 6 /FontBBox [ -549.31641 -303.22266 2068.8477 1099.60938 ] /FontFile2 264 0 R /FontName /EAAAAA+Georgia-Bold /ItalicAngle 0 /StemV 189.94141 /Type /FontDescriptor >> +endobj +256 0 obj +<< /Ascent 916.99219 /CapHeight 692.87109 /Descent 219.23828 /Flags 6 /FontBBox [ -549.31641 -303.22266 2068.8477 1099.60938 ] /FontFile2 265 0 R /FontName /FAAAAA+Georgia-Bold /ItalicAngle 0 /StemV 189.94141 /Type /FontDescriptor >> +endobj +257 0 obj +<< /Ascent 916.99219 /CapHeight 692.87109 /Descent 219.23828 /Flags 6 /FontBBox [ -490.23438 -303.22266 1796.3867 1074.70703 ] /FontFile2 266 0 R /FontName /GAAAAA+Georgia /ItalicAngle 0 /StemV 133.789063 /Type /FontDescriptor >> +endobj +258 0 obj +<< /Ascent 916.99219 /CapHeight 692.87109 /Descent 219.23828 /Flags 6 /FontBBox [ -490.23438 -303.22266 1796.3867 1074.70703 ] /FontFile2 267 0 R /FontName /HAAAAA+Georgia /ItalicAngle 0 /StemV 133.789063 /Type /FontDescriptor >> +endobj +259 0 obj +<< /Ascent 750 /CapHeight 661.62109 /Descent 250 /Flags 6 /FontBBox [ -203.125 -428.22266 1700.1953 1271.48438 ] /FontFile2 268 0 R /FontName /IAAAAA+Times-Roman /ItalicAngle 0 /StemV 108.886719 /Type /FontDescriptor >> +endobj +260 0 obj +<< /Ascent 770.01953 /CapHeight 719.72656 /Descent 229.98047 /Flags 12 /FontBBox [ -1017.57813 -480.95703 1372.07031 962.40234 ] /FontFile2 269 0 R /FontName /AAAAAA+Helvetica-Bold /ItalicAngle 0 /StemV 141.113281 /Type /FontDescriptor >> +endobj +261 0 obj +<< /Ascent 770.01953 /CapHeight 717.28516 /Descent 229.98047 /Flags 12 /FontBBox [ -950.68359 -480.95703 1445.80078 1121.58203 ] /FontFile2 270 0 R /FontName /BAAAAA+Helvetica /ItalicAngle 0 /StemV 89.355469 /Type /FontDescriptor >> +endobj +262 0 obj +<< /Ascent 770.01953 /CapHeight 717.28516 /Descent 229.98047 /Flags 76 /FontBBox [ -932.61719 -480.95703 1571.28906 1138.18359 ] /FontFile2 271 0 R /FontName /CAAAAA+Helvetica-Oblique /ItalicAngle -12 /StemV 241.69922 /Type /FontDescriptor >> +endobj +263 0 obj +<< /Filter /FlateDecode /Length1 9112 /Length 5376 >> +stream +x��Yx�~�ٙ�l� ;�k��f'�d� ��C �V�%$H�rC�E��Z,U$!�� QP�ZO�}`Ui��@��c�4�ꃤ��԰��3�� +ڞ��9;O��;3�?��.��m�a�-C��*2ඕ��u?�+��r�&o|��<�O�[׷u��ضfKk��-ͫ�']z �J��[��;+��sS��&��Ȫ5�V6�ُ~�~��ܽ>��m�l�{�/��~��M��hH1��Ѳ~��v�Ё�Fp�P�uX�UP��H�zm�2I|52�!��È�a3�X��=:��5^��(9@"�4r�� w�w�}5�t^.:�b�u�+N�ަ��v��K#��~=��<�)�3��P��"�_u��;6�/��C9��g�� B�Zbd��bP}��\2��"xd# SaA��/B@��$xy��I�a&�[T�C��/�<:se�;{k��&�s��D�` +q��E5j1�p3�c!�с5X�;��W�s��4c%�`-�?�e�U��vȐ�r7$��*�]�.��3�O��A}�aew��f��(�׉��v��Y�މ�s��^�0��rj�N�� +��@; �c��em~#fD��߈�l�yT��U`�Ĺ\�W�=��(��}ץ�~#��m��:m��<�(��dԤ>#t+��h�� h�&�- ��e�.h�h��y��zۛW ��Ɛw:�X� T5�&?�46Nϓ��݇W��a�z��P$O��ytY��0��5PU��$�쁓�Nָ��<k&w��O�s�4ky2�jwY�0P�@c_��-C��h��.��_TE2�;�2g�d�|ږ!�� Cʐ]RcM��u��-l�]㒤�<�c~-��U�� 2r=��1*�� 0�F��cHF�Y��JF�n�ʉ��$J��$��5�>�|V��R��=�S�Y�i`Y�N�_�Itk��A��A$��1�B[�1"�3��!�p��#W��,��/�!��;�6�Ao��-��$Q*,+-)�gY��L��YŽL�l��ɺq��u?�V"��K��oW~FV7C�#��Y��y��fs;��s�d��Mr('��]�Lp��)�g1GU��t��15$#�,��e��,#~DQO��*b��sp�}8�5�r�R�@��s�@#�[��N�?B�#��ဩIӚ�awؓmf��sD��y��n.��/u�Q�%�s�!a��&��G��ݗ+��|��NB��@�Y�� x��,�13�Q�қe�1�rT�)��/ �8y��)��+.��2�]�O�$)c�0�T8 #n��E[4.f�m��ЁWS��A�UKĮ��r=�-�+��^��8"�� 1��m�� +l!f]#��%T�� t��T7'2C2L! �aFd$�e�a�L:��Hq��U�|B��kR�U�B��P��J�M�1�%��A�2�i��D��vR��y��c�Z��b��}z1�B��I�1��O��ǔMJ��ʫ$�U�"��E�<ƭ�<�2�>[B2��8��lW�� L�vy�#��A$N~��DdXB2�!��@gF��GT�g��{q��a��̼S�2��5�^Y\ZB$b��1�r�$�xrBy�S.*�+p��J݅�}��\��\p�Z�!��yƊ*�Z��%&Z͌�l!��Y�.�\Wc�YۺIc*އi��LH� 90�&̅ �0�&l� ;)�ŏP��4L05i< NFb!o3�Ғ�(�')��+�=o�8哖�t*N.9��л+c�#E%7~C�Z��7CjF:�w>yT�`�D��X[��R��ܐ�i!�_Pʴ(��G�e�p��%�eXÚ?��2��$�E��`Y��Tbu�6)˟"A�ؒ�!�5�}��痯�f��iʏ��V-k,\��%M�\u�aHW[S޻��L ��_�+��S9k��$1��&͋�� I�G�x��+3ӄ9FW~&�Gu52�?�2u�h��|=D��f v�,�0�B�I��,��\�M8qNͅ��T�l8Q'��8�'��D/E�P�0�e��&2��Y.��.d��R�P��B+\�vP|7ş��1��.e�{.\�\S~Af��,VI1�B��Y�4bg�$/JX��~�Y󝪪��G��x��"��.n͜ۻ��{X9�x�rFU�V��y�;�[�f�=�3[�|��/}p_y�m��f�̾��R�J�-��Ì��r~��W��k�v�1��KXK�Me(��|��}�߄�}pl�P��C/��P�0�e��AD�Ka@�� 5�k%f��2��g),C +i�\�RL�T��D��C6)H�z�>#=��zRZ��@�Ao�J��$�L!z��f}f��9�ۭ��q�M?��O4ߒ>�iiK��|d$��\e�UK�պ��WW�K%��lS@��Y�Wrd��)S�n(*̼��ri̓�E�1��/��s|1��kӯ�K,�R��f!L�Y��!��@@�C@7�R|�S\��)*E26i�gb��0�&��%0� &t��S�ŏS|X�X5�LU��Hc.-F�aD�(��0�F��0b�wS�I��K��h1��j�-4jl0#Iz�^�r%f�|�2'�|��.ڵj��d�/ڡ|�|� I��Y"p'fT?x��s{��߳�)��r�U�׏�ŷ'a��}"Sub��|Qa� X�g��g�6�aCl�� K`Cl� =��!��0l�a��z+)��*5�fd7.�=ad7�p#n�Í9p�n�Ín��K�=?Lq��Vci +��Ad��dx��#�0�Au"dV�N��2qD5�;DdAD)D�BD=D�B�f��A��(�� S�� ++�e8C�D�J�HF:T;�"�̫�t=ߢ|0s��1B�X�7_��e��-��)��8HՒ��K��O�?��+0��5�Y->Ʈ��d�b��i�v�aG쨃K`G��=��!��0��cv��Y��ь��Ex&��dÃrx04��vx� z)��).S�~p(�:pXm��=��!��0� �0 +Nu�-c��~�P��C��)~��)>�&I�n>��<>5�E��x\D`��p"�lP�� Ў�@/��P�0�e��B@eD4��(iS{��G3pRXFRH+��aV��H2�j�Ad!�RQ� �D+�، vP|7ş��1��qA5��+)N.*t�%�h5�t�_��4��\�soqzNhq�o��ւ��U�V�M��W�O�Hi�m��U}��]�y��V��6c�_�O2/%w�� +�Ծ�'�=�k�Z�� ++A��p��>��#9ƃb��U.��l�}��|I��N� �a��a@7��^��a��?�r��j�G�j!2�uN�N� ��Z2 br��âF�6��$�Y<�_��߹�N6��<��剒<+�-��o(��g��#�V̫�3�;,�v�)�:9��q�&^SM�6Fd�nr\��ԇ��O) ��z��T>%f�k�au6Q��Mh3��I�r~b&�,��/�w��-Z��b�(ߨK�'��, �h+�&�.N��q�҄l�P��aB/��P�0�e��Y�E�:��`A,X�`A,�x?�Q�8Ňa��:Fa��L��_��>Qo�a�Y0�f�z�� +36Ì�M�')~��/�� K�(R��;7q�c}�7�D+��^"S��0�o�fQ��SB�n ƟZ��ϗ��)yp��le��n�荗y�Se��vGy��7��(,�OI��]��M�M��\��!V�W�+t �I��`L�JxJt�ne��Zk%��Z��/ȴ��V��Ȋb�?��_ݘ�g�z�h�S��ܣgH�>9y�D@�u�=~1_��Jƿ��u2�3&��f?��CDD�A��h��.��x?�Q�8Ň'��ҦqT��T�Flp�b�q��W~9�UH<�c$� 3�X�Ȕ"�V��a%]�߰u�_��n")� +-�M{��C�uM��:��?s?�(endstream +endobj +264 0 obj +<< /Filter /FlateDecode /Length1 11460 /Length 6980 >> +stream +x��z `E��S�3��d��3��$ݓ�䚙LnD2��2@0�p�Mt�X�(�x��߮v\t�~qwu=�D<AЈ�O��I}�L\P��:y��ު��z�}�� 8(��)��[�-�ʧ5�/)�:�х+[�.�!�Ip��K:�i��kq��?|��sKZ;� @��(KV\�8�?q? ��K�V�ypǻ#@B"��tQkۻ_ �H7��K��-�<�KWv�)u��l#@�V�^��r�� #+[״��@Xպr��l_��I�p@J$y�E��:��V� +,�`��P +�mf�aĠJ0`��9Y�BNt��y�D��Y#��GX�F��4K�Qi�3�J��S-�t��ץ�K��o�nT��L�'�p2OA ��.e>��#o&�1�,�P3J��aXyV��i3ҳE�)�s1��3�Y&=�\/YF֊��D-a�k8��E,�P�� K�+�a�2Ï�ŲlUDF�=��,sf��i��O~|��aZ��(s�mT�(R��yeH��Q�i�a�?;�K�\��ϓ�YT��^u=��N�]�*��z�R�)N��kT�<m��i+sd� s$oTh��h�@��yjl��YLGm4�@��y%��ͫ �eUX${��ǭ�cVc�0�p:�� + +��` �(�I/�T�sՏ��R�<� ��#9�#�`~8�:E0 ˰a5:�� er�v[�5&bV@�t,�,E':��"t�^�EhC�J+��z��N�G��g�"I*у2őr��I`d��{��}"W��zm k�BREBXCV��B|$�fr�\��*\ �,A+� ��J�`*1�tc��P�Z܎ߠ�Le��U؈m��0d<�+�P��r}�Ff��zr+��6|� �1mr�G��J�I�1OÁ�� hp%,hg��T� N2W�a|��$O`rث��a/�p��B�fzo�?��+�?"��G�-�k��߽O��g��_�j��F��sE�;�;w�|��;�w�c��G��p��,��S��O��ї��v{&{��ۚ�s��03��6ݼi�&�)�I��n�U��r��>��- +>��i�;IpgZVQpg,W��3��;��`G�&8��m~��<|��qVq�;��m8�ȃn��n~ۃn��-f��-I�4��,��q��X��/p��]�"-�ۏ3�E��7�|� ��O��g�6�O��~��i#o��og�?��y{;��+}��=��+�_ף��׬M� k��7�j��Cz��U]��ً��]��.�o�j�b��3>�!˫��4��S=��^p�y�31I�`w謶x��l�Ũ�x{Bo�f�fS�`3yΐ�8�V��k5�qZU�Z�*�ZF�'�^E�RU��5�RC�AQ�ֱL��e(EVC�� <[��1V��T��I��Ԡf�$1�[3��t1�[#j��6�rSH��̺�FQ�n�A�h*�3�i��%q�S4�I Hwύ7:��\(�M�jf4��I!1O�lH +��z;::;:::��H����awyG��R�E<�.��H��M�D�};�ދOu��z;:/��#t��:��R�Nog��c�g��|3$R�ߪ)Ԉ�#�@CÈE, #q4-�4t4=�44NF#8�L0�`��{Xa��+=�xOOz��ӓH��D�I$#�~I�;��wpɘ�~7\�[��聛~�4ӑJ�A<�d"�~�,��o�E�>dү�1^�5�ѯ�S�~%�p�+�!��@>� +�CO��GO`�� E��(�_b��QD�D ��/q��PL�@)��/�q"J�瘄��(��9�e�@)��H?C��1�~�j��O1��SԠ�~��b�觨�d�)�PM?A�� B?�t��a��4:�Fg��c��0f�� t!L�Ø��c�E#��0�~�f̢a>fӏp>��Gh��!�`�1�~�6̣b��X��,��,E�@��\��+��h��c��{}�XB�ǅ2^��}t`9}�XN��b\@��Ǳ��q\��=\�v�.ǅ�=\��{�-:��D'B��ݸ��*��o�X�5�]\�K��e�\��;�W�w��p=��Ћ.z�e��n�U�n��(n�Zzp =�[�C��V\K�b#��G� ��Q�&��Gp��#��z�6�č�m�%�ݸ��=��@��>�B�`3n��a܏M�0�m�0�ʸ wз𠌿C}�ǝ��]��{�!��G�K�ğ��G��Geq?=�~<@�1l��gl��҃x��o� ��!�d܁��؉��o�I��d�+�D_��x��ƣ�uB��7��]t/�b7݋c�C��{�.��=��t��~��e|��@��ݍ��:ݍ��ݍa�� ޤ��)�]�o�]��.|!�8B_� �M_�W8F_��2~�w��8��q�@%�k;�� ?/7'��y�23��<��''%:��x��b69�^��ըcTJ��P��o��:].W�-;�,��%�tF#�O:%��r�r��X�.+��Qy\�Y$�(��ƕ��>��~R9��b\?��M^��TH��].��'��NA4yGs�H��21FWX&[E��}��7pX��ն��Z�5�lk��X�(ٱB�mY*� +��\��z��B�[2�t;`=��w�z�OԔ5]��*�U�F�X�.�.r�� {� nih:]�0 +%�}B�| )��*�O��HB��)j��Ҙ�[%=+��ɺ� � 7�X�X��K�z{+��m�m�"�eb�Q~�qN��+�Vʸ��3��0j�%SC��H�h��_�a��:��W��<�Ju�7V��3��?�wn^��W�F�g�i�aO�Us��7V�4��F��L^չx#��s��ELl�WhL&ә:��W'��W��_�є;�1q��i�X��r&��x��Z��X��+y#0�U��5�8��v��tn^�.w=&�9?/G45�U��3��mBB��k9/�>�k�x�N�s?��uJS�Ye�d�8N� �F@��t&��7� c¹�bF�Q�� 0�p��Ͼ��q��X�3I��`�F��+y#�:�*�Hҙ%?�C�0z<�3mz��c�Ye6��U�7 �9++�LϾ��bL9��$ ŎIN��}��g�X��5&��$ !//ot��oҟ�e�<�\\�$I�h�[��!r�S.��ûe��q]F��"P�d��Kot+�o�׏\�8I#��F��qO�"V�3 +�9=�4燇��ss<UL�2f�NI+,(�ϳ1�]7�rg��3�c/m�n��2:��Q��5�t^Cpɝ��r�r=�Cw�('��96�4 �є��@�㷖�^�kc�V��{�HH�icA�^�+}��J�J�KY��d�K��y��w�T\h��J�l4��C�9�ĝMҳ��󊤩�O�Xl��$Z-IXW��+4��g��B�205��{n3��:U�ks��a�]�^�ɪ6�1W;O�Nl��9lY�� b��[O|�7,fu%�:G�k,f-�TZ̹/�T�2Z5��~�0k-l-��7c��{��x��bb��K��{�K�Da�]�ە��+�II��(_��D&M*��3��iʮ�-�-��J��"6)=]��+�&��I�5l�#0_\�/6L��ac1�xhH2q�)�8�h��^ǅՃ�9�9��1��f6��՘�/H;Uc��%�ī��ڵ��n�r��΅��Ą��tGV��S�e�}5�_�B�צ4ڃ�'��F&��՜��Q��yi�'��ֲq��)#)U�CKӌ&Em��U��̚�8-�:zL�φQ��l3�c��bO��+�/�^�{y}\�>{�}�}�}�� +��n/�=�#�� }�\��6�- U�Ӯ �sb*�=p`��7� 3�혚k�'�s_{��yy�C��(��7�ù9�=q��e{E/ꠅ�}�)��db��N��IKa�0ֈS�� +Xl�{+�}k�.ﴪRJS�k;��s~��Kr}�Ml��_<�qV(X��5�Dcu�a�*�'��`L��j��Sr�6��N��L��D�g�Y�� [�ı��dW|��G^��,4�:g��UV��p�M��)ꐺ�py�2�/a� +.&��~�w[��Ix$��z��sgQ��}E�om쓅��Ò=�yyC�N�J�A� ي��t�@(,�d�Q��G��c��w��:�(�F̆��|q�$�웾��l�L��hmp�01v��,��;>)in�zk�~SXG|��B�=vR�h��䱴C\,Y�Ə�[�w�4\UXX�vGi��X��ҩiS3�.QcucT*�Y�t��S �>;��vt�Ĥ�G�#�.��I�o��C��P~~ hF �W:\\q)i�u�t=I".��b�]�?�A�ǜ�ZPP7'?#�XF>IMJ��͎\��K&��Y�:�B{��X�m��ú�to��VW3��\q�u�-��"q�\q��c +��{FFJf��Ĺ�+/OT��M��U�c�� +�}c�R#1�HÜ>��rc�k��f�\�%=Y��㈲@��o��2�m��)x��;1��6,S|�d��I��@s~~�Y� Y��9N��ҩ�R��*�&��s��#�HA��e�D�(K~^�⮷�ӧ��e�6�r�ܝ�.��.��;�u仑O6�3oI��XＮ��w��5#�)��Ǚ�ک��=�6)�� @b�{��ֱ��4�&_ +��6��^Y5��u��Z~Iy�8�٩��۰t��y��uw�ܜ�7\��YU��%�$W�2�<��ʂ��>n��3}D��6s�ٖ�Վ��i�� +��e_;1l̗�I~r� +3֢��Xx�#�$8�Su2�А�h�q�zq}��74t��P��j��\�F�*�^p�ԩ+V ��e��]�Ґ�U�w�L��NGD�fI��?�-sw�F��y��'�I_rT1Ec�D�.��l�F�J��{L��`%[h,L-d3��{9�S{[g�@Eb��Auτmcwon� �AYU;gg��=��(��F��aSts;]��+Ɯ�,��[#�p~��m�\��Ho5�TgT��ZK<��b�7 ��4�E��)�5�[�ڒ��l��Lr`�xK~��Q�e{�J�Ĭ�5��^iKH�q��Ȏ��ˆ�2�vW��ilε�S�9�u�-�l�ZÓ9�}X>ss�i�\�h J�T�J9 O�1Sj�M�;t�e��6M�6��7��f�XfN��o��Y�Jf��+�W?[YS_YY��ek��6-��ey��ɵ���&�EVW��~?ڂ�l�%��c�gW�]��e5�iӭ^�ݗ��Ùm}V�5�n��8� �@ P`-��zRX�!�$�#K+�^�Yj*.� ;z��U/�Ud/N�9lm��$Dװ��̩S&�ӍLI^��;[g ?�[j�1�n�s;��1d�~endstream +endobj +265 0 obj +<< /Filter /FlateDecode /Length1 9900 /Length 5588 >> +stream +x��Z \TU��{��b�a�{��23� +�L�@� +"錊 +��& �� e��}XZ٧��}�n]��Z��,5ð�Ȍ�ì,8�{/�K�.�o��{��<��9��;sg�� +�z�3g҆u��Q6�s�xmQ��_��B�d� ��m?e^��O�ɕ-�8�?3 �}}sACkb��ħ]��+[?w~ +��Ў��%˗�:n ��(/knjh��c+��pKd@�Hi^Ҷ��6 ��K�7��kf1��4,oa�T�$UÒ��E@�Z��j\�b�޲��Gva{V�`�-�`!�f�C:\�`�}S��,q��B��%�v��ڻe��e�� nʊZ|��wn> ]wh�U��h��ʝ P��З��Dg \��p��X�e��k�� +3�:B�y$$��W& �3�,�P2!��a��2e:σ��B#��f��I{CVIޖ{p?��v��t0�g��_%�/p�_��3�/�=�#a�c`�YR�*T�jP&�³A��2�ZL � +ġ (� qA9q*Є�X�X�80K��&,C+b)��|d"YX�.��_��F�E��l8��,8��\�u +�c +b>�a)Z�W� �xf��{h]2��c��)�u38m��5q��]�V��u�ײk�Wr�ݯ�\�� +\�� +|��_�/��8�V�ښ�Q��-X��b��B��}� +�� +ֿJ\��" +�^q�R� +�8�=�T?O�W_��y�i��mzZ��`��ܐ(-�� &s��54��W�~E0O;�f�� h9�� @w`��{c��ަ��M��~��l�63՛_��|��lަ㶵�8��ٻV�s��ӹ;�霿=�[٩��4r�wFp��H�4+�̊N�{�Sɭ�$��&��t��/vĘ +:�u��D ��h��}��Z�f��~�9��~�z̠�13�i\= �� Ә�Y�4�c6��C?E��)\�� +z +��S�'z +�d��O��I,A#=��DOJ��I�`=��e\�fz�XDO� �� \�_�� K� ,�U��Rzף�� ǍXF��&��c�m��e��n��V��~�XN?�m��~�N\O?��~�;p#�+q=�;q3= +?��Q��.tУ��У��^܋��q��}褽��^��k��CXE�`V��0��=�ý��jzO�>z�q��=xki��e܄��xZ�_a��az��z��Qz��1z��0~��-��6��"��Ѕ��!��^�&z��4=�W�+z��z�ó� 2n�s� �� ^�o��.��[z��=�?�Ez��{x]�=�/��'�L�ß��`3}o�U�.�*��-�~l��;��N�F�c^��`��{��M��>t�}xo�}؏?�}x��6�M�6��m��ѷqoѽ�@��{q;�^|��tz��Q�{��=�{�n��t7�cݍ2��~��]��d��]8�t>�A�}8Dw�s�Ow�ӝ�Н�=t'��k�;p��G��[|D��;|L��{��q��v��t;~�I��e�)��UB��0D�.�Q@z9w��!;ˢ�� ;B��tEG��Bք��%�pd��5��SŨT�j��x��Y�Ͳ&s�à}JG�rJO��l�'�X�v��V�)�.i)aU�� y�LО;v�?�^��}��8��r�%��c�ǲ�ꉐI�2��rs +�N csu4�1�I$��,�%_��uy�i�$-�✜xOy�/��e�u�M��c�5�~C�if�0ϚQ��Zn�4�ߕ��N��Ɉ��Ta��+M��A�5��/f�ǥ�� j&$$ڐ�V�@�TF��f��R�u4;��QEۏ2L�=�FM�b\R��Z8+kQ��e�P�~�n*6M6�1-6�h +5��#_��K�k��1S��qJ*!�6�s�vB��9��o�~�t2ݥR��M�Z��\��^{�XnN�S�\_� %}ґ� ��/;��E��4!��R~ސ��ew ɡaI�:ܜ�#��M��.��ڌ��%)ΙS�6�X��~,RFFg��.�p|��dSy�S�N�T�t�I�qW ֌��f��,K��T$�L� ��]��D�L�t�71��"t��َ�H%K� �愙㧖;,`PA{C*Bհ Waa�$�W٢�!/$v]�>ڰ1Ej-및�<�Ľ��h;�yk��ނ��}E�k��s}�?�sr�9Q��<�W"�i�Yً!i�i��;C�f��X�N�O!9�8��!MX�3�o8O�{kWےL\��IИ��0SLtiFcGlb�l�%5�Ĩ9L~5��ɴ�H��R}�AQ�=45ON2��LF��n��/vJ;=a�L��#�c��nל�tSzz�� 'ܐ�$]�.[QĎ �K֍Ֆ�N��=�L��7g��ڱiUQ�gT��xwfd��<�b�f�s�o.[w� �;�^�מ��u�I{X[ؗ��t�9��r�9�ss��d��O��C�e#r�H��`��)R��g�`Z�R��`�K�rs(��:Ē�S�x䃴�Iŗl�mT'�ϙ�5�hYFfqeUu�Ɓ�>_�؜c��D��Bn�M��V�ש'�_6��ԉ)O�� I��d[��ץ��Uiz{2�Dv�EXR1��q�W��eE1�2"gu�1V5��5+�B��%w��oK.5�*�j�L��F�#E�-ht��c��0��R��U��u�Ѧ֥�j 1댡�A��GD��p:�%N6��Δ��MH �z�zl� +gN�پ}a�t��Q$,��Ƌ��Xc��Is�T6�+2�L��?ߔ��c-�G�p?ZW5;fr�˄4S��XȌ�А�J&,4"�ʔ��Z�2^s_��VWQݸ��y�9�Y�L�ݖ1&=-՚"$[x.)1�o��1F�:�&*R�R��(X��NĸRO�)�f�X,^G��Ӻ�Z�_YD��|Q��Փ.ԧ��˅�2��'DD-B��;� ��˅2Aˋ�_8�S��"�K�Ҧp�]��Bi��\h�"��,0��.�H�]��D��^�D�*��X�2��Ⰻ��[��&�zېd��Z*��EW��U|��W@�y>��Qhl��n��^�f^TX�2�"k��ͼ_��!=��V��S��.�J=wX��G�;ܢ�&Veb��̬��~��a�g��"��s�y��\氻Mt�I��a\S��EҘ��y��UM�\�� 7u7� ��n܍ ��KEW�|A�,��@��P� �� fyD�l�y-�ή��J�̃a��5U��fP)�|"?�Q�D�:N��q��''��K��C�Z��~�ijB��o ��B��/�r��^+��-n/��#��ݺ�,��fR��R��zJ��w�Z3T�V*�J)\U+yATX+��^D��5��:�$�y��H�=��&��i��E�E��U�9��c�g��c��%��6��$K��x�d��\��,�f��L�pj�1ws�HbF07 �EC��53�A�1��n�|�ks��t�_+�{QkC<��b/�Չ�Kћ.TM��Y0� �T��$%}��]Y9��EVN��٢��u��&z��.�W��*@hge�[�{�\G3�<�^X&�#��fdXx�|�`x/��~��onhV�Z��^'/b�g�`@E��|Al�z�x%�� "��u�;�e��E6��3�#v��EW��l��ef��f��<��G��,�=��,��w�@�E +��"��F6u� �`�-�2Gs�UӇ�� یG�R��;��l�C�PQ��a�9�7$""b��#��a#��Q��䍌�%�H�R��[QXTT�� +�F�%�HqP��Uj�ڟ�7\6��R�Ӎ�w��F��a�R��Q��5��)��<�3�e�a� +7��)o"�a��F��Č�w��Q�7Q�r��~^�O"�6�ͣ�)o��'��JLL%�H��[��/#�� ;l��,�(yG�C4�O�V�(yG�C`��M�f��/�a+��%�H�2K��Q�`K��Y��l�endstream +endobj +266 0 obj +<< /Filter /FlateDecode /Length1 46364 /Length 20172 >> +stream +x�� 7�TwO��L�t�}OϹ�;��Ӱ��.�.��ȍ(��g�UBT� ��_��A]<�b�OD<��J�W�!�(;��U��~��C �VwU�t?��S��0��<��yU��#Ξ�Ǟ�@9��Ʊ@@/-�[r�̾�l�8�Ȓ��,Yq�ba�,��˗.<��y ӷt��o��y��K�7>m��3��K�^s�=opp�W�\0߰3<��gϿ�O��y��Y{��@r��k�ɰ5��s��W]�:@_�4@�HQ�@m�}��u� +(�a�1{��@�Ҏ��33)�=�-��𙑉�@ΨV|��<��֐�_��>%�{�2��0Os#p@�.��Mh+�@��IU ?��|��T��i�T#Ҥi~?��Y}�~��F:}M�K��Ғ�X��֏�~(�R(�j��h�I0f��~�7+ +i�b(�*HB��q0��;��B��'��mY}gR�l|w��\��`q�L�8��̈6*�g�e�� \X �`%�~�&OZ@ TA,��p.,�e0��E�A9�C �C�@T�TXK`-��p�z�Y�2�?L�e�΅��V�bX~h"��H>��޻�0V� +��TXK`)��h�&�y,��P��Y �-`��P�r\�_A�a��M�#�fr�W��b��a��zBg��e�Ë��0��'�a��Vx^��a,��d��n� +��^��a*d��Ih�6�~c+�6jN��10�rX�"?�C't«�Z�~#@&��b��a��P�C��R�Z�F܇z��pc��t>��G��܇��G�C{��P��]r��m�Ts��%I�v9�;wUH�֭[�v8��w8��Jܵ�w1p�x&u��Hz��M�_�$�%��?��#.�W��x�Gֻ�-�� M�M�{nA�lrI?��N�R;6R�7UJ/nrI��n��܄6��H�C��~�� Һ��D��Fj`=�]߷��tq�� +��!�!�? �$��;�.��fכ��INGL�Yc��H��$�F��xN��Y��%_�GI,J�)��!!��I�<�A��V�n�HB�F��4�j��J�6�C��i3j��icӕ�� LMW��\眮=]ݝ��LO3�P0=mj�=�k9q�&w�Ԅ��MW]��J��qoza��t��;]�[��㫳)�wҞ7��P��7B�{��˽�á��;-��4�қ~'Լ�÷�p�X�z �lM|u|��x|��x|��V��^C�ou��j|��5��m�&�gm��=�ŭ�F��5 ��B<��hz,��*i�� \NsG��d?PoԔq�z]{�vH��~��`5X ?n��m.6��k�Me�jǝ��5��z�y��vJW��!�� aS��+�?~5�~�~�~�+��j�<|�Q@j%p�hA��t��+C��20(C ��Hr#��q0�Q9f0+��f�k��E�l`U�;�`W�'8��Er78��.�+��G� +|�Q� +x��$� y��_!�K��c�Q�#��Q�� F+�f��@J�#��3h#�8�|�I�&@��)�C��)L�V�S�mʧ0�)�B�W>�N�O� �'0ڕA��A�N�0Y��С�,�T��(��=+�0�*��|g��c聙�G0f)�<�R>�^�χn�#8f+��|��CX=��*��|K�W��`9�ς3��a,Pރ�a��zIy�`��"��TyV�r�=X˕wa-�� +�]��w�8Gy.��"�S��*��=8W��QXk��p�r.!��p��l��w��p��l��p��6\�Sކ�a��\땷`3lPނ-$��*�Dy��K�#p lT��V��r~��#p-�@9��2��.W��u$�6+��آ��Eyn��7�&��V��p�rn��a�~��)o��#� ��Sހ�$�nP^�]$� lW^��p��{�)ܤ�~?V~w��s�Ey ~�)��ݰCy �!ynW^�=p��*�;�W�p��*�]ʫp�Dy��+��TyH�~��)��C�s�ex��/��Wp��2�/�Gy�CZy �=�K�k�Wy �_*/��Wy ��*/�o�~�Ex�俅�E�(��Iا��A�< )��xX9��#��yد�a�r^�G��p~��q� �D�῕��<� +O(��k�[�y�=�N9�� xJ9��i�� O+��xFyނg��mxNyށʳp�W��?�A�Yx��!�Yx^T��H�!��<��3�1��<��|�)Oç�{�i�^W��?��9��<��V��/��I�+U��ʓ�5��< ��=�w0D�|�� j�5�U��e�Ғ�x��0Z ��z�.��n�Z�&�(��N�iԬ��)�(�h��T�݁@��$w��q��`�F#��.�|��c_�xr,��PS3�x��sYҀ��'��e��P˲��iaoo�5��[!}��i�BM��%ŰG�k +5�J�!��ۃZG#R�Z[��P�ї�M�4i��-�i*j%�i�e��#� m��J�Mi5y_��ܼ�2zǔ��www;J��ʬ��e�ؒb�H�g�)'��{.��e��E�^�$�@��, �,��V�7�,�,��pl��<��鳻��ݹS��m�d��>�� Xh~�;;��3��3�S�Z�+��Ƨ�޴�? S�Bi*R��E��yA-O��w��*��!��?C��?��3lD�3�bk��w��֐�us��J��!�ڼ��}s_K�? �]i4@yp�;�zewZ�]��K��S�R{��s��4�!�>K!�D��azW�ߔ�]��<��wu��1��9�a-�͋�)W0:��pfIq �?�+{�3��w��^\��:��/� J��5�)��D��ei}��A��l}��E��l�rӸ��Dc�/)N�7�!��PZ��U]�ݍ�~ј�i��)��-��(Ȟ�=)�AK�%4��T OSMӻF6�%{(�'�.��GF�O�ݵO�_>��^ +QM�c��K`Fq�p�Y�R6 @�wp@ϛ[23��e�i�[2��`f,P2]��츇��/��o�/��0�D�ߩ]�E��4L�Z��uwח@7�!�l��zXNz�ܝ*��4]��5�+��N�� gB��J�ovȃ��ߩ�߲n�#w�g��X��d{�֕��i�޼9{ +��7ovou��o�@��r��H��$U��S �G��d��O��+h��4@��9��S�A�U��vDb6A9S ��n�a)sL'��>��˪�� +�dU��a�p� Τ}��A��3��L�YP�7}��{�� VT3��*��:��N�b�3��l�q�aG��1��R]�f*˘��R�R�P�<��Nhe��s��`��S�� B�_�Aջ�׌x�3e🹏��~\��=�"��=�N��t:�N�ӿC�q=�}N��p� ��h +XlXU��T�*5˪Y �bU$�A �F��bU*�b�e4,>b�հ�)�w��/�0��Z� 2��d�YX�ͅziԬZM0�f F4�j4'0�b�1�cP��"R}�Ԟ��:=��E��c +)� +�a�l(�4 ��@�k? h8��V�j�QkF��VͲ*�F�U�J<��1��0�x��;T�ĠF�i0��N�A��M��:�N�R CO�Cӿ�AC(�F��à6�AV�iF8�h�#0�Qk�yj4��Ԝ +\'kp��A C�`0�1H�� 7�k9�vѱj�FON�V�=��AƠ��>O�h9��}�� x ��h�Z��Wk��8^+�{�j8�A�� O ��t�r:��$=Q��A5CF��0��aP�Ǡ.�<^̌� ��4�� ǩ�,��ܩ�,�d Nc�?(a��<�A�"��j�1��u5�k Fx��u�|O�V�հ�V��T��j�jV��s��'k�;�NO�ф�g2�N`�Uje�8�6��y^�j.�<=��)�� r:�J�uZ�V�au��I1x҅��0��f3�� V��aP��A�^G��ޠ7�{��8�ju,�UuZ�N�Q�8��T0h:Y��Jz��"�Ah�<�zA j�� {�Ԫ�<��X��4�� +�S�n�5ПB��ӿh�гZ�y�0Y KF�ׁN��6%�h0jtBV� � '0�㵼V�#4�:��u��A��ys:��$=��4 +8��1h�t9�'��(X�=�x�^��Y�g�z^�sZ�^'�Ne�`=Y��Jzv�=��k��bP�c�7�AA0 +f� #"Ơ-�� j5��x��`�ȟ +m'kpR��0��b��/;��1QL"��#&ў��cP��V�1�&�T��58��v��(,^C��j�`�M&��`�̢��h1��=^�sQ�1h� :��é,\'kpR��0��^� b��i�{D@0��Y��l38t��p0�̞|O��78��iD��(E^�5�©P7��t�|:��$=�ߟ�DQc��Y�� M��1��x�͌��l`qX�|O�I0 Z��qF�c6�Lz��,:ē�}G�N�ऋ��'a��Ơ�F�9�A��F��`��-�E�`4 fQg�h�&�k1�M�A�`��1��0�"�H>J�㸴-YX-`�Ys�N��.[i�{��|Of��j��6�΢�,V� �&��{ߑ�'k�8�NO�ф��(��:QG�X��ܦ��A��uF� /��|OV��a1X�O.�/�GEfW�W�O�@*��{rz�^�� P�s{=��8��pW�'kpR�x:��$��|��K9g\8*[=� +|�E��|On�Sr�<�(z�2�W��f�]�>�Yu��S��t�Mz��X��6�,xC�\$U� /H�}�� -\��/h4J�ʠ�;��7�}��۩�d "��/��(��5X`)��3o�xf ��F��m��i��Lw?aSf��B�"u�R�|��w� +I��>��݆n%�r=��7 ��d�h5Ԗ��2�>��o�##JR�3K3��0ڍ֡��дpƍ�~�V�86��n�y�3�!a�j��2&)�o��`=C��e�6 ut4J�F�P� '�_��{K�&:|1��a��и�Ҳ �&_�/4��!ֱ�@�S?��=%Q ��ʐ�:�6��,/C��S��໾J��AQ�%q �➣A֊I��RlB��Cb��ۼmJm��umG.��/Q�0�2S�!X*{b�H��]ǵY��YV��f�>p3�t\A~|�s��@�4E��y��#Y�O��( ��g^��D<޳j0n�L�{��a�P�!+�e(X%�M�F�6�� +)��TYabh95��u�y�<�;&�$?y��{w^y��l�v+��Iϭ��f�O~��;e��32�O��ꓷ3��g�1�L;hA�d��R��D~�#��S *��4�e|�KGM�[@�� (�:��j��SqJf/~Hn@yW��8Nd��L��u�$��?u<��% �I��:c]]��;ؕ��龬4�Z'>^^ C�@5�4V��i|j�P?��Ƨ2k��ԡ'n�g��fh�CCf}�)��S�z�j�9�j�e��نx��2�uj�;�K��.u2��9R/��^!��H��EWD,�؀rHv��b�b}1:) (��!�mX֋�p8��7�,b�(�,c/& +�"r��F8�f�_r|�P��X�4��Sy�I��TyG�� I��u�$��ӓ렩�.�'KKv�2�4�%q��'�_�*�k�X�'Gx(� ?%0}��C��U� ��d��Y��޾��Ɣ�Մ�� +��h<~K�d��MU%ۺ�߂^��3��ʉ��׮v�tƪ��kιj_c��:xo�˟*ݙm��)=ܹ@y_v��Tr��F��hE�0(�^�V��y^G��UO��)��*��r�"JMQ:��8} ��(� s��ʄ�lX�:,�8��.�c��#�",/��һl��q&Ρ +��j(��-s��3��r�(��g_�d�`��˱�[�fzTi�S*�_v��d�� I��q��<]J1tB��`��t�*�Q9� �+�u��xu�okE�wf�r}�y��׿ �[��S��}�s��P��xep��ϝ��h��X��A"B`��;XjSv3�tl�j�a�L<>e� (_��%\�_�g|[��9�(_Ʌ�$��&+Æ��9��:��DU��)�"��M��u�x|0�Xfϕ�� +�0�`+A� +�y\��V��>��b��az2�?~#��-7mX��>��Z��΃�cc5��,��@��\�ƶ��[�W��sj�E7n;{ٵ��nZ��-��}ɶ�v`�0=s5�g�� +�wrQU�&D��k�[�Aw0m3�ٶ��%�ٶe�՜�â=[+$x�=8��Xg1X��z�B��u\M��F;�}3��Y��1��b1ڻ9aDF�z�"F��1b5�3�S=�#Q�9��/��|6ΥE�[��k��L[e� UHG݄��k�ev��!�`�5��*2GT[��Ќ��*�k��c)�*JՕ575OC�C��Ǧ5/+\^��/va�f{�`�6�*QEUUi+�9v:��t;�Qc Azɺr�?��`��mx"�� +��p��o�%�j��F��a�% C��*N3�|)�p4�*��z��c��8��{��%�n@9D��:zt�"g��$�� 18�$�X�� җ�jcV�U*l��J�Շ�ySl��1�Y��9z�̖��u�a�\��o��f�T�Ě��[F�p�ۿ�??�a�ڻv��e[<�[��b��_pm��s�D�>��j��.�پ6V�8u��ŭk7�h��Xf��;h��S�݀`��S��p��=�A�RL��j�Ձ�� ޖ�\$��L t�Q�Qk�kb��+ML�I�r6��s��aN�!��w�w��;Le�2$�c��&s�K�%��} ��`��`eeb0�mJe�'5��榼,޳J,�P�U�,��'+͡R-E!b��vh��G��Ai��Z��-��h�s�9~�0��z��3fLoө��a�5�`��:��3G~}�{3�Υ�׏�i��9� @��U:��(Y�m]f=�q��i�Ku�� ׳ ��a�C��GJ��O��a<#e6��;�tv��4�j��*��d�r��5��ܨ;�h[��l"�)9�<%��۩۹��9"{nX�\^��=w�� D|�k�J�L+F�P*�*�Pe��Q(�,��Ș�y�3Y��1%U:�׋}��r�q3�-��7wLjs�$U{;2Kcv�\��+"S��6J��"��@��Pm��&��a��C/��:23� f��a��J��,g%�h�r�Ƨ)s�e��2��2��V�'�T14֬�6��L� +V��c��Zdg9sG.�{��b ک�<"��(��/��fL�2gBm�3��<8mHT��8t�7�G�+0��Um �L�N��Ȅ׷�3#^/&��r��b��\/��^(��k�jvh��@G`^��`yk��Dޚ�㬼5��8Yah�h��5��e`C��1�L*��=G�=��H U��w�zv��bk�g+��R��7d��ї2L(!�V,�� |=�4ft�a��(��f�֭��7�&�B��L~�?@��a�ȡ?��,�ԛKH�^L��P�ngs�}��81n,��"a]X,�4,C[�*C�6��r/&m61A֩�l+��qb��=q"�4��9�҉U��O<ӼAGue��e;�jmj��}|��f��u��6p��?�o��5%��>��h��>��4�0�F� ��z�c�-�_�}ZS��rƭr<��LT�*�i�ر-3��1��/�G�-c/a#0^$1%(�ԡӎ:�yh3/�m@�u�u�I�F�L��O�v�w��{wvX󐳒)n��O�b,c+1�}[�αq��΋�h� UV�0��^�oG��m�� x,��+IE��ġ��2�� :��̞ϟ��݇��2�μw��gy�~i�2a3Ꞁ��tK�a�}w[��Sl�,�Eo��B�%G��D_|��ە1�D�%��B�lwM�svL0�,�[!�2��k�WK~��U5cD��h�3� �ߺ +�~١�f36� k�$�ˢ�aī��^,v\�Ŭ��v�֋;��a��;a��_9��H��ſ�~C��h�d^�}�_�9D��inj��t��9S#��P��l+�Ifv��+��*��cY��|}�x��aY�X�P= ��j�j4 V.��[�[\HF�.��,Z��0ZX�D�[��:m��Hm@/p�Kn�۫��/�b)Tcu��P=�|H��JF&K��G�^0��`�� χ}��2�k�2߀��#��y�tZ��FvFik��N{�C�Mӻd��t ��u�T==ݕh��GM��˲.��R��S�C�Q�[ +[m�*x�E� +OE�R/�m��C��,p�\>�4�Eף9f7B�yfY�G�t�� :sy��|��<��~��16�g��K�N��E�}� +}Q�Y��fOkD@a\�~��ʋ� ��*16��Pce��j�O��]��s��Z_+-5Ęk�� ۈ��$��j]��p8VJ[��h +`��?��#Uը��։��T^a��VCY�P�@�\k� �*�rE�o3�nx��¦��O�<�D��ٷʣ'T��lˢⶐk�k��]v��O��M�ѡj��L� +�8�j�^2g��xʶ�W�7֙�D��n�J�y�蔝�0��:�9'��Yb!:��)[t�d4��l+�M֝�,�I�炘Dw�V��zw�ѭ��-��[݌��7�@X�nL,tX�n�n�1b��v�c�x*>/�!Μ =G� �CG� �Fe"^� l�E� �#$��p�&~��Kx}mina�'dZ�:ӳ��3b�6��ke1��|��@RP;��xN_^�\@��Et ֽ�ê!��C%rU�0\LU�&�g��/ҫ��Z`\;�t�&�M��1B�c٭��1�p%)�\�M�Vo~��Z�V��Xw�g0&��|@��,��~�_��"D�NȋH�A;�'�e@y[��+Y�AhH4�l�=;��;�̲�ӌ�ӇY��m�f�+��<��?� �g�F5�܋�3��ѡ�#�P�P�"��@֪��=�A��v�?2'*UC��Ύ{:��l9';��K��D��EA��O5��fߺt��I4��m��B ��x�iV��&��ZWs^#պ�*�Ӄ9��j�j��+{(��b�^�h�i�5KR��R��TW^��Þ��dYMq=��[�-��a.�m�� ]�r�佄�Z��8��Z�-�b�%v��Z\�r,c!�`�8-�J`��XN��K�1�А��%�u#��HQ�uwg�A�� ;��\cɆT\�"KS��Ӷf��X�޾DCG~r��`�9D��fd�cKn��ن6RϮ�V#:c��|ܙ��̒}�˹sZmvI��mH9�|��X��к��և� �Nk�z�NWQ>,֔�8ٳ�?�V�I��`J��qQ��A��ClP�4 �� ڱ��.~Vhf��pnb�vy�b�y��EA��.�iC��r(�$R��j#Exd˂(迄��p1}��wp��n�n�~9�\w!{��B��(�u�0Ǫ�d��d�+��D�\�W��|��HS�o�_��*ɡ�!:�,M@He�i�ϳDOOe�C 奄L�?��e�!��4�s0(VA�`Pt�C!-�l�J��Gu@��s�'�>��M�-]��䦚&��R��љg�E�|�R��1��*��!��i��a4��c�TfɁĥ h�2geV0ޯ��~O�#X�i�{A��{�'�0ڑaL�;�XaI�1��M��tQ[1��&��!�+��v[�VLڰT�X��l �ʂ T�D��:�j ��n�~ +��h۸�6hcnX��V5VEE�z�}Smmʶ ��H�cu�~��n�~p�4)1)5�n�J�Դ|_�'��l��֑�8��<��?41�-��\G��Z��:�@��x�<��N⑬: /41�� '��n@^x��*�yG��+�ڥ?I��L�ѳc�[��95\��e:�e|Si�hN�V$�T�􏚾��u{�1��Ɋ�YMm+"&�u|븉��S7;�K�g��'"I�h�շ�Nx��Ȇ�׼�1Z*r:�,{�i�[��S��e��D�I^N��rδ�/��"�:ubqc�m>� Y�j��]��f�b��9n�d� @L&��Gx�\(O�w��h4� +�,�d�[��Q�r݀�1y?6�:l:u��&��Б{� (��x��x�{��ē�w �R��1��ʅ�x�(�d��ƈ�h�C:��;G��g��7喖 �U��0w��r�z��=�ϙԺ,Ɔ|��ɋ�G��6L80�>l�zJ�C�Qc�f�|2ᦺ�q��Я��qʻ�-��G�T ��c�➰�HU�R� +i"5�ZIm��RX(A�bd�@5�c;'�Á��baF"Ӌ�Y�1~�ӟ��q��[�Z&��猖Me�g^m�.�8_�M��\ո(��Bg�w.3�gP�ڛ[~��:Cn4+�b2��v�gC�F�E+�1�F�30��˦_�P��JF�]z/��!��@��%TL��8��LuE�T��V�c;�媕l/��g:O2ط�d��j��vA��-3��z�� @��6?��F�62 lŉ�b:f�>��n�!v�#�Ñ��?�s�V��$.!��)z�@��yC�}��Y__7��g�V�TVZ2�:R]?��V{gj��>n��z�Mh��P��{��[뛦�@�j��.`��e�#szo s��.ȍx~��N $C��+y�!,c�ݢ>,n +��ܥP�~r��l$󆬩�3xZ��2Of˾�P,S=9E�9zt��N�w��;K�KQ��W^h��o��?of��j1Ο4�gU��+��"��fl��޶�u��2pn��.V��J��r��ƗV\�WG�j ��Q�T��-�p��|�[��ֺ`rHv�$>��,��DY/� JY�9;<� ɢݱ��y��I�6ġ��_W%:G#��v�DQԝ�M�w��UKF�6?]jm��Z-ɱ]'1�A��s �c�1�g<�Z�ǽ��;k�[��+��4�� L��`��^{ė��e�� +��żV�'�Jŕ��z!o��J�ÿ��!%�9� +�TT�8!��Qe?� Q<2�XZQ��Э^�ռ^��[�93� ��]Q�2e� �QtR��v�7��.ͰSJ�Wb٭E&�՘լ�=G��h�=j�[�e�f_��h_�ɫS��&+1X�"|39}��iT4"Ӌi:��b��ʺ�êr��16��C�X��fL�JN��,ӌ~��$��ԉ�<�!=$/�5D� +�<��}b'��4O�ϛ��[��!o�r^��Jz1��z��ٶ��d�`X�0��9��r��F�(G��0�g�@I��់WTf#�'(r{��?�j�{�X�5#$l��\]w��e�R�I�HN�B ]i��xΣ��T�x�z��`��6�Ѱ��:��]%��O�"y=�T��S$G$g4d@r��E9�+�@�+�Ĭ�RF�XE�\9�tY`a��ՕFŠM55~��P,�j1��UTԔ��e}��e}5L�_VQèԵE��w�E2��Z�� +xf��E��P\��T�H��`*.��Q�p��|��R}�>U��6y�6�js{��ǚ�l��r�YJ��|[�՝ �D��v*z≬��+58d��D�~��f坄� �$+��j��k��a7��f"�'��{{��bo^�Y��l��Fk̳ +ѽ�/ :�:�>�(nmZ��&}N�F*��"c��+��'��rn�J�l�Yf�_�y7�f�ecyp½�] ��n�ф��A��H�\�G�X��u7�%�%wr��U��i�5�jh�63��_o��Sl�e�� +��1�ڇd�'ܮ��U$v��¬��z��e${U��z�j��k��es^8e�-X6_=*�v��o��h��l��.K��]�P7��{Z��n]��-b۔j��Yb�u��J��3Q��*�%��tk`לMh�Ƶk�y�{W�jWkɦ�;�w5(G�Li$^��u"�� M��8�F�ъ�Ո�UI�I�][Nܟ�!bH�D�d�k̆��L,X��Z k��x�7F��pi�4��rY�!d��"��T�s��+��vF0W��QLA��0��p2d�1�$�:��o�ǵ�L��0�G �j ��2OAߕ]9 +:L4�<�(��t��mP��;��ܗ��=A��o$��-�h�[�xE�kO�X�D9��Z��x�Ǹʇ�T�q��b��+^GH�{�W�#�c��L��G��߳��G��WhU�f8�0B�)��aʱ]��Y�i@��~,1�F�Ur��ѕ��ʱa�rLvg=��}��Xn�F��3xL �{��ɪ�\Γ�;^'� �C�a)��渘�t�7>ڇ?eq�M:�yč�r��w�3 �_{��K�'��`p��uע�,�]k��G/xd>��}w��cG�<�w�,��K(�if��Rh��.,b!ߊ��ƭ�"��\��J�W�� p�N��B2�;r�@�r��c�s6�}��Ҩb#=�C�P��@ڔ��$��V�5��X�6�7��y�Rɋc|pW�ᮼ��;��Dt�w�]��}��*��@ǎk��8��ĳ��H��x��j�~�mx��l��1��;��>7��x�ޡW�z��s�d�4h�n�xƍ+ +�s{��)��]J/j�w��G�[&�3��t�55Q�Z[d��Z#��Tc��}�"��S2��]ouJ �חif.eD(�1��v/�s0�� Xm��q��˴ܴF�x�6��E��^��R��;�iA�&8.+�;i��%oJ��N�!��$k_��d� +��Il�h��6��/2�l�/h#0��i؆]��/�G,VUf�X�ܐ��+�O��x��.J��z��_>��P�e{b�]:��/�P;��α�=��, +K��3��`�\�)py�T��M*�3}��Lsn�� +ʔ��'�_��e�$�Q_��T3�)��B�z�@Qe�$�(��7�6��Z ��J`�\H�"sQX��nuxf��|�C��#�ٶ��Ur��mTw��ٗ��;��sl��/qb��N;��j�Mۤ)i��]��Z�MGզ�llD��5�5CӘ4��l�ZA;�M�&`h��4i�`0�!V�L��N�&E��l�ޏ��>?��0Q�$0Y�V�V�X1��$i8��V[F�2��xo��V[JR��]�3��թ�b�]�E�?k�_�*�ى��K�ZH�L{C��˹��D�8D��n��+�ɯJ��Z:�4��һT3� ��'b�V��j��Z#[k��n�8u��Ҫ��F#Ҟn�mn��u1�M�S�=ѩ�tbO��O�T0>� N�1��d�D8�R��Q&Y�(A��OX��X�<O�)'��gj��҇g�*@X��K~-�a��yg�p�y^<-�rB�Y��/��O��s99��p�X��!� �[^��_ z��3�t-�U�5��'fyq��|#jΟ�vP��2�Mw�N��'j��u��7��/ 7�>�b��6팙�[7�ֺݍۚ�qk��\��_9�9��t)��jG��>��%��7J>�.9�4��,2�s��~�R% �h��f}�(�ꦘC��PTbj��a ��A��T��f�A9 +S��4�3�(��+tED��n��3�W~��˥6�%�ku�ޖMw��*�Va��;�\ d�ީ�o��R��Q�嫷�U��綦��9�Z�ٔY�!��dOv��o�[%��H��'��i?8 ~߰ǔz��nTHJ;��g��v�XZX��V�KD��mI��k�y+ʟ� m��k��q�9��QI��G��#W�߾(�}�Mm�� X��Xͣ<3"��"'zBk��1��")r,� Y�U'�� 5�8��9d��߉ +%�p")�ݝ��}�&w�O��y��ƽZ�E��M��J¹a�s��X��Y�'`�E�r�M�O=]��p��s-��ÿnI3��[�<��G�F��}Ǒ%��B]"�`WUu�-�sN �yJ9J::5�X,Bhw�s��7�_�]y�}��@I�8��k��yy��<��p� +k�� EgDݭ@�K=I +��OOe�eɷc/=�,��⦽Ү��o�_.��|�7�rhl]��4�L]"�� ѕy?�@Nj��+O��%\/g�>�-��Ry~��.�/��/�1�WEz��{.\X��T<��tW��I�_��Ly�x�� +҃^oJ��Lo��gt�i�v��-��^��6��hl4��+1�z>��O9{� K�`�h+��Ӊ��Z��{� ��,ǉ��p�<)Q�Ѿ�~�w zp��Yl١�K�JA��~ݯS-��P;hG=۞E��NV�f�K�Y B8�C��e�8�@ш��*��N��%ޖ�6=A��H�h��u��7�!=�O��!tY��^�-��ŧ8��JN�&F��( բ:�R*��V��rh��ދ�p��[3�"�lE�$��J��:��c��S��(��׊�-Ëk�=�Zϓ ��gg�~k�y�w��1xC� ��76�eD�1��֗��yVJ6��?�'�N��M�7K� ��1lq:��4r��s��~ER��"�*'��.��c��L��endstream +endobj +267 0 obj +<< /Filter /FlateDecode /Length1 8916 /Length 5120 >> +stream +x��Y`U��ѤM��HۙtL[��郦��BK)%)�R*�j��U�B��("�v��ũ�� E� ,b|B��4��f�u�=�|��{��瞙�L��j&9�.{��;��P9�;_�~(`��͚�Ԏ�3��*g]�)�cG�\��W�Ϟ��n� +$�m�즎v$C0�<��7\�o��ȕ�zUk��O��0��Z[��}[9 `]\��-M�[��J��\�V�J�>��.��d��Xp��yM�U�t>��7�kIl,xP��}AG'M@�F��k[ڥ�;�ڳ�jt�`#� +rlFa"�p��LۧXUP��e�0��r �ٳ ��P,�� O��gG~ �N1�nT»r�� �[@�B_N�/��i�(�S�c�jtX��h*U7[ t�5�,�L�puj��p��S��򃌟$p��J�9��.�(�hV��Uz�?�s;ӳ��eq�t��U;��b�� *�YT�S��@��Y��h�-!�!�!�C� +�*�EfHW��G�L�RֵJ{Zp-:00J��`!�1h�\�٘�&��/�e�"��p�p�CQ��u��&\�֧C} +��Y�Ё� +�P�x�+؄��0 +s1�0�ъNt(�t(�-hF�V�zA�Y��1Vʺj/� 蠂+�T*��M]��υ��}�=�Ev� ��&^��*�>��5��/0~\��c`1^�Tc��1�� d�B�9&0��>�� +Ux �1�8f�ո �`8��^��E2��MG�`�7�w�.$�]U�z��]�B�*,D$n$W�;��H�AL��뱅�}�>ν+��*H��#V��G�q?��|�~5�ӹ�|g��;��;�;�w��Po}��o۞ƿ�=��S]O��\��?�6�_�*��{ 3�֭\�֭_��~�u�_��?��?�:��T~��{��W��V��Y��J�v�_�Ǹ��Jxy�˜��H�˴��7�o��trөM�95�7��/վ��%^d��9��V��BT�k�q��R��8�w�m8��$��~2��"޴r�J��S��8�Y��gԘi��=��?j��{��4�3��\��4��6��7g��'�<��n{�X�ه��i�m�Y��K�x}�Қ��ȥ��4��{e7�Q7�^��8�n��.��ŗ�Kg�ەx<�8��b�_ܾ��b��a̳��1f��\b5f�1�Ř�ǘ�3�SRӢ��#�y|jJ��j��d�㍦X�!:Ơ��2h#t�Zc�|�3��̩-��hUF��Xn�1�� FŜ(G @m��u��L�c��-fR\5��GK�ՓFKŶ�uR��Z��N��0v�O*�Iܲ C��^�P/�UL�� T��m��*d� ��Ns�:��|��z�Wj��IE��2��f��?��j��im��=Ⱥ_:(VJ�=s$� +I��K��<��Ѭ�S��i�u^g�M�n�ٮ��ѩ��1��!�;e�uv�0��*��0/e��^�] S~��L^�rjȇsہm +Xb-�VK��A�3]|/��(w)K��D��FwT�ڨ(St�4��(Ci�rRDJᎊ.�)��"��oc|<7B�O�)7B͍��t.ݭ7��c�ʜ6��"��)��/*/v�a%��XI�XR\�*-�a�� Iə,Y5��%)��t��X'��|F��c��H��{��6�Ϲg��8��:Y=��]��=Ik�;um��X�W[e�"bJ# +�bK�"KK�]+Fs�5,!yy�ܙ�e�dg9��He��lll�-v��28��6[��e~K�f�YڈL��K��J/c�EI� ڈ&-��S��ѹ�cj&��ՙ�X~Y�k��Ǧo��VPrת)�Tl?�avg繝|câ��7/{�g�k^�c�Kk�-�/7*gr��{՜z"��{�^��K��S�lq�m|aJ��F��[=abJm��Ic��0+şݖ��`B��7Lʸ�:�!Q,G��\ykˣ�r6��7X�O\떇�+��9�=��(n�'N��'An�//+=*�y9�Lu\��&Q�[�� ՝&��58��Ae4,׺��Z��l0�6[c\r��_ɣ~g\��b�M ��_�ee��,+;'��u�"�ɒ3��Sʆ��3�S,�91��'Yi�M�J��5;6}��a��,O�FX�B�{�<{a2�_]�tJ��HLt 嗖/+{��wd�bǔaK�K*��m--��-�L�g�q��]m99��,Z�j�� y�W ��t7��=��b��yV�I0<5��۬9��S#ބ�~S��oݵ�L�cN=kt��i,Fyya��Z-�n�\��l��ӫ䌩��y�\w2�&4oj3UJ��B|��|%�XK��;�f <�%l�R;�>��t��衧s�D$�C��0�9D#��!1tF�L +��Dg�X:�x��Y$ ��G"�{$!�� Y�$��"�� MA3R�ґFg��t:�L��i�Ƞ��iX̂@�!�B��;d+��K�[��J�b��[�!�� -�B��`>l� ��7�}8}-ߓ��(B>}�b8�+E}��Wp��B)��+�P�e(�/1\�pї�a�%.U�2��(�p�nGa$}�Ѹ�>G.��Q��9�0�Na��c1�Na*�$.G%�D5Õ�L�� t�1�N`�t~��t�0�>E3��OтF:��0��c6f�q��O��:�6��L�+�b�<4�1�G��a��1��E+C��:�FGq��s�|�MG��(n��7��>�M��>��Z��AG��tK��t7+x~A��0n� tݸ�>�m��>��˰�a9�!��a��w��N�L�pn�>܍�ԇ��p��p��^�N}�˨�Tp5t�c�XAa ��.:��a��X�{� �=t�q/��1��q��O*�+<@��k��4��҇� ��[��}��c��qڇ��F�I;��B;�!ޢد��M;p[i>�Vڎ>l��8�wh;>�vڎ��A��v�;��U�v�;�+��6W�S�m8��h>�� G��7|@[q�V��~ڊ�q��Az_�#z_��o��|�ô�q��>�-�Gi��s +�8��=�~İRW��g��n��m�D̲|fF�9-5%9)1!>.�d��6DE�uZ�Z�1ؙ�R��I��-��}]RYM_X$��'��R��I=�B}��J��{PuTB��$ȣ��B�<�m�g��Z��KUb�h��S��T� +��%�aGOdT�X�C��{X�eLQ�*��h�]��I��#m�{�_�b��bqإ�,A��&Hq��Z��$m��+̑�MV=��Af�m�f��J��j�9�=PY=��r=��o$�U�T�,��U�r8��U��9�a�vI_��ҫ�#�� /=��=T3�;�h��gN�� ؃hȳ8��څ��}}B@�kUBkS��*e�Wl ��I�9�;-�}�j��7��O�Q+<�P%��y��T{� ��k��RW�YrW�+��+�V�-�B�]�� x�I ��J{��A4�L�Jn�_ 0X-RW `�� z��w�!�Qe�YW�b�-fe�,��JG��ՓΟ��z}�W��g�VA��pZ��PC��0��}í��9��/�#&&�9�F�Eqh��d +cх��n��G\\\�y�͚p��Ǉ��w8�X�a3/��&&&��y��#))�9��X"�fo8�C��0��a,��N"��pa�7�D! +� ��Ąͼpb��9�Vk�)1\慓� ///�%.쎇L��cIHK��EQQQKbz��G<\.WKrf��H@�ƒv��I��`��endstream +endobj +268 0 obj +<< /Filter /FlateDecode /Length1 16048 /Length 11155 >> +stream +x��{ |յ��wf43ZG�d��5��ݲlˑ�� dOLb��@B�$�Z �-��Bh�B(�(�y��)PZ��&�kI�ݑ�,��73��{�9�s� ��(:{V��#0�x��3��ţ�=�[ޑ�_�]篹`ս�oRp�0[.�p��]�dՆ�w,x@�@�-[:�䏭��5дl��a}Qy� +u�V�߰�G�bj!��Ջ�Uwצ��PX5�a ��C� +< �Z��3��׬^��^�t�߸�}��@ƺF@w`��e��R�ѤT3a)\Å0k��*/%m��B +(��PiHs��y��xN��~�.w�99�ör��Ǳ=��c�i�Q�1��~o�+!��c��>ޏך�kׁ��7e�Z�t��z��'��P 4l�s=@�a#�B/��4��<��p!��pI�,׈~��X �ZR��?��A�3]�z_w��х�M<NM<��r��N��a��ȭ�x��^}�}/{/�-r��?(�$�'��f�f��tA]P��^��{��{|�_��x��0��m4� `>�wYx�_m��k !�͗��W�\8V�*��{�ң�keGyO��]r��}x~@F��kp��,4��S�eM��|�\8@�~X�`\��]� +�o�o��P��֠�h+z wQ54˜U�8� ��l�� n��x��?��#?�ě�aJOI�f�|eyg��π��~臩�6��n�[�)x ~ ��?C�H@f�F{Ї�KX��x3~�ZMm�v�@�Kȴ3ו.*_P��V��, ��-��`��2X)��ݰv�^xރ��!��\� 5�.t��G��0�v�0��~��Q6�I��u.uu?�}��a��R|�n)-.-)��./=W�/�Y�qy��<��`2L�0��p�L뭄�`'@��I�L��UXYŨ�A��J�q� ��'��)*d1l�w��x"|^�o�Ip6�'Ѓ��^h�F�> $HA +��! �P �@��]N��f��kM��^�i5j��XCSA��7�ɋCyZ�O�#i��'/��1��{�}��{��j�SkJÞ��Ք*5��H�� zz��O{��:g�|�'S��69��x��kY��ɣ!Oo�ot��ޡ�X��J��^ T8�Û�Y��Mj��m��޼��#�Q��%��3��ؽށX4��ߕ�E��\o�ǪX��8]i�?9/ �=��F�yh!��-�uq��K��.��C[��X噀��l=y4��Ԝ�j�" %�9�y4\� �/?}�.�n(��5�q��΍ D�ayO � +� T��>��S�>2o��l��y� /��9�9߿t�@�Y�{�0{�7/ �ǣKZc�Z��:��"��$��D��<%N�?c~~s�=/�ؽ^Oo~��}=v��@��z<��[�}�G�� QA�5?�ٞ��n��[�[�T��gHՌ�T��6O��6��v��v�@O��h��[��?�p��NG� S��3� +��% +7��-�4 +g��Q��$ +�@|K�S�jc +c�\S�D��bcWuqH�j�됡)��<EU�P��)(�;><:n̞ ��*��i�4��:��[��W�)Z��< +��o�S��8a��H_4j�7�Y˲��sU|�7�Y2 VA�,�aY��HA�V�0�eD�Ϟ�B %A��W|�8�Nw:>CQ�qܙ�U��׍��X��h ��I]A��>�q��O*QV��_�Ӝ��h��y��׍�k�G��4��A�2|�^3./D��q��JT�[CĲ��O�gN�'��$��j��Ǜ��8%�P* �R ,�'�3�辁��'��Zů��"`N�'j��Lg�W��t|MuNk��4N9@�+T*�G�:�_k��0�_I�)�|�\�?��\��k�sj<�ԎSN �V*�j�G��X�U|]�4�o�W�U�ⳕ�)@*R�`D��?��t�Q��at�l$uj�Q@��-�o@G%��Ƣm +��γ1ٳ��>��<*C#xs@�N�`'��H�Jp%&!�`�[0�͘�[0��`5ư3�s@�y�tL�y2��i�l9��:0 I9��rcn7�gЧh��҈��G�`�Qj��g�_@�4��}��h>�hz4�x4Wz4/�`:hЅ�I��_N�tO�tL״N�<�&@4��!��!M!��QHsyH3��B��.��#��]�3E�)�קh�f��)��h�S4��:��O�n�m�_@W�٦�u��<�J�;�O�ݑh��=;�.�o��ӄ��0��6C��r�Í()�PR.�dw"�.�u��s��C��J�˅+v��e��sݝŽ��~y�!��D��y��#(��T��L\��A�\�=-��1[nh��٢{]��v�M�T�+��Gҹ��f��ĳ��+-MI�g��U�t:ݝ �� =�v��J� qyx��H�@�O��g�ܾt#)�s{;�mf3l��b��3p��Ԝ��8Gs��X(,E�V�Q!��,Wh9.`�ƕ�b�8S�X�ytS�h��Ю��5T}G�t9�w�Ϛ��7�H�9П��o/�=��g/��"/�}�L9��30`��O�>��>�]^@��]>��H�g�޼4�30П�+�@}�۳f�``�^�a�a�$W��R-�~O��T; �Z��T۵%�۳+��0q�"W��I�F��'!�۳+��k)�()�;��@N(��՞&Uv5�{g�ߕn��'��♕�r��J��J��3��Z��{b��x�C�ޥC�ޡ� ��,�͋<�]W}�ꚋC�/#��K{�W�{<��{��{I��.��w��]�JK{v/�Ȏ��4�{J[׏�5r��n"`#��k{�P�K��%m��zI[�J��m��.�յ��/>wݺs��`�d��`4�Q�\�4;�%CO�(l��dV��7��u87��;K�:7M(�#W� �s�\*��{�A�W�eFG��c@�~��5� 50E_��9 +S + �8�Ve2�m�+�"R�_c5��{7̕�V�|l��ޥ=�AǴ�l:��s�7RI4H+�`;�4�~�'f�� &��k�&��}�]K��D5]Q��tt3��@�GRB�lXi��v>��!�3HLK%k�&�t��ы��@�j�z�Jc6۸zlUk^�n����n~��N$�:�t%�D7�?� fP�.Y��6`�0E��-��T^�Ȯ�["+�:r9d�f��l^I%Q��[KJ�g��c0�c�.f��[%ߏї�O��+��d��)��Z �b(`� �r��]w��1��1��K%�7J-�T��Q�QТƬ��p�Ne� �U�T��oWd�Wd�Et��]�Fa�Πob��R��ڴ�lh��z��)�7]�m߱��\��ǯ\�e�e��B��~iK_|�??�Rb&��k�ғO��&!�L��O�s.��j�:j�ֈEc��}��k +��5��t��ר)Ls&�c��ن�J>�Vk�f�shx�Ch��/s�)�u��iGƎ�#�H* #�Pk�k�&o�e�ߧ`3�4^�f�)�8��ͪC��:��v�M�P�%��Y��6�-f ء M ЅpZ^C-�w��}m��h��t�3z�ŪT!��FY�z�E�D�¨X�eL�mx�jj��9볘'fw9��A��Ŗ+�:ƈ\!�!kȒ��W�w�`5��J��QīD�2�BN��ڑfNJ+˧��y��͔� ]��W_?'�4��G��]�Z��Y)��c�5��ђG��%�k�Ϟx�Ko�.��K �(��eV@n�r��X��օ�,�Ϯ`Wp�˸G��֩�^��T��Ōk��HX�i��x�󉚚Z'��LbKX#�t1+��H|�@�� ]�d��C�׃��[�!�l�}��&��*K(e��j� M�Ơ<�U�t2 X[K��m��';n��<��"��^4;�q��r��z�3�ZW��+}u�bo�'k�_�r��Z�l��х�]O=r��x6ܼ��JM��%�>7P�1� \#�7P��(hS@h�A��D)4Ϣ��b�je�$��4��i�BRlWP�g�t��.�Vf��Ã��!��Y�� #�&��:Byk��ɯ��<�dB_v��;��ۊS�o��ch�5��3d$��X}�Ye5�m�z��a�V��e��qL8�l��|!�'��o�F��R<�{_��H�u5S��h��#q,ߕ�uwƣ}�ѯ1��0�)�\ ^��:%R^�u�e��sR^��Bu��΅�ƻ�v� K��4j)��J�N��5ZN'��C�8g��:]�P��!.��|�r=��E,��Ɉ��vE�ȊJ}�pΐM@G Hޘ06V�SB�lV_Y��"�JXM�Ih��`�[�#�ק0�8��rRo$�ۜ��<�ݵ��F۲=�/b&4�[�j�I�K�L��gg�>��-��9�tp�tg��|�%%�i%��v��29gv0��B&�9��:�� ?v�X��W5�Z�l�<�\׵��=�{\��p��|�u�q��?�> ZD<�mT��8ͅk�5ဳ��o�v�� |Շ�a�3bw:�6j��;p��#vg�>�t�-ǂ�d�X{-�� <��;�v�֚�ΰ7b��&�C^�K&'�F��s9��~2 B��D��EsN~U��9Z��YN�948�J��b�"��h9��ȅ��H�DleQ l�u��I��nlY4��t��sr��彽��WS��DQ�sQi�#?m �D��uQ�[��-U�er��hǄ��k۔2_f��L?�Ei�g^��Jᰪ̎�*��s�u\�x�W)y�J5��%χ .��eWR�jX�Z�s�f�h�u8��^�)��kC��B�T9��m��}�O]�h��q��e" +_}%|%� _}Ֆ��n�@��B�\�� +Zȱ��\E��,�*�=�&#?��$i�rR��vt��Z�յ��m�Îb1�Y]r4?�� +��h^�WYӡ�Խ��Z��c?{$�U��4=��-�iҫ�7�)�Nl��g��Cu5^�9��y�b�o�5덻��~v"+��b�A��I�t;e'�V��5B�ph\-��@�h�rX��#��(��N��:�D�S;a�v�8�sl�T�Ȯ��*�<{��n�<�X��9�!��I�B��'��6i�O�W�J��`�5�f�pӌێ=vqk�E؁hd,��g��/+�L��'Z#�/��T��.Y�Ëo>��sT��9^�?��9��W�3�#d�z��B�fhF-ʖ��K��~��C�zX�S��p�aq�A��<-�xU��4kem�M�d2SF��X�R�'��T<�J7f2 ��*�?��م��o��7�[B)�E��P�� !�)�/�f��1O2�JE <��Ţ�ͮ��͚ihx��4�8�%�YUJo��2�z��JN� -�:��!��f�-P@}{<`C�g�$�V��Qޓ}T��s��[��$��ǒ�y\4Uov��nH�tH7�!ޔMEuJP9c:��ۛT\{��U��^Jm'�w��I�m �{|]t�T�hWwq��M��Lo�s��0V1Z�YYAZ�T�CVV�;�!��8��ADZZ��v��Zs�(�P0D�AD�:��s�>ht{��$�\��&� +�� ,��u�M��=,��gf�'ɵ� 3�}b�d�_׿-Lvm,�w��oe:WL�X��4a�+� ��QS�̜{ڄ��Ȃ�cJv�p�9燺m��:˵�j&��}E�K�[k�Z{c�*�o��4m�cDw��c�3 +1h�_K�� +=z+�F��V�/��^��ȏ��D��4r �j��_��k�R��a��T��J�j��1Ow:\�zR�m�}��L1��!�G��zU>:aJfd.`x*��s_=�'ә��8pxp��}�Ί@�ѐ��;Y:e3��@�MK&�f��"dC�"4EE�nAv�"W]#dj�\k>�6��f3�E&�X� +f�I��FC��T�_M��4��;�'a��Uժ7|�m O��ŧ� ��uō��1P{[S^�ڦf�xyv�c?�lD؋��l1�>�+�|쑅a��+��Թ��8�W��rn�z��B�e�F�R�&��M�M�cگu�9��ܐ~?��]�3�w:��f�+1��@!j"�p8;k�[?P`E�F崔p�S��)?�~-һ]�-�AC��X:�;rcU�)�!K��7J�Z;0��M4dy��Ņ�FK�P�w��9\��)�Y�=䊴�rB� �demj6kQ��D)zŠ�oƋ�-^�|��S^F��o��U��d]NV�J�� P��uzl�B�^T�`��{�G��lx��-��D�E�o9��gҍ�ǽ/��@qD��OP6B+�{,2��oH��N�V�M��w5��U��M�6��6��?�9�]�b��w M�.�%j'��$��mixrפ��ag��wۺ��+�[6>l�;k��h�F׊>;� +� ��h� +��g44��$w��a*<��<��v�<�v�m�v�CI=��<��`mM�\��j��j��ҵT�50?q�88vxL8<&x�H�5��u�劣�~J��"��/K�3!��VB0��C��Ц�& �_�JP�}A_E�zH��,I�\�)*�7M��=��;��Eu"\�̎��95�� J��ۭ��--��t��qs�w�H 7<�(P3z.W��nvw�# +#-�!5��=C��Q�nm\�"��)��5��z{&� zw(0��J^r�S�.} ��ԧ�zH�SRmo-��f��8!�Nb��]@G� >½IO��#��\��H�o�S�`�L:��d��R��VQ� B0��)2; )ل>$�{i\��f^��"�ֹ��0]H^��:��]�vԁ�ċ��n��V33��I��:�;R��Ojj��y ��uk��D�j�EԾ��4��m�tqi� +I�p�[/B�oX�/լiQ;��Zf=� ��W�:Ԏ��X<�t��r��t�#�ݎ��Aٖ�ng�nχ��H�� 1A��&��eF8r��Hg(yh��rB�}|Џ䢑��H4�ye�K�L ��7�lLS{�\zT��\�>촆i�[�u�Y�ذ��s�'vT��4�W7�3j��鋺Ա�G��&O�w�8��"y�I��o2�}�o��2�RISҘJ%�>��Ȋ� +��ew8��DGKsPiJ��&��O�`�s(�9�/H[� ߩ��D�Gup�d�i͒��Сb��~yQ��e��8�Y��J�`��ئ�Z�{��r�>;�z��Z�Px�rj��Ҭ��aS�P?�44��Ҁ�6͆��>��6��鄷��:^��t:��s�:<��=��ⱀ��شAn7ul��Q6��b �H�O7��E�dқ2Dc�� 浼�#�'��0�0�A��0�!�-d]�jZ�'��ȩd�@ �N�?(VQ��A��Yu��F��\�Z�x�잖[{Js&�Js'��R?��g��M�دnp+��zR��/�b�$�p(��G�� 3ŷ'�� P�X�-��T`+�*E_�?��ݵۅ�j�(�Z0�f��l�i�,i+f iNe5�XUL�� +��d�UMWCV6�:r�s�(�55҉`�L"��:jX��`Q-m��4�`TD8qV$��AJ7�Ɉ�>�Y��͍��^E��F�K�e��{n �;QM��/_+D��}�~��t��8j��e�a�! ��&��{��C�C�#AN`�S��u��8�[�H��w;>�[ ��2��<��#�݌utTh�a� ��u!��B��:FDt��p"��Ğ�� Xƙ�V&`PVj��KK/��U5��p�y[_�ӓ-a��kuI��}��DK|�1E�dj��n��+��|y��3ZSKi��8��o�.j'�a�4u�5��;��,�=uuݎ�ǯ��J��V��vQ��:M�a��Ƴى<Τ;ow +A3)7ې-R��,�{vF��4��| � �dE��U�T�c�*V]��M^~a.�Rz � �KI�Hi)��ϟ>X��r��s�k��k,J��%m�pK�H��6�{ -+��7�]��DI��A�8��j��k6�6��a.�H�#!��D�)V@^I竷8��Hv��!)��#�HY�kЀ��~��f��ˤ�U :��lö>��Al��_%�kt�u�t��Qc:q`p�]��1��ACvlt��*��~ �پ��:��6ۘ�vxLβY�[ۘ�r�"ҎP�� +V�\ȣ�� 1.� +n:�s�W��5W9��ks��/V��ntb.X?Z�5��б��7f��º��o��{�n�ݥU +� k��?+ u]�`Ͷ�RU��pӒ��'M\��!��1<�~|0O��F�u��ʫ�ȶ�T�ҫ��\^��n��Qc�?q��[��m�/l�۠�8��ف&;��S��Tqː�b��#��0k,}u��q}�h�y��޶ C�3��oj��u��.�.��R +F�2f'�Ek�&�U_��zZ']+�b2k�ڳ�V��g9-��|��.1^g�8H�f�L5ӊ��-��bI��،&V�r\�`$1�,C�=��%c�Y��%3 v��Ƣ�Z@��AAᶨrF��@��Q(��hRd�t֖��x�(+:|1��T|r�GD�ī#��,DL?�4R��~Q�,��'�~~��'G'do��X��RI��i�Y��&''H��b��(��+Ą��F٣�x�B�U�zj��gi�P��J�,�ƥ�R�ZO�-��{p�hL��լ-h��J�۱��)�NF�x�K�S�8Q��U��#��M[�Α��!t�ڸ*��n�/sՈendstream +endobj +269 0 obj +<< /Filter /FlateDecode /Length1 3548 /Length 2086 >> +stream +x�͖mlS��^;!/��؉��IH^��$�cU�Ђ��8@ #��I��rյ&`�uR�+U�S��4� +cZa['��j��:�V�j_�:��{�A�0MӴs>�s��<�o#�2&q�{��X󑑙8�Y��mc��9�"|kۃ�Z�/��^11��𕶫s��djhp��˯��S��⃮~��ؾ�� ap<"�k϶�pC�%P>N� �/�@}�b �{pl��ꯀ,p��}s)�@�X��ޡ��}��~�q@�B-F��d��D�qLD�2{��\/�j'A� Xg�� +n܌��yT�67d��5M�syn9��ُK�#�M6"`❋M�/��l��I�y��G��W\o p�F�3(E��~�� a��]��o�CU�� i+@,�?�+7%|��{'��^'�iw�?�G�ک��pS�y��*�58�J-˸� +J��u�Ш��o�Ha�4R�gqPs�)��It?<��/�C�YwY]i��#��4hg9��?i�&^�_��N��kƝ�nǭ<*�� )�M�T�$��{B�u��k�d�>��ʈ$�z�H�h��z2�ꞾD0H�wmOz��~�%�C�d,�acb4��ބ��H�� VՎ�Nz;�#�#�� GMo�ۉ�d�/�ѕ�F�;s�7�9��3��ɮ��tf�*ݨ� ��̀��kg��.վ��˳��,�֢΂,TP�,��2Z'Q�s�%EY1��f��4�tc�0��]}�j��q�u��ߔ��sO�.kgr3��(��;��T��yx,��? (��|b:}Ѫ|��:�?�!��u>��XI�; � D�� q�䏙��&/M�!��h��+�p�ho�eo1/@�[�OK��1-1b�on��J��ػTx=FT��.g��-�V��-5�*��e�E��x[k�X(�.�g��7�|Rq��ߖ-p��y⾗�j�2Y� ��b�]��kk��t��Ε�|semY�X]��ډ�li�.XxK�+��>}u�� ݅�ڹ��v7KŞB J�.-�`�_i ,@��:.C��${�h!�� #�x��L��ɿb�o��Y��B|f_{�ʖ��ݒH�C.��K�PK��D�� "��M��ɿj��=�A'��PcL��x��J�"N[�$_ ʧ%�1I�[��c�r�$�wR�J��%N��l"�0Q$�A�?b�/��&�.�kN��)Q��&�rEs[�]�UEimY��[�m5��.�Q +��*�Y,ގ�o�Z��f�8�מ��Ͻ�~��r/��>�=��#�� +�t>��\'³�Q7��|�S��B�rk��K'�j��2��SA��c��+�-�]�,�Uh��z1e�«�d�΂�h��U�M�ڌ��≷��[�k��1r�Φ�ɢ� +_ıf|�cW/��!D.�P��Eʼendstream +endobj +270 0 obj +<< /Filter /FlateDecode /Length1 20416 /Length 12258 >> +stream +x��| |TE��9w�{{߻��M��,MiCB�dQH� VV��8�B؆Q�E�QGE�"408<�㌣�+:��8K�y�3 +�|�n�Oy~�{�﫛�Tսu��S�N��@�� +�� y��1��kY��\��B�P�oo0j7l�vC��?�,�Nf1Ę�P�,�A��T��b�#��b�m�`'�g@|/9sQ��|n7l��;�f�| +v�3�c|�@��s<��rf1�1�a��0��n<w_�O�Zn7�Mҙ�Pޏ��@��t~�F@b,WS��_��y�w�<\m`��g�A�]N&�=��&��=��@��4i��d��p~;`/p��C��6x��a�:�m�@.�0~�==��x��8�a�!��a#z��b0��=?�(��p�06BW�S= ��z� {��~jc�y��0��ݰ^�ݳ�� U0��<��=�� Q�7�+މ=�z�z^��(��&�X��1��Y��`ϟ{��0d�xh�M�솽��#` ��%� 7S1�N��Y�ڒ��A&�B-��:�0t��e��dOq��u0ZA� +m��aE�q��x?n��RY��Tuu+�=��^J��Y��g7��8)�e�ўS=o� ��Xw��p^�� ȡX�U8�� +�A�G�05��k�n�=~�_�9��Ԕ��PK�M��*=��L?@��R��\�?95�6�jOy��=��! +U0n��`�6�3��c� '�ee�]��ЈN,�z��1x-��S<�G�y�._Q@��rQ��G�cX��X�Zf$l`�1ۙ��Nf�� e��c؉� +v-��ƾξ��m��s_p��h~>�:�e8�viW��C��8��0|[`,��i��k�|� ��m0��rXK��K�w`��y��7��p'�CV�m��#�4$`��H�z��J@S��P��#�F~�?��)�`�%)�<�0J!�wS��Z%%�7�0�K +:2�h8��# <�Y>�^�� $�2��H"�k�w�=��@�l\Kq=��)\�q�~F��Q1��^gvx��}#�\�O��?9��د�5">��t��e0¸�@�M��kD| +�S��/[��)�>��`��,��~Z��p��D(��~�y�t��z��}�=ӻ'��;3A�y�}�w�'��{�s'�{�M�nv'(<蝗��;�@��%A��81&yK�do��7/�0&ysܣ�Y��f��|�b�˽�;�=��qׄ�kBGq7�,ܱ?0�{w�Ff�mI�mF� \+ޒ9"�� d�2G�&��/�#|��2�ƛ��ZA��J/w�@%xq��>��g��g �@ �=uܜ�=zr��=9�K�3RI�ļ�b�=E�T�v�P�`�� VY�*�ơ��|�1��;��[�&4�w��ؗ�-��#�HkU$R7~遶sfԴ�k��5��Y��>߾9H�/N��N�E�-��o_ی+d� �m��}0�溆}3b��bm5��S�5]�}�ZTu�ªHa�Ȼ�6]!��dO%�j"�j"��ଙ=�j�n��̮��#�Mn��Z��W3��f�XfVf8d��e��v�9v��b6 z�ZUϱM!d��7��8��C��_<��/�9�k�Ň_zO�׬��X�/>c��ԝ��;Q變��l_��u�ߗ��x�r]�\ߣ\k��gU��쫉o��:'��M�H��8kY�_-��5q��&��W+yt��ez|츆��4Yn��=��֫��O�`j3�j��!N�4ƩfR�!��e��/F/\�l��[Z� �ǚ׏HE�I�eÈ�� ԪƆ8��TuS�h�㋫�U�Y��4��0�a�3�T$"c�;b%��}�~G�,�dι&�r.��w��+��qI��#)'�n|H��5�}Ӕ��T��Z�`ݴ�4��F�ɮ�틫�ũ`s��Ȗ�� 1�:U��9��U��3��a��s�1��~ߺ/!��^��қ��_�$ �'+ql�pݦC^g��"��V��k�%��jH��xa��9�kl��I@$�.�� 76&�gU�݇A�Sr�MDmvu�s��,9'�پ�q:0�Ȋo�o��|�}�Z�Ǚ�r��o]ט�Ä�پ8\� �c�i}��s�G�a�r�5��0G)a]c^wN��V�0�!��:-�nL�e_M��؆��4��1'}5��T��P��T)�+��u�R1�?�n]�:��R��XoB��@MW�U�V��4�s�/7�ɍ��A�u.HT��f'�Da��_�p�j|U�_��1<��+��X��N@La��_��a��2�óPC��cx�%��v�G�g�.;��G��\��^�� 0<��0<��?��g'�:��"�'] � W�p��F��š?�+ +�r�o��o�|J�o�N��-WC�ԫ�|ڕ)��i��G9�|��*�w��Xȱ7�Svh�x#y�kT}. �cX�fY�,��(�f��{�T��ew(�ū��+T +��x��h�@�w`J1�?*I�E�xLi�H �~�4G�ve��%8y��hT�=0H�1]�A��r��]T0HJ��54�X��y�Hj��A��R��ǠѨgu�eT�.�b�.� ��+�S4O� ��f@| ��1h��1��q �`P+-,^��ZhZT��Ġ�Nz9�Ӌ1(}��Q��0\\,Q��Q)4 +�@A�?��t^��҇ �؛!�b`��HIԥDF�R0�D��=1\F��2M�� +�{Eh�H�Z��Պs��TQ%��bm��`0|��0��0p��H9�f��]ۋA��4�0��J�Q$��D�V��`��2m+�L��+`�(�8�� J�1p"��w�wbP�h6k.|j,�@"��aЂ^�`RZ�`0_,U%5+I +IM�NN��r�� +�$�co�V� �I��ZT�9��8\��.�i[�zUt��cШ�jM/��u@|`�\��f��Ð�)8А@��`0�b0�"�A��hx�Z��V0�j0�/��l��0`$��S�2MJ��4�)�F �y ��(�� {��z0,�p'�%��K��O�� q}��$��'@�Q�9�01�_Pd� !�l�b6&��=�ͰS� lLN�Zط�Cc*�Ae��lN�Q|0�1mV0��M�� pU$2�l}��wή7��ԴV��H�Ao�� +��5塼�q��7<��*5�e�J��'?��?��wM��܄@�NN�l�[`��Z0��b�9��:�� +��6{�ov��_-L��>��36��Lb��Y��)!K��q��9�c\Z�̻��Mb��yo��2��tmH��[��Q�'�x��l��ˮ3PY�]Q�e0F�(�Qc4j��T�?li��8�CP2jà2�at0�F�"�"�ɚ�Jh*1VbIiI�?��xIi�\h��yN��E��ߎ�4~��õA�]u3��d��N-��3�q�� s�o�P��h�kɯ{zYm�蒂IsvBA�{��TO�i͜�ss��9ܖ � �1k��_�1��X�#�� 8��4�-�jCG�tbY�|ʛ�[I��~�t��ʮ�.c4�l�g�.=a�0�R�� A�00��/��A>�<>9-��`�BU$B��W(|�\ M��P�EV��0E�{��VB"�A��h��C�~�Qc�� +Z[�Y��&��-��6{��%��!��r�[�oC|��|e��]rkղG��V�:��`��@�n��(Q� E=o�8��9^�n��/h5�� =�[4��b��Z��ߗg+��ٮ�0D��h�=7�,ן��A��E��z"Y��vo��9jm�-I��o|(�3ҟ�z>`\� +��Xd�q��&�f��ZÃZ�i�7M&�V�3�x0�D[��S�#��m�c�b�)��9�C�/�VN��]�P��U�m��9{��/ן�ބ�|��݋**H{ ^��è��aD/.� ��H� +�� +�Ԥ��L��q��)9=#�m\��V4��{p�y��C +s��>��sɮ��_ݾ��1M?�\4��^I��Ni�� VrPe�`uT3Js=5��J��5��R�F;t�JM��|�P�zZk��+��JC%� +Qc� +��Q<��~��T.f�j>m��=U��ڭ�?gW<4,�q��i�v��g�#^��w'� H`�Bx%V;'�&�MӱU5W7�tK@5R��Ϳ(�8t{��k��К�5��Z�Ph�@�Tl0d��V��>�-˜�'��%p�s沼A�E��UhJ�-��u��VZc�1�AцQ4 +a��B/FK�-�,[xF��J�]�M�� i1[��!,TR��]�d1��g*)�_��;�l��ݫV>��sY�~�֫��m,�1��߾��a��Փ[u��^Ϧ;��N�8��a�"T�� k@�;�6�&�$��T$AE��a�%�_��O˴��US�ӛ��>�+j2�d��o�a"9��{/��S\2cb��a$ڃ@��( jJ��M��ZZ�C�M�t@޼6e'uW��U��S}��4Kwʎ1��;�� HV�Nj��)��ǆ�q�qE��ˮƵ��>6&��ϳ�a�tV��Y��7xLF��N[�`��O��ǘ��f +�r}��GF �3�s�w��]g��T��R�� +%)�ԫm��N�A� +��P�� b!p�"E��"K�) �ү)�j�&HY��t0�ra�I.� ��s~_(hЗ��!��ӿJ��3dN!K'Q�V䷎��s?��K�c��~2vq�G7%_M��&y�H=q�o��GF=�&.�_��e�O�m�c�#��qO��_~Y1 �=��c`��X��sp3�e�V��A��1R��!N��fK��7\B%Qk�h��<'��`�&�`0��Jg$��ğ��]�p��$<2��>�=�֥��KFӭ�M��+��j��)U��;Ț�E��4�$l��7��"#2�j5Ō��RT�֔�e�Zj�F��4R��Hs�b�s{GL#zi��V#խ�� ي��3��h�~��vj`��0ڢu��Ө��BJ��{?E��l}�nfyg;�:�CӢ��i�IV�l��-f��ߊn܉��l:��>�;dN3��-�2�wrJ~7��C@�؞��OAi��u��v��En�`��*�rI��H��]yΌ��sAʻ�.�&�U� Z��B0��mjK!�(K!:ĴB00i��bQ��j Z1%�F��z n6�4��}Ov&7'�9��L�K��r&��?Тe?��k�C�{�wqf��o~�K�D+��9��N +�{�b�b?�<ȇdlJ�.�K��rmpjp��\��F�,��tZ�N�Hi�e_m�ˋ��ifpD�O�Z�� H!vz��s>"��e]g{��h�Fɞ��#��.jZ��X�5xA��T0'��A:"��,!�n�7i{v�a"� +IH��D��d"�1�"谺"��\�W�G��`�+��C�<$��A�B�C��f��2�3D��"z�AӾYp��?;�¸ h<�GvTW��_{u� ɇ��ێ4U��7��[� +9��7z1��/�*�?��B_~^z��γolX�7F"�zw�{,�M��P��[��E��z��Zͪ��iǀ�a�?"o^p%A�h"��vAC��:P�%��A�/G�NnHnX5��;��9�<3�az��S��/��߱��c�Ɖ�͂ +o嗪n��q5��bUM�`�*q��.�D��O�/I�i&?KZK��W�k��-�f~��~��-��!X�I<' �H ׄ��P�%��D�,�f�D��YnuL��"�:�B\�vh6N�Gc�g��Ѩ�;�=9�]�-6�\��>��=��n��n�錈��Ǔ��F��)PK��\;��O��qK ��U��aXZ)P�R�! ��qe׳˺�B�]�X.t�㢦��дȤ�"�'�Џ�h��h��N��z��ד+�c�g��ảsC��g|��:(� +�]�,+E��v��F�g��`T��B>C�֫��*7��P9U^�0�yVp��m��mn/r�J��X��+*\f>3kW�shZ�k�.T�2��d8�[��|��O�Uvu�y�-�d0Fs�r��y�-5·KJ-逎��d�{�d��2��PJ��t�d��6��o��Ԣ�P�\�m�E�6��@�ÒR�,C�9��Ԅ�Ecnl�"�*�7�`v��Z��rY��cm7�j�!+;ؔeU��z��cG��;ﵸ8�ƕ7o��97L�5��#Fl��J��Uj��1�5�7��w�z>d� 0��rw�O��q��C�67�D�[��!��s��s1��]>֤�J��3}C~eW�!jH�g7Z9�ʙ�h�A�� T� (��Td T d� +�,��>��f-j�[�x�K_uz�u�ѝԌ{��m��'��J�M&��?��u�>��c�5f8�k��d,�In��v��Q�a��͜��k�f)�OK��CF�C��C6�˝@��h�E UQ��^.��]�C�-b�&}��w�A`��I�%�&:�&*;D9Y0{��^�P��s�Wf�%�E�ن�x��Ol{��xzT��7�A�.fL��iw��ȣ��孓X�:q0�� n��jk��O2O��L��z�n�$Rn��r��+��]^T�~��3[�iw�D@�KAP�� +��&��mWB��6��Pd�X@&Q_��Gw.[��?䙟U>=�@�/>�?��W^x�T� O��f��i ��͟q�0��4�dF�2 ��ҭ��'�4��t�٢5�,�:f2�X'�O��w��To{��n��/�2�2R7��nugD9��n/��R��z�u�� Xu�ռAҹC�3��ˇ�`�MygS�T'%�ov+~He~��''� +e��ϰ4K��2�7h��&�Y�p�@zZF|��ǭ��A�,� j�~�D+Q��A�� @䂦Q�'+��6��&"Bd��S>9"@� ��0�Sf�T��e%F��/�{��|�>�ڂ�K��R��h�z��gn�Ţ��{��F�챓M%��u�яRX��<��b��(�ؠ>�ݜ�Qg�Z5�Qt�lN�Ɛi3�F�֫��<�^�� +M��Ŀ�fWA��x��6��8��qQ�s��C��w�߼��tE=N��n��hR�r��N�a�G2d�A�'��3��A�Uk�� 2��Se�t�� k��S��SƉҪ��%�H�>��xgł��U{Hr�obA�d��g��?�Տ��M��6;�:S5�?��|��N�xbD�⾉�?��F �&=��;�n*��W�&V��x^��6!ĄL7�7�IC�,7�[Ԣ&St�ђ V�͞@�<��8Q�;LT(�$��k(��^$��0(e��wwĊ&�� 9�=�v�'�?'Gk�i�8걶҆�ow�DF��w�$��?%��E��ip��]��'p�� +Y�(��>t��!&��1�\�C��3cZdm6'R�@;v�%�..�k;:z��wp��*V��֦ !mH7�1Q�)��;D��d�ͭ�x��\�$�m6�!�L~O��$?+�� 8B�j��{F��l��ieE}�2C��W$\[z��]�A��A��~lP��c�3*~���s��M��&�p��~��.��Fd�� S��[��Y�eq�@�D.�fԆ&fL +�Ș\�^�Y�m�/�XX��ٕm�A��9L� ��4��n�1�u�l!( P�t��DL�]nϸs�G�<^��S<��yN��jن��|(�,�zC��u��v�t~T�[�yd�� n)g�h̡��3(k�2��t6��f�2��^��v}�t�t�F�2*s�LF��%_�� r�� }�M�3@��Ο�B�z��mCB��%��è�lp�3fׄ��r�j�{~q��C8vr��I?�ɰ�3ҳF��g )��0��k��k�=B} ��/(;�0�9M�x\� ��k�FsZ��ٜ4r�`�Zt��V��y^��@^��w�Q�*��@�)XL�]�� Z +4�wX��NN��$!�Q%��I�� 6�Ô��%3�2 f��as��VN#�n�d��vɪ�3ՒӉ��o\PI}��EgqE�pA#]��*VX(Xl`�3��W:vS�A37}:!�8��5��=��O�d]��Ի��Q&��Q��VQU8�)��U�/�N�E�e�E�eI��Z��R�Th�J��u�H�j�[�V��#aʱ�� ,8�e!��,v4�)R�N p�ڡ!�Z�â��2wESݸ�15�i�x2�j�Ķ�#��H?��/x��y&Y��a��i�CR�%��H�~y��mϵ+��}��/� �.\�Ҩ�4␐Џ��/�x�#��ӿ=�|��;�A�>��=�:=��L�V�?f�@?��8W��8]�qu^\�6��`�G �(񎧡M��wx��'��ϒ_%�gp)��lǥl��f��H�%��/c?\gYc�N�+3�06g�з��`+�ͲպնvY�#��Rk{��T�/�T;�v��.�f��%�[Вaǚf�9Y��6�}��'V�3�͔q�s��J�j}w�Бg�� ںn�Ҙ�b�u��f��v$΅d9�8��)E�(��%�C��S��V=��`�'/K_��g�j�K~�^d�f&�M��Nx\��v��f��m��G+�T�,�S +��vhG F@��Fc�h21Fh�Io�P�,yA�p��Q{�Z�h]�W��q6W�qZL��ĭ� kN��V�*�+��*�SʚO�W�H�wj=]Y��+*��}��ڕA��/��16��tw��;��wYc�(��!;��^�H�F�|��bF��b��z(Sebj�9�a��j�0��JV�U�i��C��V�ܴ;@|^��`��=t{]��fU�G�#wD�;�+-�pݐ�С*v^��qL�.��.,s��]�!EL�w�� +k]_>�S�7 ��`$0��u2�Ċ!r8Co�[�l&�BnUA��e��'�'�'�l.��8b��+"��,2�%��I��>�Z�OB��P��Y��[+�,KN )�_�f+-)�x��[o}V�ˑ?��hb��:�RT��ӟ�l޵�ƱY��bUy�|�}��k�UKb��ߴp�{��˞��Լ�}��h3��N�nr��|O��≓}��0!� RS݄��o�ybѬ�Oe��G��-�� e�9�]LE��;�ׅo�ޒo\�a�l�Y�s��Hr�K�7�4� +(0��V�]\ԡ�=6�2��pe�8�sz��<�p{ސgV ߾0��蛐WbA��7aj>��Xd��K::��ؿy��!��!�v�R� ��߲�f�0�]��}��endstream +endobj +271 0 obj +<< /Filter /FlateDecode /Length1 11748 /Length 6213 >> +stream +x��ZtTչ��y�3�L��̀�E+XT�[�_��iP�rmM3[�&��ƕ-ſl��=�{��;�3 ��S�5�%�� +�kW��r��_�� _h��Y��:� ��j�Wb�2@z�ڍ[�Xusv@�׬[Y�Xr`�й�fK��[�gĻj�֑��s1��u6�� .�~}]��l�V��>< +��6��x� �LH`j�M��P^�u��aفl�H��vu��-��!5|�<��}w�M�f�z��߼?�k��4>�K�`}�W��c|@��W�"�a��q4߳~Dl��9D~"�<9�\>�G��)�$�DxQ�YX�4a �ac0�~�}%��.l��9ԟ!G��|��N�9� 9s��ez� +�K��,,�U�?��h�O�_{ +��Q6��x��.fg�d��:�Bz� +��$�1��_��l�'uhA�<؋��@Ƒid>�J�O��f1�c63[�K�\v9{�3v��@��6܁�؆'Џ��/��XI�H�� +��Is��9r�g~K>$��+�g��h&��c:�~� ��>��1�a�c>C�Mm��R�a.�D ֡��؋8��+8�W��CX��7�@�I +)$s�2��#��^rV�� 0:��X+��e�2��L+��3�e�I�${�}��^ay.��Up3�=X��qG�v��{�+�'�s�%|+��_ɾſ�o� +��,�l�:�M^�9�x�B:9�B܅�('��QL#�I i�F��.�9i&Y� +nc+�q�q�ŷP�C؊��=�_��>� ��(�8�6܏qȉ��>��92�A�y� 7�)\>�Y8,�%��I璗I/�P'l0�#�er�o�� 7Y�P�Z#�L^�L�YH�T�F5� �(�X�1)7�z�YpX��D�֦�r3rL�U0>WAn��$��#x��S�X4�ݸ��d7#��]/#��CvwrZO/�)d��o�lO��?U��C��I�!Ɉ�$�3��z�y�`'�Ҽ�Jd�C��ث��14�ch%�B�*}'�XgZ��,q!�#��L�q�E�'y��t�� +G:�?NS≳��~��`�Cǳ��c�5�}i +9�iĦ�i8|�dS�}m��U*>�Q��贗��c��8��m�󝊖x�<�l{N�O��6��n��O�Ͷ�ټ� 6��9NGy�3c��Y��+��m*��|�N� �J��!@P��Cs�<ϳ<��]?�S�G̭Q 5�r��["��v��ڿ��M�u#�֪�R�� + �Ѱ��*�r��t�f�:�P��p@�~��j9�V��j#�q��0� 5�̭��[�&�a��o�o�V��t:�rQ�Uns��爸cn��hnď�ۨrGHc�bz��z^�W��wBl��c��i��}��r�q�Qq1�'�?⎻5�0�[F�bG��&�i�P��(��qFu� q��q*w��4n�9n��X�11BL��$��]V�0!��qk�pkOvNv�33#]��i6kjJr�%q��x�)�h��X��啦W�rf��eJ��y�^��̚(C�,ֈ��ud�Z�&��jz�5=�k�8 ��\�W埔K�B�-�I��`��?��9j�!�S.��<�(z��E�T�^yzKc��<�Ez<4��\��%�1�fkh=�4Z�+�H�^9Y*W16�[�J��-Ou8�y.Qf3�$�#��D��n�*i�nŃ�jZ�Y��TS.s�l��e˽'ݸ,y�D�2�1��m��]��w5{*�\��}2��Ɏ��aw�k|F�jQ�IeRc��Q�B_g�'E]�e��u&{�՛W��\�T?��}.t)(��SPfY�[Se��-|'9�smm�mt�� A�J�fx�:_�Z%G��Cr�S~��m�Y�#J��[+\��ۥ�XU��kR�t$ +O��Wx�K�D��7�y +O��[+�Vx�K�GU��kRx�H.��R� +W��[+<3Z�Y.3U�gM +��sG��R0�*��! +/��·G+�إ�vU�%_��KG��oD +��Wx�K��*|�u�=�2�n�IP|�/��pew��O�h��9*��+��hM +pQ轩,9sYL=Ш%֤@�+(��O�O�!�L�G��=��rY1��'��qEf��iv�˸��}_NS�9WN��O�D�TPo0��儀��a��-��Z�-��8�ȋ|��;��[��_}{5{_ƕ~��+Pv<{�~;�|W��cp��~y �J͘�_ףּ6��'��?C25$��' +�T��Q.O�ᇀ&`��$Ȃ��lT)H6��H��nQ#��<��z�T�b4��0Ǘ�+ ��4�Hc�2��JNA#H�33!�]�(��7�g��{=�>��0�Hqi�tf��V�֥��=p��S��d�� x+�e��<{�?��<�s��E��I�ŷ;��(�@ B�x7F��~��T��8�J��U$᚝��>^��]��y��|u��]��h�" �p!�R�\Ȥ��(HW�O�F�r��Ht��]�̔�(�$jb�&��h�H��M!*�J��;��%L{{��և�Λ�(űl��e��vo`��J�z�p��k�ۋ��oϞ>ʑ0i��^k��௸C��a"���n'��;3��F�n<5�ON�/OMU�A-��*,�`3��!�@?»C�}��K4�T��x�� +�X(3�wa��HO�$��,XP*a�4-�`W?��C��/��*:�8/*ȡ��^FZW�Q�.�Bb��A#$�bw��8S�4RqzVH,*,vkb�4ƙ馗L�d�&��\�lzZ�]E�u�D��n֍ZrR��I�3}��w��=��ˬ��ݷ��M��\3�<<�*��X��.cFi]Iy��Y�o�WV�v�:�ew�2�S��]��@�ɦ��Vn.R�=jItL��F-��QC��T��p�L!�� I��%�P NM!��p�7M�0`)h��a��~ � ��!|�~�b.�&�Kzhf�"*��-8i�GQa"�p�_��*�H�i~��3?�)�D��>f`�w��>u�Wl~�<)�DR��u��L��p%Z�8*E\�|�R�5J,35��B<��W`USx$�|N�æp4B��?��E,g�/�EB�DC:�j8'Z��,�.FQ!S��vnF ��D`�i��/�;�o��7��W_�׋��i�ۗ��/�@�R��p�"��V9�J�%�(jN�6�"85 �R�ֈ�t>k�nhQ-�B�h�Z��B��TT�KЪI#��&�]U#�:�)��:��@ ��>8��]!�`o�J�b��"�I�8hx��E��ƪ4�IPc��yy��ᾓ��i l�$��v��o{^�e��,��s[b�Ǟ�� ݹ`�̧�9_U\1��&"�0�웁�7M��t�ۧv!�SXp{��_��(��La�cȋ�U��5)�� Ƥ`��q��0Ѣ��)�H�� Y�eNO�R� 3:Ӻ^�|e��t+߲�a~Đ��ｺ�}bo{3�W`"�*W;��h?ӨiQ��9w��R��Y�4u�t&:�"!a�f8a�fT��0�fl�;C��~4�w��CK0G�1��i��!!�S��!!" "� �"*!��"v��!�=�+!�Du��9�8�� '*��R8�'6É�!�@=»C��:q N0j�Y�E��c2�E�%k��f0A��ʴy3w�WO�6��w��*�m}��o��m�>�� o�jͽ��i;��V:'=��H�d睇^�;�r��/ +��ߏD��+�ٺ1�j0De��0zp� �6G��qd�1�I)u�1# fd��Q 3�*u�V�.��RBd��ڴp��?��x��{��_��G�3�^[�<��z��fC� %;�gu��Yd��W[�H�F"�ݿ�&%ݮn.��;|�� P�&�1��J �#�qW�B}Z��0W��a��y��vJ��z�&�"z��Ӡ�&$��,�P*a�譙L�ȭ�^��HC�X}��)c��hS�6�k��]\��\�4ռvҪ�Ĳ�w7]{��>�.w��#��jt#k�0�KA��;�LW��w'��N|�9l�]!�`oW�(��8u��LT��t��=�� =*��R��=6C��!�@?»C��U�KЇ�k$錌w3pDv,��{�5��2��9��K��ȒO��~t��vv|D�:�ϓܫ$�x��/��,�Ż�*Eo��K ��Ԋ�� +�ԂC8l��!�@?»C��*��7�T�$�[n�LK��Q��SXG��-V��JX��b��Q5��9�$ә�l��A5-�v�hI��-`��R�Wu$��Y��:�H؞?��t\k��ظI�̻��ebZ��o>,��_6n��;�鉶�193�s��QX1�3='9��~�{�}�λ��3��ː�ea��C&� �-�`~�&"?݀�#In�E��AW�H�5��Ԓ�^�t�;��$Ĥ�Nsx�ۦ$x��0�,��kޒa��i`��I�,�endstream +endobj +xref +0 272 +0000000000 65535 f +0000000015 00000 n +0000000227 00000 n +0000000526 00000 n +0000000765 00000 n +0000000824 00000 n +0000000929 00000 n +0000001577 00000 n +0000001615 00000 n +0000001720 00000 n +0000001778 00000 n +0000001984 00000 n +0000002191 00000 n +0000002385 00000 n +0000002564 00000 n +0000002759 00000 n +0000002954 00000 n +0000003149 00000 n +0000003351 00000 n +0000003549 00000 n +0000003756 00000 n +0000003952 00000 n +0000004147 00000 n +0000004338 00000 n +0000004538 00000 n +0000004723 00000 n +0000004913 00000 n +0000005095 00000 n +0000005298 00000 n +0000005486 00000 n +0000005634 00000 n +0000005831 00000 n +0000006025 00000 n +0000006204 00000 n +0000006390 00000 n +0000006592 00000 n +0000006798 00000 n +0000006927 00000 n +0000007132 00000 n +0000007320 00000 n +0000007533 00000 n +0000007654 00000 n +0000007846 00000 n +0000008032 00000 n +0000008233 00000 n +0000008434 00000 n +0000008635 00000 n +0000008834 00000 n +0000032455 00000 n +0000032495 00000 n +0000032574 00000 n +0000032725 00000 n +0000032874 00000 n +0000033023 00000 n +0000033167 00000 n +0000033311 00000 n +0000033459 00000 n +0000033610 00000 n +0000033756 00000 n +0000033910 00000 n +0000048784 00000 n +0000071966 00000 n +0000075956 00000 n +0000076059 00000 n +0000076945 00000 n +0000077484 00000 n +0000077873 00000 n +0000078278 00000 n +0000078646 00000 n +0000078950 00000 n +0000079276 00000 n +0000080141 00000 n +0000080550 00000 n +0000080816 00000 n +0000081143 00000 n +0000081645 00000 n +0000082039 00000 n +0000082346 00000 n +0000082672 00000 n +0000083311 00000 n +0000083755 00000 n +0000084209 00000 n +0000084577 00000 n +0000084929 00000 n +0000086290 00000 n +0000086660 00000 n +0000086794 00000 n +0000086916 00000 n +0000087013 00000 n +0000087105 00000 n +0000087196 00000 n +0000087395 00000 n +0000087578 00000 n +0000087761 00000 n +0000087886 00000 n +0000088059 00000 n +0000088185 00000 n +0000088311 00000 n +0000088443 00000 n +0000088568 00000 n +0000088694 00000 n +0000088821 00000 n +0000088948 00000 n +0000089083 00000 n +0000089210 00000 n +0000089337 00000 n +0000089472 00000 n +0000089600 00000 n +0000089728 00000 n +0000089863 00000 n +0000089991 00000 n +0000090119 00000 n +0000090247 00000 n +0000090375 00000 n +0000090503 00000 n +0000090638 00000 n +0000090766 00000 n +0000090894 00000 n +0000091029 00000 n +0000091157 00000 n +0000091292 00000 n +0000091420 00000 n +0000091555 00000 n +0000091683 00000 n +0000091818 00000 n +0000091946 00000 n +0000092081 00000 n +0000092209 00000 n +0000092344 00000 n +0000092472 00000 n +0000092600 00000 n +0000092735 00000 n +0000092863 00000 n +0000092998 00000 n +0000093126 00000 n +0000093261 00000 n +0000093389 00000 n +0000093517 00000 n +0000093652 00000 n +0000093780 00000 n +0000093908 00000 n +0000094043 00000 n +0000094171 00000 n +0000094298 00000 n +0000094449 00000 n +0000094624 00000 n +0000094759 00000 n +0000094894 00000 n +0000095029 00000 n +0000095157 00000 n +0000095285 00000 n +0000095413 00000 n +0000095541 00000 n +0000095669 00000 n +0000095797 00000 n +0000095925 00000 n +0000096053 00000 n +0000096181 00000 n +0000096309 00000 n +0000096437 00000 n +0000096601 00000 n +0000096765 00000 n +0000096893 00000 n +0000097057 00000 n +0000097185 00000 n +0000097313 00000 n +0000097441 00000 n +0000097677 00000 n +0000097812 00000 n +0000097963 00000 n +0000098098 00000 n +0000098233 00000 n +0000098433 00000 n +0000098561 00000 n +0000098761 00000 n +0000098889 00000 n +0000099017 00000 n +0000099145 00000 n +0000099345 00000 n +0000099473 00000 n +0000099601 00000 n +0000099729 00000 n +0000100145 00000 n +0000100296 00000 n +0000100431 00000 n +0000100566 00000 n +0000100741 00000 n +0000100916 00000 n +0000101044 00000 n +0000101171 00000 n +0000101407 00000 n +0000101535 00000 n +0000101663 00000 n +0000101791 00000 n +0000101991 00000 n +0000102119 00000 n +0000102283 00000 n +0000102411 00000 n +0000102540 00000 n +0000102706 00000 n +0000102872 00000 n +0000103038 00000 n +0000103204 00000 n +0000103607 00000 n +0000103698 00000 n +0000103790 00000 n +0000104025 00000 n +0000104125 00000 n +0000104216 00000 n +0000104316 00000 n +0000104497 00000 n +0000104740 00000 n +0000104840 00000 n +0000105022 00000 n +0000105265 00000 n +0000105365 00000 n +0000105547 00000 n +0000105790 00000 n +0000105890 00000 n +0000106072 00000 n +0000106323 00000 n +0000106423 00000 n +0000106605 00000 n +0000106848 00000 n +0000106948 00000 n +0000107130 00000 n +0000107373 00000 n +0000107473 00000 n +0000107655 00000 n +0000107898 00000 n +0000108006 00000 n +0000108099 00000 n +0000108192 00000 n +0000108285 00000 n +0000108385 00000 n +0000108567 00000 n +0000108810 00000 n +0000108926 00000 n +0000109019 00000 n +0000109112 00000 n +0000109205 00000 n +0000109298 00000 n +0000109398 00000 n +0000109580 00000 n +0000109823 00000 n +0000109923 00000 n +0000110105 00000 n +0000110348 00000 n +0000110448 00000 n +0000110630 00000 n +0000110873 00000 n +0000110973 00000 n +0000111073 00000 n +0000111173 00000 n +0000111273 00000 n +0000111529 00000 n +0000111780 00000 n +0000112031 00000 n +0000112278 00000 n +0000112525 00000 n +0000112762 00000 n +0000113018 00000 n +0000113268 00000 n +0000113528 00000 n +0000118992 00000 n +0000126061 00000 n +0000131737 00000 n +0000151999 00000 n +0000157207 00000 n +0000168452 00000 n +0000170626 00000 n +0000182974 00000 n +trailer << /Info 2 0 R /Root 1 0 R /Size 272 /ID [] >> +startxref +189276 +%%EOF diff --git a/tests/gnarly_pdfs/edgar.pdf b/tests/gnarly_pdfs/edgar.pdf new file mode 100644 index 0000000..b9a1a1e Binary files /dev/null and b/tests/gnarly_pdfs/edgar.pdf differ diff --git a/tests/gnarly_pdfs/edgar_image.pdf b/tests/gnarly_pdfs/edgar_image.pdf new file mode 100644 index 0000000..f670a71 Binary files /dev/null and b/tests/gnarly_pdfs/edgar_image.pdf differ diff --git a/tests/gnarly_pdfs/failing_anchor_pg4.pdf b/tests/gnarly_pdfs/failing_anchor_pg4.pdf new file mode 100644 index 0000000..2a1555c Binary files /dev/null and b/tests/gnarly_pdfs/failing_anchor_pg4.pdf differ diff --git a/tests/gnarly_pdfs/failing_pdf_pg9.pdf b/tests/gnarly_pdfs/failing_pdf_pg9.pdf new file mode 100644 index 0000000..8b869d5 Binary files /dev/null and b/tests/gnarly_pdfs/failing_pdf_pg9.pdf differ diff --git a/tests/gnarly_pdfs/form_on_later_pages.pdf b/tests/gnarly_pdfs/form_on_later_pages.pdf new file mode 100644 index 0000000..4d0e895 Binary files /dev/null and b/tests/gnarly_pdfs/form_on_later_pages.pdf differ diff --git a/tests/gnarly_pdfs/guidebook_failed_pages.pdf b/tests/gnarly_pdfs/guidebook_failed_pages.pdf new file mode 100644 index 0000000..5323680 Binary files /dev/null and b/tests/gnarly_pdfs/guidebook_failed_pages.pdf differ diff --git a/tests/gnarly_pdfs/handwriting_bad_ocr.pdf b/tests/gnarly_pdfs/handwriting_bad_ocr.pdf new file mode 100644 index 0000000..305ff88 Binary files /dev/null and b/tests/gnarly_pdfs/handwriting_bad_ocr.pdf differ diff --git a/tests/gnarly_pdfs/horribleocr.pdf b/tests/gnarly_pdfs/horribleocr.pdf new file mode 100644 index 0000000..699be02 Binary files /dev/null and b/tests/gnarly_pdfs/horribleocr.pdf differ diff --git a/tests/gnarly_pdfs/instructions_and_schematics.pdf b/tests/gnarly_pdfs/instructions_and_schematics.pdf new file mode 100644 index 0000000..c9e792d Binary files /dev/null and b/tests/gnarly_pdfs/instructions_and_schematics.pdf differ diff --git a/tests/gnarly_pdfs/large_prompt_hint1.pdf b/tests/gnarly_pdfs/large_prompt_hint1.pdf new file mode 100644 index 0000000..2635fe6 --- /dev/null +++ b/tests/gnarly_pdfs/large_prompt_hint1.pdf @@ -0,0 +1,4779 @@ +%PDF-1.4 %�� +1277 0 obj<> endobj +xref +1277 739 +0000000016 00000 n +0000022410 00000 n +0000015383 00000 n +0000023825 00000 n +0000023971 00000 n +0000024014 00000 n +0000024057 00000 n +0000024100 00000 n +0000024143 00000 n +0000024186 00000 n +0000024229 00000 n +0000024272 00000 n +0000024315 00000 n +0000024358 00000 n +0000024401 00000 n +0000024445 00000 n +0000024489 00000 n +0000024533 00000 n +0000024577 00000 n +0000024621 00000 n +0000024665 00000 n +0000024709 00000 n +0000024753 00000 n +0000024797 00000 n +0000024841 00000 n +0000024885 00000 n +0000024929 00000 n +0000024973 00000 n +0000025017 00000 n +0000025061 00000 n +0000025105 00000 n +0000025149 00000 n +0000025193 00000 n +0000025237 00000 n +0000025281 00000 n +0000025325 00000 n +0000025369 00000 n +0000025413 00000 n +0000025457 00000 n +0000025501 00000 n +0000025545 00000 n +0000025589 00000 n +0000025633 00000 n +0000025677 00000 n +0000025721 00000 n +0000025765 00000 n +0000025809 00000 n +0000025853 00000 n +0000025897 00000 n +0000025941 00000 n +0000025985 00000 n +0000026029 00000 n +0000026073 00000 n +0000026117 00000 n +0000026161 00000 n +0000026205 00000 n +0000026249 00000 n +0000026293 00000 n +0000026337 00000 n +0000026381 00000 n +0000026425 00000 n +0000026469 00000 n +0000026513 00000 n +0000026557 00000 n +0000026601 00000 n +0000026645 00000 n +0000026689 00000 n +0000026733 00000 n +0000026777 00000 n +0000026821 00000 n +0000026865 00000 n +0000026909 00000 n +0000026953 00000 n +0000026997 00000 n +0000027041 00000 n +0000027085 00000 n +0000027129 00000 n +0000027173 00000 n +0000027217 00000 n +0000027261 00000 n +0000027305 00000 n +0000027349 00000 n +0000027393 00000 n +0000027437 00000 n +0000027481 00000 n +0000027525 00000 n +0000027569 00000 n +0000027613 00000 n +0000027657 00000 n +0000027701 00000 n +0000027745 00000 n +0000027789 00000 n +0000027833 00000 n +0000027877 00000 n +0000027921 00000 n +0000027965 00000 n +0000028009 00000 n +0000028053 00000 n +0000028097 00000 n +0000028141 00000 n +0000028185 00000 n +0000028229 00000 n +0000028273 00000 n +0000028317 00000 n +0000028361 00000 n +0000028406 00000 n +0000028451 00000 n +0000028496 00000 n +0000028541 00000 n +0000028586 00000 n +0000028631 00000 n +0000028676 00000 n +0000028721 00000 n +0000028766 00000 n +0000028811 00000 n +0000028856 00000 n +0000028901 00000 n +0000028946 00000 n +0000028991 00000 n +0000029036 00000 n +0000029081 00000 n +0000029126 00000 n +0000029171 00000 n +0000029216 00000 n +0000029261 00000 n +0000029306 00000 n +0000029351 00000 n +0000029396 00000 n +0000029441 00000 n +0000037869 00000 n +0000037919 00000 n +0000037967 00000 n +0000038017 00000 n +0000038065 00000 n +0000038113 00000 n +0000038161 00000 n +0000038210 00000 n +0000038259 00000 n +0000038308 00000 n +0000038358 00000 n +0000038406 00000 n +0000038454 00000 n +0000038504 00000 n +0000038552 00000 n +0000038600 00000 n +0000038648 00000 n +0000038696 00000 n +0000038774 00000 n +0000038822 00000 n +0000038860 00000 n +0000038908 00000 n +0000038958 00000 n +0000039008 00000 n +0000039058 00000 n +0000039108 00000 n +0000039158 00000 n +0000039208 00000 n +0000039258 00000 n +0000039308 00000 n +0000039358 00000 n +0000039408 00000 n +0000039457 00000 n +0000039506 00000 n +0000039555 00000 n +0000039605 00000 n +0000039653 00000 n +0000039701 00000 n +0000040825 00000 n +0000041682 00000 n +0000042593 00000 n +0000043481 00000 n +0000044443 00000 n +0000045340 00000 n +0000045581 00000 n +0000046293 00000 n +0000047168 00000 n +0000047370 00000 n +0000047548 00000 n +0000047765 00000 n +0000048408 00000 n +0000048974 00000 n +0000049244 00000 n +0000049431 00000 n +0000049690 00000 n +0000049868 00000 n +0000050139 00000 n +0000050308 00000 n +0000050575 00000 n +0000050806 00000 n +0000050972 00000 n +0000051153 00000 n +0000051337 00000 n +0000051548 00000 n +0000051823 00000 n +0000052034 00000 n +0000052624 00000 n +0000053414 00000 n +0000053719 00000 n +0000053906 00000 n +0000054170 00000 n +0000054348 00000 n +0000054590 00000 n +0000054756 00000 n +0000054952 00000 n +0000055144 00000 n +0000055331 00000 n +0000055524 00000 n +0000055714 00000 n +0000055910 00000 n +0000056149 00000 n +0000056321 00000 n +0000056502 00000 n +0000056689 00000 n +0000056882 00000 n +0000057081 00000 n +0000057286 00000 n +0000057467 00000 n +0000057681 00000 n +0000058360 00000 n +0000058728 00000 n +0000059090 00000 n +0000059365 00000 n +0000059546 00000 n +0000059821 00000 n +0000059999 00000 n +0000060263 00000 n +0000060456 00000 n +0000060655 00000 n +0000060854 00000 n +0000061056 00000 n +0000061244 00000 n +0000061470 00000 n +0000061782 00000 n +0000062319 00000 n +0000062681 00000 n +0000062956 00000 n +0000063146 00000 n +0000063421 00000 n +0000063602 00000 n +0000063877 00000 n +0000064052 00000 n +0000064314 00000 n +0000064467 00000 n +0000064648 00000 n +0000064921 00000 n +0000065502 00000 n +0000066273 00000 n +0000066628 00000 n +0000066836 00000 n +0000067050 00000 n +0000067255 00000 n +0000067454 00000 n +0000067647 00000 n +0000067834 00000 n +0000068012 00000 n +0000068178 00000 n +0000068337 00000 n +0000068542 00000 n +0000068714 00000 n +0000068880 00000 n +0000069059 00000 n +0000069225 00000 n +0000069403 00000 n +0000069593 00000 n +0000069786 00000 n +0000069988 00000 n +0000070193 00000 n +0000070404 00000 n +0000070609 00000 n +0000070820 00000 n +0000071269 00000 n +0000071568 00000 n +0000071761 00000 n +0000072034 00000 n +0000072221 00000 n +0000072477 00000 n +0000073070 00000 n +0000073372 00000 n +0000073553 00000 n +0000073804 00000 n +0000073973 00000 n +0000074189 00000 n +0000074348 00000 n +0000074533 00000 n +0000074689 00000 n +0000074861 00000 n +0000075048 00000 n +0000075235 00000 n +0000075463 00000 n +0000075656 00000 n +0000075923 00000 n +0000076128 00000 n +0000076401 00000 n +0000076665 00000 n +0000076840 00000 n +0000077066 00000 n +0000077228 00000 n +0000077410 00000 n +0000077566 00000 n +0000077732 00000 n +0000077916 00000 n +0000078106 00000 n +0000078302 00000 n +0000078501 00000 n +0000078713 00000 n +0000078930 00000 n +0000079185 00000 n +0000079402 00000 n +0000079996 00000 n +0000080581 00000 n +0000080798 00000 n +0000081008 00000 n +0000081207 00000 n +0000081403 00000 n +0000081590 00000 n +0000081774 00000 n +0000081940 00000 n +0000082127 00000 n +0000082293 00000 n +0000082529 00000 n +0000082707 00000 n +0000082968 00000 n +0000083149 00000 n +0000083423 00000 n +0000084005 00000 n +0000084783 00000 n +0000085232 00000 n +0000085597 00000 n +0000085871 00000 n +0000086058 00000 n +0000086316 00000 n +0000086491 00000 n +0000086697 00000 n +0000086856 00000 n +0000087022 00000 n +0000087200 00000 n +0000087387 00000 n +0000087580 00000 n +0000087779 00000 n +0000087984 00000 n +0000088196 00000 n +0000088401 00000 n +0000088751 00000 n +0000088944 00000 n +0000089209 00000 n +0000089414 00000 n +0000089687 00000 n +0000089898 00000 n +0000090173 00000 n +0000090384 00000 n +0000091178 00000 n +0000091759 00000 n +0000092124 00000 n +0000092388 00000 n +0000092566 00000 n +0000092812 00000 n +0000092981 00000 n +0000093180 00000 n +0000093373 00000 n +0000093562 00000 n +0000093813 00000 n +0000094030 00000 n +0000094621 00000 n +0000095212 00000 n +0000095803 00000 n +0000096103 00000 n +0000096284 00000 n +0000096546 00000 n +0000096715 00000 n +0000096934 00000 n +0000097093 00000 n +0000097278 00000 n +0000097434 00000 n +0000097603 00000 n +0000097790 00000 n +0000097977 00000 n +0000098200 00000 n +0000098472 00000 n +0000098641 00000 n +0000098875 00000 n +0000099031 00000 n +0000099218 00000 n +0000099365 00000 n +0000099534 00000 n +0000099721 00000 n +0000099908 00000 n +0000100104 00000 n +0000100300 00000 n +0000100496 00000 n +0000100707 00000 n +0000100894 00000 n +0000101117 00000 n +0000101434 00000 n +0000101794 00000 n +0000101956 00000 n +0000102156 00000 n +0000102325 00000 n +0000102506 00000 n +0000102693 00000 n +0000102883 00000 n +0000103076 00000 n +0000103281 00000 n +0000103459 00000 n +0000103679 00000 n +0000104070 00000 n +0000104357 00000 n +0000104556 00000 n +0000104831 00000 n +0000105027 00000 n +0000105302 00000 n +0000105483 00000 n +0000106147 00000 n +0000106515 00000 n +0000106880 00000 n +0000107239 00000 n +0000107512 00000 n +0000107690 00000 n +0000107923 00000 n +0000108089 00000 n +0000108242 00000 n +0000108420 00000 n +0000108610 00000 n +0000108803 00000 n +0000109005 00000 n +0000109210 00000 n +0000109421 00000 n +0000109627 00000 n +0000109838 00000 n +0000110206 00000 n +0000110568 00000 n +0000110864 00000 n +0000111051 00000 n +0000111341 00000 n +0000111519 00000 n +0000111764 00000 n +0000111926 00000 n +0000112117 00000 n +0000112279 00000 n +0000112454 00000 n +0000112635 00000 n +0000112825 00000 n +0000113024 00000 n +0000113238 00000 n +0000113452 00000 n +0000113666 00000 n +0000113853 00000 n +0000114031 00000 n +0000114200 00000 n +0000114344 00000 n +0000114755 00000 n +0000114927 00000 n +0000115164 00000 n +0000115342 00000 n +0000115608 00000 n +0000115964 00000 n +0000116323 00000 n +0000116962 00000 n +0000117176 00000 n +0000117362 00000 n +0000117564 00000 n +0000117766 00000 n +0000117968 00000 n +0000118161 00000 n +0000118351 00000 n +0000118526 00000 n +0000118704 00000 n +0000118873 00000 n +0000119277 00000 n +0000119430 00000 n +0000119657 00000 n +0000119829 00000 n +0000120070 00000 n +0000120254 00000 n +0000120522 00000 n +0000120709 00000 n +0000120977 00000 n +0000121336 00000 n +0000121889 00000 n +0000122346 00000 n +0000122560 00000 n +0000122738 00000 n +0000122940 00000 n +0000123142 00000 n +0000123332 00000 n +0000123519 00000 n +0000123703 00000 n +0000123872 00000 n +0000124012 00000 n +0000124354 00000 n +0000124504 00000 n +0000124768 00000 n +0000124940 00000 n +0000125195 00000 n +0000125373 00000 n +0000125640 00000 n +0000125830 00000 n +0000126102 00000 n +0000126464 00000 n +0000126844 00000 n +0000127052 00000 n +0000127230 00000 n +0000127432 00000 n +0000127631 00000 n +0000127830 00000 n +0000128043 00000 n +0000128218 00000 n +0000128448 00000 n +0000128614 00000 n +0000128862 00000 n +0000129012 00000 n +0000129374 00000 n +0000129530 00000 n +0000129760 00000 n +0000129935 00000 n +0000130166 00000 n +0000130350 00000 n +0000130584 00000 n +0000130777 00000 n +0000131027 00000 n +0000131226 00000 n +0000131474 00000 n +0000131673 00000 n +0000131926 00000 n +0000132272 00000 n +0000132480 00000 n +0000132658 00000 n +0000132854 00000 n +0000133041 00000 n +0000133225 00000 n +0000133397 00000 n +0000133575 00000 n +0000133725 00000 n +0000133906 00000 n +0000134096 00000 n +0000134392 00000 n +0000134790 00000 n +0000135010 00000 n +0000135195 00000 n +0000135400 00000 n +0000135599 00000 n +0000135798 00000 n +0000135988 00000 n +0000136169 00000 n +0000136344 00000 n +0000136506 00000 n +0000136907 00000 n +0000137060 00000 n +0000137287 00000 n +0000137459 00000 n +0000137698 00000 n +0000137876 00000 n +0000138132 00000 n +0000138325 00000 n +0000138587 00000 n +0000138783 00000 n +0000139055 00000 n +0000139417 00000 n +0000139631 00000 n +0000139833 00000 n +0000140032 00000 n +0000140225 00000 n +0000140407 00000 n +0000140594 00000 n +0000140790 00000 n +0000140977 00000 n +0000141164 00000 n +0000141333 00000 n +0000141480 00000 n +0000141666 00000 n +0000141819 00000 n +0000142046 00000 n +0000142215 00000 n +0000142488 00000 n +0000142669 00000 n +0000142944 00000 n +0000143140 00000 n +0000143415 00000 n +0000143614 00000 n +0000143895 00000 n +0000144292 00000 n +0000144512 00000 n +0000144717 00000 n +0000144910 00000 n +0000145100 00000 n +0000145287 00000 n +0000145468 00000 n +0000145637 00000 n +0000145784 00000 n +0000145982 00000 n +0000146144 00000 n +0000146404 00000 n +0000146582 00000 n +0000146811 00000 n +0000146989 00000 n +0000147262 00000 n +0000147621 00000 n +0000147986 00000 n +0000148354 00000 n +0000148746 00000 n +0000148960 00000 n +0000149174 00000 n +0000149373 00000 n +0000149566 00000 n +0000149747 00000 n +0000149922 00000 n +0000150084 00000 n +0000150271 00000 n +0000150433 00000 n +0000150664 00000 n +0000150839 00000 n +0000151129 00000 n +0000151316 00000 n +0000151612 00000 n +0000151974 00000 n +0000152342 00000 n +0000152706 00000 n +0000153023 00000 n +0000153246 00000 n +0000153432 00000 n +0000153646 00000 n +0000153845 00000 n +0000154041 00000 n +0000154240 00000 n +0000154439 00000 n +0000154632 00000 n +0000154816 00000 n +0000154997 00000 n +0000155163 00000 n +0000155430 00000 n +0000155596 00000 n +0000155867 00000 n +0000156045 00000 n +0000156306 00000 n +0000156493 00000 n +0000156761 00000 n +0000157327 00000 n +0000157973 00000 n +0000158190 00000 n +0000158368 00000 n +0000158570 00000 n +0000158769 00000 n +0000158959 00000 n +0000159152 00000 n +0000159339 00000 n +0000159508 00000 n +0000159661 00000 n +0000159988 00000 n +0000160154 00000 n +0000160426 00000 n +0000160604 00000 n +0000160872 00000 n +0000161146 00000 n +0000161327 00000 n +0000161602 00000 n +0000161964 00000 n +0000162332 00000 n +0000163017 00000 n +0000163234 00000 n +0000163418 00000 n +0000163623 00000 n +0000163822 00000 n +0000164018 00000 n +0000164205 00000 n +0000164386 00000 n +0000164564 00000 n +0000164736 00000 n +0000164941 00000 n +0000165125 00000 n +0000165351 00000 n +0000165669 00000 n +0000166203 00000 n +0000166565 00000 n +0000166840 00000 n +0000167030 00000 n +0000167305 00000 n +0000167486 00000 n +0000167761 00000 n +0000167936 00000 n +0000168195 00000 n +0000168348 00000 n +0000168577 00000 n +0000169663 00000 n +0000169757 00000 n +0000169932 00000 n +0000170206 00000 n +0000170612 00000 n +0000171069 00000 n +0000171707 00000 n +0000172285 00000 n +0000172825 00000 n +0000173496 00000 n +0000174022 00000 n +0000174621 00000 n +0000175114 00000 n +0000175674 00000 n +0000175747 00000 n +0000175811 00000 n +0000178482 00000 n +0000178549 00000 n +0000178628 00000 n +0000178695 00000 n +0000178762 00000 n +0000179616 00000 n +0000179692 00000 n +0000179762 00000 n +0000180424 00000 n +0000180509 00000 n +0000180712 00000 n +0000180896 00000 n +0000180963 00000 n +0000181036 00000 n +0000181100 00000 n +0000181954 00000 n +0000182015 00000 n +0000182659 00000 n +0000182726 00000 n +0000182790 00000 n +0000183461 00000 n +0000188921 00000 n +0000190232 00000 n +0000238970 00000 n +0000239016 00000 n +0000239062 00000 n +0000239136 00000 n +0000239196 00000 n +0000239256 00000 n +0000239338 00000 n +0000239421 00000 n +0000239495 00000 n +0000239552 00000 n +0000239593 00000 n +0000239650 00000 n +0000239720 00000 n +0000239765 00000 n +0000239795 00000 n +0000239836 00000 n +0000239906 00000 n +0000239951 00000 n +0000022219 00000 n +trailer +<<3f57ffbce6d8a944be18becf4608848e>]>> +startxref +0 +%%EOF + +1279 0 obj<>stream +x��]X�ڞ�b��,@HB�6J�E��&6��tB'z�QE�" +6@��*��O��w��u��כ�s�̜2g�y��{�� @�� m'��ԧ �8� +1B�K:.c��H�:kTd��I�Q�!�A��P[]GI[�Y��j('dd3)��JA�"�&��!Y��'1(�I$� �8=v!"�]$g�UM�0d9�̙D ��Eʏ�3��DY��קH�p�x��(T��R0�ʔLGM+�@,.��=��DK|!��K�Ċ�.{�O�;�A�8&�?�|umK�[�6y�@�M��珷�3�N��c��4�oJ�Vbm�pF�6��I��ͷ�g~ޕ�V�C��6W��2Bs�ES`�-H��6��a�\&V��-;NKY��-�h�7m��+ ��H��3��k#K븍zֺھ^�U>��2��"��<��.ث׷�!'��>ǆ�hr?�jz/vV��^��LL��6L;̬�;7gz�3�Hs��ڜ�f�1~��M� U��QS�!eS��JfO��2_;�ߡ�t��C<��)r��ٕ��gC��{��x��>Z�;�Hߖ;�\͒?�<�M(��%`cB��h1B�Z&6��(�ٱM�/��Ⱥ�7�ּ�zh��m�%8:�8zo�p��|'Х[�3f��܁�9+7�8�Z��ݻ�$l��{:�^&=v��}ޏ�G�}��P��$'c��&T�K�܎��H��*u]"�ǆ�U��>�r)�>K;3�|��H�~�^9�=!ιs,�nys�9^�s�5[R�{�R��&��-mK��yQ��]�`$\|�u<��ґGG�� +-�J� +9Z�X�$_V��RT�X�W��[��d�<Z<�;T4/�u�pn�?Ԧ�ǌh�İ6��Kԟ:%=�� #�{��ĥ32KWe&>�pVv=��s�{�oI��J��"D��Ѣ��|�b�ј$tw�Z*6阇!,{�0,�D\�bY�h�ס0�#ڃ�� +ޤg�6��lM}9��1��X��v�l��e��&+4oN��k��N�ͬ��HMs��&Y�;�I�Rf�ٞ��vx^�[�3��e��%3��W_m�-~��]��%)[L߂díc^�_�+��X�#�.��[�o��P�ԥ��F��}�G=�=L��UL��Ͳ.�ޱ;k�e�ͯ�ۊ4�`�/��7��얗#��z��,��*�V��Sy5�*��ky�_(p�q��S~9�q��,Q.�V>�ط�y1�y]�x�g��EV`яZ�A�[-Fb]%��\}��]�nHp]�%�I�a�|��j~y�ɚ��r�Z?��ֱ�<�a�Э� �R�7�)rĔ��k;{��l_��H��Hg��Q�87�iB�t�t�E�"{��]��de��^-�'͖��Y3�C�O��d��A�K��,�N��H��6ʪ�rȴϩ��[�!�5�F=�3�&��7�)�K1��j�SF�Ӟ��Qeugcn?��X��\O{P�D/2�B$a�%(rŉDф&�4qM�?kX�I�*��Xܶ��S��;��x��R��r�/xnз� ��b�� M0֗�χ>Q��%VR�?}��~(��V��Pà �� ?��tz��Y�1*��F�VK��Nǲ�9f�X��@��륭xQ��& +�B�zU�@�o�%��6l~}��E ��YuN%Ǜ�+rə�ݴ�30�®=P�Ei x�3j`��]3=�fY�,z��YC[��D/��7�\ K��c y�Ŷ�e�X�~Wg�ݎ�13w޽��Ű'?�|W\�렗Z��|��۾�@��}�Sˁ^#��:�:�{��b��M�<��YU�ؾ�j�gMs��𬊉�p5��}lN%�m�Kpsz��W�1��n=��'�P��T��6�� K�'��&j��\��8�]�~�ط��E��G!N�^6$�q�nF��u��7��N� +�M��㧺C6%L��фi�\�4��1��k��|��k� +D�8\��H�Q��K +}�E�wr9�V��8�;#��6?�r��ԇ��K�>Y��,S�a�*E�D��G{�U��c�h��Uڻ7��*>�F�k�t7��՟i8�t�l�~��=n9��K��pv��߿�O��v�tl��]&1SKm|6�7��C�]��'Ȑ[��w��EG��>��n��Tf��UV-��f��c'��M?o�3ɢ1��V'cX�M;�$�F+Ú�A3��{f\c7 +dd��wؓ�%u��4� UUKydc�0�۾��͎�Vc?�\��Ӯ�#�w!?<�Mʁg�9/�3��v9I�� +�Ɖ�9��[� 6.B@&�'�� +$�6]:��Xɏ7^/U�m=��k��k��?��:�0��J?�p�9�v6�{��3�񾮩��M��=6�'\�l�iRI-M��K�Gޝx�0�1�ɭ��''Q��I�I�5z�[ug��vƌ��q&g��N��kUL��M�"kXf��b��ᄒ��q��)��ʩ�>��^nO��x�LR�5�\�Ah4�Q��Dȣ[��>^7�X?{.�qB�R]�pti-y"��\��[]C>�_*�Z{�D.�b��!ׇW��[�N��~��P��S�ӗϜj]��x�'䍳 �x:�5.�H\�,�L��)j�f�3p�@@TȦ��l�m�\LO�A>��C�N"u�+2p�*�[t�-H��z�N��'Du��W��8�nh��ͩ��L��1,w�b�E��8M��s��8M^��,�k��R�[`K��)�*��=R蕟�{p}�x��=��⊽�m�=J��[�xI��v��e��ohaͷ�nj��,eh)�N=�]�T��`�࣭�}�U9�FT8��ߨ�mK�w��F��˧�_$��m�c\�'�厼>望=u�)��>�SHՐ��$�B +��q~��X�l��u � {�<]RO�섥Kj��%O��w�,��U�o�4��O� 0ɡHHs ��u�;�Jy�&�Z�z��#��gN�3`밻Fe�<Ѫn�.��r�|��K<��.ɀ�KN��K��% �a钲�Kr��t��K��X>A�qX>��>�B��%މ�C�� Y�,��V�`0��$(��6�K&��S�1��Ǜ� +��s�z��J{0��J�k�-�3?�s�6��ֈ�(��o�1��zV�}��fcx��<Ý�Kq�#I(s��s<޳�k�+��{�9��gB/��w"��o��̝M��F.x��S�/�˯�˳P��)#��_��|��^��!V"�^X?C�15�_��cl��Po`��c��r{i��$\R�B�jZ�"�^f�b�WA�boHI�k|Eݒ�-�A�c��Sҭ��z�� Λmow��>�9�¿m8J��En(�$��?)f�d�%(��0,h�,,Dq��p��rS-�"��S�P +)�� +xn>Z��L-��w��hio�ӄ�4��4��Ʌ��K&< �ӎ)� D/��!ċ$,��6*c0��)n�ha ��R�=y�룟�$e�ݔO��`�$��l��Kr2��_~Y��FYlo��{z|R�L�8ԕ��K/�U��wĹ��M�;��l�8��ᅠ��A��^�� x!(�b�BPΓᅠh� +AA�� +A �|��g%^ +'��$X!(BPV��z��x�B��P��"Y(��*A�,G&H�#D�-�(1��(��4H��h�gC��L\b�Kr�A��ii�J�� +&��?Q��T��8��:>6�L�͆'S`��d�fx2��R^�!��X�?��3�2K�%SJ��Ȕ� +2e� +_��#�W��ޟK�d~�̖C�p�~�8ԐZ.6��<Ӄ��pq���a�+�ɕ��Ű_�b��QƘ:'��t^�M�}��r�R~H�O�zT�/�cB��f�w2/��]�LH��TH��Ι퀣��ӄ�`D�R'��0�&� � N��q�jT,U!�p�P��'��LPm +`EXI}�a��au��:ªH��s��'�۾Q�7��tf�g1ۘ�l:��Kf�3��+hC�s1=n.�8M�f��ǺPڨ��O�� ߍ�: ��=�I0(�n2u��_#-�l^=�*�Zo䓭�M��=��]ߧ��5ąc"i��:�?� %|�O�jx�� pXH'2�Q��P��Jt� �x�A��y��F#//:Jt��0�a��D��&:L��]��a�OU��ן7��Fǈ��_��l�P��oG��RA�qAA��p[��6�~��&��Kݾ��z�� E=�z��z��080T1�C��PD�� =��+��{��d��\�j�V�f�<�=��/��aPRU��J��4P��G�A��D��v� �@J ��@#~��RN��c �)��E�BMP��( +쥁�� @�J�O#H�J�jN#�\��G`쭠M!Z��xj�� BP,��(c'C�+�@@)r��Ъ�(��s��)�|ηz�π�΂�PF��u_�� z�/�}BJ��(�T +Fx�@��p��/�M��rw�~1�˱��L)E� ��5ԃ +�<A��8� +�pE��;�aFB�n�q�vgv�!X�b#(�z,��,��P7 +��P�@�.�?G�!��C�(��H��g��^��/AM6�)P�D�YF�4:�P�B"is�/e�2�3��z�\=v�O@��' O\,�TBA�HO۩-�ܩ4��8]X6R{*ǈP,"� �zӠ��zO�SpN�Qk'�rA��8�m�^o18N�W�=�� m��F�zfP7��0.c35�"�Qk�PGY � \��^S��C +"}�`RF}��c)Pg} XFG��P��:6��Fp�!�x��'m\�-FQH��"B��.H�32 +� +��d 4��@�2�Ar�$2� +�=b)��(~U&�u<��\�,�ћ eH��Cl I��7b<�Ȩ��%��z�@�2 +�� V3&�g� A�/u�o�b $ +endstream endobj 2015 0 obj<>/W[1 2 1]/Type/XRef/Index[951 326]>>stream +x�bbb^��8J� �q�h��(1�4% �Y +endstream endobj 1278 0 obj<>/OCGs[1280 0 R 1281 0 R 1282 0 R 1283 0 R 1284 0 R 1285 0 R 1286 0 R 1287 0 R 1288 0 R 1289 0 R 1290 0 R 1291 0 R 1292 0 R 1293 0 R 1294 0 R 1295 0 R 1296 0 R 1297 0 R 1298 0 R 1299 0 R 1300 0 R 1301 0 R 1302 0 R 1303 0 R 1304 0 R 1305 0 R 1306 0 R 1307 0 R 1308 0 R 1309 0 R 1310 0 R 1311 0 R 1312 0 R 1313 0 R 1314 0 R 1315 0 R 1316 0 R 1317 0 R 1318 0 R 1319 0 R 1320 0 R 1321 0 R 1322 0 R 1323 0 R 1324 0 R 1325 0 R 1326 0 R 1327 0 R 1328 0 R 1329 0 R 1330 0 R 1331 0 R 1332 0 R 1333 0 R 1334 0 R 1335 0 R 1336 0 R 1337 0 R 1338 0 R 1339 0 R 1340 0 R 1341 0 R 1342 0 R 1343 0 R 1344 0 R 1345 0 R 1346 0 R 1347 0 R 1348 0 R 1349 0 R 1350 0 R 1351 0 R 1352 0 R 1353 0 R 1354 0 R 1355 0 R 1356 0 R 1357 0 R 1358 0 R 1359 0 R 1360 0 R 1361 0 R 1362 0 R 1363 0 R 1364 0 R 1365 0 R 1366 0 R 1367 0 R 1368 0 R 1369 0 R 1370 0 R 1371 0 R 1372 0 R 1373 0 R 1374 0 R 1375 0 R 1376 0 R 1377 0 R 1378 0 R 1379 0 R 1380 0 R 1381 0 R 1382 0 R 1383 0 R 1384 0 R 1385 0 R 1386 0 R 1387 0 R 1388 0 R 1389 0 R 1390 0 R 1391 0 R 1392 0 R 1393 0 R 1394 0 R 1395 0 R 1396 0 R 1397 0 R 1398 0 R 1399 0 R 1400 0 R 1401 0 R 1402 0 R 1403 0 R 1404 0 R]>>/PieceInfo<>>>/LastModified(D:20031211134210)/MarkInfo<>>> endobj 1280 0 obj<>/PageElement<>>>>> endobj 1281 0 obj<> endobj 1282 0 obj<> endobj 1283 0 obj<> endobj 1284 0 obj<> endobj 1285 0 obj<> endobj 1286 0 obj<> endobj 1287 0 obj<> endobj 1288 0 obj<> endobj 1289 0 obj<> endobj 1290 0 obj<> endobj 1291 0 obj<> endobj 1292 0 obj<> endobj 1293 0 obj<> endobj 1294 0 obj<> endobj 1295 0 obj<> endobj 1296 0 obj<> endobj 1297 0 obj<> endobj 1298 0 obj<> endobj 1299 0 obj<> endobj 1300 0 obj<> endobj 1301 0 obj<> endobj 1302 0 obj<> endobj 1303 0 obj<> endobj 1304 0 obj<> endobj 1305 0 obj<> endobj 1306 0 obj<> endobj 1307 0 obj<> endobj 1308 0 obj<> endobj 1309 0 obj<> endobj 1310 0 obj<> endobj 1311 0 obj<> endobj 1312 0 obj<> endobj 1313 0 obj<> endobj 1314 0 obj<> endobj 1315 0 obj<> endobj 1316 0 obj<> endobj 1317 0 obj<> endobj 1318 0 obj<> endobj 1319 0 obj<> endobj 1320 0 obj<> endobj 1321 0 obj<> endobj 1322 0 obj<> endobj 1323 0 obj<> endobj 1324 0 obj<> endobj 1325 0 obj<> endobj 1326 0 obj<> endobj 1327 0 obj<> endobj 1328 0 obj<> endobj 1329 0 obj<> endobj 1330 0 obj<> endobj 1331 0 obj<> endobj 1332 0 obj<> endobj 1333 0 obj<> endobj 1334 0 obj<> endobj 1335 0 obj<> endobj 1336 0 obj<> endobj 1337 0 obj<> endobj 1338 0 obj<> endobj 1339 0 obj<> endobj 1340 0 obj<> endobj 1341 0 obj<> endobj 1342 0 obj<> endobj 1343 0 obj<> endobj 1344 0 obj<> endobj 1345 0 obj<> endobj 1346 0 obj<> endobj 1347 0 obj<> endobj 1348 0 obj<> endobj 1349 0 obj<> endobj 1350 0 obj<> endobj 1351 0 obj<> endobj 1352 0 obj<> endobj 1353 0 obj<> endobj 1354 0 obj<> endobj 1355 0 obj<> endobj 1356 0 obj<> endobj 1357 0 obj<> endobj 1358 0 obj<> endobj 1359 0 obj<> endobj 1360 0 obj<> endobj 1361 0 obj<> endobj 1362 0 obj<> endobj 1363 0 obj<> endobj 1364 0 obj<> endobj 1365 0 obj<> endobj 1366 0 obj<> endobj 1367 0 obj<> endobj 1368 0 obj<> endobj 1369 0 obj<> endobj 1370 0 obj<> endobj 1371 0 obj<> endobj 1372 0 obj<> endobj 1373 0 obj<> endobj 1374 0 obj<> endobj 1375 0 obj<> endobj 1376 0 obj<> endobj 1377 0 obj<> endobj 1378 0 obj<> endobj 1379 0 obj<> endobj 1380 0 obj<> endobj 1381 0 obj<> endobj 1382 0 obj<> endobj 1383 0 obj<> endobj 1384 0 obj<> endobj 1385 0 obj<> endobj 1386 0 obj<> endobj 1387 0 obj<> endobj 1388 0 obj<> endobj 1389 0 obj<> endobj 1390 0 obj<> endobj 1391 0 obj<> endobj 1392 0 obj<> endobj 1393 0 obj<> endobj 1394 0 obj<> endobj 1395 0 obj<> endobj 1396 0 obj<> endobj 1397 0 obj<> endobj 1398 0 obj<> endobj 1399 0 obj<> endobj 1400 0 obj<> endobj 1401 0 obj<> endobj 1402 0 obj<> endobj 1403 0 obj<> endobj 1404 0 obj<> endobj 1405 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 0/PresSteps 1999 0 R>> endobj 1406 0 obj[/Indexed 1425 0 R 185 1990 0 R] endobj 1407 0 obj[/Indexed 1425 0 R 2 1989 0 R] endobj 1408 0 obj[/Indexed 1425 0 R 255 1988 0 R] endobj 1409 0 obj[/Indexed 1425 0 R 3 1987 0 R] endobj 1410 0 obj[/Indexed 1425 0 R 6 1986 0 R] endobj 1411 0 obj[/Indexed 1425 0 R 4 1985 0 R] endobj 1412 0 obj[/Indexed 1425 0 R 33 1984 0 R] endobj 1413 0 obj[/Indexed 1425 0 R 40 1983 0 R] endobj 1414 0 obj[/Indexed 1425 0 R 10 1982 0 R] endobj 1415 0 obj[/Indexed 1425 0 R 191 1981 0 R] endobj 1416 0 obj[/Indexed 1425 0 R 5 1980 0 R] endobj 1417 0 obj[/Indexed 1425 0 R 7 1979 0 R] endobj 1418 0 obj[/Indexed 1425 0 R 255 1978 0 R] endobj 1419 0 obj[/Indexed 1425 0 R 4 1977 0 R] endobj 1420 0 obj[/Indexed 1425 0 R 4 1976 0 R] endobj 1421 0 obj[/Indexed 1425 0 R 8 1975 0 R] endobj 1422 0 obj[/Indexed 1425 0 R 4 1974 0 R] endobj 1423 0 obj<> endobj 1424 0 obj[/Indexed 1425 0 R 3 1972 0 R] endobj 1425 0 obj[/ICCBased 1973 0 R] endobj 1426 0 obj[/Indexed 1425 0 R 6 1971 0 R] endobj 1427 0 obj[/Indexed 1425 0 R 157 1970 0 R] endobj 1428 0 obj[/Indexed 1425 0 R 143 1969 0 R] endobj 1429 0 obj[/Indexed 1425 0 R 170 1968 0 R] endobj 1430 0 obj[/Indexed 1425 0 R 155 1967 0 R] endobj 1431 0 obj[/Indexed 1425 0 R 194 1966 0 R] endobj 1432 0 obj[/Indexed 1425 0 R 160 1965 0 R] endobj 1433 0 obj[/Indexed 1425 0 R 173 1964 0 R] endobj 1434 0 obj[/Indexed 1425 0 R 183 1963 0 R] endobj 1435 0 obj[/Indexed 1425 0 R 129 1962 0 R] endobj 1436 0 obj[/Indexed 1425 0 R 110 1961 0 R] endobj 1437 0 obj[/Indexed 1425 0 R 63 1960 0 R] endobj 1438 0 obj[/Indexed 1425 0 R 30 1959 0 R] endobj 1439 0 obj[/Indexed 1425 0 R 13 1958 0 R] endobj 1440 0 obj[/Indexed 1425 0 R 194 1993 0 R] endobj 1441 0 obj[/Indexed 1425 0 R 3 1992 0 R] endobj 1442 0 obj[/Indexed 1425 0 R 4 1991 0 R] endobj 1443 0 obj<>stream +H��Mo�6��<Ȉ�C�Phѯ99[�ك�I�C��;��]q8 +0V+�/_�z��x|�c:>�&��o�j��|7O�j��9��>�1��3DT��~��_��Q�Ǐ�o�8M��I��?Lj��uR_޽��p��[5�v��|�}��C�&x��'�<��kuxQ�v�6W�rx�:҈?�_>��*� Ϋh|�a�|֎D��L�h�컱}�,�Aƪ!$��`�BzhQks��Zp9gZ��}MJ�`t,�i��ڪ��0��Շ��%��,�G!��,�2�+�,~�:Gw*��$��;[H��1��k]��m;o�!�6��@�~^?ӺVK;�g[ 1m��2�حZ��UI/) +��N��f��e�"��{��:.�L��Ew�%��.U@�ʎ>/��^��z�63a6��kg�a�� +��R�2%�~�4��(P��&a8�Mm)1��Ƌ%�.NV ��Dj�lL�o�s��;�=_�ܳ�w$>CS��qw��@N[M��nr�a.��N�Ƹ5k%+_{=WY�� D0��n�vz�n:��p�!|��]bb��V¥,i#��N*��-�h��E%fM%8 �*z�z��e]N�.�p7��w�&K��I_�ʢbE�Yc��,x� +�e��d��,�|��0��MHW'K�YZ��N�D�Hsu��{�-�]�i��Kx��2��;�|�n޾4޵��(\�=m�dY��e꠹��X*Q&�M�� $ʥ��N)L܂2�X�.$4;��晻�>�.�w6�1HBI��S!�V�� ȵY9V��U⮲\n��[F��ݜ7ڮ��0u�q[xq)�@�-�"��?�A�@��J)�q��P��#��b |�t[B)_��X��a�P�5x��RI��ؼH��f�P!M��JY��kzm�mRx��@��5� 7 +UQ��X��o�cc�� +endstream endobj 1444 0 obj<>stream +H��Mn[1��>Ż@��#��.��R6,�IDL8��ɓg8D�|�!49�^�C+("��]��D�g��/�B��pG)U�H�P +� +�RHu�Af�C� +)�$�e�V��d�$3�r�_"ҝ�A��;/��U�� j{Q��5$lV<��W(�b�%'�Q��f��cv2� +��O\�x%�A��y��7��]�Q�R'4��J��cjb�|~�;��ҙ �p*�n��8 �A2�4Yű��̦��yfw朇�@��.�1�ܢ��oy��I��$ `� +��QTu��;|2;��t� +TR��#�wt�YFHQs-�v��p(ə��C��G�@�g��mK��b�,r��,~[PH⠛r�q_��?й�� +endstream endobj 1445 0 obj<>stream +H��Kn1D�:�\�-6��A��EVF��@z��R��K��n��?8��˯۷��w��ss[L$j?ۛ�"�۸eV�-d*�9��? +�=S�UOҠxIY)課��L�Ze��,��_��.��pS�3^�g��=�2K'�fJ�S�ɕ��Lh^:t$I/3�e��3��$�'�%�D9��@:��[iظdo{Eٛ#�(;��B�C��A��c�pE'a��Ȇ�j�8Q��9�p��b�3�ھ��V|��jYI��Vб�"HP�� w/4hE'�\��gwWtR��ʲ�݁b�_��7��i)Ӫ�a}'��;�'n��[�%�k�q}��Ed4N ؏�T�w�l��&��U��^�E��V (�ɑ@�X��̎՟P8�H��Uщ�U��?�D�MF]ȧ�?�Q:��f�V8n2�V&F(��T�'��Q9c�'�H��U��T�6��7Ә��\�I��[6@�!��֙�GMF�;� +6��S�c��pU��]ߟ��G]��:��e�\��&5��Ӵ��⨫uo��M��Ƅ�X��+/��~A�P��[~A[~�Ɗ��W�S�� \n�';�ʵ5}z��֧ж�hkA��r!��ns��5��e��cтߟ8�� +�{L��y�k�Xx�X��5/�~�)Vj��KH�:�u��u�L^&�nw�9�q��l _%;��c��u��w ,-GIO��q� �gI�}�2<��aRT��Ԇ��㷿Gs� +endstream endobj 1446 0 obj<>stream +H��Kn�0D�s +] m��r�,�2� ��R<?%P�l�MI��w��}��x��m?�<�ƞo��o�SV�ٔ��s��x��l�ِH[�)��&�r�F��H�W��D(�g�G��}�kѲ� ��-�,�@�#Vi8Y��;��.;?��>stream +H��͎�0��9#mo��;�� iA��Kt2ob��ao#͔��T.?~~��.ݞn�o��P��/��bFV��Ju��B��J�40%��E +�x�&��ڤ��L��ڛ�M(�J��)Y:칀=��f�}+IZ�5��澑j��k$ +��I��B9_��Z"=ƪ��7��\a_�� Hz�eIs�L��Q�H��%��l[��Y�}�W��׀�Z�(P&�*e��-a]d�Ul=��OQ��"d ��KP��YWmA�=��`=�f�6\� \F��w�� +��h�.�ؗC�+�W�p�O��=�SA��G��8�1ߤ��R�>\v��/�:R��{��y��Yv�JR��#��e��l@�3W�#%B�^�7��;?c\��C(��Xwx r��=g�$*��X?Z��>��5�;v�4@t*�l�Ӣi+UH|�D|1�� +F#�>�4Y��m +�&��k�V7��^��͉��KA|A��ݺI��I�@��M�G��>n�~��tN�.��͑��`7m�]��޵�.�M�� +�y�.��`�bߴ]��7�b/��;�&7�oz�b�9t��n/��)i\a�/�� za0��.lA�;P�L�R��Y��VvWq��k�<л�t�,�J��h��DQ��禾��F|A��I�J.��X�.��k��xŵ뫷 +��d�i[C��!��&�o��7Ϡ�W~=��޿�ɓ:��-�P��6iO��:8W�:��ڤ�3j)�s-~�'�Px� +endstream endobj 1448 0 obj<>stream +H��1n1D��iI�(��r�� p?�Zi6+ĕa�y��壄��ǯ��oa��hĺ�*��K�Tj-[2 !m�?�_>8��LV��Tҁ#��/��d7q^7^O$��BlYRf�X��ʜ2�.��x�ϭ3^O�["�'��f(]��HY��%��m��F�B��x�� +HW +e��(�H��f��]sx%�Px4�4%��B5��^r�R7��eB�C�j�t�Jr��/�h��Cc�� +m$�C31 ��^��UJqUX��W"��e0P|��x%6�Vϴ��di��SVC4�$�N�`*��rw�/tR^�i��?��cq�Rey�E�ƒ}��e 7!�]& F��A� &}&��z��IQ��=�_��Б��x�I�w!/�폋9>�~�� ?�aC��C]>��ū�ܾ*v�]�]<) +0V�X�w��=��u�z�n4��[�`'��ol�;�B�6t�ق^��~%g;�｝��<�=�,;��J�x�_��Ǝ�n�s�{*\�s��c��}�pě<C��f�9��50�GV��츀�x�#�� +endstream endobj 1449 0 obj<> endobj 1450 0 obj<> endobj 1451 0 obj<>stream +H��A��0D�9�/P~��$r_��(��($��4R�t��e�|��+��߾��޾|�Ͽ�p�x| +T��L��ʑ +U��no'�g�# ֘�c��J!��B4�Y��@��7ϸ%R !5\w��Ai�٘)��/��+��d#1bM��ZW)e��f�R��e�Э��3�YT�P�w݆κ��˘�n,�ι��9��u-m�VJXˤz�]��T��u��0�*��n� iy��F�JkW��O��t��k�n�$�6��]`�)�"i�ʂ0�F*$��(+`�,�z��t��V�-6Q��>�J9�_)�e�;�ҝ�PM��J��L�,18��%�� /\�Q JE�Xkի�=�%nT�GJD��E �fdW+E�/�l��͇��DPZ�p�'�uH�Le?�q�t��(�d?��#q��Ǔ�c��uF8k��K�u@��M�;�F�\u��P��Ӡ�d'ݵ��g�~*��tAx�x>��2��n�5��(]j��r�{��(]2ɥ_�%�{��v..�hq��tcG��΄�}:��%B��Z��h�*x`O�s�E�� i]К_��Sn�t��K��o��e�8[^��v�]�i>��?0�֘��u��>stream +��һ��p~rYnXm\M[RB_TCgZMcaOjdKfbNXXHTTBIJ;9:1891BB8c`M�� + +endstream endobj 1453 0 obj<>stream +H��)��P4�2nq5@�'= + +endstream endobj 1454 0 obj<>stream +��·�ű��t�ezq[hi]cg]eaMc`KTSECD9<<11-$+%-*"33,JK?��o�� + +endstream endobj 1455 0 obj<>stream +H��)��-��C�� e ��,��$��(��$rfK��߾}��ͳ�/޷a��u��uh��]�{��K�֞ؾ`Ղ��ܸ��0Oog77skcCMm5Y q y1-e)#C-oo�9�+߽zL��K`�/�FF@7�>stream +H��)��!�zK��K�&�E�F�&�VeF�d��f��{�ĺY��9�X�;�zz:::��ژ��F�Ŧ��g�vO_x��O^~��;g��]پ��272!%!��}��'�^�w��>��_��{�n_��o֔� �I} ��!~�vN�z:ʚ�2�*2� +�� +��*&��i ��o��1�k� +�J �L�� +��tq�sv��w�657��5��3�26�33׷��3��q�ϊn,��-�o��<�c��9{��ߴ�g��9�I]ya5�a%�e�A�� &7��_��7 ��}��on]ݻsӌ�=��{�&u��U�7��'�G8�{�::��XY깸��;�ϟ7��0�� + +endstream endobj 1457 0 obj<>stream +H��)��&�F��8�ڙ�i��뛙��Z:Y;٘�ٛ�8YƹZ��ڵf��EO��X�[�[ړԑ�ג�ۚО�Y4��d2M��]�S`�@3� + +endstream endobj 1458 0 obj<>stream +��ͯ��m��o��q��n��i�|b{s^wm]�~d��Ƭ�� + +endstream endobj 1459 0 obj<>stream +H��)��J"�sB�B�½��CC�"�}��=�=m�lLܭ�}��<�LÜ͓��R�]S#ܻ��&W�M,�EF�dYM(��/��^��rr=@��5� + +endstream endobj 1460 0 obj<>stream +��н�˰�� + +endstream endobj 1461 0 obj<>stream +H��)��!��J�z":2�;"�"Jb�"}¼��LL�� 4��U��%ՕelM4�-u\퍝,��L�\-�M�l=l�ݭ��l�B]�C\��|�2�R+� + +endstream endobj 1462 0 obj<>stream +�� + +endstream endobj 1463 0 obj<>stream +H��)��!��D��ɢ��֌�D��D��xߚ��hϲ(��ה�W+#Gs}[cm}U9#m55y=eueY-]u}MS#-+osO;/�r�.� + +endstream endobj 1464 0 obj<>stream +H�b`@�L��,�@��/ ($,"*&��H0KJ��4��,��<^��S#@1@� +� + +endstream endobj 1465 0 obj<>stream +�� + +endstream endobj 1466 0 obj<>stream +��ϻ�ɶ��ʸ�� + +endstream endobj 1467 0 obj<>stream +��ؿ��}lTTFDD9AA5>>2KLFa^R{zh��}�ֺ�� + +endstream endobj 1468 0 obj<>stream +��׿��zxr^i\KbYCi`LvaSteOebLdaMYXFLL>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�tZ�mWxbR�k\yoehge_``_ca[`VNSM>stream +��Ļ��r��j��k��i�|cqkUXUIOPDPQABC866-00(,( 0/&13+I=4UQCvs[Ż�� + +endstream endobj 1471 0 obj<>stream +H��)��zK��#K�b�� +��c��R��3#<2b��㯞]x|�ȍ+o]=t�҉��N߻��<�q֊�u͵��a�q^� ^�q�~��.�N�.��&��扡��Q{�Oz��> + +endstream endobj 1472 0 obj<>stream +H��)��K"��½}]�|�K�|��|�2��rB��zJc�+��*�'��vV��ŀQl_ylGE\OYL[YL/X��,��$�� 3/�9տ9ݯ%÷37�>7��r��E�f�X�l��ew��p��ɧ^�x��w�ޱh��ζ��T��Xϸ�o�O��hWˌX�֚��k��:��W��⨮��?'GKSWs]_;�@'��`�P��8��x��h��8��h��X��X��8��☶��v,�R5wbewOmCcuCkK��V�\��ٻ��߻v��[W�\=��=�o�w��c��,j�ܕ�Ӓ�Q��R�Z��V�؜ِٜ^�Z��X�ז>cB��/�>��ޒت찈 w7GkS-MuE%%E%iu %imEIC5YMMy3=Kc- r�5��L�+-.�o흱z��G/\:x��c;��8m��ܢ��̜��y�Wnܻ��գ��]y��ʫ�^?��۽'O�|��߾=z��⻗��3r�Ԟ쮦��洶�Ԗ��ֺԎƴڊĞ��3*�]�� ]� + +endstream endobj 1473 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�uZ�kU�mV�s[ocVYKBM=3>7+9;000&+,&,/),/)))!,.&.0);=9ITURfgeqtl��z��ʨ��i{p\mYElVIti`�wm�¸�� + +endstream endobj 1474 0 obj<>stream +��պ��z��s��m��j��m��o��p��w�� + +endstream endobj 1475 0 obj<>stream +H��)��!��DQ_I��Ж��D��$�Ҕ��$��0�� ̌��Ą��Ъ��HϢϦ܈Μ� e1ӫ�WL�0=9� + +endstream endobj 1476 0 obj<>stream +��Ӻ�ӻ�Ħž��ħ��ȭ�׿�� + +endstream endobj 1477 0 obj<>stream +H�� -�R�@�AJ!:��`�>/'��/�Py*��xm��%�F�%��C��@��(A�F(}�� +0�}>� + +endstream endobj 1478 0 obj<>stream +�� + +endstream endobj 1479 0 obj<>stream +H�b`�A�� A�+V�8��j,��H3 �^�j + +endstream endobj 1480 0 obj<>stream +H�b` 021�H0h6)�$�G�(C2U@ + +endstream endobj 1481 0 obj<>stream +��ּ��tcdU]^NbcSmk\��ѿ�� + +endstream endobj 1482 0 obj<>stream +��ʲ��}hXUE;:/,) *'*')'+)!BA7okZ��w�վ�� + +endstream endobj 1483 0 obj<>stream +��§��o�z^]ZG54,&&&&)#/.';;1FG8SQ=woV�պ�� + +endstream endobj 1484 0 obj<>stream +��Ἱ��k�|dibN?>1.(!/'66.B@3HF5JI;KK=HH:CC9DD7OO=��u�� + +endstream endobj 1485 0 obj<>stream +H�b`��YX��98��yx��ED��%�)�IFIFF)iY9yE%e|�a@�ATՈQ ��)U `�L + +endstream endobj 1486 0 obj<>stream +��ȼ��Ϳ�� + +endstream endobj 1487 0 obj<>stream +��̢��wkiW[SH\WL[]MaaQqlZ~x_�� + +endstream endobj 1488 0 obj<>stream +��v`_LDD60.$(%(#*%-+$67/JK>gcP��z�׿�� + +endstream endobj 1489 0 obj<>stream +��̱��j_YHDD9.-&(,2,$?=1ED4LL=MM=NN?LK:pm^�ô�� + +endstream endobj 1490 0 obj<>stream +��ʼ��lphQ`TBNF:FE7UJ;bM>aXF__M]]MYYHMM=BB8>?3AB6okWǿ�� + +endstream endobj 1491 0 obj<>stream +��Į��p�}cslWpiVvmWkfRbaLebM\[IRSCGH;:;20-%*%.-&67.�i�� + +endstream endobj 1492 0 obj<>stream +H��)��P�[=�:aIo%@��>k + +endstream endobj 1493 0 obj<>stream +��˿�ɶ��x�zeyjWdbS\XITRDMOBA@532)+&(!( (,) 34,LL=�� + +endstream endobj 1494 0 obj<>stream +H��)��:*��ޒh ꃑ`FToqd���Ԁ��9k�ܹp��kW�޻}�Ɲk'�\�t�荻gO�?~|��l\��Y\S��fic��,�./�� D�@��몺��޵�Ƿ�D:��Ȓ�$�� ^�/��/��P3�,��8��"{#��CZ��c=�#��"=+c}K��*��R3B�3�[ʒ�n]�r��S�]��С=/]yt�ڃ�O_^�fi�WN~��P�Af + +endstream endobj 1495 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�t[�kU�iUtfZl`Tj^JfZM`SM_SF`SHWH@QB7IC6GF6A90?4-<;1=6.FE>GJCLVVbkuq��̬ԩ��nwt`kkfrspz��Ծ��f��e��f�za�t\fbN[VBSQC<;276-64+<;0FG:gdP��n�� + +endstream endobj 1496 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�tZ�nY_ZPJLE@B:9<413-)($&"($*+%+.'*+%))"+,$,.%-.&13*CDBPY[akqr��y��z�xb`KsUFiP>hYDlVG{eT��u��Ӿ��n��e�{`��g��k��c�z`xr]ieQkfT|t`��~�ï�� + +endstream endobj 1497 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�tZ�s[kdWRZSPVSRSVIJJ>?6;B<9>7:;0;<2;<19>;885;=:=?<782;<8>=<964=B>=DA851,( /+"//'.0)02*64-E:1EA3WL@wna��{ + +endstream endobj 1498 0 obj<>stream +��ǽ��y��r��o��p��r��y��պ�� + +endstream endobj 1499 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�tZ�q\~rbzi[u]Rk]�q\~q^}nZ}kT}lU~lUueWnZPwdUr\PjSIWQDNN?IH9A=-;5*5/%0,"1,#0+#64-75-65-43,23,>stream +��Ҽ��ż��ǲ�ӻ�� + +endstream endobj 1501 0 obj<>stream +H��)��!�zK�&�D�E�� @ YA@��HqdoaxwqTWItgqTgQdGad[ADk~xK^Xu�_U�OY�Gq�{��SJ�SF�Un䔪��U�>stream +��⽳��mfaPBA4-) (!+' )%*'+,&:;/KJ7xs[�� + +endstream endobj 1503 0 obj<>stream +��Ĵ��mkbOJF>stream +��ٲ��k~v_`VIC>4;:/<8)CB5RP@^YD][HZZJSTCEF;AB7LI@xsd�� + +endstream endobj 1505 0 obj<>stream +��ñ��|axatlZldPf`JfaJ_^KgcMidN\\LQRJBB;97.44,9:5VQ@�� + +endstream endobj 1506 0 obj<>stream +H��)��'*�ޒh ��T�-�0'�= + +endstream endobj 1507 0 obj<>stream +��ɲ��ĺ��w��a|jUa`O\\K^^MVWHOPC@>33/'+%'*$43+FH:��w�� + +endstream endobj 1508 0 obj<>stream +H��)��-��-EG}% �D� ��0��0�+/�=;�-'�%'�>-��p�֕�O�:ux��S��\��ж��{��<��旻;�휴jń��PWO/[;{um59 E u9Q q �Q�31��]8��04xog + +endstream endobj 1509 0 obj<>stream +H��)��!��D��BPL_YL?��w�I �͏�� i+m��L�-M�N�� +�,��k̉�X��lr�� L�*jo)o��5g��GW��q�͋�=x��ݣ玭۲i�� uIM� ueI�A>�V&�Z��j�bj�" +⚊@�./��db��|��??�Q �G ��WITi��ִ��tjL�oH�� +ʋ�I��u��s� +q�t4 t��NHM*k��>a�ek��ؽk��]��jF��Ɩ��%��>��֣��_?��K�o��}q��S.��v��Y��;��Z��ZkSK��"ݽ\�,,��tT�54��ĀHSE��L��֠��0��; + +endstream endobj 1510 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�pX�qY�u\}eQucTsiXjdZ]VKSKCLE>E>5A:1C;2A?431'12-.2+14-@FD[fpi��x��Ǣ��fwjaccalqr��n��f��f�w^wnS`SCMD8DE887.10'0/&65,FGA�vbƼ�� + +endstream endobj 1511 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�t[�lUsbP]WGOJECDE8:4,,$%'#)'+,$-/(13+23-44,::1::5EFFYb^hu{{��~��x^�qW�pV�pX�qY�sZ�pW��v�� + +endstream endobj 1512 0 obj<>stream +��ȿ��o��k��k��m��j�{^xoUkfQfcPggWsnU�� + +endstream endobj 1513 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�v[zlU_XO\VXSRKIH:D@2CF@?A>;74>>;68213+01*67/>?6FG=OD=WF;TC5^A2nI7rL:pK:tR@�kS�qX�rY�rZ�rZ�sZ�tZ�pU��u�� + +endstream endobj 1514 0 obj<>stream +��¦��u��s��r��m��r��ƪ�� + +endstream endobj 1515 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�u[�kX}bS}aP�mV�iT~cSvhTjYJ[NAJC9C;2;/%9*!4)!3) 8) /&*$)$-'+#0,#60'M;1:4*SB7UA8S@7WB8eP?i]E��x�� + +endstream endobj 1516 0 obj<>stream +��ʳ�§�ʹ�é�ʮ�Թ�ɴ�� + +endstream endobj 1517 0 obj<>stream +H��)��!�QITE��_ݑښԖܜД�_�]��i�b�`�oo�e�m��f��h��bh�ea�omkdae`ec��lA˳Ǘv)4; + +endstream endobj 1518 0 obj<>stream +�� + +endstream endobj 1519 0 obj<>stream +��ů��y��w��n��g��k��p�� + +endstream endobj 1520 0 obj<>stream +H��)��!�J��ʢ;sÚR��C�b�l"|��}� 5m�ll��,��Lm��L��\�<<�ü�"�Rc�rR��J��˧�w��6�6&�7�tF�EN��[�_`fm4� + +endstream endobj 1521 0 obj<>stream +H��)��bZS��ݜܝ<]-]-�ܬ�=l��+b<�}�|�R�S�S;�( D6%�u >u��m��E}�ӫ�W�毙Y�ff��ͳg�uu׵tu�O��3󪃧w��r�ڥk�O]�t��C��x��Ϸ>}��g�,�W�Ӛ�U��U�Q�V�PSSU�V�P�^�^W=��_��~?BQadufPB��Y��e��e��ML�oVZdq^jUiiK��I��9��ٝ�v�۾�s^oQOaL~�_EfDmjHOO㺽��\:q��O�<}v��x��O~�~��ϋ?�~��ջ[��n۶�V_��ꄉ�Y�M��5 =�S'-�Yw��//��3~� + +endstream endobj 1522 0 obj<>stream +H��)��!�zK�;rB��B��#��#��"sb#2½�"|�B�==m�l��tm��mtTltU��V�<ʹ�̴� �}m},t��M�],�K�=�R�f�d]��/�z��̣��:x��sǷ^>��}gNl=�w��n��,� K�� ssw��644�24�22�26�12��7а��L��Z��7� �iQDyZTavJnnv|Rt\�lLPL�_��*�4�Q1�Q�U�֔�R�0�V�Q��T��S��U��S��~��CNjP[Kٺ53�n��fլek�/X�lݺy[�n<|�ཻG.��t`ݔ�S*�K㊲"�3Ê��S"�½m�ܬ]�ͭ��, �,�-Lu��]��|=,⢜;��=��b"�--0#28,:.&"27.�2Ƴ&½!ѯ1ɷ5;�+?��0��0�� mMjI �O��O��,�t+sN�s�s�J�+��n+��k(��hm)��nh��=��ٛ��߱{ω��.��ra��7��rhǢ��'O�)l�Jl,�n-��.�.N�K�vu� pv�pv2�s� ��pHO�j�J<�iƇ�� `�#. + +endstream endobj 1523 0 obj<>stream +H��)��!��D��Fw��w�tfwFvG�D4F5g��d��d�bE��=��@FnX��֝�_5�"nBy�Ċ��?_��矿��ۗ��>�zt��۶oZ�rY��Y�z�:�s� +c��#��=}}휜��5t5�Ue�U�u4��,��ʪv� + +endstream endobj 1524 0 obj<>stream +��Ѻ��f~v_zt]rn[mfQngKd_H_\IZYHKL@==2/,#'')#00)]XDŻ�� + +endstream endobj 1525 0 obj<>stream +H��)��!��D��ɨ��h*��+��<��4� �^8�6vJU��j��>� + +endstream endobj 1526 0 obj<>stream +��Ժ��tulR_SAJD:KL<]Q?bWFjeMleM_]KTTCII;75+.*"10(89-��q�� + +endstream endobj 1527 0 obj<>stream +��ƭ�~j_WGJH:<>141)1.'B?2TQAa\H^]JPP@JJ:DE5LM?ys_�� + +endstream endobj 1528 0 obj<>stream +��ªzxeUVHNN=?@4,*"("*)!99/IJ:RSBRRBPN=��|�� + +endstream endobj 1529 0 obj<>stream +��Ϻ��vqfZWHDD8:;2893;;1A@9USFtsb�� + +endstream endobj 1530 0 obj<>stream +��ȸ��Ǭ�� + +endstream endobj 1531 0 obj<>stream +�� + +endstream endobj 1532 0 obj<>stream +�� + +endstream endobj 1533 0 obj<>stream +H��)��! ��PoI4�Q)�E6H �QI4D#V3'W�/�0��>� + +endstream endobj 1534 0 obj<>stream +��ۼ�� + +endstream endobj 1535 0 obj<>stream +�� + +endstream endobj 1536 0 obj<>stream +H�b` 02��xT5�U6 + +endstream endobj 1537 0 obj<>stream +�� + +endstream endobj 1538 0 obj<>stream +��ӽº��Ŀ�� + +endstream endobj 1539 0 obj<>stream +��ȳ��zmlcBB>880::-883JK@gdP��w�غ�� + +endstream endobj 1540 0 obj<>stream +��պ��|v\\WHDB:0.')$)$/.'>?2JJ9QRCVUCoiV;�� + +endstream endobj 1541 0 obj<>stream +��ϼ��nZSCPA693,.0'=;.OK9YVBXWFTTDNN>stream +��̲��og\JWI?QOAWL@fXIoiUriQpfKdaMVWGEE887.65-+,&mfP�� + +endstream endobj 1543 0 obj<>stream +��͹��n�{d{r^lgQmcJlcLc`Nd`L`]ILK<=;0.)!( (!,+#@@3��w�� + +endstream endobj 1544 0 obj<>stream +H��)��!��D�BȒ��R(ꇢ0�r��% ԋM�NX�_`�S>� + +endstream endobj 1545 0 obj<>stream +��Ǿ��«��pzrZkdT`^PVWGTUCNO@AA8,'%(( (.+#46.qcN�ۿ�� + +endstream endobj 1546 0 obj<>stream + + !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNIOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|<}~��=�� + +endstream endobj 1547 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�t[�mVqZIYR@LM<971-)#)%(#*$+'*%.,$//(12*66.==2DD>HHILRTfhgmt|z~��{g�hRaM`Lw^Ktne�� + +endstream endobj 1548 0 obj<>stream +��а��p��h��i��i��p��q��j��k�~k��m��|�� + +endstream endobj 1549 0 obj<>stream +H��)��!�zK�{ +";�ê��>stream +��ع´��u��{��u�� + +endstream endobj 1551 0 obj<>stream +H��)��!��D��F�E�F�GQoitcZp[VXO �T��W��W�ߞV�Ж�YјО�Uљԙԙ��E%��'W�/�0�<� + +endstream endobj 1552 0 obj<>stream +H��)��J��s��b��-,� �4 ��L4�,t�ʹ�M��"��b=�}�}�}�R�\r#|s#��C�Sj�ӺJ��Z�f��=�O�Z�o��;��׶w6tL�0q�k�/?x��+�n�9v��ӛn��_o��Ͽ'��]?sx��i=uI} )��L�͞ԓ?�?Ro��7m��驟??��#�Du�5��'�$��89Y�j�j+�+��)˨�+��ȩ�ɫ�ʪk)�h��:8�;�:�EE%��5u�L_8qͺU;V��2��)+'W�hȨ+I�M�.�Ϫmh��rف#�O_8z��gN<�w�ٓ�_~=z��ߟ��?��ر~��Ių&�M��oK��]8�;��tʄ�Y�+��?yt��W~��?�� + +endstream endobj 1553 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�t[�qY�qY�v]�v]�gSuYHp`KeRAZNGOKAKF:ME;FF?E?7LIEPQLOQQMORX]U]b`dghhkkooprrsqf_tfUpaMwYFyZGy]L|bS�x`��|�� + +endstream endobj 1554 0 obj<>stream +��Ե�� + +endstream endobj 1555 0 obj<>stream +H��)��!��D��ɾ��R ��/�Ҙ e1�١ME��9��U��eYq�9q�iA�q�5�� I>M�~M��ͩ�]�P3�N��_�_`�E<� + +endstream endobj 1556 0 obj<>stream +�� + +endstream endobj 1557 0 obj<>stream +H��)��-��+��]�a]9!��JA"�y�}ő}e1Y�J4��+��U�f>p + +endstream endobj 1558 0 obj<>stream +�� + +endstream endobj 1559 0 obj<>stream +H�b`� A��ČG1UM�j�&�1 + +endstream endobj 1560 0 obj<>stream +�� + +endstream endobj 1561 0 obj<>stream +��±�˻��ѿ�� + +endstream endobj 1562 0 obj<>stream +��ν��wddXWXGQQBFFAMNC`_V��л�� + +endstream endobj 1563 0 obj<>stream +��麼�x{jXUHNNBFF775(1/&12+@@2NO>\ZInob�� + +endstream endobj 1564 0 obj<>stream +H��)��!��D��BQ_IT_ ��/��De1 T3�,fbȳ�I�@4��/��+��\�� + ��?: + +endstream endobj 1565 0 obj<>stream +��û��lb\JEE<25/13*0+%32*J@7SL>NL>stream +H��)��!�QIT_ITiTWnh{FPGVpgVpGaTKQDKaX}^xmnxmj`ufpUvhyF`IJ`MJ`a�_A�our@q�_u�Q�OM�M�_E�oE�w[F��UZ�:� + +endstream endobj 1567 0 obj<>stream +��³��zg^LTH>J>4E91QJ:]Q?bXF]\KOO?RG9JD7CD6BB4@@;zra�Ͷ�� + +endstream endobj 1568 0 obj<>stream +H��)��!�J�{�#["�C�3C�r�K2��b#�b#��|��̭l��LuM��Uu��4��4TdՔe��Td�5��5lMtݬ��-�}RB��uW�- + +endstream endobj 1569 0 obj<>stream +H��)��!��D��Fv�O*��XӓޚҔԐZ�S�Y�P�Y�\�^�R�R�\�ԘМܜ�ۜ�ה�ߖ4��4b��}�]_;� + +endstream endobj 1570 0 obj<>stream +��˱�Ȯ�ʩ�Ͱ�� + +endstream endobj 1571 0 obj<>stream +H��)��!��D��F��e1@4�4��$��( +��/��P=��jJ��@( +��J��͜\�� + ��>� + +endstream endobj 1572 0 obj<>stream +�� + +endstream endobj 1573 0 obj<>stream +H�b`�1��q)�FUS� �#�/ + +endstream endobj 1574 0 obj<>stream +�� + +endstream endobj 1575 0 obj<>stream +�� + +endstream endobj 1576 0 obj<>stream +��Ģ��ummakhXroZrwh��Ѻ�� + +endstream endobj 1577 0 obj<>stream +��ٯ��ig]WKDE=556*54)33*96,FD>stream +��ٽ��iiVPJ;MC567-'&.(66+KA7TN@RR@PPEkgT�� + +endstream endobj 1579 0 obj<>stream +��ǯ��}][JUL;HA68;1A82D=2JK:XXFZZJQRAJJ9DE=EE=OOC��t�� + +endstream endobj 1580 0 obj<>stream +H��)��!��D��F��F�� &��#�p�"�4�&W�/�0�>� + +endstream endobj 1581 0 obj<>stream +��ư��l\TDNH9TF;]Q@lbHjaId_IgbL\ZGKK;>>332)-(21(DB3~q�� + +endstream endobj 1582 0 obj<>stream +H��)��!��D��PtitWnX{FpGVH[FPSnHKVp[z�]��ל"�@de�oe�oU�_kz P�� X;�<��,��4��$zJU��j�o�<, + +endstream endobj 1583 0 obj<>stream +��ɰ��j�}b|lW{kWxp[kdN`]KYXI[ZJ\[KKK<89/11*,(!.,$//(78.SQ?��m�ǭ�� + +endstream endobj 1584 0 obj<>stream +H��)��!�J��J��J"�3��KӢRb��#��m�m��,t�M��u��-uU,�5�u�M��MuՌ��Te�4m�TLu\�}\�S�2��z�"z��}��}��ww߾��g��ܾr�ܩ�;�.ٱi��5SW�iY:�~��֚��H_g?g'gKS3=}} } }#-cC-S//��$ߥ�˟�?��"<Օ֚Y]��jom�g��b��d��h�)ob�ee�cj�kh�md��jd��)��)�g�a`�D&f��֦Nv��n9!��2��;��ϼ|z��n�{q��z�Ȧ��>}t��/��+H��sq ��r2wt�p�1��2��U|�[SeҾ-s߼��o��D + +endstream endobj 1585 0 obj<>stream +H��)��J��B�b��cSSS��3��K��}+b}�|+�3�*Ӄ�dY�_m�oI�oV�gv$Ey��y��$:�E:�F8��yx��9��z��$�V�͟?a��[��Y�f� kW�޷��n�zq��̓��ݸ��9�� (+�0ͯ.䒘(רp�Po�7k_/+wOs?7�w�� &<{x��} +�lITSqlTT��{��c��mi�[u�oM�_kvHkvpQd_Q�Ē�ނ�p91'tbvHwfPgfPGf@K�WC�[M�{i�{^�GiRHMYꄮ��[W.��W3��|��=��ٵ{Ņ�/\�p��[�_ytϢݛf��2s��ɭy]��M � Meq�q��IM�qey�9��I^�ɞ�Y��%q�śu>�y�篏��0�� + +endstream endobj 1586 0 obj<>stream +��˵��n[TDLG8QD9ZO>i_Gi`Hc_IgaK][GLL>stream +H��)��!��D��P_iL�,2b��@89�� @Ԃf��}U�6>� + +endstream endobj 1588 0 obj<>stream +��˳��_]LWN=IC67;1?60@:/GH8VVEYYIRRBKK9EE>GH?TTG��~�� + +endstream endobj 1589 0 obj<>stream +��ĳ��llZQJ>stream +��۵��omcXNHF?778,76+55+97,FE=njV�� + +endstream endobj 1591 0 obj<>stream +��ɮ��vwkurb}zd}�s��ս�� + +endstream endobj 1592 0 obj<>stream +�� + +endstream endobj 1593 0 obj<>stream +H�b`��ŘYp(� +FU�[5@�A�K + +endstream endobj 1594 0 obj<>stream +�� + +endstream endobj 1595 0 obj<>stream +H��)��!��D��F��}�1�e1�c&�� +#'��GM(��XE@6��D��u�L@2sre��*��R>� + +endstream endobj 1596 0 obj<>stream +��׾�Ī�� Ũ�һ�׼�� + +endstream endobj 1597 0 obj<>stream +H��)��!��D��tDvf�N��C��j2�K"=��=JR��܋��J��K��+2�*3��s��}��Z3��ɕ�K��:�; + +endstream endobj 1598 0 obj<>stream +��u��t��m��e��i��n��~�� + +endstream endobj 1599 0 obj<>stream +H��)��!�zK��ʢ�sB��]�]�=��l��,t��-, ��lm��l��M�m�]�>stream +H��)��bZR��ݜ�]��,C�,�=lS��C\��]k|�S�S�[Ӄ�2��2��a�%ſ+=�!ѻ!ٷ-=�=#�#7��(bRY��S�VϨ�<�u��ξ ��ݝ�f��\�x��.�t��}i�{�_>9��o<q�䱵�fVu7�u�&u��u�ƶT�ԖFUE��&�f�Ve��F�w��2�˧��G*��)��H� w��154��V��U35�07Ҳ1��7�5 r��v��v��q��I�(�I�,.l��8o��C[��zh��˺O.�X�PP�Y��ݴt��3g_�z��]|��߼��ß?��ۏG��6.��S0�5��6��&qbgVwSzKur{uBOS��¹S��]�� K�� + +endstream endobj 1601 0 obj<>stream +H��)��!�zK�;sC{�Ck�«2"*�#��³ý��#<�|흜�\, �� 4��H��@��H3�Z/�Z��D;��4�� ,��<��&-ܵ*ާ%-`vk��;�9��'N޻��ֵ}gO�8ud�#.��y�ؖ�{��2c�ܶ��܈�8��(�Pww'+k#C#mCC-CCMc##-=}uss]ww��X��˺>zMЛ@�WY�Y��lkfb��b�ce��l�ce��DzZJ:��j��jrb� +�� +ںj:@��bg��e��^�y��[foZ7w��K6,_�nޖ��>��k��0u��ꆲ��Ȣ̰�̰�p�po�@K;3KKC3s}s]S3}+C7��{7w��p��F��fF�D�&�E��$�V�xVE��%��'�4e��t��Dt凵其�4$�'�U��TF{�E�W�y�E��F��9$z�E9YĻ[F��F�U�7775�w�wM�0gVߒ�37l��m��/^>x��C;lZ5��y��w�6V$4E��WF��$G��;F��;8��X��[��٧�y4��^3��k��0��!� + +endstream endobj 1602 0 obj<>stream + + !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKHGLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~��;�� + +endstream endobj 1603 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�t[�lUnXHVP?JK;75/.(!+%+$-%.) -&1-$20)32*76-==2ED=IGHMQRefelszx|��µ�«��q�mW�dO~_Ku\Kpmf��˫��~��o��h��j��h��n��p��h�i�|i��j��x�� + +endstream endobj 1604 0 obj<>stream +H��)��!�zK��#ZsC+Rr"|S��c=��=m]��L�L��L��\mݬ}=݃��#�B�s��rJR+�ӛ+��[3�}Z҃��[3�{K��J�&W%��o0��5� + +endstream endobj 1605 0 obj<>stream +��ָ��}��q��x��}��s��Ӹ�� + +endstream endobj 1606 0 obj<>stream +H��)��!��D��E�D�E�Fv�Dצ7g�vGզT��T��d�V'�7然��'��w�wdvdtd��*��#��+��Ux0<� + +endstream endobj 1607 0 obj<>stream +��׷��Կ�� + +endstream endobj 1608 0 obj<>stream +H��)��!�D��DCPE��F�P E!s{��+��U�>� + +endstream endobj 1609 0 obj<>stream +�� + +endstream endobj 1610 0 obj<>stream +�� + +endstream endobj 1611 0 obj<>stream +��İ��γ�� + +endstream endobj 1612 0 obj<>stream +��|xl]ZKFF:<>5;<6==4CB;YWIzygþ�� + +endstream endobj 1613 0 obj<>stream +��Ʈ�~lVWJOO>AB6.,$(#*(!88-GH9RRBRSCTQ@�� + +endstream endobj 1614 0 obj<>stream +��˲��n_YHKJ;>?231)/,%>;.OM>_ZF^\IPQ@KK;DE6PQB�{g�� + +endstream endobj 1615 0 obj<>stream +��wumS^S@HB9HI:YM=_TEidLmeMa^LUVDJK<97-/,$22)::.��v�� + +endstream endobj 1616 0 obj<>stream +H��)��!�@�ޒh0��-��+��R ��zK�{A +�{�QI4��)U ��>� + +endstream endobj 1617 0 obj<>stream +��һ��e}t]wqZokYlePogKe`I`]I\[IMNA??41.%( ')$00)_ZFȿ�� + +endstream endobj 1618 0 obj<>stream +H��)��!��DQwAxoaDOaDw^hWNhWQdWIT{Qd[qt{^x{NX{n8V�[ D��=}@vAM(��\7�<{t��;6/Z��o��ʮ��Ā�(��O??{ss}} uy-e) ��P45� +�_��0E�w� + +endstream endobj 1619 0 obj<>stream +��oc]JEF=47023*.*#1/(H>6RK=NM=ML>stream +H��)��!�G��DQiTw^hGVpgvHWvpgaT{Qd{QxSADc~DcFp}Nh}^xMvHUFpCzpyj`YJ�Q�X�T��ߐؐX�P��4gzM��*�|R;� + +endstream endobj 1621 0 obj<>stream +��ȸ��}i`NUI?I=4A7/MG8YN=_VE][JPO?RG9KE8EF7DD5BB<�yg�ѻ�� + +endstream endobj 1622 0 obj<>stream +H��)��!�J�zK�Z#3B�B+s��2� +#sc��"b��]|��\ݬ��-M4�L4�T�t�͵�u�4��͵�u5�uT ��M�ܭ��̒ܭS��*�ftW8P-� + +endstream endobj 1623 0 obj<>stream +��ľ��}xr`gZI_UAe[Is^PteNgcLeaMZYGMM=AA355-32+22+77/VSE�� + +endstream endobj 1624 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�tZ�nWzcR�kZzncigc_aa`ec\aXOTO>stream +��Ǿ��q��h��h��g�}csmWZWJPQERSBDE:77.00(,( /.%13+I<3TQCvs\Ż�� + +endstream endobj 1626 0 obj<>stream +H��O��?�V�@�E�8��N�N�x�;��9%D^�ͭ�li�Q�Z+�Z�J��i�Բ�e�oM3L��7�F��y��p��0,YM�ތ��*�F�6C�]'��*�a�&��pЇ��!5"��a�� b,�Z�}Q��7��\�hI��Z��Jm��Г��)�y�"tFcqyE��T~~.�H$G%'��1Y^n��Jy��cqv$�N^�� b��j�%j`�F\��E��0�S� �9,��AC��&x��@d7�F�2�aq#1��`� c�i-��ik6 +~��hrn��+3ߖ�/�[|3�z��ө��[s�G'�^Iv7��4��4YJS�$Q� +iIj�4�@"��J�9�r�2r��6XP��T�N-Tk�j�!�m��X�'""�/�v/�V�xV�'�^2$RQ +D,��<q�a +X��6�T61dk��{-�;p�;��?��`x��ǯ�/'F��=��y��8�q�M��\=�3��;b:��,��J)\$�ET-"��uv��?�L&�O��W� + +endstream endobj 1627 0 obj<>stream +H��)��z�#;K�ӣ}�|�m��,��B��ݬ3��#܊�<��S��|��R��hL��K�q��uo��7��(^3�i鄲��==��M�Mm-�S��/_�f��Sw��}p�ʅ}wo�v~��sۯ\�q��8�#��9��*��,��.��"�97�1;�.#�6-��»[Ҧ��۵�ߟ��M�+��.�-� ps��01TWW�WT�VR��R��W��S�6Ր��V"kCu[S r�3�� +�Ȍ�.�ϭ�o��h��;O��q��2k�®�ڢ��K��ޱ��Ϟ\~��Gg�?<��ÍǏO��x�ϟ�_��GG�l�>gb��T��Z�R[j��d{CZMybOk�i��~��`�3�L + +endstream endobj 1628 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�uZ�iT�kU�pYi_RRG>G9082'57,-,#()#),&*-&'')-%,/':<8ITVRfffrun��}��ê�Ա��o~t_mZFhTHpg`zsl��е��v��q��l��i��k��m��m��r�� + +endstream endobj 1629 0 obj<>stream +H��)��!��D��Fɉ�1�� 9aE ��y ��q>�A.I~��A��iqi�1��Yq�)a�Q^�1� 9��Y��Ӫ�O�0�x9* + +endstream endobj 1630 0 obj<>stream +��Ѹ�ʲǽ��ż��Ĩ�Լ�� + +endstream endobj 1631 0 obj<>stream +H�� )y�)u�PH�Ee��"�R��]��h�-�~衁�^�L� ��q��-D�rT�4/]��>� + +endstream endobj 1632 0 obj<>stream +�� + +endstream endobj 1633 0 obj<>stream +H�b`�A�� (�ƎU10��b�� + +endstream endobj 1634 0 obj<>stream +H�b` 021�H0h6)�$�G�(� �@ + +endstream endobj 1635 0 obj<>stream +H�b`�`��!��I��F0br� + +endstream endobj 1636 0 obj<>stream +H��)��!��D��FQitw^XGVHWvhgVHk~X{.�љҐؖ'k��R�S۳��@-=y��&��N(��S5�*aI5@��h<� + +endstream endobj 1637 0 obj<>stream +��ε��k�{ayhTxgUvnYldMc_L[ZJ\[J]\KMM=::011*,( -+#//(67.QP?��p�ʱ�� + +endstream endobj 1638 0 obj<>stream +H��)��!�J��J��J"[�B��*3�3ₒ�=b}Mݜ�\l]-� Ԭ 5��m��,��, 4�4��4\�t��,��L�=�3��B\�S��#��"��?>�G��?x��7��>stream +H��)��zK��B3b�b32��ӓ�BҢ�s� +R��b��#=�c��|kR�K��da�Oe�w~�wj�{Z��E�%9��E�9F�9��{��:��{��$W�e.^��w�>��@7/OO_�p_�,_��H��D�� Q��Q�Jc��"��"@da��ܰ��!�Y��YA�ف-�� q�5�%��١n�1%�q�93��-��3��jJW��u��Ϟ�}ے��V��~��Ww]<��Ȯ;�Mݺ~ʢ�-�M��I�e��񍥱�e��U��EY�>��)��YiA�E1��޹��0�-�' + +endstream endobj 1640 0 obj<>stream +H��)��J��sÓc�=<�,,�� 5M4,Mt��=��|,�"�c�Mc�M��-�}��S��2\�#|�#}�˓��+sLm^�t��\z`��uK�fNi��j��?{�U�8r�ʥ�g��NX1oJY_[f{CZOmRC��i}9S{�'u�M�͝$��M�^�zʳG�|��ߏ Tՙ֘��hkoj��*/��$��&��"��*��"��g�ifa��d��S�_TR��2u��kW�Z{`��f��<�=��2� -6� ��i��Żn9~��['��;��α'�N}�q�ݷ��~��?��ׇW/�>�}`� 3'Θ��ߞ�ז6��`bwNOWɔ �3'W�_�y�Ȫ�/��3�� + +endstream endobj 1641 0 obj<>stream +H��)��!�zK�{ +#� +"z�b��bkRBݲb=��lC��=��\�=�,��lL<ܜ||=}�|��c�b��S3�2s�Jʊ˫J�Sʳ"��#�*"��c|*��*��Լ��ϏO~~��۟�?��`"�P& + +endstream endobj 1642 0 obj<>stream +��ۿɻ��ͮ�� + +endstream endobj 1643 0 obj<>stream +H��)��!�D�QqT_ITI4�QL_ILoI4�e�$4e�5�FT��Ņeĕf�է�ǅVe��&�4$��'�5&t�wEvE�Qq�ʸ�}��>stream +�� + +endstream endobj 1645 0 obj<>stream +H��)��@Q_it?��ΜЎ�� Q}%Q]9�}E�}�1�^ B5dre��*��>i + +endstream endobj 1646 0 obj<>stream +�� + +endstream endobj 1647 0 obj<>stream +H�b`� �@��̂G1UM�j�5AD + +endstream endobj 1648 0 obj<>stream +�� + +endstream endobj 1649 0 obj<>stream +��˻��Ƴ�� + +endstream endobj 1650 0 obj<>stream +��þ��~ii^[\LUVFJKFSSIgf\�� + +endstream endobj 1651 0 obj<>stream +��ĵ~�q[XJOPCGG8:7)31'12,??1MN>_[Kuxj�� + +endstream endobj 1652 0 obj<>stream +H��)��!��D��PIT_I4�QitȲR ��P3��@a@H�8�0�2~q_@��a? + +endstream endobj 1653 0 obj<>stream +H��)��z��K�:sÚS�"�rÚr#�2�˓�ks"��*R��r�Ӄ}��Z҂z��J�z�"Z��}��=k�j��|��Z3�{K�'W�/�04�:� + +endstream endobj 1654 0 obj<>stream +�� + +endstream endobj 1655 0 obj<>stream +H��)��!��D��I �/�� +�)��Pӕd��AP)P XKq� �P��0��Q�+��U�c>� + +endstream endobj 1656 0 obj<>stream +�� + +endstream endobj 1657 0 obj<>stream +H�b`@�L@̌!,��Dq�QՔ�0G +Y + +endstream endobj 1658 0 obj<>stream +�� + +endstream endobj 1659 0 obj<>stream +�� + +endstream endobj 1660 0 obj<>stream +��rn_ohWtqg��w��v��Ǳ�� + +endstream endobj 1661 0 obj<>stream +��Ͳ��lhVPN@?@3:;.67-66,::.GG8RRBolU�� + +endstream endobj 1662 0 obj<>stream +��ç��xheRNO=<<100**( +)"00)==2JJXZM��r�� + +endstream endobj 1663 0 obj<>stream +��y`\R?LI:@B472*E>4QO<[XD[ZGOOFLL@JJ:FF@LL@�� + +endstream endobj 1664 0 obj<>stream +��Ұ��ktmTeYCeZJfcRpdRreOrjRleP^]KWWGII<;;265-99/^UD�ǭ�� + +endstream endobj 1665 0 obj<>stream +��ʺ��w��w��i|wc{uewnZpiPniYjeU^[IHI><;132*.-%+'.,%?A5xr]º�� + +endstream endobj 1666 0 obj<>stream +H��)��?*��-��( �R�� w>q + +endstream endobj 1667 0 obj<>stream +��ĮƼ��z��hzr[qhUkhbbaWa\ERQA;<300),) .-%.+#,'/-%770JJ=��z�� + +endstream endobj 1668 0 obj<>stream +H��)��!�zK�{J��4'*�-��8�;?�+/�37�-/�-?��(jb}֦��|�І��6=��#��u�ཇ'�?8r��k&͟��^P[��Z��m��fngcdh��o�ih�i`��fck��myﳇ'�� `�yqA + +endstream endobj 1669 0 obj<>stream +H��)��zK��{ +�{"�rB�rB�sú�� +v��CPo~�`q��Ў��d��X��$��ߢ��ʴ��Ҹڊܚ��ꊼ [�-�1k墅�m�x��s��/�޺s��e��&��ק6U&�֤tԧf��8��;{��[:��ZZ��Y�Z��?��0��n� + +endstream endobj 1670 0 obj<>stream +�� + +endstream endobj 1671 0 obj<>stream +H�b`@�P��YXXX��؁��0��ު8e + +endstream endobj 1672 0 obj<>stream +�� + +endstream endobj 1673 0 obj<>stream +��ϴ��}~qyyfsthro_��j�� + +endstream endobj 1674 0 obj<>stream +��rYXJ:9111)0/$0/$11(=>4RP?wp_�«�� + +endstream endobj 1675 0 obj<>stream +��ʢ��y_YWF76+& &% +&20'AB4MM>stream +��ͺ��|bfYGFA721*3/$BA4OL>stream +��ǿ��arjSbZGYUH^ZHhaHlgRpiSdaLVXHMN>=<275.88/FE7��g�� + +endstream endobj 1678 0 obj<>stream +H��)��h4�*aq_5@��>g + +endstream endobj 1679 0 obj<>stream +��p��j~u_xp]ojYidPf`I`^K[[KSSDCA363)+%'-*#:<3b^M�δ�� + +endstream endobj 1680 0 obj<>stream +H�b` +021��sprq�� + ��KHJI��+(*)�6GUM]CSK[GWO��~Fƀ��@fff + + +Ve ��OHLJNIMK��/(,*.)-+��G2��I��L�6}��Y��̝7��E��,]�|��U�׬]�~��M��lݶ}��]��ݷ ��!J� + +endstream endobj 1681 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�v_xfTjUGiTJcL@_RDSRCPPIQPGKG9MK=GGAHF>stream +��°��k�{a�~c�~e�x^sVtnSiZHZPBZWE_\JgdOwq[�� + +endstream endobj 1683 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�tYztfS\ZEHF?CC250*-&'%)( (&&"&#&"#&"+-'-/)-.'35-K[\gs}q|�u��~��nl[KTF8D;0CB4NK>stream +��ʭ��h��e��g��n��q��s��o��i��i�� + +endstream endobj 1685 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�s[{m\f_QSLDPKFNG?FA8HH>CC9AA7>=3?>3A?6A<3JC>MFBNGCSNHXUP[XTWQOWQHOIDKEAG=5LB9B:/:6.<80D?5HA5IC9WS?~pT�� + +endstream endobj 1686 0 obj<>stream +��պ�ϰŷ��t��y��|��}��Ƕ��θ�� + +endstream endobj 1687 0 obj<>stream +H��)��-���Ҙ ��@�W6�9��=��zx�̃{��y��sw�:rݶM��.�[<��!��"�� >+�<3$!�7��.��B��L��<��17�{� o�=��@� wAG + +endstream endobj 1688 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�u[�eQo]M`_X[YWTYTVXURVSOZZRVMJIADBACB8AB[]T��}��ëü�׻��í��j�}a��e�{`�uY�t\qmXg[JaLBXPASQ@[XGb^Lvn[�� + +endstream endobj 1689 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�tZ�pXw^KWPDBD:67.35.,+#/+&/*$+'0/'23*22*22*12+12+33+??;ENLRZ\dkuk��jqg]RPLHOSiqw��ռ��v��j��m��q��p��q��p��l�}`�|f��l��|�é�� + +endstream endobj 1690 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�u[�rY�ePsWEYIBIF?@9/50&22(..%.0*+*%,#21(68-87,:<2>@9MQM]ccprysw��lt`I`UEMIERKJ\KAa[L~q^��z��Į��y��s��x��{��|��v��v��ϰ�� + +endstream endobj 1691 0 obj<>stream +H��)��!�J��J��K�;�C;��;r�j3�3�2�S�}S��r�<�|�3��R�3j��}��}*c>stream +��׺�γ�δ�ǰ�ɱ�ϴ�� + +endstream endobj 1693 0 obj<>stream +H��)��!��F��D��F��L(��X3��&��N(�@��2�-��-��$H{B�$��8jre��*��? + +endstream endobj 1694 0 obj<>stream +�� + +endstream endobj 1695 0 obj<>stream +�� + +endstream endobj 1696 0 obj<>stream +��²�Ÿ��ɶ�� + +endstream endobj 1697 0 obj<>stream +��͸��triFGC==5?>0;;6OPEokV��ܿ�� + +endstream endobj 1698 0 obj<>stream +��za]YJGE=31*+&($.-&<=1II8QRCXVEwp[�ƫ�� + +endstream endobj 1699 0 obj<>stream +��ҿ��q\UDQB693,+-%96*JF5US@WVESSCNN>stream +��Ͷ��rh]JVH>OL?SI>bUFmhTriQpfKdaMVXHGG9::077/,-'pjR�� + +endstream endobj 1701 0 obj<>stream +��Ѽ��l�waxo[kgQnbJodMeaPfaMc^JOO?@>3/+#)!(!,+#CC5��~�� + +endstream endobj 1702 0 obj<>stream +H��)��!�BPIT_i4��/! ��%訧$ +��ܩU ��i>� + +endstream endobj 1703 0 obj<>stream +��ȿ�ƾ��q|t\leT`_PVWFUUDPPBCC9-( &(( (.+#36.qcN�� + +endstream endobj 1704 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�t[�mUu_Svh[tdPk[Rk\QiYIiZKj[KgWGbRI]MH_OIaRD`Q@]M@WG=UF:ZI=VF:]MEbRCgWFtcXzjc�vj��y��z{�{��ͻ�Խ��ŭ��o��g��g��c�z`pjZa]KUUBWOBSN?KM=\P>bYIyrW�� + +endstream endobj 1705 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�v\�cM]NIFLJDF<:=658034*34*56,34*11+14/25002)01&/1)-0(261AGFO^c^ktm|�s��nskejojbinqy{��˿��s��l��m��k��g�{byq[zt^uoZvoY��c��w�β�� + +endstream endobj 1706 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�t[�kUwbO]ZLGH?9;436321.+("&"%!&!*)!,-&'$%!,-&8;6JB@RYUmttsmmzw}|��u~kXt`N^RCLB?THDWJEXKEeYH�� + +endstream endobj 1707 0 obj<>stream +��д��{��l��l��l��o��p��r��z�� + +endstream endobj 1708 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�t[�v\�rX�oVs_Qs`So[OcWK^ZL]XJZUC[VFYSB\VG^YLdRDkTJz`N}iRlQ�iP�gO�fM�aJ�cL�eO�dN�eO�fP}bM�gQ�hQ�hQ�mV��t�� + +endstream endobj 1709 0 obj<>stream +��ïƾ�� ¹�� + +endstream endobj 1710 0 obj<>stream +H��A� ��-��@[.�� +F�&��dw��-�FK/��H&ITVe9'�-8cg�s}�q�d"肯3�5_^o�? + +endstream endobj 1711 0 obj<>stream +�� + +endstream endobj 1712 0 obj<>stream +H�b`@�L�,��ʀ��E��Q��V `�|� + +endstream endobj 1713 0 obj<>stream +�� + +endstream endobj 1714 0 obj<>stream +��̼��ų�� + +endstream endobj 1715 0 obj<>stream +��Ο��hi_XWHGH@?A=BA8CB:XYLsp`�� + +endstream endobj 1716 0 obj<>stream +��Ű��p[WG>?40.%+)*'*)#34+@@2MM=[WF��s�Ű�� + +endstream endobj 1717 0 obj<>stream +��ٻ��k_[EDF855,*(!+'62'EC8PQBSSCPP?II?MM@iiX�ī�� + +endstream endobj 1718 0 obj<>stream +��ռ��qldPXU=ND8MD8NE9_SDidLkeMcaNSTCHI9BB4==4I>4xlW�� + +endstream endobj 1719 0 obj<>stream +��Ⱦ��s��bxpXphQldLmcIldPmgXlfQ^\GSSEDC530&)#+&31*IB6��w�� + +endstream endobj 1720 0 obj<>stream +��ð��z��h{r`nfQfaJ\\KUVFRSBBB4/,#'') (,+#=5.{r^�� + +endstream endobj 1721 0 obj<>stream +��Ƭ��ofeUEC5--#*)")) 22)780idQ�� + +endstream endobj 1722 0 obj<>stream +��ī��}yfml_zxf��r�� + +endstream endobj 1723 0 obj<>stream +�� + +endstream endobj 1724 0 obj<>stream +�� + +endstream endobj 1725 0 obj<>stream +H�b`�F&fV6vN.n^>~AA!!!�4��GTL\BRJZFVN^AQIY�AUM]PC��OK[KKKGWI1�� +�� + ��OHLJNIMK��f@�9�y��E�%�e��U�5�u� �M�-�m��D�;��}�:��{z��'L�4y��i�g̜��f +�񟫣=o��/Y�l9n��b�U�9֬%F1��3a��@Y + +endstream endobj 1726 0 obj<>stream +��Ͻ�;�̺�� + +endstream endobj 1727 0 obj<>stream +H��)��J��bc�<�<<�}�l +"�겂k2��[�C'�DO��RD�A(vRy�Ĳ��&C��ʸ��ah:� + +endstream endobj 1728 0 obj<>stream +��©��å�� + +endstream endobj 1729 0 obj<>stream +H��)��J��rÊ�=��}ݝM,��4�� 4lL�, #�l�C�C@(%�-5�-;. 7>�#?��8��0��0�+?�� 0��0hfoqT_ �+ct��43 + +endstream endobj 1730 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�u[�qY�mU{aPmUH]SCHD><7240(24,*/)'+'&+%*.(/1+34.ACCJXZorr��{{qg�u]�oV�lW�nW�mW�pZ�z`��~��Ž��s��n��m��v��kgbOQO?EB7?<3HF8qn\�� + +endstream endobj 1731 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�tZ�u[�tZ�sZ�u\�jV}aOz\Im_L`YOXNINC;J>4A@6BECFGJKXYYdliw�v��Ʒ��~�|qjaOZWOggg��ؽ��g��h��n��guiQPM:31'(#&&($--'JJ=�� + +endstream endobj 1732 0 obj<>stream +H��)��6�-��b�IEv�7$��%�U�zW'xWe5'Ll�9s|��N��w��mǎ�?��k�_>{��ܳ�^��t��Û�/�X�Y�Z��g��fik�o��*��(�./�� $�uU]��i��|�W�W3�4zbYt_aDoAxw^Xgax{AtE�wI�ka�KB�gb\@zRHaNZWC��i�3'VN�)ij*kn��0y�ڍ�Ϟ��_�y��v�>��sz�+��R�<��5U�ā>��V��T�А��P06��q�>}h��I�N (.��>��(fbYLwnXGVH{FPMVHQjHF�Oj�G��S\�cTBtJJLNazYSU]W��YSvn_�ˬM�[��kmk��h��nm��>}Ŷ=��\��O��uq��s�.�zp�-�f�6U%t5�N��n��O�rww0�֐W��D�w +bZ�R�:��Sz�=>��{>stream +��ɹ��µ��rvn\keRdbOecOf`IPND?>831(.,$+$0.&>@2��x�� + +endstream endobj 1734 0 obj<>stream +H��)��h*��Z�� (�=$ + +endstream endobj 1735 0 obj<>stream +��ҽ��v�d~v`vmYpiWjeSpePpbPofVwoYpiS^\KPQAFF::91>=4JI;�� + +endstream endobj 1736 0 obj<>stream +��ջ��q{sZjbKYU?>@276):.$A90KKNN>stream +��í��j|s^_[I;<0)%$&("*$-&20&:;1@A4MM;�� + +endstream endobj 1738 0 obj<>stream +��̳��rvr^OPG:;2-+$' &(!(#(&,.)ON;�� + +endstream endobj 1739 0 obj<>stream +��|�zd`]KDF942(.*"-,$34-KG8oiW�� + +endstream endobj 1740 0 obj<>stream +��ػ�ĩ��t��~�� + +endstream endobj 1741 0 obj<>stream +H��)��P4�2nao%@�&�= + +endstream endobj 1742 0 obj<>stream +�� + +endstream endobj 1743 0 obj<>stream +H�b`�F&fV6vN.n^>~A! ,��WPTRVQUS�� MM!-md��:�z��F�&�f��V�6�v��N�.�n��6�2xy��GDFE��'$&%��1��tQь̬�ܼ��¢�Ҳ�ʪ�ں��Ft�` �M�-�m��]�=�}�&N��a��9< + +endstream endobj 1744 0 obj<>stream +�� + +endstream endobj 1745 0 obj<>stream +H��)��J��B�Rbb�R�bs�*��[�b�ˣ�V�M��X;�4��$��𙉊&W�-�0��
>stream +��͸��ж�Ӹ�� + +endstream endobj 1747 0 obj<>stream +H��)��J��ª2#��<�l�\]�l}lbC��K��rC��"{��$��8��J��J��F�+��V̽8� + +endstream endobj 1748 0 obj<>stream +��ø��}��m��x��̸�� + +endstream endobj 1749 0 obj<>stream +H��)��zK�{K�[2� +�Ó��Ll5U�45Ut4M�t�,�m� =�-C�m"\��ܫ2��"��z�{ +�z�"zK��b� �4drq��y��3� + +endstream endobj 1750 0 obj<>stream +��ڽ��{��p��p��l��j��c|v[~{e��{�� + +endstream endobj 1751 0 obj<>stream +H��)��&�F��&�G��z:9[��j+�j*�k��XZ��ؚ;Y�8�$e��e�vG�G��w�u�QAXWAxWaDwQdo1M��]�[`1� + +endstream endobj 1752 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�q[�r]qg_]\[e_SVRHHDBFC@<:366-/1+.0+.1+./)/1)8:7BDFHTWZgnx��ν��kwnesg^d`Qkga��v��ӽ��o��m��l��i�u_siT\YHB?443*22)97/FE9�{h�� + +endstream endobj 1753 0 obj<>stream +H��)��!��P_i4��Fw��E��wf��F��6&��e�xf{�y��:�D{'&��g�gf�֖�uU-�ӱxF�i�s��N��?o��sfnڱ��[��=��GO��xj�޵S��(�hHn�L*��N��SU��Q��V�T��T41�M��:rh��o��ZIdoItOIL_YloYlkFHmVDMALiZHj��W��{x��wdlLF~i[kǢ�;�9t�С�;�X�g��ڻ[�J��*ʊ{gN��b��V��s��sW�޿}��C�O�|f��k��pd��UsfN(�Й>�5��<6;�+��LOIYV^ZTSQBCA�eueͬ8��f~��֔ + +endstream endobj 1754 0 obj<>stream +H��)��6�-��D}0�[�TV�S�]�[>�5�֭��^yr�ʳ�ݾt�⑋]�z��#WO�>{r��S�Xvh��:fO)k�M�J �� p��q��51�U�P�P�S��R��5�H�?wd��(�QO �_�@/DQIdoA�PADGAXsnHcvPCf`Y�oE�wmR`]jp{VXK~|O{��ͳvm_�g��M;Vo9��;�ڇ�7߼��'Wv��:c��ܢ��''SSuMYU)%Y1UYQME 2�Wwv�h��sc��?{+� + +endstream endobj 1755 0 obj<>stream +��ǰ��s��q��m�|f|r^tn_ni\ojWtm^ifWZYIOPBA@564,/-%22*CC7�~i�� + +endstream endobj 1756 0 obj<>stream +H��)��P4�*~Q_@�'= + +endstream endobj 1757 0 obj<>stream +��縲��k�{bvoV^Q>ME:HG9JF8WRA^YDb]F_]J]\LRRAFG:=>4>?2c]D�ȭ�� + +endstream endobj 1758 0 obj<>stream +��ͷ��vv\\YFEG842)%& +'1-";9->?3BB6EF7QN9}{h�� + +endstream endobj 1759 0 obj<>stream +��丯��tup\OQB=>1*' '&&'!+)"1-(C@1��d�� + +endstream endobj 1760 0 obj<>stream +��ΰ��tp_RP@77/-+#*$-+$00)99/YPE��x�� + +endstream endobj 1761 0 obj<>stream +��ȴ��vfbPWXJhdQspb��q�� + +endstream endobj 1762 0 obj<>stream +��л�ǵ�� + +endstream endobj 1763 0 obj<>stream +�� + +endstream endobj 1764 0 obj<>stream +H�b`�F&fV6vN.n^>~A!aaQ1q I)iY9yEd��o`hdlbjfnaiemckg��? 0(8$4,<"2*:&6�`rv�OHLLJNIMK��/(Ī +��KJ�J�+*+��k�)��ɩ��0��H;��(� + +endstream endobj 1765 0 obj<>stream +�� + +endstream endobj 1766 0 obj<>stream +H��)��"S��"J󒳳�S��#��b�ͣ�"��J�=��S�j3�2�겂�r�;�":��sÀ�Ĳ��IqS*�&W@P�rJͬM\6� ��9n + +endstream endobj 1767 0 obj<>stream +��ֽ�� + +endstream endobj 1768 0 obj<>stream +H��)��zK�[S�c"�}C|��=��=l��B�3�}�|+�B�+�W'L��Z7�2nrE�� e�@�_�[�W�ˊɕq{+�,9� + +endstream endobj 1769 0 obj<>stream +��ʽ��z��{��ä�� + +endstream endobj 1770 0 obj<>stream +H��)��J�� +"S�S��ܜ�l,��ttT�t5�Lͭ-� �\-��m=m��bB|��Sc}��#: +#��sC��;rC��p��}%�p+&W�-�0B82� + +endstream endobj 1771 0 obj<>stream +��θ��n��k��l��u��l�{^ih\]ZKTS@[XK�� + +endstream endobj 1772 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�t[�rZ�rY�hTo_NYRIOE?D:091%3(0/*03-(,&&)$(+%.1+574:<9GNNWecdort|{v~��{��v��t�{c�pW~kR�rX�tY�sY�y[��v + +endstream endobj 1773 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�tZ�t[�v\�nW�iS�gR�mT{eOjVKbQFXE9PI9H>4A>3?@;>?;IQQQceeu}|��ú��e�v^wj^{pb�zn��ظ��p��j��g��g��dznUWTD>:-+''#)%)( 46,l_O�� + +endstream endobj 1774 0 obj<>stream +H�b` 0213��spprq�� + ��KHJI��+(*)��7CUUUMMM]CSK[GWO��j��A�!�a��Q�1�q� �I�)�i��Y�9�y��E�%��e��U�5�ƨ��7465��wtvu��O�8i�ӦϘ9k��/X�h�˖�X�j��oظi�۶�D + +endstream endobj 1775 0 obj<>stream +��ƶ��v��d��g�z_rlUpmYslVmhXlfRrjQskQebNTUEFG988/34,9:/^[F�̰�� + +endstream endobj 1776 0 obj<>stream +H��)��P4�*nao%@�' += + +endstream endobj 1777 0 obj<>stream +��ϱ��l��i|s]a^L@B741%8.:5(GE5SP>SRBRSDPP?AA5781TP=�� + +endstream endobj 1778 0 obj<>stream +��˵��p��hpiTBB5)'!$%' (!,'1.&<;/EF6E>0ujU�� + +endstream endobj 1779 0 obj<>stream +��δ��s�~afbO?A9--&)#''*$)&*)"@@3rk[�� + +endstream endobj 1780 0 obj<>stream +H��)��-��}�1}e1��@rbȲ�X 9�>stream +�� + +endstream endobj 1782 0 obj<>stream +H��)��D��Pt_)��Qgvh_It[fPKZ@cnHsF`Yj@y�oi�y�OQ�_j�wz�OvJ@Bb$@�J�8� + +endstream endobj 1783 0 obj<>stream +�� + +endstream endobj 1784 0 obj<>stream +H��)��!��D��D��F�GM(�r�Kcӂk�+c��}"=��]-ݬ�l �u��tM��-��]��mܼ<��}c�"�2* + +endstream endobj 1785 0 obj<>stream +�� + +endstream endobj 1786 0 obj<>stream +H�b`�F&fV6vN.n^>~t g� ��KHJIc(CW+#+'��i �ZU5u M-\*Q�j��]ˇ_-��3.�.�n��^�>�~��A�!�a�@)))��f��ظ��Ĥ�Դ�̬�ܜ��"'��tW��WTVU��7465��wtvu��)�,� + +endstream endobj 1787 0 obj<>stream +�� + +endstream endobj 1788 0 obj<>stream +H��)��zK�'��t�d��8ؙh��*�K��*�ۛD�;��4g��FQo �Ѥ��]��6e + +endstream endobj 1789 0 obj<>stream +��ָ�Ͱ�˲�Ѻ�� + +endstream endobj 1790 0 obj<>stream +H��)��!��FO��o�Ɉ��5�VQV�WUS��P47Tww� �s,Kn��+�)�$��Iq�J�O5 + +endstream endobj 1791 0 obj<>stream +��ź��~��z��z�}jvve��n�� + +endstream endobj 1792 0 obj<>stream +H��)��!��D��L*��̌��vr��SUU�U�UQ��4 ��)-��.��\N��U +`�5� + +endstream endobj 1793 0 obj<>stream +��ƹ��w��n��m��p��ic_MA<.40%22(::1QO?�� + +endstream endobj 1794 0 obj<>stream +H��)��!M(��P�^U��ane��]\�4{��-;�mX�jA��i�9��y��%Q�%��ő@s&U��*0�:� + +endstream endobj 1795 0 obj<>stream +��Ѻ��w��q��p��l�y_HF:% #&'+&33,GG9��p�ػ�� + +endstream endobj 1796 0 obj<>stream +H��)��!��D��FO,��/��)�nH ,N�J��q��*�,�3o�ֽ.\�}�ȪM��(�k� �Ɏn)In)��)�\5�5 ��?>- + +endstream endobj 1797 0 obj<>stream +��ؿ��s��p��l��o�~cQK=*'&( ) ( ',)!78/NL;��t�� + +endstream endobj 1798 0 obj<>stream +H��)��D��F��L(��mN�-L��,�l+M�1��؉�7Oo�~d�� յ�uM�-�{6mZ��Gw.>��?0ܓL� + +endstream endobj 1799 0 obj<>stream +H�b` 021��sprq�� + ��KHJI��+(*)c��;^3<<��}|��CB��#"��c�c��SR��32��srr�� +��ĤI��L�:m��3g͞3 �[�2� + +endstream endobj 1800 0 obj<>stream +��ż��}|w_ohSc_N`\FXUCTTDRRANN>stream +H��)��84�"n~W @�r ;� + +endstream endobj 1802 0 obj<>stream +��Ǭ��g��e{qW^ZFJKOOASREhg\�� + +endstream endobj 1803 0 obj<>stream +��̽��w��nzs\VUEHI8GGAHHIPOCT[V^`Uxse�� + +endstream endobj 1804 0 obj<>stream +��txr\e^Le_KicOzxj��Ϲ�� + +endstream endobj 1805 0 obj<>stream +��Ưķ��ƫ�� + +endstream endobj 1806 0 obj<>stream +H��)��84�"n~W)@�r +;� + +endstream endobj 1807 0 obj<>stream +�� + +endstream endobj 1808 0 obj<>stream +H�b` 02��yT-D-@�� + +endstream endobj 1809 0 obj<>stream +��̸��t��o��t��{��meeVC@71/&01*:;1\]M�� + +endstream endobj 1810 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�tZ�u[�mV�gR�fQ{`M`M?M9-E6,B6/<:203-.0)/1*24,BDDYfllu~{�x��~l~p^}`LoUCz\I|`L�gR�nU�rY��q�� + +endstream endobj 1811 0 obj<>stream +H�b`� A3+;''7/��6.�ttt�� ML��-,��ml��]\��=<��}|��UCBTC��#"��_PXT\RZV^QYU]S[Wߠ��4�Y��ݜ��޾� 'M�2u��3g͞3w��-^�t��+W�^�v�� 7m�ݲu�v�;w�޳w��87N + +endstream endobj 1812 0 obj<>stream +��϶��lSo\H^_LZ\L`_Lb]HKJ:<=241'.*"/*"0.'CD8�� + +endstream endobj 1813 0 obj<>stream +H��)��ꅐ%�@�T7�� r�;� + +endstream endobj 1814 0 obj<>stream +��η��o��g��kyarjPodJhbJgcLeaMcaM``M[\JPP?DE:?@8CC8ibN�ռ�� + +endstream endobj 1815 0 obj<>stream +��ʲ��p��ghbOVUDJK>stream +��ĭ��n�{aa]KIK=32(&$&%("*&98-QN;�� + +endstream endobj 1817 0 obj<>stream +��ű��t�{bfcQKK?66-+(( ,( .-%.-$<<1��k�� + +endstream endobj 1818 0 obj<>stream +��Ͳ��ksp]^^NPL;DB5@@5JKBpn_�� + +endstream endobj 1819 0 obj<>stream +��̴�� + +endstream endobj 1820 0 obj<>stream +�� + +endstream endobj 1821 0 obj<>stream +H�b`�F&fV6vN.l��ū�ABRJZFVN^AQIYEUM]CC��AKK$ �4��2hi��[XZY��;��3�=.�n��^�>�~��A�!�a��Q�1��q� �jyy��RR��32��sr�� +��KJ��+*��z9ɱځ��ZH��sN�8i�S�N�>c�,�j!@��fP��yH8I + +endstream endobj 1822 0 obj<>stream +�� + +endstream endobj 1823 0 obj<>stream +H��)��#�K�r3�#�ݽ��=Ãs�볂r"��V�M��X�[��\hRE��R�ŋ9� + +endstream endobj 1824 0 obj<>stream +��ʹ�Ҹ�� + +endstream endobj 1825 0 obj<>stream +H��)��z�#�J��ғbݼ< �5��4Td��}2"��#=J�|z��&W�M,��X�_�D=P��*��w��76? + +endstream endobj 1826 0 obj<>stream +��ɯ��è�� + +endstream endobj 1827 0 obj<>stream +H��)��!�z�#'�D�V$T��e�D��:��ʨȋ+)J�i*�k*9�y;�z�ZDy�ŷ��Et��udv�w�uF��<�"n~W)@�1�2� + +endstream endobj 1828 0 obj<>stream +��Ȼ��|ycWTEKJ9CC4GG:��h�� + +endstream endobj 1829 0 obj<>stream +H��)��!�G�E��D��6g��G�gF{��ػX�jj�ʪi*�j�h�k��V66��Mj�>stream +��ϵ��~��{��nicN<90' & )%..'PE8k\I�� + +endstream endobj 1831 0 obj<>stream +H��)��!��Q_ItY��ߢ$ߤ@�Pg�`G�@'��:��O�<~r�ʅ ��L.�o� �ʎ(Έi�)��d��{W�>��ٝ�� L/ + +endstream endobj 1832 0 obj<>stream +H�b` 021��sprq�� + ��KHJI��+(*)6DEUM]CSK[GWO��Q��n fbd�a�� + ��OHLJNIMK��˗-((,*.)-�jRyEeUuMm]}CcSsKk[{GgWwOo_��&O�:m��fϙ;o��.Y�l��:y + +endstream endobj 1833 0 obj<>stream +��ȼ��{��x��v�|ivlXkdQieThgXfeVhcO\ZHPP>@@456.44-89.JK<��o�� + +endstream endobj 1834 0 obj<>stream +��Ţ�~��k��h|r^lfQcUD[NBQPDSQEYVDYXHPQ@FG;?@7::2DE6YU?�� + +endstream endobj 1835 0 obj<>stream +��̵��|��tvoWPO@64,&&,'/-%2/'74+980??3PN?��p�� + +endstream endobj 1836 0 obj<>stream +��λ��oleLSQA98.("%%"$++$KK@��u�� + +endstream endobj 1837 0 obj<>stream +H��)��[ DS*��rf;� + +endstream endobj 1838 0 obj<>stream +��ѹ��u~za`^IKK>B>4?<2<:1SQE��ȸ�� + +endstream endobj 1839 0 obj<>stream +��ɽ��ufcPMN<<<1/0)*(!-*#22+??3JJ>stream +��ʯ��}hdRMK=>>188,44+44*88,FG8QQAheP�� + +endstream endobj 1841 0 obj<>stream +��ʵ��}�vkgXibQlj`xxn��l��Ŀ�� + +endstream endobj 1842 0 obj<>stream +��ɻ�� + +endstream endobj 1843 0 obj<>stream +�� + +endstream endobj 1844 0 obj<>stream +H�b`@��@�! +L�X�q�QՔ�0&�1 + +endstream endobj 1845 0 obj<>stream +�� + +endstream endobj 1846 0 obj<>stream +H��)��!��D��I0#�;?�3'��(��4��)��/�郠�h��R��>�2*��`œ+��U�>� + +endstream endobj 1847 0 obj<>stream +��ؿ�� + +endstream endobj 1848 0 obj<>stream +H��)��z��J��[�{#{��[�#k3�+�� +ëӂ�Z�B�S��S��2��&�E��D�g��4%�6&�4�H��Ԁ��` ّ�_3�2~Qo%@�Y3;� + +endstream endobj 1849 0 obj<>stream +��ָʽ��x��|��̻��ӿ�� + +endstream endobj 1850 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�s[o[j`QVMEVNISKBKE;LLAHG=EE;CA6DC7FD:E>5ND?QHDSIEYRK\WP^XRXPLXOCPG?LC70A:2IB8MD6ME;[VApU��~ + +endstream endobj 1851 0 obj<>stream +��ӷ��i��f��g��o��r��t��q��k��j��ʭ�� + +endstream endobj 1852 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�tYzsfS][FIG?CC150)-''$)( '%&"&"&"#&#,.(.1+./)35.K[\er{ny�s��y��ycaSDKA4?8.?@2KI;VM;�|e�� + +endstream endobj 1853 0 obj<>stream +��Ŷ��m�}a�c�f�z`�wYxqVm^L^TF^[Hc`MlhR|u_�� + +endstream endobj 1854 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�v`ueTfRFeQI^I>YN@MM?KKEKLCEB5HG9BC>C@7>:,>=/78.33,66.>stream +H�b` +021��sprq�� + ��KHJI��+((�*)�2EUM�I]CSK[GWO��Yc@` �ZPP333Kp+kChXxDdTtLl\|BbRrJjZzFfVvNn^~AaQqIiYyEeUuMm]}�Y�M�-�m��]�=�}�&N�>stream +��Ƴ��r��k�v`yq^oiXgbOd_I^]JYZJQQCA>140&*$'-*":;3_\K�ɯ�� + +endstream endobj 1857 0 obj<>stream +��¹��~�ariSc\I\YJb]JjcImgSnhSc`KVWGMM=;:053+66.DC6��d�� + +endstream endobj 1858 0 obj<>stream +��ȴ��zz`eXFF@733+73'GF8TP@XVDTTDMM>stream +��Þ�|}v\VUE54*%& &"-(53)BC4MM>stream +��λ��kTSE76./.'/-#.-"//&;<3OM=oiX�� + +endstream endobj 1861 0 obj<>stream +��ϳ��xqrenn\hi^heU~|f�� + +endstream endobj 1862 0 obj<>stream +�� + +endstream endobj 1863 0 obj<>stream +�� + +endstream endobj 1864 0 obj<>stream +H�b`@�HF&f&VVV6vNN.n^��j�U9hG + +endstream endobj 1865 0 obj<>stream +�� + +endstream endobj 1866 0 obj<>stream +H��)��!�zK��K"��"z +#��#�K��A"��W��K��h ��订�΢Ȏ�䀺D��X��h��`��ל�ڼ�)U ��i<� + +endstream endobj 1867 0 obj<>stream +��Ӿƾ��ü��ï�κ�� + +endstream endobj 1868 0 obj<>stream +H��)��!��DCPī��e1�ʁ(zby섲��R��#��4��4��C��ɕ��? + +endstream endobj 1869 0 obj<>stream +��׻�ε�δ�ư�ɱ�ϵ�� + +endstream endobj 1870 0 obj<>stream +H��)��!�J�zK��b:sú"��"��+�C�}��C�b|JR�sB�s��2��Ӄ��}�S��S��|�k�<�S}��|k}�S[҂�2��J��J��T�/�0�:o + +endstream endobj 1871 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�t[�sZ�gQwZH^LDNJBF=3:4)66+01(02,,,'.%43*:;0<;/<>4?A9NQK]a_ootssz}��}ep\F`TDOIETKI`NBe^M�r^��y��ȳ��|��u��z��}��~��x��y��ַ�� + +endstream endobj 1872 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�tZ�qXy_LZPCCC966-14-*)!-(",'!*%/.%11(00(00)00)00)33+??>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�u[�eQo\L]^WXWVQWRRVSNSPLXYOSKGF?BA@BB7@B<@A8@A4<=435./1)/0(./&43.9>9EGEae]��®�˲��۽��|��ɴ��ī��l�~b��f�}a�w[�w_vr[k_MdPE[SCVTB_[JeaN|t`�� + +endstream endobj 1874 0 obj<>stream +H�b` 021�+;'7/��<� +P�(�� nn�@��4��? 0(8$4,<"2*:&6.>!1)9%5-=#3+;'7/��a�weUuMm]}CcSsKk[{[GgWwOo_��&O�:m��fϙ;o��/Y�l��V�Y�n��6oٺm��,��I/ + +endstream endobj 1875 0 obj<>stream +��º��s��c{rZriSmeNmdJlePlfWjdP\ZGPQCB@31-#("*%10)H@5��u�� + +endstream endobj 1876 0 obj<>stream +��к��okcOYV>QG9PG9RI>stream +��ֶ��h^YCCD755,+)!-)!:6*IG:SSCTTCPP>II?LL?bbQ�� + +endstream endobj 1878 0 obj<>stream +��¾��}kYVF=>3.,#*')%*)#55,BB4NO>YTE}k�� + +endstream endobj 1879 0 obj<>stream +��ɘ��`aXQQAAB::<8=<4>=5QREigX�� + +endstream endobj 1880 0 obj<>stream +��ʺ��Ĳ�� + +endstream endobj 1881 0 obj<>stream +�� + +endstream endobj 1882 0 obj<>stream +H�b`@�P��.�̂�UMo�K�Q + +endstream endobj 1883 0 obj<>stream +�� + +endstream endobj 1884 0 obj<>stream +H��)��!�G��F�ɒ��Ⱦ��X �+��+�� H�J��^ *ł��!�@(zju��j��y? + +endstream endobj 1885 0 obj<>stream +��ʷ��˪�Ū��ȿ�� + +endstream endobj 1886 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�t[�v\�sX�qWwaRxcTt^Ph[Md_Pc]N`YFaZJ_XEb[Jd^OiVGoWK~cP�kS�mQ�kQ�hO�gN�cK�eM�gP�fP�gQ�hR�fP�jS�jS�jS�pX��u�� + +endstream endobj 1887 0 obj<>stream +��պ��o��m��m��r��s��v��ç�� + +endstream endobj 1888 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�t[�lV{eQb\MKKA<=6796542.*$($&#'#+*"-.')&'#..':=8LC@TZTnsqtki{tv|{��zn{fSq\J_PBMB>WIDZJDZJCgXF��|�� + +endstream endobj 1889 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�v\�cN]NHEKIBD;7:325-01'00'22)01'//(02-14.//'..$.0(-0(372BIISbhanxp��u��|�|ilfbfkg_floww��Ļ��u��n��m��j��h�}c|s]}w`xr\zr\��g��|�չ�� + +endstream endobj 1890 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�t[�lTs]SsfZqcOgZRgZPdXHeYIfYJbVF]PHXKGZMH\OB\O>XK>QD:OB8UG_S@f\Lw\þ�� + +endstream endobj 1891 0 obj<>stream +H��)��zK��z��{ +":�B:��(��('�#+(؝֕$�sA�n�� +�� +�� +n�lKhH��q�H�J�/��o�L��.��ί��k�r��5SW.��W�ѐ�\��V��٘^��c��a�jobomdkg�hedoc��n^�~�� hpn� + +endstream endobj 1892 0 obj<>stream +H��)��!�zK��+�!R$R�Sٕ֑ܞܒڜ�Y�W��j��;W�޵|��5��o�p��^?��?�f��[�j�� +2�Js"�SÂ�ݽ��<�-�m� �� 5 ��tU��2��?�~V�p� + +endstream endobj 1893 0 obj<>stream +��ǲ�ŭ�íǽ��y��gxpYpgSkgbaaW_ZCPO?:;20/(,)!/-%.+#,'/-%770IJ=��v�� + +endstream endobj 1894 0 obj<>stream +H��)��;�-��)��K��>t + +endstream endobj 1895 0 obj<>stream +��ƴ��x��x��k}xdzuevnZogOmhXidT[YGFF<:8.21).,$+'.,%>@4rlX�� + +endstream endobj 1896 0 obj<>stream +��έ��jtmUg[Eh]LjeSsgTsgQrjSleQ^\KVWFGG;99144+88-ZRB�� + +endstream endobj 1897 0 obj<>stream +��Ҽ��x_\Q?MH:BD6;5-IA6US?^[F\[HPPGMM@JJ:EE?IJ>��{�� + +endstream endobj 1898 0 obj<>stream +��խ��i}u_`WID@5>=2@<-FE8VSCa[E^]I[[LSTCEE;@A6JH?rm^�� + +endstream endobj 1899 0 obj<>stream +��kj`NID:.+$#' *%/+!:8,==2HH9II:JK<|uY�� + +endstream endobj 1900 0 obj<>stream +��ݸ��ib^N@?3,(' *'($*&++%:;.JH6qmV�� + +endstream endobj 1901 0 obj<>stream +��з��wueNN@>>4;<1::.EE?XTIon]��w�ո�� + +endstream endobj 1902 0 obj<>stream +��λ�¯¼�� + +endstream endobj 1903 0 obj<>stream +�� + +endstream endobj 1904 0 obj<>stream +H�b`@�L�,�l��\�<�|��B�"�"b��R�2�r��Y^^AQIIIYEUM]CSK[GWO��b$mL@]��f�J��V�6x#��l�`gO�b0�oZM + +endstream endobj 1905 0 obj<>stream +�� + +endstream endobj 1906 0 obj<>stream +H��)��!��J�� +#:�C��"��b�K�|R½��L-Lu��Ԍ 4� ��4Ԕ�U�%5�d,T��5��U�� 퍝��#]-��-\��]-"��2R�8+ + +endstream endobj 1907 0 obj<>stream +��˱��;�� + +endstream endobj 1908 0 obj<>stream +H��)��z�#��B3C]R"}��#"��<�|<�l>stream +��Ե��o��o��q��p��m�e�xbtb��k��ε�� + +endstream endobj 1910 0 obj<>stream +H��)��&�F��y�;;�ٛ��[�� y9��;[&��w�M��V;�4��(��0�'?�3;�5˿-;�#/��0hlI�R�V�� #4O + +endstream endobj 1911 0 obj<>stream +H��)��!�zK��J�'�E��W'��$�D�Y�;��{�y;Z8�[�y8�ٛY�X��E�E'fe�TM��/_��k�O�?��q^snqFhlBLZ~i�̅;O��v�ҫ�7޿��͏_o�|��;?��^޻iݤYS�'4%�5��F�:��;8��j��j�ɩ+H�+J�騚��>�}��-3�,9,5.,��P[_C�PK�@��T��\��ߥ$;��<��pBK��E�&n]>stream +H��)�� -��+�FF�@T�Wbe��"��"zK"g4�ݻw�ٳ�ϟ޽y��oZ�v��m�w��z�ƥ}�Nm�tl��m��in��ˎIIq��qpv6�4�1�S�R�QW�"y1-e)#C-o��Sk޽zL��K�doI4�_�EI��`d�G��5&4&��'�T�z%ze�8%y�EzTg��ΨM^ә�lB��sf�L�1a��9SV,_pߩ;g��8s�Į�GVٹ`Ò�I�y�9q��a�N�fzڪj��jr�@}��(D@��Az��@ZԔF��E�E��FQoADGad{^xuB@i�gi�W^�sV�G��s��gl�bfjyEag{��y�-�ۿnʮE�+�u-��?��a�� ڰi�嫷�>stream +��ű��w��gzq\gi]af\d`L`^KQQCAA799//+"*#,)!33+JK?��o�� + +endstream endobj 1914 0 obj<>stream +H��)��P4�2nQ_5@�'= + +endstream endobj 1915 0 obj<>stream +��ͼ��mrZ�oXp^N_UEbXGj^PfcPleLfbNWXHSSAHI:88/77/@A6`]K�� + +endstream endobj 1916 0 obj<>stream +��ܶ��h�{ch`M?=11*#3*!9:1FD6LJ8LL=MM>II:CC8CD7OO>��p�� + +endstream endobj 1917 0 obj<>stream +��m�x\ZWE22*%&''*%10)<=2FG8OM:ogO�ҷ�� + +endstream endobj 1918 0 obj<>stream +��ê��ydTRC87-*')&)&(%)'=<3fbR��n�˳�� + +endstream endobj 1919 0 obj<>stream +��ж��lZ[LTUFXYJbaR��y��ȵ�� + +endstream endobj 1920 0 obj<>stream +�� + +endstream endobj 1921 0 obj<>stream +�� + +endstream endobj 1922 0 obj<>stream +H�b`�F&fV6vN.n^>~A!aQ11q I)iY9yEd�Ĩ��o`hdlbjfnaielmckǀ�U��]\��=<��}|��C0�@hh��sXxDdTtLl\|BbR2n�P��(+;��gdfRu��Rhhvq�� 7o�(0M� + +endstream endobj 1923 0 obj<>stream +�� + +endstream endobj 1924 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�iU~`Mwe[klhjd]aa\\c^WXLFIA77165-30)=72C=:D>9D@;L?7K8/eP>o^Jn^KqaNrbOxhQyiR{lU|kU~dP�fQ�w]�x]�v\�}a��| + +endstream endobj 1925 0 obj<>stream +��̽��Ƽ��ĸ��ǥ�Ѹ�� + +endstream endobj 1926 0 obj<>stream +H��)��zK�;�C*��R�C�||�ܽ�ܽ|�<<]|}��}�|�l|�}]��s�=+�#�2"z#'�FO��Q7�"vR9�L,��B� T�_�ere��J�T�7 + +endstream endobj 1927 0 obj<>stream +H��)��:�b�k3�*�sbڊc��)/�2-�8:�&=�,>0/�3:�-8��J��D��P��P�DO��L��\H��蚘��9ZDE{u�&/�0-�21 + +endstream endobj 1928 0 obj<>stream +��ȿ��~��u��q��r��u��}��پ�� + +endstream endobj 1929 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZmfXT]VSYVUUXMNNDD:@GA>C;??3?@4??3?GB<@>;;8>?7829:5;:852.8;78=93/*($,) ,.'-/).1*43,B80B?2ULAwna��y + +endstream endobj 1930 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�tZ�oY^ZQIKC@A98;4/2,'&!%!&#)*$+-'*+%*)"+,$,.%..&24+BC@PYZajoo|�v��{��o~mXtWEkPAeM=gWCkTDybP��r��ô��q��f�{`��g��l��e�}c~w`njUpkY�{f��̸�� + +endstream endobj 1931 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�t[�jT}gTpdZh^Re[HaWK[PJYNCZOEQE=K?5D@3BB3=7.;2,88/:4,EE?GKEMYZep|s��ǦУ��hso[jhcsso}��Ǳ��f��e��e�{b�w^kfP`ZEWUE@?5:9/97.@?3JK>stream +H��)��6�-�FF}@T +$��Rő=E݅]�@�3?�1ٯ='l��N\�r��7��t�⑋��:s��߳��V�$'�Ex��x�XXhj��k*J�ˋi(�k��.��Y)��,��y��-��_�EI􄲘 �`2��(�� 0�'?�+7�9'�1ݯ<ʽ8ܵ(ܳ<ƿ41�:?�!3�8;�"'��"u��U�&-\2u��+��w��ׯ�;��lZ�>F� i�1%Y��A�66z�j�*��o*�kȉ�h*�踺[f�]8��OD�1 +���#� +#{��]١��-y�-E�ui� ~��A�)�9�.��A�I>��!1Q�I��%u�'u-�;m��I[�O_0��s��%�q��7o��{��/��z��g�޳l��ھ֔�� M��19�^��Φ&��F��JRJ��T��R�51� +��Kپ~曷w��`]B + +endstream endobj 1933 0 obj<>stream +��ɽ�Ǵ��w�xdwiVbaQ[VHSPCJM@>=31/'*%( ( (,(34,JJ<�� + +endstream endobj 1934 0 obj<>stream +H��)��8�[�W=�:aIo%@��1>m + +endstream endobj 1935 0 obj<>stream +��ë��r��fvoZskWwmXjfR`aLdaLZZHPQBEE888//+#)#.-%56.~xc�� + +endstream endobj 1936 0 obj<>stream +��ǹ��jphQ`UBQHeQ@dZG``M]^NYYHLM=AA7==2?@4fcPǿ�� + +endstream endobj 1937 0 obj<>stream +��ƫ��g_XHCB8-,%* /!5.'BA4HG7NN>NN=MM>JI9igY¼�� + +endstream endobj 1938 0 obj<>stream +��Ẵ��n\[IBB5.,"'#&")$-+$780IJ=a^K��u�ս�� + +endstream endobj 1939 0 obj<>stream +��Ŝ��mcaPTLAUOESUFWXIgbQtoV�� + +endstream endobj 1940 0 obj<>stream +H��)��h4�2~q_@��>c + +endstream endobj 1941 0 obj<>stream +��³�� + +endstream endobj 1942 0 obj<>stream +��~c�zbxo[pgRhbLfbJ`^LfbMhcN[[KOPH@?975,32+894UP?�� + +endstream endobj 1943 0 obj<>stream +H��)��#�-�FF@��{+'�= + +endstream endobj 1944 0 obj<>stream +��й�ǵȼ�ƴ��x��`zhT__O[[K]]LTUGLMA=:01,$*#'*$32+FG:��v�� + +endstream endobj 1945 0 obj<>stream +H��)��-��-EA}�1}%�@�_=�4�� '?�#'�5#�%3�9'�>=��p��l8�k噣�_��Μt��/�=��;/o��lioWsAU~Tj�`��(�"uyoo��j߿��_� gn� + +endstream endobj 1946 0 obj<>stream +H��)��!��D��BPL��t�we�v�u�PWNh{fPsnpcN`e�i�OQ�Wr�WF�waLPka��53*��j�0�vb[^c}I}ۄ�y�9��K�_�z��ޫ��n�?uh��u�'N�k�Ih.��/O�J�u��0��TR�W��P�T�"uy1-%3��X�[�6��J�>��+�� O�Ƕ�7�QcJ@m�OM��1^1��^��6��N�!A��ɑ�9Yu�S�]�o��Xi׬�'-�T��]��=e��;Ϝ�s��o�^~}��ͥGwNݿz��m7.�=vb��M]��-��u�ͥQ�ў��VV��j@Oi*K�ɉ��ᔉU�_��/@��< + +endstream endobj 1947 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�tZ�oW�pY�t[zcPp`SnfVd`WWRHNHAGC=@<2=8/?:1=<20.%01+-1)14-BIH`kwn��|��ھ��y�|bth_dc`nqq��ǿ��p��f��f�y`|qVeWFQG;GH9;;133*32)98-IJD�j�¬�� + +endstream endobj 1948 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�t[�mVraO[UFNIEADD682**"$&")&+,$.0(23+34-66.<<3<<8GHHZb\fquyz�}{��u[�oV�pW�rY�sZ�tZ�oV��t�� + +endstream endobj 1949 0 obj<>stream +��ǧ��r��k��l��m��l�~a{sWpjTkgSlk[{v\�� + +endstream endobj 1950 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�v[|mVbZQ^XYVTLNL96@@=79303+.0(45-<=4DE:MA:UB7R@1Z=/kF4nI6lH7qP>�iR�oW�pX�pX�pY�qY�sY�oT��u�� + +endstream endobj 1951 0 obj<>stream +��ũ¹��w��t��s��o��t��Ͳ�� + +endstream endobj 1952 0 obj<>stream +�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�sZ�u[�lWcR�bP�nW�kU�dSziTn[K^OBOE:H=4@1'=-#7*"5) 9*!.')#'$*%'!,) 2-$G7-50&M>4O>5M=4Q>6`L>stream +��м�ǭ�ӹ�ǭ�γ�ٿ�Ϻ�� + +endstream endobj 1954 0 obj<>stream +H��)��!��D��P42�Pӕ֑ҙږܚT��W�[��i�b�lhc�b�io�a��l��nl�mec`�`lecdck��|a�Ǘ��5# + +endstream endobj 1955 0 obj<>stream +�� + +endstream endobj 1956 0 obj<>stream +H�b`��YXX��98X9��yx��E�)�kQQ&1q1q 1II)i|�с�,)��FUS� � + +endstream endobj 1957 0 obj<>stream +H��n1��)| ��4�G��h�(U��@i��!��w��=��6R7�ʗƙ�=^}��>��խ��ۦ��8�HD�HCLJ+O�E�j��[e<ָ��@�̺�Y;��f� �D@a��^�E� 1�KU뵐��瀟Q�5P�e�� B{�I`)�Y�hں��y��&=׵m��K��ry�4�u��rA�k x�%i^V�ۺ�TW��o�,�y3 ��TZ��ᕴ��I�)��(�p$��l�1f�� ? +M9H7X��_�q6��I��v3ri#I��B�'��$�,�u�[�k�3@q��P\Kϛ��i��^:-W7u%)�Rr�*=��4�Å��.�`�K��7��2��R1�K�"Dsh�V��V�[� �/�ׁ��mc#�(��6��'��p2��@�m��t��)��w�f �O*��%��!W��,D�T��E9(�j��Nh�o�]��Shu��+��p��ZL�Ϩ��@�� ?�a��\�k�+��̈́�f��mN�N��r�E�j�ۖ��0�h�Kk��]S�Ͻ�"?�R��~�q�yk�#i$Y��[��6W_W/I�Z��U�m�|E/'�Sy��Ri�(2�1\�)�|�9p��t@}x7�z�D�5�4܎��΄��MX�BE�p�(Zy*FݶQt�T�Z��x:��u�F��b>�6�ebs�k�F��;�� +��p��h��h�͒s#��T��V��oޜ��g��5�]�ku�j��`��UG�&_ ��6�Ψ[5�u�w��UQ��N��u^~T��T�a��|��\�G�%=�d�,7zR�}��>��G��~�N�lnכ��]/�e��옗V�5D��h�6��cK�g��ṃ�5�)��>�n�w7��?w��a[G��0�� +endstream endobj 1958 0 obj<>stream +�sZ�tZ�tZ�sZ�t[�t[�u\�u\�v\�v\�u[�tZ�y_��z + +endstream endobj 1959 0 obj<>stream +H��)��-��+��+��P�[�W�_�[ў^�[�U�S�Y�V�Q�]�W��za��[{�K��&G��DM��_�W9�"~Qo��*��@*X + +endstream endobj 1960 0 obj<>stream +H��)��)��-��-��-��)��+�j� +iLhH�oH��ʎ�� +uM��w0s�1r6�5�P4��0V��P��Ӱ0Բ4ֱ�6�q��+��+��+��/��D�]��]�a�Y!Y��u)�)�q>��.� �E�nQ�5)u��U��c&��N*��/��-��P3�&a�Ěޒ�ɕq{+&�H� + +endstream endobj 1961 0 obj<>stream +H��iO�`��3M��c�y��g��(@6ms�v��c_Z4��KWlCqL�:mZsњ�۩��Iw��M�*�|!�I�1��ⅼP��9 +��;{,@��8 'H$�賓z�R��!��ecҳ �`��&;�-k�]��[��_�LO�չ\�M��K�"�h8 d��#>�s!~�wa8��d��Σq��Pv��3Ug��㠢�j��pl2�Nd��:��"�2l)_� m�U[W�M՞IoS�Yk=i��^�H�̥�X(ۥQ��**h��.�Z��e��:��e��^� 0��k� + +endstream endobj 1962 0 obj<>stream +H��N�P��VM�K��+I��2e)q��K��Ze��5�\m�n�A�� `��1�Ԕ�S��*赥� 8�V29:�$Ȱ�<��0��2�@��`�a6��V�\&�8Z=*td�s��E��|�(��n��]�9��^�� yS�eQ�J��ϥ�S��NRDh�ւx �q��e�w��qy�.��&�P�h<��I&�@��c &�<1�Ok>RF��ب�/�S��V��A��j�x�q��d�$K'"T��El�ȝ��L,��mř�sق��8��UW2t��"ICPj��c�P*��cǂ��Y`fA�/:��F��bL-�3Dӻ��׹�`�X�? + +endstream endobj 1963 0 obj<>stream +H�(��sZ�u[�u\�t[�rY�sZ��|��ȼ��ǹ��r�|anhX^_T]]N_]MYWFKI;;:0.)!)#-*!/0(@A5��tZ�u[�tZ�sY�hR�eN~iP�dN}iT�q_��iwoZjcK_]PPQG?=163,-'( *")!(*#21*EG:��t[�t[�qX�qX�pWbM_Ly\Ik`MjbQi_Yfd^xtj�q��v��ý��̲��u��iu_yo]gbSLN@21'%%' * -+#02+fbL�ʮ��v\�u\�w]�rY�pW�oX�sY�nW{aOpWFjQAh\H`RFURNUSR`]Qmlit{~��µ�Ƨ��}��s~yw��к��l��g��i�v]{lXd]L>;.+' +""&(!0*#\SB�̭��r�_ + +endstream endobj 1964 0 obj<>stream +H��N�`@_b&˂3JA[@)��2D%�Z��׋е�"8hiK�/s�X@��.��s��'8��\�ޒ��sE3�<�(��^=��P��/��O�8s��N,�[�e(��w�ˮl�}��*A��_;��Ƚm��S�dU�`л�m��ϏM��]ij��Y�;�Ɩ�`«~��k��O%��2��&��Q$��78��-%�Hv�r��|�<%\�pU�UIO�ou��e�voʯ߇��&�Z`�I��OFK�+b��b:�O��xۈ��C�b0�Zy��_#�(�� ,G[��v��M��c�w�'�[��6�;��h�~[ԴF]�8ȓ5n�T,�R��D�Pl��D,��I�2�\[~[)e&&�ث�6٫��H��-��B�8�F��F�S{i�,VJ}�1��ڝ�1�o��`f�m�k�u�N�uٛ{�Ѯ��׎?�8��}!��y� �� + +endstream endobj 1965 0 obj<>stream +H��r�P@�B놴6%$��1�BY�@@�%{( Ț�\��:*e,$h��>�/�Μ8s��6P�6��q�%��f�{,��>Xi��D��_3��T�L$g"�P�m)�l�6��~� �݁�e��@'rQ*D��Hb�R�_��E��LqƢ�?s��;f��u�ZO�t�l�,N�Q܇x^��A��ܱc��q-m��&H��-XMOEf1K�[j܏�sW��P�6��k�/��T8B�!Ab�:�}��x��{�O�_��4��~�8�9 "��m%s��x� k�c4~i�W��H� +�%��P�Ĕ +�|6��)� +G>ć��z^��ˇ��T(�LQL��cGZ��v�:&go�Gҍ�^��Y��T*�UZ�IȊܹ��s唔�Il沐l+4�&L�ҁ�y��w��G��\y��V�I + +endstream endobj 1966 0 obj<>stream +H�I��sZ�sZ�u[�u[�u\�t\�t[�nV�iQ�hO�lT�rY��nĹ��´��}�rYi]IVRDNN?IH<<;144++('!( ',+#EF8yp\�վ��tZ�t[�r[�t[�u\�tZ�iR�eO~`JoYG|]H|]I�}l��p��k��v��}}��ɰ��ǭ��s�y^}u]{s^rkUQP?-)"'' ') +%22*CDYE9ME7LI;TQPbjn��;��tmM>bC6mRBe`Y��Ϲ��}��p��p��cukSEC5%( )#+(!--$:8.|vf��"�5\ + +endstream endobj 1967 0 obj<>stream +H��S�p�񿣰i��@6p��Ķ�o2��E�#/�`�a��?��}w/zM��2 +O��t#��o�,m��oN��el��4��l��X��gps,��V��E��Cj��U��p��z�r�?w�z$�o��#ii��6?��6;��O,W�K�T4��K" ,�K��$�:Gy��G\�w��N~�\o�ή[�M?5�謡�ǧN��H:��"�#;/��/��P<�U܏�i��=:�P�@�JA��*FW�f �_�-l��0��}�y�e��+� +�!��F��h2C�7��W�N��v#[�K��L,�![/�E��a�q��7hO�Xc5H�'d�!�*��~��T1�"�ɢ�l��&��7��4v��?M�wj>stream +H��sZ�u[�t[�v\��o��ź��Ӿ�˵˽��uxqYicNZYJRSDPQAHH8FF;;<31-"+%)#-*!..&AC7��sZ�tZ�sY�qW�oV�nV�eO�kR�v]��{³��Լ�˱Ȼ��{ctlWriQnfM`\FDC82/',&*%'(( *#10)LK<��u\�tZ�qW�pV�nT|_L�cP�|e��y��x��|��~��{��Űķ��l��f��h��j�za_[I88-& '+&)!&,*"89/mgL�ջ��s[�u\�w]�w]�rZ�oX�oW�jS|fOqWEtXItiU{u`��λ��ʰǫ��s}wkouvx��Կ��v��z��l��u��p`]I41%$)"0.&;3*74,YVB��q� + +endstream endobj 1969 0 obj<>stream +H��QW�P��A�0i`˜G<�� XS6��ݘ��v�]F�9e��=�G�s��=`�`a�N�ըW^�3/�ɓV�ˎ!�c��+�BW `��_X��KAJ� �<� +Dm9-w�nvzV��h�O��SBG q�־xd4��c=8J`�w�i��eod~��ci��am��?XC�PGM��%�-��'.=�0��҅�� +��b��w�e�)�]��׾��+��٠9Jk�>��f�n&(��৫�c�k��Y�8�P�L��&��6��& +L�4��t�<7�+��BԪ�V ��T��@�n+K��nCG��3�^�(��=6G1�J�Ɑd:�=KP�;;[��~>��х��!��?��7�^��4M�8 + +endstream endobj 1970 0 obj<>stream +H��%��sZ�u[�t[�x^��u��Ƹ�̷�é��p�v]i]HSVHRTEPQBVVEUUETTCNN=LL:AA633,66.<=8NL@�~j��sY�pX�gR�iS�x_��ƨʱ��{aqhTf_Md^Nf`L[ZIWWGQQCJJpjU�ư��sZ�t[�rY�pW�mTaL�mV�td�zb�yb�jT�t^��{��Ǯ��n~t^�wa��e�~d~pV[VADD6==3<<5==6::3;<4;;3BB7MN@ztY�׻��v\�s[�u\|iV�hQ�iO�hT�ra��q��naw��Թ��ҵ��q��o��m��leZJ=6.--&0/'1/'0/&54,770EF:wwe�ս��A�� + +endstream endobj 1971 0 obj<>stream +�sZ�sZ�u[�|a�u\�}f�� + +endstream endobj 1972 0 obj<>stream +�sZ�x^��w��u + +endstream endobj 1973 0 obj<>stream +H��yTSw�oɞ��c [��5la�QIBH�ADED��2�mtFOE�.�c��}��0��8�׎�8G�Ng��9�w��߽��'��0�֠�J��b� + 2y�.-;!��K�Z� ��^�i�"L��0��-�� @8(��r�;q��7�L��y��&�Q��q�4�j��|�9�� +�V��)g�B�0�i�W��8#�8wթ��8_�٥ʨQ��Q�j@�&�A)/��g�>'K��t�;\�� ӥ$պF�ZUn��(4T�%)뫔�0C&��Z��i��8��bx��E��B�;��P��ӓ̹�A�om?�W= +�x��-��[��0��}��y)7ta��>j��T�7��@��tܛ�`q�2��ʀ��&��6�Z�L�Ą?�_��yxg)˔z��çL�U��*�u�Sk�Se�O4?׸�c��.��R� ߁��-��2�5�� S�>ӣV��d�`r��n~��Y�&�+`��;�A4�� A9�=�-�t��l�`;��~p�� Gp| ��[`L��`<� "A�YA�+��Cb(��R�,�*�T�2B-� +�ꇆ��n��Q�t�}MA�0�al��S�x ��k�&�^��>�0|>_�'��,�G!"F$H:R��!z��F�Qd?r9�\A&�G��rQ��h��E��]�a�4z�Bg��E#H �*B=��0H�I��p�p�0MxJ$�D1��D, V��ĭ��KĻ�Y�dE�"E��I2��E�B�G��t�4MzN��r!YK� ��?%_&�#��(��0J:EAi��Q�(�()ӔWT6U@��P+��!�~��m��D�e�Դ�!��h�Ӧh/��']B/��ҏӿ�?a0n�hF!��X��8��܌k�c&5S��6�l��Ia�2c�K�M�A�!�E�#��ƒ�d�V��(�k��e��l ��}�}�C�q�9 +N'��)�].�u�J�r� +��w�G� xR^��[�oƜch�g�`>b��$��*~� �:��E��b��~��,m,�-��ݖ,�Y��¬�*�6X�[ݱF�=�3�뭷Y��~dó ��t��i�z�f�6�~`{�v��.�Ng��#{�}�}��j��c1X6��fm��;'_9 �r�:�8�q�:��˜�O:ϸ8��u��Jq��nv=��M��m��R 4 � +n�3ܣ�k�Gݯz=��[=��=�<�=GTB(�/�S�,]6*�-��W:#��7�*��e��^YDY�}U�j��AyT�`�#�D=��"�b{ų��+�ʯ:�!kJ4G�m��t�}uC�%��K7YV��fF��Y�.�=b��?S��ƕƩ�Ⱥ��y�� چ��k�5%4��m�7�lqlio�Z�lG+�Z�z�͹��mzy��]��?u�u�w|�"űN��wW&��e֥ﺱ*|��j��5k��yݭ��ǯg��^y�kEk��l�D_p߶��7Dm��o꿻1m��l�{��Mś� n�L�l�<9��O�[��$��h�՛B��d�Ҟ@��i�ءG��&��v��V�ǥ8��n��R�ĩ7��u��\�ЭD��-��u��`�ֲK�³8��%��y��h��Y�ѹJ�º;��.��!�� +��z��p��g��_��X��Q��K��F��Aǿ�=ȼ�:ɹ�8ʷ�6˶�5̵�5͵�6ζ�7ϸ�9к�<Ѿ�?��D��I��N��U��\��d��l��v��ۀ�܊�ݖ�ޢ�)߯�6��D��S��c��s�� 2��F��[��p��(��@��X��r��4��P��m��8��W��w��)��K��m�� + +endstream endobj 1974 0 obj<>stream +�sZ�t[�ya��~�y_ + +endstream endobj 1975 0 obj<>stream +�sZ�tZ�t[�tZ�tZ�sZ�y_��z��{ + +endstream endobj 1976 0 obj<>stream +�sZ�t[�u[�y_��z + +endstream endobj 1977 0 obj<>stream +�sZ�tZ�t[�y_��z + +endstream endobj 1978 0 obj<>stream +H��sZ�t[�tZ�rX�rY�t\�jU�cQ�dP�cQ�dQ�dQcP{_Lz^KbO�cP�eQ�rY�rZ�rY�sZ�rY�pW�mT�lW�~s�ù��ĵ�Ѽ��Ȯ��nzs^ldOc`OZ\QYXGHG::;233,.,$.,#+$) +'..&F92wlW�Ҹ��u[�pW�rZjXkb\lh]ndPb\WQNFFC4>8-=7.93*60';5,?8/@90D>0HB3IC3SB6S@5Q=2Q>3`F7aH9fO>oXGwaStbTqtl��Ƽ��Ż��t�~e�t^vlYwlXgbNXPBQD8A?3;<29:235-0/(--&24+?B8SC8�{_��tZ�mUy`SgeaXkk^gi`_UT\PRWPMNC@A;;:085/;8-96/85184*41&63&30&-*",( ($&#12*574@IF[c]etpz|o�k��s��q}��Թ�ɰ�ǳ��s��e�x]�s[�mTkbK[TETF;KG8H@5C>3BD8C;3L>3\VCuoS��|��eP`QKENOIPIPH?_G;dPDePBhO>fPDfPAeOChSCYH?TE?TE>stream +�sZ�tZ�t[�tZ�sZ�t[�y_��z + +endstream endobj 1980 0 obj<>stream +�sZ�t[�u[�tZ�y_��z + +endstream endobj 1981 0 obj<>stream +H�@��sZ�t[�u[�t[�t[�rX��r��μ�«��o�|cmgU\[KVVFPQBOP@GG9891("$(.+#670TQ@��u[�v\�v\�v[�t[�lU�iQ�mS{iR}jT�lT�~h��ë��s�}cohTc_M[\LRSB<>100)11)45-3/#,%*$+$-+#34-c]L�ɰ��tZ�tZ�u[�u\�rY�eP_Jz]IhZLl_LgWHn`Tnk\qtdvtb�s_�}k��u��{��|��l~p[}o[~q\qfPWP>K=4@;135,`^J��wMQK-,%22+9:/AB6woW��sY�rY�qY�qW�pW�u\�u\�t^�t]�t[�q[�s\�sZ�qX�oW{aNsWGpSAoR@pTBaUAePBdYLe`[noj��zyp��e��d�x`}oZndObUDQM?3MM=d_M�� + +endstream endobj 1982 0 obj<>stream +�sZ�t[�tZ�tZ�t[�u[�u[�u\�v\�y_��z + +endstream endobj 1983 0 obj<>stream +H��)��-��+��+��/��P;�"��0�7?�� '/�;/�'?�%+�2�+?̣<�3/�-7ԥ (ʳ0ʷ*3t��gv��͉�+��2J�z��(��8��4jJu�� 5=%Q�*��VN��_�W`%7� + +endstream endobj 1984 0 obj<>stream +H��)��-��)��+��/��/��D屝�aY�m�A�i�Չ~q��e1^y�.9�� Y!�e��}e1��c��z +"z"g�$,�X3�2nao%@��+� + +endstream endobj 1985 0 obj<>stream +�sZ�t[�w_��s�x^ + +endstream endobj 1986 0 obj<>stream +�sZ�tZ�sZ�tZ�y_��z��{ + +endstream endobj 1987 0 obj<>stream +�sZ�tZ�y_��z + +endstream endobj 1988 0 obj<>stream +H��sZ�t[�tZ�qW�pY�t\�jUbQ~bP}aP}bP}aO|aO{`Nw[Jv[I|`O}bO�pX�sZ�pX�pX�rY�qY�pX�nV�kQiU�{r��ҿ��Ǹ�Կ��ƫ��kyr^ldOc`OZ\QYWFGF99:122,.-%/-%+%)!+( ./'H92ymX�ӹ��u[�oV�qZ|jZha^ig]kcO`[WNMEBA;?;1:5*94+50'2-$72*:5,50(;6-?:-B>/C>/M>3M<2K9/L;0\D6]F8cM>oYIyeXyfXtvl��Ǽ��t�f�u^xmZymYicNYQCRE9B@4<>29;245-10)-.'35+BD9UE:�}`��lUw_ScdbVjj^fia_TU]QTXQQODDC=@=3=93@;0;74;7,73(95(72(0,#-*!(%($34-8<:FRPdnho�}��z��q�� t��s~��Ը�ʹ�Ϻ�ư��t��f�y]�s[�nUncL]VFVG;LG8JA6F?3EG9G>4O@5_WDxsU��eP_PJDLMHMFMD:[C7aL@bL>dK;dNAdM?cMAfQBWF=SD=SD;SD:VG>UE=VF@ONDOQNadbruu��˳��j�{iyuv��ȿ��v��k��d�}c�|a�u[xiPneLmbKg`LhcLthSzo]�}a��_j + +endstream endobj 1989 0 obj<>stream +�sZ�y_��z + +endstream endobj 1990 0 obj<>stream +H�.��sZ�x^��y��Ͻ��|��hzr_leQe`J[[JSTEPQA??2-) &')!(,+#;5.xo\��sZ�u[�v\�u\�u[�tZ�qV��j�ʵ��ɷ��Ůֿ��pyp\keU]]LUWGSN>C?456//,%*$+%,)!792a\M��tZ�sY�sY�sZ�tZ�t[�tZ�t[�u[�rX�dO�aN�bNaM{]Iy^JocOpeSvi\�t^��پ´��o}t]vn\sk[lfRXWDPI1;;235-00)1/'.*".*#98-CD7vpY�ɫ��pY{fTbQ}bS{bT}cTy`Qt\NrdOt_LiPEcMASJ?LG;MH;JE:LHdMCaYIg_MwiV�}r��z�o��{��ɮ��}��e�ybq\�kSyjUc_NYM@QI;GF7>9098/9:1?A4MC9UPBrdN��i� + +endstream endobj 1991 0 obj<>stream +�sZ�sZ�tZ�y_��z + +endstream endobj 1992 0 obj<>stream +�sZ�t[�y_��z + +endstream endobj 1993 0 obj<>stream +H�I��sZ�u[�u\�t[�u[�t[�rX��p��ͻ��o�{bjeTZZJUVFOO@MM>DD867/'!$(.+#UR@��sZ�tZ�u[�v[�sZ�jT�fOkRxgPziS�lU�}g��ɰ�ò��ʾ��r�{bogTe`N\]MRSA;=1//(00(77/84'.(*%+%-+#44-d^M�̳��tZ�t[�tZ�u\�u\�pW�cN|\HvZGdXKi_LgYJpeZpoawyizwe�t`�~m��t��|��̺��~��lq[q\r]rgQXP?M>5B=345-b_K��|OSN--%33,;;0CD7yrY��sY�qX�pX�oV�nV�t\�t\�s^�s]�sZ�s[�oZ�r[�qY�oV�lUv]LnSDjO>iN=kQA]S?dPCd\Pgeatxr��y~zr��ı��g��e�w`~p[pfQeXFSN=II8TN;\\TGJE77.@A4ON>gbP��!� + +endstream endobj 1994 0 obj<>stream +��Adobed�� +$$''$$53335;;;;;;;;;; %% ## ((%%((22022;;;;;;;;;;��&�"��? + +3!1AQa"q�2��B#$R�b34r��C%�S��cs5��&D�TdE£t6�U�e��u��F'��Vfv��7GWgw��5!1AQaq"2��B#�R��3$b�r��CScs4�%��&5��D�T�dEU6te��u��F��Vfv��'7GWgw��?��q�j"8<��3k�o�ࣾK��%�j4U��1�s�Σ��k�ָ�q�t�pR�I��@�R��h �]� %�$��y" �&T��$�:�1;�iݸ�c_�G�Y�C6�I�ȝH� ˠ}��̍�&C�I@<꒙��:�i�Rw?BD�pS4<��`jG#�J�j��QR�'`�E �sZ��8�E��0u�Y �E89�4�Ω)�O��:Ɯ�'s�$K�3Cˉ&�r;��ަ�?�@,�vƐ�]�H�߀��A�x>��%�L�JJY��O�x& >çp;BMx٬Op�{��D��Jf��0I��P�75��Y�Z]5��Pc��čAh)Nk ��5T��7�u��Q�%�Nc��v��i�5$�mk��3��)�浮"\d�9��vG:��r��Б.�.$��z��qT�� OB�w�"~��l�L�� i2u))f��?��l$�� 5�f�=�u�S?qF�i)�KL�1'O�Bx�ևol&g�hLt�H?�A�kO5y��9�67��DpyShf�� #�G|�9�;'�ۃ��K��h��1�s�Σ��k�ָ�q�t�pR�I��@�R��h �]� %�$��y" �&T��$�:�1;�iݸ�c_�G�Y�C6�I�ȝH� ˠ}��̍�&C�I@<꒙��:�i�Rw?BD�pS4<��`jG#�J�j��QR�'`�E �sZ��8�E��0u�Y �E89�i�^L�$ȝ5*\D��LKK�И>=�*S� �� 8�Gb��G۴��$��< �JJQ#R=��P�N�#X�R�G5��zbH$�~�$�A��'R<�*-{ѫO�O�('q� �2�k:��,`;p�g�x��a>��!%,"A>CX%;�}ژ�摭��L+cH#A#]|e$�\�l<�JM�8XG:��9$��3�L�44�͈䤅5�D ��:��d��'��X��$�H��vEKm#i.q|�D�� I�:jT��#A�☖��0|{T��A�8�q��i��H%/`x��F�{~��&F��68� k'�0�ĐI��IL�� &N�yTZ�5�V��PN�@3�*e`�&u)�X�v�5��5*�2�<�}D7�x�E��@�^|BJXD�|��Jv5��1��#[�O�?�VƐF�F��I\�7t�y0 ��=�p��u��rI� �gR��hi��I +k~�홎u �ɇ��N+n��>I��슖�F�\� ��y2@�"tԩqF��1-/#B`��(�O66� ��q��I�>n�;��J^��'@ ))D�H��>@;�L�c�Hlq@�O�a� ��=��L�H�<��kF�?�>(��p�g�T��L�ST�-��k��jU�e�S +>ݤw ��N�RR��| +�2w��ǒ�,��8��A'P{��%2h&�:��yQk�֍Z�|QA;��ό�9��#X�Ԧ�`[ۀ��|ԫ<�Ʀx� � �i�U��y� )`q ��)��k��o4�l/�>�� +a[A ��)%r��a��Rh�9��9�> �a'p&A�Jf��~lG%$1��~��$��JH��oT��:>i>!؟�֓z�d��=��~?�mx2Ii�އS��'��D��i��=�䌾G�a�?�mx*H��ޏS·�35']�z��oڙ�5��z׃$�اޏU��8��i�j��|tlϥ�߉��$��>��[+ۻ��h��b�?�iLul�WN��Y�1|��T��n�=7��ԑ��6}��L^?�)x*I}�}ݝ[*]��1|?��ڹz�F̉��k��K�S�-��ٜ��O�ݨ;��!�#3̇bOo��%$4�S�-�zGG�'�;�z�oT̗�3'�� I <��y��ȁ��~?�m!�:��8�� I<��y��Ff��O��M�S7f�#6g��7�\�ώ��?�׃$�ا��ke{wt\��L^'� )��swk1f/��^ +�?b�zV��g��!��z�=[;x�"�Ϻ?I��/I/�O��eK��\�u�&/��O�W/Y�ّ#��>_��x:I}�}�W;~�3��I�۵u<��?�fy��I��vׄ�� +}�S/H��b�ZMꙒ��Fd��a��!��zO?p��?;�� :�P��2��#��z=O:�H�ԝwa��i�jf�פfǎ�?��^�_b�z=W3��G��~v��ѳ>��~'��d��l�n}�?I��1ղ�]:.n�f,��K�RG�S�O�ٻ��\ߤ?�b��RG�go�\��G�1x�إ�%�)�vul�vދ�ο��I�j��=2$~~'��կI/�O��oӣfpb_�?�v���2�=��O��e�4��O��I�S2]�H̞��?�6�$4�S�C��s"�a��ᴇT��F_#��6�$t�S�G��C��=?�m7�Lݚ��݇��k��K�S�G��{��޴ߵr�>:6g��o��[^�_b�w��s�G�1x��4�:�_��E�ݬŘ�?�ix*H��}��[7q��L_��H�l��>��&/��$��>�έ�.��sy��i?�\�g�fD��ڵ��%�)��\��tl�K�'�n��C�'��^�x)��L�#��i7�fK��݇��׃$�� +}�u<��zNd@��?�6��C��H��~vs��ׂ�� +}��<�w�#3Ru݇��ͦ��^��;��x2I}�}��\�o�3��֛��^��F��Z��k��K�S��.a��&/��V��t蹻�� /I�O�?�f�3�s~�� I��sg��b��ا��ղ��z.o:�� '��lȑ��/�V�$��>�ޫ��N��~$��ڃ��^��3<�v$��k�RCO>�ީ��t|�|C�?��&�L�w�#2{��d��O��OÏ��S�{�|��Î��T��O��$fjN��ٴߵ3vk�3c�v��I/�O��|ϡ��֒�D�� + +endstream endobj 1995 0 obj<>stream +��Adobed�� +$$''$$53335;;;;;;;;;; %% ## ((%%((22022;;;;;;;;;;��&V"��? + +3!1AQa"q�2��B#$R�b34r��C%�S��cs5��&D�TdE£t6�U�e��u��F'��Vfv��7GWgw��5!1AQaq"2��B#�R��3$b�r��CScs4�%��&5��D�T�dEU6te��u��F��Vfv��'7GWgw��?�Dli-q�φ�V��;�@:��F��4�e��2��Y��^��0��b��`L�i��H<7��y$� [_��IH��x�p� �S�s&7 �j{�Ĉ#�?�R��xN�%,�Qs@��ǸpKt�!�Ƣ=��T��8|��D33&��5��L7�2�mk�$� ��ld�q��|Jp��i��;8��{v��㔓:']��g��I)z��+0a��LH�Q�ՑbI��B��h$�D�A��sbkH;�� ]��PD��>��OX�0��N��o��t���v �h %�n�bHE+�4��?�"L:jʁٵ�p'i� ��'��A0N��b(e�z��F�w��3_%��;��q*!�VK]$��{|��;I�^'�J\�6^'� �`ꐳ�7��{�3�5��,�0Iì�ln�4��7��sOkg�yI8��n��$��yo�� +od��G<�i�wI?F�7��7v��F�&|��$R��#g��8?�@�'��$��G�m-�o L{��;�043�f$r�rI?F�q`�x��ߕ��G�]<{j�tI%?GO��ǔ��I|�J�� + +endstream endobj 1996 0 obj<>stream +��Adobed�� +$$''$$53335;;;;;;;;;; %% ## ((%%((22022;;;;;;;;;;��"��? + +3!1AQa"q�2��B#$R�b34r��C%�S��cs5��&D�TdE£t6�U�e��u��F'��Vfv��7GWgw��5!1AQaq"2��B#�R��3$b�r��CScs4�%��&5��D�T�dEU6te��u��F��Vfv��'7GWgw��?�C��O�S�� ;�!8p�@��3^�\ q��*̋08�;S�BR`i�y3�ԡ��|��m�Ё��JX�7��8H +�� +Gl�Z�L�X��$�5�;@.��u��$� �mh4�mgv�C��h��4m#S��P�� ϹN\��x�H650{xT��i4�� }�v�*&6�sA��cO��'_��Pi�r;��p斾>� ~� ��Jf�� N��Pm�Z ��(��tA�h��=�� X:Z /��=n`�s�q�� =�5�%��N�ɘ��/��<}�r��D7��"A��;H��#��k��Q1�k��vH�ocx$�I:��J�cH?K��^�4��xK�pHopx�S0�$48ju<� +�m��p�D�C��D(��MR��I~��s˝��8m�A��i.8�A�R��-��Д�t�L�5(a|��k4 i#�⒖%��'N��w��$��S4V�"8 )Mk��4|$'��y�)�jG y$[Yݬ��u"�!�H��7yqA��S�0G�!�^) � L�.��@ >��Hs]�J��\�}òG`�{{��%(��I��?�TA�\��9��k�_��C{���A!��S��hTnֈ/�� >�g�!E�4�l�Bj��K�?�O[�&\�x��h-o�{Ip�ƺ��`qhv��Ӧ�f;��C��Y�I��,n��:p��K��$�2��IJkXv�]��!8 �S�INP�8�h#�"��d�k��1h�F�4��ˈ�r��=� ��H�l j`��(�w��i��j@��82TLm��;��ƞ )E��N�1�R��wׄ��-|4|#^��t��: �O#B��v�A�� +/5�{gBT�t�_��jz��2�|��'Ah{x&k�K�5�x��s�C�0|4%&�7�1��J_'�xx��H��cw�.Iӄ��]��p�v�%��A��`�JSZô� _ �n��bJpچ�ƃAI�wk$;]F��cF�50;�� \@�|��o��D�aS��EK�v�O�G�R��i��cn�4p��,��4�IJ..�u��Ɛ~�#��'ik��;��a�Hhp��y۵��D��Qy� #�:��O�S�� ;�!8p�@��3^�\ q��$��Z��)04鼙��jP��>��6�h@�G��%,K��rN�$e��Å#�I-B�h�DpS ��ޒ'�W��I"��k!�X&$�K$�N�gO&l @'M +��l�h<!2��^Ii#�&�v�v?� �[��;�� w�(>�wa�H�Uۄ�N��iԐ$��1'�7v1O�a*D��_��ӎ��a;�v*@Vlx�x )E��@5��l��A3i�@d�0��:� l>��8��w�}6� +g?��v��c��,��`�v�$�@�T��C��KBLF�:'�s�tO��Q>�\�\�%2 ��@x��{�,��&�\��%�=�RkX7��L5�ӬyI��@'Q3��6��L6D��K�/$�4��H;s�Ǆŭ��D;ʔo�� E*��\'_�[[��HI��'�K��0�"X��/�~��i�e d��ݻ +6< �<��ւ�à�S��6Z� ��4� 2H�x�R�6LDwIIJy�;�>�y3��i�X�R��q�H��wQ0[�dF�k�J^�C[!�Dy%�&#p��9�:'�x(�M�v�x��ÉF��{G�o �s�� 7kjV8�2��&wl�H� uIK��kd:�$�$�n�y�;GD��S �q ��P ag��<ự�rx�D��3�S�~�H{��JdZ��H�d��8Q h 4��icu�&x=��fd�Ԕ��dv��+cO��Ø��<�p��ο|RS(l�>�HEI�{��F�S��|�a}�aѦ��'l� ��>%<�P�q�Nv��J\�}�\��q%��G��ky��T��I�]��a�⒙n�&��Ƕu�q;̃��)3xf��|%*gp �p�;(��xR.:L�:@�5� ��׷t��` ��{8i�"H�ؕ��{�@O>�$�H�1��Z;�L��G +!�A$�8�� ,n�$��p�̂�Z��;`,��x�S�`Li�<��s��N�y��Je �'٩ȩ4�q�H��Jv��L"O�:4Ӕ��햁��għ��`�<�I�nѴ�K��˃�R�$��@��o?��ycI:뷺b6�<|RS-��$��Σ�N'y�x0e&o�sO��L��e�{��O +A��I�H&��}��8�?�O`� ;��I��y�v� ��d��>��kC'q"I��D5�$��|��D��N��Z KRS'l�y��`� �<��cws��I��:�A�IL��5 y&�� ~i�IN��<`�)�I��F�r��6��=C Lǘ 8m�6�`q)sa��p|B +]Ėh�ޢ�}��R/,i']v�LC��Je��33�?��vI��2��ᚎc��T��6=�t��p w� �H8��3��87O�^��R'�\'�I��x|�#�bT\O#Y��<��u#��ޒ�-hd�$I2bu(��㏂@4��ȓ<��@32D jJd퀲;A��O��1��Ps��n�v�i8{}�_�>))�6x�f�"��=��#_�)�Z�>E0�>�0��NRS�Z��y��G�d �8�' �F�L%.l>�.�AK��#��\O��*E�$��x�c��IL��f|Tg�c�:��8��A��3Q�q>��3�ǸN��_��<)&| pT��ۺJD�0��=��4��$~lJ��k=� '�u�N�v��Rŭ�ĉ&LN��ր �@q�H�7Yg��8hfAh�-IL��A�h<~)��&4�j|9��'o��%2��ԁ�T�G��k�%;K@�Ȧ'�i�Jbv�@�~�3�S��0G`$�h� �ĥ͇��)wX}�{�z��H��u��1v>))��\@`�ϊ��{gQ�'��<2�7�j9�'�QR�w�� Ӳ��=��P'� ��σ� +�\�`@�>�{wIH�p��'�S��$�͉Qq<�g�d��IԎ�zJX��$ɉ�p��@h�> ��"L�{' ��-%�)��<��?�Vƞ�Aχ1��y��~ ��P��}��<��H�?4�~$�ih0��$�`ãM9IKE~_F��O�}�SVƎt:@(�ct��<��3 ��N��;{A$抗��1!�#];(��4��k�q��>>0��i�5�^R�F�6�5$ׂ�;;�D @�y� I&'��EKz��H0;y�H��']IM0捺�ξ)�� 䚥9ߣ��F�Ss�H��~��Ė�3�N$��9��Jd�=�0t>i�KZu<�3��@0e-][` u��Mt�{��R.�sK~��S�!�A��$x�s�虎ɶMp�G�*U1��xi�RSVƎt:@(�ct��<��3 ��N��;{A$撗��1!�#];(��4��k�q��>>0��i�5�^R�F�6�5$ׂ�;;�D @�y� I&'��EKz��H0;y�H��']IM0捺�ξ)�� 䚥9ߣ��F�Ss�H��~��Ė�3�N$��9��Jd�=�0t>i�KZu<�3��@0e-][` u��Mt�{��R.�sK~��S�!�A��$x�s�虎ɶMp�G�*U1��xi�RSVƎt:@(�ct��<��3 ��N��;{A$撗��1!�#];(��4��k�q��>>0��i�5�^R�F�6�5$ׂ�;;�D @�y� I&'��EKz��H0;y�H��']IM0捺�ξ)�� 䚥9ߣ��F�Ss�H��~��Ė�3�N$��9��Jd�=�0t>i�KZu<�3��@0e-][` u��Mt�{��R.�sK~��S�!�A��$x�s�虎ɶMp�G�*U1��xi�RSVƎt:@(�ct��<��3 ��N��;{A$撗��1!�#];(��4��k�q��>>0��i�5�^R�F�6�5$ׂ�;;�D @�y� I&'��EKz��H0;y�H��']IM0捺�ξ)�� 䚥9ߣ��F�Ss�H��~��Ė�3�N$��9��Jd�=�0t>i�KZu<�3��@0e-][` u��Mt�{��R.�sK~��S�!�A��$x�s�虎ɶMp�G�*U1��xi�RSVƎt:@(�ct��<��3 ��N��;{A$撗��1!�#];(��4��k�q��>>0��i�5�^R�F�6�5$ׂ�;;�D @�y� I&'��EKz��H0;y�H��']IM0捺�ξ)�� 䚥9ߣ��F�Ss�H��~��Ė�3�N$��9��Jd�=�0t>i�KZu<�3��@0e-][` u��Mt�{��R.�sK~��S�!�A��$x�s�虎ɶMp�G�*U1��xi�RSVƎt:@(�ct��<��3 ��N��;{A$撗��1!�#];(��4��k�q��>>0��i�5�^R�F�6�5$ׂ�;;�D @�y� I&'��EKo�� &�|?7�$�4L��J��#i @��Ma-vѦƂ}Ra{�� Ls��r\Z��u:��Kn�2K�3j;lq>�#C� ��3�:x��th�9�T�$s_r��^��%I�$��Ө��@:?4��$�{��5ԁ~ ��LI: AK�<4O!��Ns��D�6�2;�w$hc�JS�� mn�c]tL�f�D��0.�{q=�R��K�'��6v��ԉ�L�|k��NR�<��Ϛxs\��x$3B`@�+\��us��HIL@sD��o:N��6��$��mlh ��&��:蒗%Šn�'S��$�1棶��"41��:�JH\wF�lӞuO G0u�!�!��~BT��K�1 :��AK��h�$��L}]΂N��9��]H'��$ē��s�D��L��8��I�h�#�'q�F��:��;xy��v��5�D�q6hdH��'��%.M��|�JSgn�~A �H�d��6��0 @��,�q L��5��hG�Q�C4& 0R�ͭ��W;�t��4L��J��#i @��Ma-vѦƂ}Ra{�� Ls��)r\Z��u:��Kn�2K�3j;lq>�#C� ��3�:x��th�9�T�$s_r��^��%I�$��Ө��@:?4��$�{��5ԁ~ ��LI: AK�<4O!��Ns��D�6�2;�w$hc�JS�� mn�c]tL�f�D��0.�{q=�R��K�'��6v��ԉ�L�|k��NR�<��Ϛxs\��x$3B`@�+\��us��HIL@sD��o:N��6��$��mlh ��&��:蒗%Šn�'S��$�1棶��"41��:�JH\wF�lӞuO G0u�!�!��~BT��K�1 :��AK��h�$��L}]΂N��9��]H'��$ē��s�D��L��8��I�h�#�'q�F��:��;xy��v��5�D�q6hdH��'��%.M��|�JSgn�~A �H�d��6��0 @��,�q L��5��hG�Q�C4& 0R�ͭ��W;�t��4L��J��#i @��Ma-vѦƂ}Ra{�� Ls��)r\Z��u:��Kn�2K�3j;lq>�#C� ��3�:x��th�9�T�$s_r��^��%I�$��Ө��@:?4��$�{��5ԁ~ ��LI: AK�<4O!��Ns��D�6�2;�w$hc�JS�� mn�c]tL�f�D��0.�{q=�R��K�'��6v��ԉ�L�|k��NR�<��Ϛxs\��x$3B`@�+\��us��HIL@sD��o:N��6��$��mlh ��&��:蒗%Šn�'S��$�1棶��"41��:�JH\wF�lӞuO G0u�!�!��~BT��K�1 :��AL��F{��ByI1s۾;��Qc��t��s��:t��R�f�8�c_��~ �Oy��@��*fH�7`�>>)��3��K��$�I��C ��%( p�&:(8��"8�8��4ӄ�HтLs�5Lv�� k� |t?�S�ˣ�Ϛ��/3Ğ�)g �@��ۯ��'7�Ƣ)mpy�'a�IK �q��-'ݯa�MPdɞ݁�)��^��}��p i&Ij|B}E��A�q�ܢ睌&I#B~!2s۾;��Qc��t��s��:t��R�f�8�c_��~ �Oy��@��*fH�7`�>>)��3��K��$�I��C ��%( p�&:(8��"8�8��4ӄ�HтLs�5Lv�� k� |t?�S�ˣ�Ϛ��/3Ğ�)g �@��ۯ��'7�Ƣ)mpy�'a�IK �q��-'ݯa�MPdɞ݁�)��^��}��p i&Ij|B}E��A�q�ܢ睌&I#B~!2s۾;��Qc��t��s��:t��R�f�8�c_��~ �Oy��@��*fH�7`�>>)��3��K��$�I��C ��%( p�&:(8��"8�8��4ӄ�HтLs�5Lv�� k� |t?�S�ˣ�Ϛ��/3Ğ�)g �@��ۯ��'7�Ƣ)mpy�'a�IK �q��-'ݯa�MPdɞ݁�)��^��}��p i&Ij|B}E��A�q�ܢ睌&I#B~!2s۾;��Qc��t��s��:t��R�f�8�c_��~ �Oy��@��*fH�7`�>>)��3��K��$�I��C ��%( p�&:(8��"8�8��4ӄ�HтLs�5Lv�� k� |t?�S�ˣ�Ϛ��/3Ğ�)g �@��ۯ��'7�Ƣ)mpy�'a�IK �q��-'ݯa�MPdɞ݁�)��^��}��p i&Ij|B}E��A�q�ܢ睌&I#B~!2s۾;��Qc��t��s��:t��R�f�8�c_��~ �Oy��@��*fH�7`�>>)��3��K��$�I��C ��%( p�&:(8��"8�8��4ӄ�HтLs�5Lv�� k� |t?�S�ˣ�Ϛ��/3Ğ�)g �@��ۯ��'7�Ƣ)mpy�'a�IK �q��-'ݯa�MPdɞ݁�)��^��}��p i&Ij|B}E��A�q�ܢ睌&I#B~!2s۾;��Qc��t��s��:t��R�f�8�c_��~ �Oy��@��*fH�7`�>>)��3��K��$�I��C ��%2��$��S`�4)��gM|e3�hp�> 뱄=��:bt�C3� �c��t�k��)z�|7�2u��-䈎<� )gO��>IK��lnt��d8�<��N6�.��IIZP4?�� -�!3ZK�h�G��@" 9�y��ϔ�ɦl"G�N�4��x�ވ.y~w�IK�[<� ��ݼƣS�(�7�o�]?�}�d��<$�3ԇn�;��w�h'S��E��I�5��Rk��IK��v��⶝��x��o2?3X�q�k��~��%(X=F��v9��_L��r��z�a�s��$�!Θ�9��f��;�!��^�� {�Ky"#�=�JY��w��R��y;��#O �#Ӎ�K��RV�� w� jKd�L֒�Z4��n�A%2�&��#�i��ӧ�8g��#�7��@��Rƶ��'Bdy�'�o1�� +%��O�lY0;� )L�!ۄk��8��Z ��Qhn�G��~0��$��R�{��]�;�2��n��=��D̏��5r��6��IJQ�#A❎qt��S>�܁��A3�� )�s�'Nt3>��<��Hv��◨7�y�'_��R�H��p��qt�]�䔻l��A�+N�&C��(��h��5C�(��25��V�$~/t"��IL�I��H�f�$~t��N�q�A'��M��~D��Й~ ��j5=��c}&�u��L��JS=Hv��N'y��u?w�Z��#_�&�I;��?$��q�o��+iۻw�n<�#�5�G��R��l�x�c�A�5�Ϲ��w x�'��L�<|Jb�ӝϸ&m�#�� `��"8�3�$��]?x�%.�!��xJǓ�I��4� +2=8�$��%%h @�� +'x��L��i.գI�K��S/�o>R;&��:|ӆ{�`I� 뱄=��:bt�C3� �c��t�k��)z�|7�2u��-䈎<� )gO��>IK��lnt��d8�<��N6�.��IIZP4?�� -�!3ZK�h�G��@" 9�y��ϔ�ɦl"G�N�4��x�ވ.y~w�IK�[<� ��ݼƣS�(�7�o�]?�}�d��<$�3ԇn�;��w�h'S��E��I�5��Rk��IK��v��⶝��x��o2?3X�q�k��~��%(X=F��v9��_L��r��z�a�s��$�!Θ�9��f��;�!��^�� {�Ky"#�=�JY��w��R��y;��#O �#Ӎ�K��S-��vI-��BI)N��d@�c�I��g��8�jI=%�:[虔�4�ѩ�d��i�nԒ5�Rk�$ȃ� ݥH9�Ls��$��v�<�'��;H5��>#]�p�wӠ�eg� >�I�t�i0g�'.|h"u� + �8Q�K�q�D��qs�� O��bA.��qЉy�� wk��"��~�Q��S��x�쬶�p�<ӆ�PA�甔��Zt�|��'٠�>I��[��Ƃ;&��%3i��ƦGȡ=��h04�Qe��>��Gp"�-i2�F��T��5��Rk��h'Y�(��%��AOIy�Ǻ&e%#�.tjgY>)�F�l�$�F�� 2 �wiR{��1 %c�ݸ$ �m�N� b&fO�H�a�#]Ĵ�4��Y��O�|C ]'�DL� ˟ȝ|��k��jR��`G� $�m�\� u��(~��K�a ú\t"F�aGc}By��r�@�G0��?��F�({+-�f5�4��y$��%)��?$�s��h;ϒf8��ɀ�ywIL�fe��(O`!�� 5�Yw�O�x��Ƞ��ZL��w�.�lo�D f��f��S�^C��IH�cK��O�v��I#Q��&��L�:��T��D�?�BIX��n��x|��X�ə��5�w�q-: !&V}2��I��|�r�ƀ�'_ ��5��7�H )�w;H�t�J��$�H0��Q��P�gv��*\��5�)9��;��`� c�8h�I0�yIJk��O�� \�}��䙎%��#]�p�wӠ�eg� >�I�t�i0g�'.|h"u� + �8Q�K�q�D��qs�� O��bA.��qЉy�� wk��"��~�Q��S��x�쬶�p�<ӆ�PA�甔��Zt�|��'٠�>I��[��Ƃ;&��%3i��ƦGȡ=��h04�Qe��>��Gp"�-i2�F��T��5��Rk��h'Y�(��%��AOIy�Ǻ&e%#�.tjgY>)�F�l�$�F�� 2 �wiR{��1 %c�ݸ$ �m�N� b&fO�H�a�#]Ĵ�4��Y��O�|C/w�RQ��㿂I)A��I~��s˝��8m�A��i.8�A�R��-��Д�t�L�5(a|��k4 i#�⒖%��'N��w��$��S4V�"8 )Mk��4|$'��y�)�jG y$[Yݬ��u"�!�H��7yqA��S�0G�!�^) � L�.��@ >��Hs]�J��\�}òG`�{{��%(��I��?�TA�\��9��k�_��C{���A!��S��hTnֈ/�� >�g�!E�4�l�Bj��K�?�O[�&\�x��h-o�{Ip�ƺ��`qhv��Ӧ�f;��C��Y�I��,n��:p��K��$�2��IJkXv�]��!8 �S�INP�8�h#�"��d�k��1h�F�4��ˈ�r��=� ��H�l j`��(�w��i��j@��82TLm��;��ƞ )E��N�1�R��wׄ��-|4|#^��t��: �O#B��v�A�� +/5�{gBT�t�_��jz��2�|��'Ah{x&k�K�5�x��s�C�0|4%&�7�1��J_'�xx��H��cw�.Iӄ��]��p�v�%��A��`�JSZô� _ �n��bJpچ�ƃAI�wk$;]F��cF�50;�� \@�|��o��D�aS��EK�v�O�G�R��i��cn�4p��,��4�IJ..�u��Ɛ~�#��'ik��;��a�Hhp��y۵��D��Qy� #�:��O�S�� ;�!8p�@��3^�\ q��$��Z��)04鼙��jP��>��6�h@�G��%,K��rN�$e��Å#�I-B�h�DpR��@h:�HNu��S��6�4�H��Y!��4ELC6��(n��ܧ.`�tC|�R$�=� +*]�}�?ڐ>�N�v��d��`��1��JQqt��T�4��8sK_ ׄ�G��%3�CC��S�Ш6ݭ_�@}�: ϴB��@i�Є�,-��ڞ�0L��8� Æ�Z� ��t!%,��L I��M��w�R��)��B�<>))bX��K�t� +,.�w�)�Ih:e3E`#��ְ��A��Bp��6��q��G�E��Q�*bѴ�L�iCw�$>�9s{�� @��QR��ԁ�5�pd��۵��;$v��<R��|c��@1�� ÚZ�h�F�%�8$7�<�)�t5:�F�A��h��x��!�}�^jH�΄&�`�h$�S��e��ǈN6�� L״�k�� )`��j`�hJL :o&c��0�O��M��4��IK��\�� Yat��H�KAЃ)�+��e>/�oǷ��>��H�f�i� �<�� I�A�� B��&"ZAL��@H�I��9ݏ��b��i�"�J��w��@v�.��-��u$ $�LI� ݌C��%��E +�,D��?A��㲄�XN�݇��JQakApa� ~)��-l�Lǚg�$L�{G�o �s�� 7kjV8�2��&wl�H� uIK��kd:�$�$�n�y�;GD��S �q ��P ag��<ự�rx�D��3�R%�Ȑ<��vP�K �۰�R�c��0��IJ,-h.:��>��e�� L�$��y�)a��GqĔ��s��鷐S9��5+wd�Gu��A$j�� 5�G�Z`7 �<Ý��|��j�g� )�l8�Dj�(��If�)4�r�n�-��Z��0D�Բa��`��M,�:��D�ms�p3��6H"5�XY�$�O�X9s�H��I�`�"C�jY0�CN�LI�&�I�DΞ*L��N�?a0��xBHe,��G�M �!��sHw1@�*P}��þ�%�� p�~)mnө I'�bO�n�b� .&�(T��+��$��2w$�1:�CZ�H q�� X�dI�d��%2v�Y�� +��y�9��7w;{4�=��*3�1�Gd�N� �`�L��8� EJ��c�'N�/��}@��>�8*Ms��t�u��%"p�t��Nw�Ȓ?6%E��5��Ϻ�'R;L}�)bֆN�D�&'Qk@I�8�$K��3��43 �@��N� �?��Xx5>��oa擇��u��⒙Cg��j@�*M#�@��5��x��S��4�%1;e�m?G��)�z�@�#�0p۴m��R��,>�=�?�E��[��^X�N����;��p. 0fg�FF=��쓉�dI��5��(�S;�l{��E��@�qq�g��I�p0 n�n��N�O��)�N��G�Ĩ��F�ݲy�Y$�Gi��%,Z��H�d��8Q h 4��icu�&x=��fd�Ԕ��dv��+cO��Ø��<�p��ο|RS(l�>�HEI�{��F�S��|�a}�aѦ��'l� ��>%<�P�q�Nv��J\�}�\��q%��G��ky��T��I�]��a�⒙n�&��Ƕu�q;̃��)3xf��|%*gp �p�;(��xR.:L�:@�5� ��׷t��` ��{8i�"H�ؕ��{�@O>�$�H�1��Z;�L��G +!�A$�8�� ,n�$��p�̂�Z��;`,��x�S�`Li�<��s��N�y��Je �'٩ȩ4�q�H��Jv��L"O�:4Ӕ��햁��għ��`�<�I�nѴ�K��˃�R�$��@��o?��ycI:뷺b6�<|RS-��$��Σ�N'y�x0e&o�sO��L��e�{��O +A��I�H&��}��8�?�O`� ;��I��y�v� ��d��>��kC'q"I��D5�$��|��D��N��Z KRS'l�y��`� �<��cws��I��:�A�IL��5 y&�� ~i�IN��<`�)�I��F�r��6��=C Lǘ 8m�6�`q)sa��p|B +]Ėh�ޢ�}��R/,i']v�LC��Je��33�?��vI��2��ᚎc��T��6=�t��p w� �H8��3��87O�^��R'�\'�I��x|�#�bT\O#Y��<��u#��ޒ�-hd�$I2bu(��㏂@4��ȓ<��@32D jJd퀲;A��O��1��Ps��n�v�i8{}�_�>))�6x�f�"��=��#_�)�Z�>E0�>�0��NRR�_�џ��G��A$��#��=�t6�{L��0>*%ߣ�q%�蓉/l�|{⒙9�Ob�m�֝O�5L��KWV�H�|B*S]2��y��ߤ4��Hz�{q��k1s �*b��O�0��\胺&c�m�\=��ʕLn�x�ŕ��#��=�t6�{L��0>*%ߣ�q%�蓉/l�|{⒙9�Ob�m�֝O�5L��KWV�H�|B*S]2��y��ߤ4��Hz�{q��k1s �*b��O�0��\胺&c�m�\=��ʕLn�x�ŕ��#��=�t6�{L��0>*%ߣ�q%�蓉/l�|{⒙9�Ob�m�֝O�5L��KWV�H�|B*S]2��y��ߤ4��Hz�{q��k1s �*b��O�0��\胺&c�m�\=��ʕLn�x�ŕ��#��=�t6�{L��0>*%ߣ�q%�蓉/l�|{⒙9�Ob�m�֝O�5L��KWV�H�|B*S]2��y��ߤ4��Hz�{q��k1s �*b��O�0��\胺&c�m�\=��ʕLn�x�ŕ��#��=�t6�{L��0>*%ߣ�q%�蓉/l�|{⒙9�Ob�m�֝O�5L��KWV�H�|B*S]2��y��ߤ4��Hz�{q��k1s �*b��O�0��\胺&c�m�\=��ʕLn�x�ŕ��#��=�t6�{L��0~i��[��]9�١�#�c�L�l��Ot��6� ��)M��"u�'_��5��,�h53��8�F�И$�J�6�\��S�3;�Γ�*gx��5��F�=�I��:��1κ$��qh�I��-�D�.�y��ȍF�&s8μ�⒒Ѡ4�S��}�pC�kAxnߐ�'��CN��PR��Z& �|��Ws��Ngp�R �&;��1$�%.\��<�4)9�6�El�4�� l��=��)N�x��ݵ�A�u�3�M�;�?$��IK�a.�/��ۿ��H;R'Y2u�M��k�i��s�>Z�Tl�� L�skpq��]!%1�3��:��w��H;�|�XK]�i��T�^��J\��4�N�|ۤL��Lǚ��O��jg0��Ξ))!q��Ny�<�I��܇<�� R{�.l�4�?/�//%�`��1�w: :��w u @��c�8�N�PR�� sA2��g�tQ&ͣ@L��꒔��>~i��[��]9�١�#�c�L�l��Ot��6� ��)M��"u�'_��5��,�h53��8�F�И$�J�6�\��S�3;�Γ�*gx��5��F�=�I��:��1κ$��qh�I��-�D�.�y��ȍF�&s8μ�⒒Ѡ4�S��}�pC�kAxnߐ�'��CN��PR��Z& �|��Ws��Ngp�R �&;��1$�%.\��<�4)9�6�El�4�� l��=��)N�x��ݵ�A�u�3�M�;�?$��IK�a.�/��ۿ��H;R'Y2u�M��k�i��s�>Z�Tl�� L�skpq��]!%1�3��:��w��H;�|�XK]�i��T�^��J\��4�N�|ۤL��Lǚ��O��jg0��Ξ))!q��Ny�<�I��܇<�� R{�.l�4�?/�//%�`��1�w: :��w u @��c�8�N�PR�� sA2��g�tQ&ͣ@L��꒔��>~i��[��]9�١�#�c�L�l��Ot��6� ��)M��"u�'_��5��,�h53��8�F�И$�J�6�\��S�3;�Γ�*gx��5��F�=�I��:��1κ$��qh�I��-�D�.�y��ȍF�&s8μ�⒒Ѡ4�S��}�pC�kAxnߐ�'��CN��PS-�ў�'�ОRAL\��T�ٹ��D�h�;��S�e��&�'�芙�%��Ϗ�}d��&iC��û�I��1�� +[\F�I��rRR�q�d�t��hk�I�k�rA�2g�`�v4��8��ׄ�,A�`ir{\ZI�F��Qc��A:i�i�(��c �HП�AL��ٹ��D�h�;��S�e��&�'�芙�%��Ϗ�}d��&iC��û�I��1�� +[\F�I��rRR�q�d�t��hk�I�k�rA�2g�`�v4��8��ׄ�,A�`ir{\ZI�F��Qc��A:i�i�(��c �HП�AL��ٹ��D�h�;��S�e��&�'�芙�%��Ϗ�}d��&iC��û�I��1�� +[\F�I��rRR�q�d�t��hk�I�k�rA�2g�`�v4��8��ׄ�,A�`ir{\ZI�F��Qc��A:i�i�(��c �HП�AL��ٹ��D�h�;��S�e��&�'�芙�%��Ϗ�}d��&iC��û�I��1�� +[\F�I��rRR�q�d�t��hk�I�k�rA�2g�`�v4��8��ׄ�,A�`ir{\ZI�F��Qc��A:i�i�(��c �HП�AL��ٹ��D�h�;��S�e��&�'�芙�%��Ϗ�}d��&iC��û�I��1�� +[\F�I��rRR�q�d�t��hk�I�k�rA�2g�`�v4��8��ׄ�,A�`ir{\ZI�F��Qc��A:i�i�(��c �HП�AL��ٹ��D�h�;��S�e��&�'�芙�%��Ϗ�}d��&�܁��A3�� )�s�'Nt3>��<��Hv��◨7�y�'_��R�H��p��qt�]�䔻l��A�+N�&C��(��h��5C�(��25��V�$~/t"��IL�I��H�f�$~t��N�q�A'��M��~D��Й~ ��j5=��c}&�u��L��JS=Hv��N'y��u?w�Z��#_�&�I;��?$��q�o��+iۻw�n<�#�5�G��R��l�x�c�A�5�Ϲ��w x�'��L�<|Jb�ӝϸ&m�#�� `��"8�3�$��]?x�%.�!��xJǓ�I��4� +2=8�$��%%h @�� +'x��L��i.գI�K��S/�o>R;&��:|ӆ{�`I� 뱄=��:bt�C3� �c��t�k��)z�|7�2u��-䈎<� )gO��>IK��lnt��d8�<��N6�.��IIZP4?�� -�!3ZK�h�G��@" 9�y��ϔ�ɦl"G�N�4��x�ވ.y~w�IK�[<� ��ݼƣS�(�7�o�]?�}�d��<$�3ԇn�;��w�h'S��E��I�5��Rk��IK��v��⶝��x��o2?3X�q�k��~��%(X=F��v9��_L��r��z�a�s��$�!Θ�9��f��;�!��^�� {�Ky"#�=�JY��w��R��y;��#O �#Ӎ�K��RV�� w� jKd�L֒�Z4��n�A%2�&��#�i��ӧ�8g��#�7��@��Rƶ��'Bdy�'�o1�� +%��O�lY0;� )L�!ۄk��8��Z ��Qhn�G��~0��$��R�{��]�;�2��n��=��D̏��5r��6��IJQ�#A❎qt��S>�܁��A3�� )�s�'Nt3>��<��Hv��◨7�y�'_��R�H��p��qt�]�䔻l��A�+N�&C��(��h��a��Kk��JS��5��Rk��h'Y�(��%��AOIy�Ǻ&e#�.tjgY>)�F�l�$�F�� 2 �wiR{��1 %c�ݸ$ �m�N� b&fO�H�a�#]Ĵ�4��Y��O�|C ]'�DL� ˟ȝ|��k��jR��`G� $�m�\� u��(~��K�a ú\t"F�aGc}By��r�@�G0��?��F�({+-�f5�4��y$��%)��?$�s��h;ϒf8��ɀ�ywIL�fe��(O`!�� 5�Yw�O�x��Ƞ��ZL��w�.�lo�D f��f��S�^C��IH�cK��O�v��I#Q��&��L�:��T��D�?�BIX��n��x|��X�ə��5�w�q-: !&V}2��I��|�r�ƀ�'_ ��5��7�H )�w;H�t�J��$�H0��Q��P�gv��*\��5�)9��;��`� c�8h�I0�yIJk��O�� \�}��䙎%��#]�p�wӠ�eg� >�I�t�i0g�'.|h"u� + �8Q�K�q�D��qs�� O��bA.��qЉy�� wk��"��~�Q��S��x�쬶�p�<ӆ�PA�甔��Zt�|��'٠�>I��[��Ƃ;&��%3i��ƦGȡ=��h04�Qe��>��Gp"�-i2�F��T��5��Rk��h'Y�(��%��AOIy�Ǻ&e%#�.tjgY>)�F�l�$�F�� 2 �wiR{��1 %c�ݸ$ �m�N� b&fO�H�a�#]Ĵ�4��Y��O�|C ]'�DL� ˟ȝ|��k��jR��`G� $�m�\� u��(~��K�a ú\t"F�aGc}By��r�@�G0��?��F�({+-�f5�4��y$��%)��?$�s��h;ϒf8��ɀ�ywIL�fe��(O`!�� 5�Yw�O�x��Ƞ��ZL��w�.�lo�D f��f��S�^C��IH�cK��O�v��I#Q��&��L�:��T��D�?�BIX��n��x|��X�ə��5�w�q-: !&V}2��Ô�}1�Fx��JPt�_��jz��2�|��'Ah{x&k�K�5�x��s�C�0|4%&�7�1��J_'�xx��H��cw�.Iӄ��]��p�v�%��A��`�JSZô� _ �n��bJpچ�ƃAI�wk$;]F��cF�50;�� \@�|��o��D�aS��EK�v�O�G�R��i��cn�4p��,��4�IJ..�u��Ɛ~�#��'ik��;��a�Hhp��y۵��D��Qy� #�:��O�S�� ;�!8p�@��3^�\ q��$��Z��)04鼙��jP��>��6�h@�G��%,K��rN�$e��Å#�I-B�h�DpR��@h:�HNu��S��6�4�H��Y!��4ELC6��(n��ܧ.`�tC|�R$�=� +*]�}�?ڐ>�N�v��d��`��1��JQqt��T�4��8sK_ ׄ�G��%3�CC��S�Ш6ݭ_�@}�: ϴB��@i�Є�,-��ڞ�0L��8� Æ�Z� ��t!%,��L I��M��w�R��)��B�<>))bX��K�t� +,.�w�)�Ih:e3E`#��ְ��A��Bp��6��q��G�E��Q�*bѴ�L�iCw�$>�9s{�� @��QR��ԁ�5�pd��۵��;$v��<R��|c��@1�� ÚZ�h�F�%�8$7�<�)�t5:�F�A��h��x��!�}�^jH�΄&�`�h$�S��e��ǈN6�� L״�k�� )`��j`�hJL :o&c��0�O��M��4��IK��\�� Yat��H�KAЃ)�+��h��u<Ĕ� ��<�-��Hv��Sƍ�j`w�J�� )˘#��/��o��|� �p��Ӄ%D�ݮh>��#�X=��i��\]$��*� �.G}xN��G�5�/��!��wILà��4* �kD��q�3��PG�t!5KKA%��.w�� ~� ��Jf�� N��Pm�Z ��(��tA�h��=�� X:Z /��=n`�s�q�� =�5�%��N�ɘ��/��<�{G�o �s�� 7kjV8�2��&wl�H� uIK��kd:�$�$�n�y�;GD��S �q ��P ag��<ự�rx�D��3�R%�Ȑ<��vP�K �۰�R�c��0��IJ,-h.:��>��e�� L�$��y�)a��GqĔ��s��鷐S9��5+wd�Gu��A$j�� 5�G�Z`7 �<Ý��|��j�g� )�l8�Dj�(��If�)4�r�n�-��Z��0D�Բa��`��M,�:��D�ms�p3��6H"5�XY�$�O�X9s�H��I�`�"C�jY0�CN�LI�&�I�DΞ*L��N�?a0��xBHe,��G�M �!��sHw1@�*P}��þ�%�� p�~)mnө I'�bO�n�b� .&�(T�`2$�9��%��wv��wa�H�Uۄ�N��iԐ$��1'�7v1O�a*D��_��ӎ��a;�v*@Vlx�x )E��@5��l��A3i�@d�0��:� l>��8��w�}6� +g?��v��c��,��`�v�$�@�T��C��KBLF�:'�s�tO��Q>�\�\�%2 ��@x��{�,��&�\��%�=�RkX7��L5�ӬyI��@'Q3��6��L6D��K�/$�4��H;s�Ǆŭ��D;ʔo�� E*��\'_�[[��HI��'�K��0�>G�$��> $�E��ĉ&LN��ր �@q�H�7Yg��8hfAh�-IL��A�h<~)��&4�j|9��'o��%2��ԁ�T�G��k�%;K@�Ȧ'�i�Jbv�@�~�3�S��0G`$�h� �ĥ͇��)wX}�{�z��H��u��1v>))��\@`�ϊ��{gQ�'��<2�7�j9�'�QR�w�� Ӳ��=��P'� ��σ� +�\�`@�>�{wIH�p��'�S��$�͉Qq<�g�d��IԎ�zJX��$ɉ�p��@h�> ��"L�{' ��-%�)��<��?�Vƞ�Aχ1��y��~ ��P��}��<��H�?4�~$�ih0��$�`ãM9ILN�hO��|Jy��&��6�A08��|�>!.�K�p��Qq>��4��{�!�a��%2��L��Q�яl�;$�w��Rf��G1��J*T��:vQ|�;��\t��t��Rk��ۯn�)��.�� +pӼ>D��*.'��l��}�I:��c�IK�2w$�1:�CZ�H q�� X�dI�d��%2v�Y�� +��y�9��7w;{4�=��*3�1�Gd�N� �`�L��8� EJ��c�'N�/��}@��>�8*Ms��t�u��%"p�t��Nw�Ȓ?6%E��5��Ϻ�'R;L}�)bֆN�D�&'Qk@I�8�$K��3��43 �@��N� �?��Xx5>��oa擇��u��⒙Cg��j@�*M#�@��5��x��S��4�%1;e�m?G��)�z�@�#�0p۴m��R��,>�=�?�E��[��^X�N����;��p. 0fg�FF=��쓉�dI��5��(�S;�l{��E��@�qq�g��I�p0 n�n��N�O��)�N��G�Ĩ��F�ݲy�Y$�Gi��%,Z��H�d��8Q h 4��icu�&x=��fd�Ԕ��dv��+cO��Ø��<�p��ο|RS(l�>�HEI�{��F�S��|�a}�aѦ��'l� ��>%<�P�q�Nv��J\�}�\��q%��G��ky��T��I�]��a�⒙n�&��Ƕu�q;̃��)3xf��|%*gp �p�;(��xR.:L�:@�5� ��׷t��` ��{8i�"H�ؕ��{�@O>�$�H�1��Z;�L��G +!�A$�8�� ,n�$��p�̂�Z��;`,��x�S�`Li�<��s��N�y��Je �'٩ȩ4�q�H��Jv��L"O�:4Ӕ��W��g��$��g�I0elh�C��7Oh�� 12�/ӷ�A>h�N�"5Ӳ�kI>ᯆ�� +Z��#^5�)�kn�PBMx.p��$@�$ �Rt�b{x�T��=�4��D�q��uԔ�h۩��|L�>I�S��9 $nA�78��?��w��-�Il�:$�K�-��Nx�؃C�t��S��S8��R�ն'_��L��~"�74�� 50<��pc�(%��A�:h��;C0��G�)l�: �d�tp@�r�S��%1elh�C��7Oh�� 12�/ӷ�A>i)N�"5Ӳ�kI>ᯆ�� +Z��#^5�)�kn�PBMx.p��$@�$ �Rt�b{x�T��=�4��D�q��uԔ�h۩��|L�>I�S��9 $nA�78��?��w��-�Il�:$�K�-��Nx�؃C�t��S��S8��R�ն'_��L��~"�74�� 50<��pc�(%��A�:h��;C0��G�)l�: �d�tp@�r�S��%1elh�C��7Oh�� 12�/ӷ�A>i)N�"5Ӳ�kI>ᯆ�� +Z��#^5�)�kn�PBMx.p��$@�$ �Rt�b{x�T��=�4��D�q��uԔ�h۩��|L�>I�S��9 $nA�78��?��w��-�Il�:$�K�-��Nx�؃C�t��S��S8��R�ն'_��L��~"�74�� 50<��pc�(%��A�:h��;C0��G�)l�: �d�tp@�r�S��%1elh�C��7Oh�� 12�/ӷ�A>i)N�"5Ӳ�kI>ᯆ�� +Z��#^5�)�kn�PBMx.p��$@�$ �Rt�b{x�T��=�4��D�q��uԔ�h۩��|L�>I�S��9 $nA�78��?��w��-�Il�:$�K�-��Nx�؃C�t��S��S8��R�ն'_��L��~"�74�� 50<��pc�(%��A�:h��;C0��G�)l�: �d�tp@�r�S��%1elh�C��7Oh�� 12�/ӷ�A>i)N�"5Ӳ�kI>ᯆ�� +Z��#^5�)�kn�PBMx.p��$@�$ �Rt�b{x�T��=�4��D�q��uԔ�h۩��|L�>I�S��9 $nA�78��?��w��-�Il�:$�K�-��Nx�؃C�t��S��S8��R�ն'_��L��~"�74�� 50<��pc�(%��A�:h��;C0��G�)l�: �d�tp@�r�S��%1elh�C��7Oh�� 12�/ӷ�A>i)N�"5Ӳ�kI>ᯆ�� +Z��#^5�)�kn�PBMx.p��$@�$ �Rt�b{x�T��m��|�ALsD��o:N��6��$��mlh ��&��:芗%Šn�'S��$�1棶��"41��:�JH\wF�lӞuO G0u�!�!��~BT��K�1 :��AK��h�$��L}]΂N��9��]H'��$ē��s�D��L��8��I�h�#�'q�F��:��;xy��v��5�D�q6hdH��'��%.M��|�JSgn�~A �H�d��6��0 @��,�q L��5��hG�Q�C4& 0R�ͭ��W;�t��4L��J��#i @��Ma-vѦƂ}Ra{�� Ls��)r\Z��u:��Kn�2K�3j;lq>�#C� ��3�:x��th�9�T�$s_r��^��%I�$��Ө��@:?4��$�{��5ԁ~ ��LI: AK�<4O!��Ns��D�6�2;�w$hc�JS�� mn�c]tL�f�D��0.�{q=�R��K�'��6v��ԉ�L�|k��NR�<��Ϛxs\��x$3B`@�+\��us��HIL@sD��o:N��6��$��mlh ��&��:蒗%Šn�'S��$�1棶��"41��:�JH\wF�lӞuO G0u�!�!��~BT��K�1 :��AK��h�$��L}]΂N��9��]H'��$ē��s�D��L��8��I�h�#�'q�F��:��;xy��v��5�D�q6hdH��'��%.M��|�JSgn�~A �H�d��6��0 @��,�q L��5��hG�Q�C4& 0R�ͭ��W;�t��4L��J��#i @��Ma-vѦƂ}Ra{�� Ls��)r\Z��u:��Kn�2K�3j;lq>�#C� ��3�:x��th�9�T�$s_r��^��%I�$��Ө��@:?4��$�{��5ԁ~ ��LI: AK�<4O!��Ns��D�6�2;�w$hc�JS�� mn�c]tL�f�D��0.�{q=�R��K�'��6v��ԉ�L�|k��NR�<��Ϛxs\��x$3B`@�+\��us��HIL@sD��o:N��6��$��mlh ��&��:蒗%Šn�'S��$�1棶��"41��:�JH\wF�lӞuO G0u�!�!��~BT��K�1 :��AK��h�$��L}]΂N��9��]H'��$ē��s�D��L��8��I�h�#�'q�F��:��;xy��v��5�D�q6hdH��'��%.M��|�JSgn�~A �H�d��6��0 @��,�q L��5��hG�Q�C4& 0R�ͭ��W;�t��4L��J��#i @��Ma-vѦƂ}Ra{�� Ls��)r\Z��u:��Kn�2K�3j;lq>�#C� ��3�:x��th�9�T�$s_r��^��%I�$��Ө��}��g�I��'��S=��;Ih�.�O��Nlqpk�L�8�ӡO�)�ns��f5�Q;>��G��y?� ��:"�d�sv�3��Y0!�;��(�?4��"Op�d1� �|�R��`s��c��@p"#��-M8K�}$�=�T�k˚Љ��"�C��?��8��G��y?� ��:"�d�sv�3��Y0!�;��(�?4��"Op�d1� �|�R��`s��c��@p"#��-M8K�}$�=�T�k˚Љ��"�C��?��8��G��y?� ��:"�d�sv�3��Y0!�;��(�?4��"Op�d1� �|�R��`s��c��@p"#��-M8K�}$�=�T�k˚Љ��"�C��?��8��G��y?� ��:"�d�sv�3��Y0!�;��(�?4��"Op�d1� �|�R��`s��c��@p"#��-M8K�}$�=�T�k˚Љ��"�C��?��8��G��y?� ��:"�d�sv�3��Y0!�;��(�?4��"Op�d1� �|�R��`s��c��@p"#��-M8K�}$�=�T�k˚Љ��"�C��?��8��G��y?� ��:"�d�sv�3��Y0!�;��(�?4��"Op�d1� �|�S/��?�RM��I1Q�#A❎qt��S>�܁��A3��s�'Nt3>��<��Hv��◨7�y�'_��R�H��p��qt�]�䔻l��A�+N�&C��(��h��5C�(��25��V�$~/t"��IL�I��H�f�$~t��N�q�A'��M��~D��Й~ ��j5=��c}&�u��L��JS=Hv��N'y��u?w�Z��#_�&�I;��?$��q�o��+iۻw�n<�#�5�G��R��l�x�c�A�5�Ϲ��w x�'��L�<|Jb�ӝϸ&m�#�� `��"8�3�$��]?x�%.�!��xJǓ�I��4� +2=8�$��%%h @�� +'x��L��i.գI�K��S/�o>R;&��:|ӆ{�`I� 뱄=��:bt�C3� �c��t�k��)z�|7�2u��-䈎<� )gO��>IK��lnt��d8�<��N6�.��IIZP4?�� -�!3ZK�h�G��@" 9�y��ϔ�ɦl"G�N�4��x�ވ.y~w�IK�[<� ��ݼƣS�(�7�o�]?�}�d��<$�3ԇn�;��w�h'S��E��I�5��Rk��IK��v��⶝��x��o2?3X�q�k��~��%(X=F��v9��_L��r��z�a�s��$�!Θ�9��f��;�!��^�� {�Ky"#�=�JY��w��R��y;��#O �#Ӎ�K��RV�� w� jKd�L֒�Z4��n�A%2�&��#�i��ӧ�8g��#�7��@��Rƶ��'Bdy�'�o1�� +%��O�lY0;� )L�!ۄk��8��Z ��Qhn�G��~0��$��R�{��]�;�2��n��=��D̏��5r��6��IJQ�#A❎qt��S>�܁��A3�� )�s�'Nt3>��<��Hv��◨7�y�'_��R�H��p��qt�]�䔻l��A�+N�&C��(��h��5C�(��25��V�$~/t"��IL�I��H�f�$~t��N�q�A'��M��~D��Й~ ��j5=��c}&�u��L��JS=Hv��N'y��u?w�Z��#_�&�I;��?$��q�o��+iۻw�n<�#�5�G��R��l�x�c�A�5�Ϲ��w x�'��L�<|Jb�ӝϸ&m�#�� `��"8�3�$��]?x�%.�!��xJǓ�I��4� +2=8�$��%2��d��$��lo�D f��f��S�^C��AH�cK��O�v��I#Q��&��L�:��T��D�?�BIX��n��x|��X�ə��5�w�q-: !&V}2��I��|�r�ƀ�'_ ��5��7�H )�w;H�t�J��$�H0��Q��P�gv��*\��5�)9��;��`� c�8h�I0�yIJk��O�� \�}��䙎%��#]�p�wӠ�eg� >�I�t�i0g�'.|h"u� + �8Q�K�q�D��qs�� O��bA.��qЉy�� wk��"��~�Q��S��x�쬶�p�<ӆ�PA�甔��Zt�|��'٠�>I��[��Ƃ;&��%3i��ƦGȡ=��h04�Qe��>��Gp"�-i2�F��T��5��Rk��h'Y�(��%��AOIy�Ǻ&e%#�.tjgY>)�F�l�$�F�� 2 �wiR{��1 %c�ݸ$ �m�N� b&fO�H�a�#]Ĵ�4��Y��O�|C ]'�DL� ˟ȝ|��k��jR��`G� $�m�\� u��(~��K�a ú\t"F�aGc}By��r�@�G0��?��F�({+-�f5�4��y$��%)��?$�s��h;ϒf8��ɀ�ywIL�fe��(O`!�� 5�Yw�O�x��Ƞ��ZL��w�.�lo�D f��f��S�^C��IH�cK��O�v��I#Q��&��L�:��T��D�?�BIX��n��x|��X�ə��5�w�q-: !&V}2��I��|�r�ƀ�'_ ��5��7�H )�w;H�t�J��$�H0��Q��P�gv��*\��5�)9��;��`� c�8h�I0�yIJk��O�� \�}��䙎%��))bX��K�t� +,.�w�)�Ih:e3E`#��ְ��A��Bp��6��q��G�E��Q�*bѴ�L�iCw�$>�9s{�� @��QR��ԁ�5�pd��۵��;$v��<R��|c��@1�� ÚZ�h�F�%�8$7�<�)�t5:�F�A��h��x��!�}�^jH�΄&�`�h$�S��e��ǈN6�� L״�k�� )`��j`�hJL :o&c��0�O��M��4��IK��\�� Yat��H�KAЃ)�+��h��u<Ĕ� ��<�-��Hv��Sƍ�j`w�J�� )˘#��/��o��|� �p��Ӄ%D�ݮh>��#�X=��i��\]$��*� �.G}xN��G�5�/��!��wILà��4* �kD��q�3��PG�t!5KKA%��.w�� ~� ��Jf�� N��Pm�Z ��(��tA�h��=�� X:Z /��=n`�s�q�� =�5�%��N�ɘ��/��<}�r��D7��"A��;H��#��k��Q1�k��vH�ocx$�I:��J�cH?K��^�4��xK�pHopx�S0�$48ju<� +�m��p�D�C��D(��MR��I~��s˝��8m�A��i.8�A�R��-��Д�t�L�5(a|��k4 i#�⒖%��'N��w��$��S4V�"8 )Mk��4|$'��y�)�jG y$[Yݬ��u"�!�H��7yqA��S�0G�!�^) � L�.��@ >��Hs]�J��\�}òG`�{{��%(��I��?�TA�\��9��k�_��C{���A!��S��hTnֈ/�� >�g�!E�4�l�Bj��K�?�O[�&\�x��h-o�{Ip�ƺ��`qhv��Ӧ�f;��C��Y�I��,n��:p��K��$�2��IL6S��{zH��^�y$��a��`��M,�:��D�ms�p3��6H"5�XY�$�O�X9s�H��I�`�"C�jY0�CN�LI�&�I�DΞ*L��N�?a0��xBHe,��G�M �!��sHw1@�*P}��þ�%�� p�~)mnө I'�bO�n�b� .&�(T�`2$�9��%��wv��wa�H�Uۄ�N��iԐ$��1'�7v1O�a*D��_��ӎ��a;�v*@Vlx�x )E��@5��l��A3i�@d�0��:� l>��8��w�}6� +g?��v��c��,��`�v�$�@�T��C��KBLF�:'�s�tO��Q>�\�\�%2 ��@x��{�,��&�\��%�=�RkX7��L5�ӬyI��@'Q3��6��L6D��K�/$�4��H;s�Ǆŭ��D;ʔo�� E*��\'_�[[��HI��'�K��0�"X��/�~��i�e d��ݻ +6< �<��ւ�à�S��6Z� ��4� 2H�x�R�6LDwIIJy�;�>�y3��i�X�R��q�H��wQ0[�dF�k�J^�C[!�Dy%�&#p��9�:'�x(�M�v�x��ÉF�))�6x�f�"��=��#_�)�Z�>E0�>�0��NRS�Z��y��G�d �8�' �F�L%.l>�.�AK��#��\O��*E�$��x�c��IL��f|Tg�c�:��8��A��3Q�q>��3�ǸN��_��<)&| pT��ۺJD�0��=��4��$~lJ��k=� '�u�N�v��Rŭ�ĉ&LN��ր �@q�H�7Yg��8hfAh�-IL��A�h<~)��&4�j|9��'o��%2��ԁ�T�G��k�%;K@�Ȧ'�i�Jbv�@�~�3�S��0G`$�h� �ĥ͇��)wX}�{�z��H��u��1v>))��\@`�ϊ��{gQ�'��<2�7�j9�'�QR�w�� Ӳ��=��P'� ��σ� +�\�`@�>�{wIH�p��'�S��$�͉Qq<�g�d��IԎ�zJX��$ɉ�p��@h�> ��"L�{' ��-%�)��<��?�Vƞ�Aχ1��y��~ ��P��}��<��H�?4�~$�ih0��$�`ãM9ILN�hO��|Jy��&��6�A08��|�>!.�K�p��Qq>��4��{�!�a��%2��L��Q�яl�;$�w��Rf��G1��J*T��:vQ|�;��\t��t��Rk��ۯn�)��.�� +pӼ>D��*.'��l��}�I:��c�IK�2w$�1:�CZ�H q�� X�dI�d��%2v�Y�� +��y�9��7w;{4�=��*3�1�Gd�N� �`�L��8� EJ��c�'N�/��}@��>�8*Ms��t�u��%"p�t��Nw�Ȓ?6%E��5��Ϻ�'R;L}�)bֆN�D�&'Qk@I�8�$K��3��43 �@��N� �?��Xx5>��oa擇��u��⒙Cg��j@�*M#�@��5��x��S��4�%-�}�_�I?�|�AL[8i��x�9��7@�ClG��L��;��O�*_ӬĆ��t�Z��O�k�)��׍xAJi�۠��^��$IA攀]$��%-�h ��"\v�`�u%4Ú6�G:��;��j��~�CI��aM�-#��]�0Kw[ Ή8��@�Ǳ�))��$� ��-i��|#T�>�x��um��I��"�5�-��H��-�CML4��J v�G2�"�.��4�� +[%Έ;�f;&�5��>�T��ᧁILY[8i��x�9��7@�ClG��L��;��O�J_ӬĆ��t�Z��O�k�)��׍xAJi�۠��^��$IA攀]$��%-�h ��"\v�`�u%4Ú6�G:��;��j��~�CI��aM�-#��]�0Kw[ Ή8��@�Ǳ�))��$� ��-i��|#T�>�x��um��I��"�5�-��H��-�CML4��J v�G2�"�.��4�� +[%Έ;�f;&�5��>�T��ᧁILY[8i��x�9��7@�ClG��L��;��O�J_ӬĆ��t�Z��O�k�)��׍xAJi�۠��^��$IA攀]$��%-�h ��"\v�`�u%4Ú6�G:��;��j��~�CI��aM�-#��]�0Kw[ Ή8��@�Ǳ�))��$� ��-i��|#T�>�x��um��I��"�5�-��H��-�CML4��J v�G2�"�.��4�� +[%Έ;�f;&�5��>�T��ᧁILY[8i��x�9��7@�ClG��L��;��O�J_ӬĆ��t�Z��O�k�)��׍xAJi�۠��^��$IA攀]$��%-�h ��"\v�`�u%4Ú6�G:��;��j��~�CI��aM�-#��]�0Kw[ Ή8��@�Ǳ�))��$� ��-i��|#T�>�x��um��I��"�5�-��H��-�CML4��J v�G2�"�.��4�� +[%Έ;�f;&�5��>�T��ᧁILY[8i��x�9��7@�ClG��L��;��O�J_ӬĆ��t�Z��O�k�)��׍xAJi�۠��^��$IA攀]$��%-�h ��"\v�`�u%4Ú6�G:��;��j��~�CI��aM�-#��]�0Kw[ Ή8��@�Ǳ�))��$� ��-i��|#T�>�x��um��I��"�5�-��H��-�CML4��J v�G2�"�.��4�� +[%Έ;�f;&�5��>�T��ᧁILY[8i��x�9��7@�ClG��L��;��O�J_ӬĆ��t�Z��O�k�)��׍xAJi�۠��^��$IA攀]$��%-�~��$�i��4�S�3;�Γ�*gx��5��F�=�I��:��1κ"��qh�I��-�D�.�y��ȍF�&s8μ�⒒Ѡ4�S��}�pC�kAxnߐ�'��CN��PR��Z& �|��Ws��Ngp�R �&;��1$�%.\��<�4)9�6�El�4�� l��=��)N�x��ݵ�A�u�3�M�;�?$��IK�a.�/��ۿ��H;R'Y2u�M��k�i��s�>Z�Tl�� L�skpq��]!%1�3��:��w��H;�|�XK]�i��T�^��J\��4�N�|ۤL��Lǚ��O��jg0��Ξ))!q��Ny�<�I��܇<�� R{�.l�4�?/�//%�`��1�w: :��w u @��c�8�N�PR�� sA2��g�tQ&ͣ@L��꒔��>~i��[��]9�١�#�c�L�l��Ot��6� ��)M��"u�'_��5��,�h53��8�F�И$�J�6�\��S�3;�Γ�*gx��5��F�=�I��:��1κ$��qh�I��-�D�.�y��ȍF�&s8μ�⒒Ѡ4�S��}�pC�kAxnߐ�'��CN��PR��Z& �|��Ws��Ngp�R �&;��1$�%.\��<�4)9�6�El�4�� l��=��)N�x��ݵ�A�u�3�M�;�?$��IK�a.�/��ۿ��H;R'Y2u�M��k�i��s�>Z�Tl�� L�skpq��]!%1�3��:��w��H;�|�XK]�i��T�^��J\��4�N�|ۤL��Lǚ��O��jg0��Ξ))!q��Ny�<�I��܇<�� R{�.l�4�?/�//%�`��1�w: :��w u @��c�8�N�PR�� sA2��g�tQ&ͣ@L��꒔��>~i��[��]9�١�#�c�L�l��Ot��6� ��)M��"u�'_��5��,�h53��8�F�И$�J�6�\��S�3;�Γ�*gx��5��F�=�I��:��1κ$��qh�I��-�D�.�y��ȍF�&s8μ�⒒Ѡ4�S��}�pC�kAxnߐ�'��CN��PR��Z& �|��Ws��Ngp�R �&;��1$�%.\��<�4)9�6�El�4�� l��=��)N�x��ݵ�A�u�3�M�;�?$��IK�a.�/��ۿ��H;R'Y2u�M��k�i��s�>Z�Tl�� L�skpq��]!%1�3��:��w��H;�|�XK]�i��T�^��J\��4�N�|ۤL��Lǚ��O��jg0��Ξ))!q��Ny�<�I��܇<�� R{�.l�4�?/�2�g��w� �$��n��E��Z>��ӛ\�S1�.t�A��EJm��{��|TNƏ��h�'�=�^O��ir~��"\݀jL��Ll��`� +'O�.�ȓ�&��`�{�4��%� ��X��@��"��@�N� F 1�d�1��4"g�s��O�.� G>j.l��{$��\&dC�t?n�;�t��0��n$��%%,7�N�@;HJ��v��!4A�&{v��cH;c�΍xNR�\f��'��5��$i�� : ��Ɵr��v0�$� ��n��E��Z>��ӛ\�S1�.t�A��IJm��{��|TNƏ��h�'�=�^O��ir~��"\݀jL��Ll��`� +'O�.�ȓ�&��`�{�4��%� ��X��@��"��@�N� F 1�d�1��4"g�s��O�.� G>j.l��{$��\&dC�t?n�;�t��0��n$��%%,7�N�@;HJ��v��!4A�&{v��cH;c�΍xNR�\f��'��5��$i�� : ��Ɵr��v0�$� ��n��E��Z>��ӛ\�S1�.t�A��IJm��{��|TNƏ��h�'�=�^O��ir~��"\݀jL��Ll��`� +'O�.�ȓ�&��`�{�4��%� ��X��@��"��@�N� F 1�d�1��4"g�s��O�.� G>j.l��{$��\&dC�t?n�;�t��0��n$��%%,7�N�@;HJ��v��!4A�&{v��cH;c�΍xNR�\f��'��5��$i�� : ��Ɵr��v0�$� ��n��E��Z>��ӛ\�S1�.t�A��IJm��{��|TNƏ��h�'�=�^O��ir~��"\݀jL��Ll��`� +'O�.�ȓ�&��`�{�4��%� ��X��@��"��@�N� F 1�d�1��4"g�s��O�.� G>j.l��{$��\&dC�t?n�;�t��0��n$��%%,7�N�@;HJ��v��!4A�&{v��cH;c�΍xNR�\f��'��5��$i�� : ��Ɵr��v0�$� ��n��E��Z>��ӛ\�S1�.t�A��IJm��{��|TNƏ��h�'�=�^O��ir~��"\݀jL��Ll��`� +'O�.�ȓ�&��`�{�4��%� ��X��@��"��@�N� F 1�d�1��4"g�s��O�.� G>j.l��{$��\&dC�t?n�;�t��0��n$��%%,7�N�@;HJ��v��!4A�&{v��cH;c�΍xNR�\f��'��5��$i�� : ��Ɵr��v0�$� ��n��E��Z>��ӛ\�S1�.t�A��IJm��{��|TNƏ��h�'�=�^O��ir~��"\݀jL��Ll��`� +'O�.�ȓ�&��`�{�4��~��{�s�rALE��l�x�c�A�5��dk3�<�䊑�:bt�C3� �c��t�k��)��x2u��N\��_��qt�]�䔻l��A�=��q��D@�\��JJ�@��N� Il� +-�چ�'��L"77D��ϔ�ɦl"G�N�4��'i��*;k/y��O�J\��L�?��5��@��&�'N��<W�yR��!ۄk��8��Z ��Pf��2N��=��|Vӷv��x"��#�k��66��JP�z��s�3��2��C��R��ft'�|��R0�LN��f}�3lyΐ��>��N�� ˟�Pc��R�.��|��m��<'��Y� �P��H�K��IIZP4?�� -�!E��P��DF�蒙~�y��4̈́H��4=��:��Gme� �}�i��IK��[<� ��ݼƣS�(_��y�� BJ]��;p�`q�'��A:��ZcB;|#fI��\u��#�C/��v��ۏPdp�v��F��IJQ�#A❎qt��R{�p`v> +U��΄��JF�ӝϸ&m�#��w{A��!9s��y|BJY��w��R��k3!�> +��r~ )+Aj��Q;�5%�d(�7j�<��q0��S/�o>R;&��:|҆��S�x��ϻM<>))s[Kg��2<��j{+��;�4�=^��IK�ԇn�;��w�h'S��A�KLhGo�Dl�:�R��K��|we�[N�ۼ{q��̎��#��|{��I)B��6h^I)s�'Nt3>��<��Hv��q��'_��ۨ1�� )gO��>IK��lnt�]�̇�(no�D��$��D�@Ԗɐ��ݨn��~ ��"#stIL�I��H�f�$~t��J�v�Oq⣶��[>�4��m-�N��O��cQ��n�t�<��{w�!%.�R�F�8��e��O��m-1��$��IJs�.:��!��m;wn��ǂ(28f�L�?ޣg�m��$��N�8�:k�)=�8 0;*��gBx��%#t��Άg�6Ǒ��|S�;��d��u<�!%,��)v� �΃�{�� ?��a��M�q��S�� + +endstream endobj 1997 0 obj<> endobj 1998 0 obj<> endobj 1999 0 obj<> endobj 2000 0 obj<> endobj 2001 0 obj<> endobj 2002 0 obj<> endobj 2003 0 obj<> endobj 2004 0 obj<> endobj 2005 0 obj<> endobj 2006 0 obj<> endobj 2007 0 obj<> endobj 2008 0 obj<> endobj 2009 0 obj<> endobj 2010 0 obj<> endobj 2011 0 obj<> endobj 2012 0 obj<> endobj 2013 0 obj<> endobj 2014 0 obj<> endobj 1 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 1/PresSteps 11 0 R>> endobj 2 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�˭Ŏ��+�1 �U�$�!!J�C�E�qL�d0ɯϮ>}��O��5��Z��6��CÛ.�E:4:�eTp "��9d<��Y<FՁWMZ�J"/��U��-#�Ndf(�To�v��m��c$�xk�m?4`��ԈB0S>�g�R;4| +(va�� o�b�|�W��x*2��!@a�`�@V �ˡ>�xt��fO�"��(��i��A��!�nY�;"8� +�<<6>��P�6t�� tȸU�k.g�&V=��*/񺺰��hm��V��|J<Ĥ�m^ͫu�Tf�mN�qJxWs��]aF��)��U��xh��1\�F(��H]��v��x�iFN �iv�F%G��Oe�!��쌌I'G�{N��2a�cү��&^��E�NM]��4)c7��pG*=9��S��ٕ�� 8,K4 ��n'e��uT��D�W�c2��1�͘��E�Ǝ��u��"�x�8�u;*��W�s4/ɩ>vNf��$t� �S�@F'�'� ��MY�KG��a[E�%��C+:{@�I��is��bL�E��P�0��mu��P!dh<�8Mg��]pw +8��7�c�9�M kTiL��6�T|PPt��0g�W��h ��66�;ύ�]��S��^�O,��U�º��{�͹��I��O��H�C?�� (��.]�@2��ޠg�p�N�@�DSgJ�S;u�:'ʛS��S��s v��9U��{`�� 1kJ��O�NQgA��Xp��~.�� [5�=�V�ʈ�{<�m��C5�c(f�O��K�o (�(43�߰��Ώ0��85�Sa�:��[JEH��luM��x�d�V��mU�H�l�$��P��ΑŤ��T<-��u�$Fe��T�\�偙��'Y;%��u8��zw:L�P�k +�D8��:D��gNw�p:�)ح��I�N �t�+:P&ԉ��ΦS��ZpU$�U�],�G�� d>��<��D>O��J�q�V�ׄÙ��S��9�3hj��2��>%��S�m�A��r\C��]ƭ��B!�?%zs>m�&�r�ꈕ�t֐p�2>/d��k:��b�tH� �C��8�5��=�E$�!i|h�O�}��ć&�� C�O\S!�0e<"�99&�'Dj�P��jA�]p8��ow +M�b:��x� j��IW�pLp�x��Z�~�jh�b�Ô�$)'u�j-&��sR?�@֊ +��w}WAH��U2�I�0t��(a#�qj�ΐ'y5%<�j�m�N�gy55��v��F�G�C��C#|�� ]Gǳ\��ծk�'�Ք�vG��#�\Ȫm�#4 4�UsZ��q�Y2��~��y��exc��ީ ��vk$(|�8�5�R��)��N +�4ך�#S�s��rW��"�.�?O�b�m�sH��Y�)�>v%��?l<�ަ��n�X>�_�S��.e>w��1�`*;h�x6v��S��U7��٘�<VS&��F�B͘˦��=��m�u:M�i�Դ�4��i'�J�@)��*H��pr�-(��A�H��P��&�O�u +�q�줩a�� #�M��8i��f2��0�� T��UA��MR� �~M>�M�0�C�d�i��5dju��g�j)/��)a�8�W�M��ϔk��Xg>��0��i�晬i�t�>v��~��]�}��D��|�35�ǋh#N�&�E��΢�F�S�)\��2N�&�?�T�9��@��0��f�M o�`��H~5�&a�jt0�Y!גΧ!"�2��2%�Oa�3��+|J��%o#�>��35�3Np�� Ш` +��ܦ��'@�X��ͨ��g`��$��f� +� 5ø�� *'T��E��H�kN��!��i�/��U��pK��d�4�H��~3��K� ��;4<��"�� B84�Lɮl8��$ܡ�sC6{h��x�C��$��͐]�p�O�C��4�=�]ۈLX��wju� +tB:�f�7(S��ʲ��e�D��!�hV��ݩ��q��q��$�v��Fg��ޛ�S��f�͕`eɔ��*��*�JX$��_�~��!]��@�s�x��E�M ��+�1��]:��bG��^S�~��2U�0��Ge��!�V�Jmu��k8��4��q�dq[��C#�zC��@��_�:��|N1lH��Pmfd)� ��ʧ)�Q�i�#�t� ��4f��BT�DLsM^puuzs��p4R C]=��)�@W�hLB�&LhzЗv�7fGa��=�BQS�S��9��^b� ��)o�%ã᫖��p^kvH%Y]Ǔ��j�n,��g� ��.�ة�#8��5��A9v��:GuY70��X��C�%;��,�s0��q��Ae�ť�a�Jvv�FŲ��SÚ8;;��x�KT;T�� 5��/ ��n��՛wny�n��y��_��[޽��|3/�B�-o��i��?��sW��g��O�f��&[��Ȗ�o ��і}M�ܮ�c�mM�$z�}M�z�}M�4z�}��DG��O��ɖH��ɖL��ɖJ��ɖΎ��q��dK��dK��dK��dK��dKg��k�%��k��o�,��^}�|�ū�o��4��_��j|�o��_��[}��\Mُ�>4/�?��3��=�0��0��_��.ϟ��W6��o_@(wo�K��o��/��O�_~��}��>��3>��_��c��Ӻ�9uǡn_��0,�G$�C�=��3^/� �@��T��i]%W@)r-��r�LUR +��?_��E��4��c�T��|��]�z_V��4�)�m|�Ɖ��^T* ��yn� ��\��ؽ��j>V�~5I�~�|,�_Za�Ej��;v�e:^�� P<��#'?��^i* �H�V:̽Ap��j��Ý�o�y&51<��/�'��^r* �P�ൕ�II�x��v ��p�N�No�M��i��1��K2�e��z��~��ś��wx�?lzW� +�4��j/כo��(�J +endstream endobj 3 0 obj<> endobj 4 0 obj<> endobj 5 0 obj<> endobj 6 0 obj<> endobj 7 0 obj<> endobj 8 0 obj<> endobj 9 0 obj<> endobj 10 0 obj<> endobj 11 0 obj<> endobj 12 0 obj<> endobj 13 0 obj<> endobj 14 0 obj<> endobj 15 0 obj<> endobj 16 0 obj<> endobj 17 0 obj<> endobj 18 0 obj<> endobj 19 0 obj<> endobj 20 0 obj<> endobj 21 0 obj<> endobj 22 0 obj<> endobj 23 0 obj<> endobj 24 0 obj<> endobj 25 0 obj<> endobj 26 0 obj<> endobj 27 0 obj<> endobj 28 0 obj<> endobj 29 0 obj<> endobj 30 0 obj<> endobj 31 0 obj<> endobj 32 0 obj<> endobj 33 0 obj<> endobj 34 0 obj<> endobj 35 0 obj<> endobj 36 0 obj<> endobj 37 0 obj<> endobj 38 0 obj<> endobj 39 0 obj<> endobj 40 0 obj<> endobj 41 0 obj<> endobj 42 0 obj<> endobj 43 0 obj<> endobj 44 0 obj<> endobj 45 0 obj<> endobj 46 0 obj<> endobj 47 0 obj<> endobj 48 0 obj<> endobj 49 0 obj<> endobj 50 0 obj<> endobj 51 0 obj<> endobj 52 0 obj<> endobj 53 0 obj<> endobj 54 0 obj<> endobj 55 0 obj<> endobj 56 0 obj<> endobj 57 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 2/PresSteps 67 0 R>> endobj 58 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�˥G��$��!Y^L�HӒ�²�`��YY��햻&��ʺ��w3*l +e��7]6�tht(˨�<@D��s�xZe�x"��B�D^ ��,��[(F�;��P<$>��>z��H*��~h.�-��`�|��4vh�P��D}-��@�L��T4§�>^�Q�C��fq�ʡ�'��J�2�Ŝԓ�5D([b �h��.8�ȷ;�&`1�BK< �5Xդ+V8&��v-v?@ 5�J��aJx��:m��Z��9��f kE��W_鳾� ${�*��u�$P:EVT ��85Ugȓ��W5�6M'ȳ��T��@��#�!w�>v��Y.N ��j�5� +��j�x^U;��F.dն��9-{�ʇ,`�P?d�~��exc��ީ ��vk$(|�8�5�R��)��N +�4ך�#S�s��rW��"�.��Ţ�J搨5L1H�:�wS"}�J��x�M n�Ʊ|��4��V��-�+\�|��ccj�TvД�l�vi�~�M�n�i��1�+x*��LXٍ��1�M K{(A��t��>(�i�iP�5G5a{(��Nn�7�R4I�U�L��P[P��i�`�Zɡ +�M�.��IS�: f?�F��3�q&�NM��5�dv%T�a�9Aک�}��ݛ�@h��|<1�*ad�j�:<Ӱ�k��.O��R^��S®q��l�z��)�N K��|�3��a�c��.�3�3Y�4�}�4-� +?3��gj��F�MX��!�Ec�:p��S��㥍�e.��yM(`��s +v�Fag͜��B�.��j�L�n�$�`��B�%�OCD.e�A�eJ8��g��)}jSßZ��4��n��r�O��>6��\�Ӱ��h�$ s��Ӯ�)�%��+2�r�n ��pJ�qJ��8��ƾ�@��)��'6;�r�&.�� $� ��Y�C3��s�p��'i6o��c|x�2��Fd� +F�S��LX��y5��A�b�Ed�P��7�,�&:��E��P�NM��'I�k4:�g��p�b��4�O�+K��'V��XP�T�"Y��7p� +��5R�S�S��м.�n�`�\�I,��u�x�;z�@��o�q�:��<*��jWrh��e\éϦ&��$��j��G�]��ձ��p�aC*E�j3� �HI�hPU�V>M ύjNL)�C��4 o�fߥ1D�P%�:$b�k򂫫Лl��ɐj��Q_O��rDc�5aBӃ>��@<1; +S��A҈��R��ƈ��M �Lyc/ _��צ��Z�C*��8�ր��P�t;�`Y��8��vHx4uy�NM�!<� t�/�ȱ;d%�%�9�cȺ��=�2\�,��Ngɜ�Aeݎ� *�8�(.M U��4*�e��qT��^��ءz�Mn�AW�~i�\vӍ�7��~��y�xMޏW b��z��Wȶ��◟��?,��[~X�|�� +�0��r_W�[��L��d��r[�-ap=ڲ�ɖ�u|l��ɖD��ɖB��ɖF��q�5��q�5��q�5ْ�q�5�R�q�5��q�8nr��l ��lI��l)��li��l��]7{z�}M�Dz�}��r��/˗_�y��g~^�G~�Vn��r5e?^}h^>��ӛ��G� Y��z��-/��Ə>�7��! �� n�PG�8��/O�y��w��??|��o_~~��o��=;z؏��/y�/�I��t:ߎ;�b��2�ȱ�}�@�<�5��3^5��@��x�� 饧��U�� Npv��)!R��63�;��2 zt@�Z��y8��J��qw��= u|�a��۞|&C>ɰ� ��]�sǸ+E��Ӂ��n�\�#p7�)��q$�[C'��:��7P�}��y(�p��PN!�m�2��]�L��qO�|��R�>�[��/c%�m��7��Z��Q��BޣnSAO7��]�]�j��i W4�V��Ey�3�8�w4��W�T�7|��⇍:��4��,��A�� +S\Q��1f��A+�}yK��t�ÇA�Pd��ݹ�}��#}��`�[�4�s�}��#��앂sS +J�vХ3��͞k��1��[Y�"��ؒ��k�L��d��8��T��5va=��h��A%;L��}_��'/V�� +�QCf8~X`��<{ߚ�8�P��R��<}�ɼa-h�Ep�� +endstream endobj 59 0 obj<> endobj 60 0 obj<> endobj 61 0 obj<> endobj 62 0 obj<> endobj 63 0 obj<> endobj 64 0 obj<> endobj 65 0 obj<> endobj 66 0 obj<> endobj 67 0 obj<> endobj 68 0 obj<> endobj 69 0 obj<> endobj 70 0 obj<> endobj 71 0 obj<> endobj 72 0 obj<> endobj 73 0 obj<> endobj 74 0 obj<> endobj 75 0 obj<> endobj 76 0 obj<> endobj 77 0 obj<> endobj 78 0 obj<> endobj 79 0 obj<> endobj 80 0 obj<> endobj 81 0 obj<> endobj 82 0 obj<> endobj 83 0 obj<> endobj 84 0 obj<> endobj 85 0 obj<> endobj 86 0 obj<> endobj 87 0 obj<> endobj 88 0 obj<> endobj 89 0 obj<> endobj 90 0 obj<> endobj 91 0 obj<> endobj 92 0 obj<> endobj 93 0 obj<> endobj 94 0 obj<> endobj 95 0 obj<> endobj 96 0 obj<> endobj 97 0 obj<> endobj 98 0 obj<> endobj 99 0 obj<> endobj 100 0 obj<> endobj 101 0 obj<> endobj 102 0 obj<> endobj 103 0 obj<> endobj 104 0 obj<> endobj 105 0 obj<> endobj 106 0 obj<> endobj 107 0 obj<> endobj 108 0 obj<> endobj 109 0 obj<> endobj 110 0 obj<> endobj 111 0 obj<> endobj 112 0 obj<> endobj 113 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 3/PresSteps 122 0 R>> endobj 114 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�˥E�� H_�0 Z�^��K� �_��xV�hНə<}��#N~�QaS(;4��Y�C�CYF�".�C��*��aTxդ� �Hx\ey��C1B��Df��!��I��n�l��+%�M��7�t�,&�O��i��S$1*k��R,O�T�<��)�/��Ի�a��_SP$��4�!�T�4! O�l�ШJ}� +��)�Y��1A=� ZPC��%��V*�ñ�|{Ph�)�ē@MP�UM�b�c�+~��l�b��`PC��'I9��fPk1��i�VT}��>� +B�'��X�M��SdE�@��SSu�<ɫ)�yU�m�t�<˫��O�� 4*S#|��0�dh�Z�h�,kԁ;�/m�,s!��kB�K��S�;4*�8k�� +vi��Ws.`v�&a@��r-�|"r)c�.S��8�ݴ|�O�S��Ԣ�m��4x��O�w�Ԕ+|��5%E�&I�cm�vN�-��^�q��vA|\�SB�S��o� 6�La\<��4�pq�h��Up��d7�LV8�f`75��]�@|A��T�ȔqIv��z9$T��(E�+�=�! wh�ܐ��q5��P�;4�Y��Β9�ʺG9T�qQ\�6�dg�hT,�Λ<5��㨸��A�C��P��z�Ҁ��_^��~��]n?C'��'��l��޿�P��!��r��Wo?._��ݯo߽|y��<�g��%�x?��8gpՇqQx�H8��G��-��jo�4��@�Df��u�)�i難hµ��P4S��)x��K<��j��Z��L�(�n��9O��ͣҖ��]�-,P}��&�_�Gj�ۢ��W��:w<�n��l�9M�U$uYj�`]& �5��no$.��-��,� +K�U�b��py�~ZiD��X�ϗYX�� ^!q2�-MqB�ć�X��t�|��Y�tX�E�d|��A��ӫ��̘>@��D��/�Y�x�#��y� ��ѴO{\��\A��ՙ��yG��[�>ZRuM�W$Y��5�5K/�� )�{��H +��=��`��!�B�B� +endstream endobj 115 0 obj<> endobj 116 0 obj<> endobj 117 0 obj<> endobj 118 0 obj<> endobj 119 0 obj<> endobj 120 0 obj<> endobj 121 0 obj<> endobj 122 0 obj<> endobj 123 0 obj<> endobj 124 0 obj<> endobj 125 0 obj<> endobj 126 0 obj<> endobj 127 0 obj<> endobj 128 0 obj<> endobj 129 0 obj<> endobj 130 0 obj<> endobj 131 0 obj<> endobj 132 0 obj<> endobj 133 0 obj<> endobj 134 0 obj<> endobj 135 0 obj<> endobj 136 0 obj<> endobj 137 0 obj<> endobj 138 0 obj<> endobj 139 0 obj<> endobj 140 0 obj<> endobj 141 0 obj<> endobj 142 0 obj<> endobj 143 0 obj<> endobj 144 0 obj<> endobj 145 0 obj<> endobj 146 0 obj<> endobj 147 0 obj<> endobj 148 0 obj<> endobj 149 0 obj<> endobj 150 0 obj<> endobj 151 0 obj<> endobj 152 0 obj<> endobj 153 0 obj<> endobj 154 0 obj<> endobj 155 0 obj<> endobj 156 0 obj<> endobj 157 0 obj<> endobj 158 0 obj<> endobj 159 0 obj<> endobj 160 0 obj<> endobj 161 0 obj<> endobj 162 0 obj<> endobj 163 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 4/PresSteps 172 0 R>> endobj 164 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�˥G��ȉ�CB,�a��1��%/��A,!��}#+3�>��i�5��OTֽ'��QaS(;4��Y�C�CYF�".�C��*��aTxդ� �Hx\ey��C1B��Df��!��I��n�l��+AI�M�o�:�X:�f �*�,/<�\��RN:GF�M�]"l�t0�-z�C��02��GC��i:CF��5��`F�[_!Rd�A�$65�Qa�1}�[ R�AU@�iJ8 +af\�G�5�?��0�<7�tnO�3�[-��X �׫��e��_O?�_��O>~�f �8~��On� �_�E� ��u$#-�)� �q� ��t$N4u�D8�S'�� q��9%�`�:u +:b'I�S��Vp�T$faII�� )�,��NY��h�p�V�KϪ�2"u�G��B>}SM��Y�S#>��R�+� +��7l8��#Lj&?N �T��ά��R&($[\S$<^=�D�՞�� S"�T:}SM��bR�T�*��_�:E��@`h�x.��LEɓ��:�NK�;&I(�%E"�NS"L��3�]"�N{ +v�4y�S6�F:Ċ� ub�&��锵�\ ud��bj&��{�B�'5��iS�U�=�A��p8�;{jx8'{�M�Z\&�ߧ��{�M3�7�T�K��e�*{۪ЛQ-��S"�7G��h+��XO�a ��!��B6�!��3�/��)�Mׁ�� 84<��Ysi�SZDR�ƇF�4��0J|hB��,�Q9��%R S�#��c�z�A��eM,��T��c��Sh�)�ē@MP�UM�b�c�+~��l�b��`PC��'I9��fPk1��i�VT}�>�]!�V�l��*��)��J�l|h��Ʃ�:C��Ք�ɶi:A��ϧ��7t�rqjx.V��Tp|�WS��N�0r!��f��4�8�W�i��U>d��!�� ,ã��_��NM��["A�C�i�ٕ:�`lf0N o�vR@ՠ��#��waq��t-�Z:0�D��`�A�š��cW,��c�mj�s�v0��85<��]n�^�R�k��]S�ASƳ�ۥ��6U��ੰ�2ae7h.j�\6%,��m��ir�N��U�A��Մ��(>�N;�Ur\J�$�VA2��CmA�6��Ej%�*69|��S0��f'M �$��Owh�N�0ǙH;5F�p\D0�ٕPut��i�� +Jv��M��k��h��%��L�..!S�]&��]�0_�6� +?S��>0`��<�gZS��ۧ]�g�g��i��iZ�~f>��gj��F�MX��!�Ec�:p��S��㥕�e.��yI(`��s +v�Fag͜��B�.��j�L�n�$�`��B�%�OCD.e�A�eJ8��g��)}jSßZ��4��n��r�O��>6��\�Ӱ��h�$ s��Ӯ�)�%��+2�r�n ��pJ�qJ��8�nc_A�Q�� L�M GO��Nq��Q��8�IvA�d�j�qS�(م �TN�H%�Lg�d��ʩ�CB��_ ��$�p��i=$޹�7�n̼�.�4&|N��N&�d�W{�C +�ЈB0Q$��p�� ,�W�E�CÓ4�7Cvuű><��l�Pvm%2a#ߩ�.V�t^ͮ�P�XeY�e� (˨�(�C�V�Rd�632��t�U�j�Ӕ�ܨv�Ĵґr:T�A��n�]�AT U!�C"��&/��9�V8��.��x�+G4&!Q&4=�C��0ŕ�D!��)��Li�h�1Y��W��UKymj8�5;��I` `�i �J�� U�� m��GS�W��Ú@�� !+�,��QC��p}��dG;�%s�u;�r4��8��45lT�Ύ#ШX��7;xjXgg�Qq+/{� +`��q7��]��s�]7��<�w��78 ��x"��oO��|��o?��[w��sW��}]�3��'_3�r_�-�Bd�}��h˶&[��&[=�&[ +=�&[=��[��d��dK��dK��dK��dKg��nk�%��nk�%��nk��nk��nk��vq��q�5��q�5��wϷ��7�}�ٛw�_} ��o�?��?�u��?�/WS��g��o�xz��B��#�0��x�c��#|��b�֍�@A/�<��O|�ߟ~��痟��|��_��<�gGۑo�K��~6��q��X�%u x�#��Y� !�3��s��}��&�q�W|ܭ��۱��'H �lf��8_�0�(Y��V��T� R��ا{f΁U�y=�Y癏��#�͝}�u׹��0I��O�00�2nC��5��\�5��Ά�%��C�pL�ݧD�y��p�:v�7:�vlt��Z7��=�/�4�,�J�[۹c��l�>;�[�gu�:dxr��S�-r*VA��b��$�"<2�gd��}��s��L!��*"Bx�n3Z��0�� +a?��>��J+4d�t�^�"��7.��xt3,�~�� [��б��:�*��`��Ս0�J�X%X>Jp$�2��&T��P/fxa3,��82��n>)��Ug\��8 ��2��E�F��#2 �j[�U?�ȍ$%� ��\��p��.��߯��w�E@��"=�kDb��ۡ߁�U��]K�9�/y��U@w*Q�>$�V��VW�� +y�#��x�v|`5%tɇ�K��|�O�2��0��H��|aH��Q#�';��.��p��N�@BjE9\��/��Y��{�׉X��*��<�@�T�ҳ9��Տ<��W�Xl�c��CH4�"�� F��%�Q�*�6*1η�0)Q�u�^-��@,w;��5V1��2o�>�S��vg�cӟ��g�4<�E��\U"P��.ꂞ��#gOR��p/�'�nJ��8�$�Ӻ4q�w��ʟ϶��ɋ��N +.�KZk�b9�i��_Ô�<[��Ŭr�K�P�p�_a��$K��s5ʋ 9��@��{�!U��s�o?)(�e��ck�?[��KlBФ��+��t�n��=�:-;�f�mi��6��tO�O��.��UD��I��D��rM]��[ۤ�0Wz��b4)k �Ȣs�t^��j�w�αmd�0C��y��>c�=+�N��ElR��y��kʺ�ʁ��2em��9��*a ��Ѓ`�u��L��d�|�0�#c�!D��eQF�U*�Jc9�E�$�t&!Z�y)�@@iL�LsXkS��9_v�v� �;Q�f�g� �e1�,��qT��Bf�c�#�P<��uC��O4w +�� +endstream endobj 165 0 obj<> endobj 166 0 obj<> endobj 167 0 obj<> endobj 168 0 obj<> endobj 169 0 obj<> endobj 170 0 obj<> endobj 171 0 obj<> endobj 172 0 obj<> endobj 173 0 obj<> endobj 174 0 obj<> endobj 175 0 obj<> endobj 176 0 obj<> endobj 177 0 obj<> endobj 178 0 obj<> endobj 179 0 obj<> endobj 180 0 obj<> endobj 181 0 obj<> endobj 182 0 obj<> endobj 183 0 obj<> endobj 184 0 obj<> endobj 185 0 obj<> endobj 186 0 obj<> endobj 187 0 obj<> endobj 188 0 obj<> endobj 189 0 obj<> endobj 190 0 obj<> endobj 191 0 obj<> endobj 192 0 obj<> endobj 193 0 obj<> endobj 194 0 obj<> endobj 195 0 obj<> endobj 196 0 obj<> endobj 197 0 obj<> endobj 198 0 obj<> endobj 199 0 obj<> endobj 200 0 obj<> endobj 201 0 obj<> endobj 202 0 obj<> endobj 203 0 obj<> endobj 204 0 obj<> endobj 205 0 obj<> endobj 206 0 obj<> endobj 207 0 obj<> endobj 208 0 obj<> endobj 209 0 obj<> endobj 210 0 obj<> endobj 211 0 obj<> endobj 212 0 obj<> endobj 213 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 5/PresSteps 220 0 R>> endobj 214 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?��eG�㈢��Z�\��6��0��i�@�I��ޑ��Y�!6s&o��ꈛ'2*l +e��7]6�tht(˨�<@D��s�xZe�x"��B�D^ ��,��[(F�;��P<$��>z��H*��~h.�-��`�|��4vh�P��D}.��@�L��TT:�PM��bR�T�/*��_�:E��@`h�x.�򄙊�'Y;%��u8��zw:L�P�k +�D8��:D��gNw�p:�)ح��I�N �t�+:P&ԉ��ΦS��ZpU$�U�],�G�� d>��<��D>O��J�q�V�ׄÙ��S��9�3hj��2��>%��S�m�A��r\C�_�]ƭ��B!�?%zs>m�&�r�ꈕ�t֐p�2>/d��k:��/��b�tH� �C��8�5��=�E$�!i|h�O�}��ć&�� C�O\S!�0e<"�99&�'Dj�P��jA�]p8��ow +M�b:��x� j��IW�pLp�x��Z�~�jh�b�Ô�$)'u�j-&��sR?�@֊ +��g}WAH��U2�I�0t��(a#�qj�ΐ'y5%<�j�m�N�gy55��v��F�G�C��C#|�� ]Gǳ\��ծk�'�ȫ)�yU��^'G��U۲Ghhʫ��*�d�YC��Q�]Sï�fw��S��HP��q�kv�5��S��P5h�53G��H��16�:��'E�]X�<]�E��!Qk<�b�fu��D�ؕK�d��65��u;��85<��]n�^�R�s�;S��gc�K;�l�XuO��]�Sa5e��n$�\ Ԍ�lJX�C �6^��0��1@IM�L�z�9� �CQ|0�vr��Iڭ�d*'�ڂmN��JU8lr�tY�`�N��I0�!�02��a�3�vj:��Ḉ`&�+��C� �NU�k/X��$�@��T #;TK�ᙆ]\C�Vw�xzv��l~�v��|e��+�L�vjX��u��4?3КF?0��>�?�<�5M�.��N�ү�3�k�O��:��|�F�x�a��Є��Y4֨w +8�+=^��Y�B�)�ׄ��*?�`whT�q�̹��-��ɯ�\�$�VMF9+�Z��4D�R�4]��)�q�i� +�ҧ65��E��Hs�Oi�L �()W��)�cs=��> kJ��M�0��>� +�B[¹�"�(��m�+4*�¸xb��)�i"�� 3�).v3��'5�n ��pB��0njx%�� +� �d�)�,�욃S9�rH�x��"�t�d!�.��Y:M8��;��?��٥�Ƅ�I��D��cO��B84�Lɮl8��I�C��l�Ќ��܇"ܡ�I�͛!��><��l�Pvm#2a#ߩ�]&�� 鼚]ߠL��"�v(��P�QP�xȢX�n +�w��V�͇ǍC$۵ +�Ɋ3{oj�O�kN��'W��%S��KO,�h*a�,��~�� C�ce��)�-<4ϋ��@8�%W��$`v��:d��_��ǫ��˗7�=��Bcy|;!�#��͈�-��.�_��`��+��NH�t>��~~��û��?.��?��Ϗo?}Y�}��~��w�7㍿)o�a|in\_�'�� qz��Ah�ږcg��,��Qm&�X�d��=�>J��*N�"��#��H޼�{��T�M��<)��4�eLL%&#��: � T��R鱡[[x��:ÞǏ9�{I��v�4�i��N\L��&��V�d�j�!�;�ˎ��8+�t r�b��!�q@~s��UV��jN�7�'Նy,�3;��-��sa� {��xS�"v֜EgZ]-&j��)�:��lÖ�{��O~��1N��/"��[�(�K��N��9��n��h�̒ɍd��$ip��:��?�޴_�t��1[&�`��S�{�F��r<��I� <�7i�}M�� >�I��wS0�˧�>��S'�@rQ�ݎ,��BcE�Y�A��e"ijqV�Hֱ�ʪ�i��FS�&V�ݤ7*�g7yD0?��S��L�H�,�H�ae�-C�X��]�Z��n�n:'t3��E��9��&��?��n��"��=y��M��{�r��sM��[��M�"��}ü[�Dw��z�F��S��A��C�ۗ��*��g$�2~��=�y鄂 +��o��[N�D�II�B��7@Rf�~y�I�>J1� /��6Mb,��\{{K�<Ϩ�.�"�o��B��A?��z�L��*v�ũD��e�M6Q��SOσ�^�)r�?Res�ǵ� /�x�,�nP&K��,�K�س��Y��O_#L��`L)H^��f��'�'��8=˼ �<��+�Z&�ĳM��Y�錨@k�*�6��h�ƙ֛X��q~sd�7�"�J��`*��5i騏��|�>W<��E��t~�Oq��O�x%q�G�TG(�7T��Xh�"�9��<�>Ց}�8�a�dMut��d��o�2 G� R��>��)�2�%�If�"�"�けqy��m�g�v��W%�_�3'�� +endstream endobj 215 0 obj<> endobj 216 0 obj<> endobj 217 0 obj<> endobj 218 0 obj<> endobj 219 0 obj<> endobj 220 0 obj<> endobj 221 0 obj<> endobj 222 0 obj<> endobj 223 0 obj<> endobj 224 0 obj<> endobj 225 0 obj<> endobj 226 0 obj<> endobj 227 0 obj<> endobj 228 0 obj<> endobj 229 0 obj<> endobj 230 0 obj<> endobj 231 0 obj<> endobj 232 0 obj<> endobj 233 0 obj<> endobj 234 0 obj<> endobj 235 0 obj<> endobj 236 0 obj<> endobj 237 0 obj<> endobj 238 0 obj<> endobj 239 0 obj<> endobj 240 0 obj<> endobj 241 0 obj<> endobj 242 0 obj<> endobj 243 0 obj<> endobj 244 0 obj<> endobj 245 0 obj<> endobj 246 0 obj<> endobj 247 0 obj<> endobj 248 0 obj<> endobj 249 0 obj<> endobj 250 0 obj<> endobj 251 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 6/PresSteps 259 0 R>> endobj 252 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�˥G��4��XL��X��X�,�0X �H2��7��"�3+��q�Df�{�6��]Û.�E:4:�eTp "��e<��Y<FՁWMZ�J"/��U��-#�Ndf(��To�v��m�?�c$w�xk�m?4'`��ԈB0S>�g�Z�5| +(va��v o�b�|�W��x*�2��!@a�`�@V �ˮ>�xt��fO�"��(��i��A;��!�nY�;"8� +�<<6�>��P�6t�� ˸U�k.g�&V=��v��*/񺸰��hm��|J<Ĥ�m^ͫu�Tf�mN�qJxWs�ѝaF��)��U��ظk��1��F(��H]��v��x�iFN �iv�F%G��Oe�.��쌌I'G�{��2a�c�/��&^��E�NM]��0)c7��pG*=9��S��ٕ�� 8,K4 ��n'e��uT��D�W�c2��1�͘��E�Ǝ��u��"�x�8�u;*�;��W�s4�ɩ>vNf��$t� �S�@F'�'� ��MY�K{��a[E�%��]+:{@�I��is��bL�E��P�0��mq��P!dh<�8Mg��]pw +8��W�c�9�M kTiL��V�T|PPt��0g�W��h ��66�;ύ�]��S��V�O,��U�²��͹��I��O��e�p�QNP$��t��s h L oP��M8U�c q��3%©:y��)aps֡S�9;I�*�X�=0Sa��%%E��'�C�� LNG,8e d?��aϭ��U+�eD�=�6 +��΁1�ԧF|Rǥ�W�A�xS�o�p\�]�G��L~�Ʃ0uY�-�"$LPH��Hx�z2�L�{¶�S$L6V�O�x�t�PM��bR�T�*�_�:E��@`h�x.��LEɃ��ڝK�;&I(�%E"�S"L��#��D8��Vi� h��m:�t�(��MDg�)ki)�*�*�.��#��L2��aE��|N"��MYBT%�8M��k��L��4�kq��~��6� ތS9.!��.�V��V݀ތj��9 +�6FX9�Uu��x: KH�Tw��Y�5�I}IƇ�K�M�m��Ԇ��y�͚K��"��4�5§�>^�Q�Cr��fq�ʡ�'.��J�2�Ŝԓ�5D(kb �h��.8�ȷ;�&`1�BK<�5Xդ+V8&��v-v?@ 5�J��aJx��:m��Zw��9��f kE��_黾� ${�*��uY%P:EVT ��w��85Ugȃ��W5�6M'ȣ��T��@��#�!w�>v��Q.N ��j�5� +��j�x^U;��F.d��9-{�ʇ,`�P?d�<��2<��aj��n�_�Ԅk�u�%�e��]��A �f��kT �k�̑��9��`��@��+a�I�x��k��ҁ9$j�S�,�ݔH�`�6{oS��[��q,�/ Ʃ��rK��2��F�� 0�4e<�]ک�`SŪx��l�� +�)Vv#��f�eS��Jж�:�&�鴏JjZe�{�QMX_��鴓[%��M�n$SA89��hsZ!X�Vr��a�ç�:øhv�԰N��q��s��S�at �E3�] UGgz�v�z_z��d�*)�Z�&��J١Z��4��2��۳s��W��k�+ۦ��gʵS��#��@kj��c��L�4 @�`;MK?��>�n��tx>��E4�'C��FcG�X��)��xi�g��p^�X��A�QA�Y3禆�P�K{$��s��[4 :嬐kI��Ks�t�Χ0ƙO��>�omj�[�� +�V�Rd�632��t�U�j�Ӕ�ܨv�Ĵґr:T�A��n�]�AT U!�]"��&/��9�V8��.��x�+G4&!Q&4=�K� ��0ŕ�D!��)��Li�h�1Y��W��UKymj8�5;��I` `�i �J�� U�� m��GS�W��Ú@�� CV�YR��:��C,��Β}�t��9T��8�Ѡ��԰Q%;;�@�bYv��aM��Gŭ��*� ��t��e7��<��ͻ�~ ��o��u��?�/WS��͗_?�<��>��^!4.�`�9��t��r�:q|�V�O�\ِ��}9�,ܽ��Oz��巏/>^��^~��~��OO_]��/��o>�g��?Ƒ�~b��@��|�/c�g��p��g��va(�*~��ܬF�aڢ��پ� S&��]�Z��mL�|,��5��yl�5X��iR��܅;z�;��n�:aDO#7�l�x�p��7®[<�ڊň��P�� +A��2O7�!)Z�u?��nH��c'-q#h� +O{"��Uq�\̻ +�=??CҸ��_a��ϋ-��L�tuL�}�w{-Y��ႢۃR�'=ГSw1Z�A�e�P��O�ˬiy�?Z��LQ}%�M/̯=�U�ށ��<H[�v!&ӟ!&�o�Q'��WV��L@��q��+�H��%�ez�wbQ�΂^/-��Nd��uƒ��O��c�a�^��Ѷ�*�d�5�팝��o�Jxw��L�1��N�ͨzO/�[��W_p!TCbf(�H��8Y��Y��ۤ��X!c�{tg�SmV��]�ͺ�f3Wҧe�['�+黛�y�h��|��A�j��+�`I4q4�!-��N��o�� +endstream endobj 253 0 obj<> endobj 254 0 obj<> endobj 255 0 obj<> endobj 256 0 obj<> endobj 257 0 obj<> endobj 258 0 obj<> endobj 259 0 obj<> endobj 260 0 obj<> endobj 261 0 obj<> endobj 262 0 obj<> endobj 263 0 obj<> endobj 264 0 obj<> endobj 265 0 obj<> endobj 266 0 obj<> endobj 267 0 obj<> endobj 268 0 obj<> endobj 269 0 obj<> endobj 270 0 obj<> endobj 271 0 obj<> endobj 272 0 obj<> endobj 273 0 obj<> endobj 274 0 obj<> endobj 275 0 obj<> endobj 276 0 obj<> endobj 277 0 obj<> endobj 278 0 obj<> endobj 279 0 obj<> endobj 280 0 obj<> endobj 281 0 obj<> endobj 282 0 obj<> endobj 283 0 obj<> endobj 284 0 obj<> endobj 285 0 obj<> endobj 286 0 obj<> endobj 287 0 obj<> endobj 288 0 obj<> endobj 289 0 obj<> endobj 290 0 obj<> endobj 291 0 obj<> endobj 292 0 obj<> endobj 293 0 obj<> endobj 294 0 obj<> endobj 295 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 7/PresSteps 302 0 R>> endobj 296 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�˥7��{��Y�ܪ�0��E�( ^$^��°��b$�>�z��ﳧi�Y��S�ͨ�)��t�,ҡѡ,�� !�i��0��j� +UyQ$<��!�t"3C�x��zC�c6o��#�xh�S3o��c��F��>��ء�S@��4vhx3�#<�4�S��.v +�@�J�^��Aǣ��45{JFAM3 ��: ��uȊ��@Vp��؅ +��3\EE'�Cƭ�]Cp9{5��tw�D�Ty��ŅUFk�D��x��S�!&n�j^�C�2#ms�S�C��3W�� +3H�M�j�]��C#|� ��6BA�FꊌgW�K<�+�H3rj�O��4*9�l�|*��'�fgdL:9��sj�˄�I��u��x5�Z�:5u�jR�Ӥ��J6�y��kfXN O�fWhښ�ܦ�,�4�߻��l�QIP��_ݎ�Ԗ��d7crJ��!;&S�)��)�� +HR��_��Iм$�F��9��%&HN I�l��$�&g�7eR,]��mu��G ��)'�#�˦�.6�:�h�=ѡ�a��B��x��4�!��v��Vp0�ُ��)2� s�֨0Ҙ>ŭ��*��4%a�03��#��Q�}lw�q� +��W,��U�²��ӏ�s�󓆏��9��~��u��W9A��~o�>��t��H@K`Jx�z�m©:�M�)N�� H�(oN �3��N��΁�I�T��3�YXRR$<}�?u�:��tƂS�@��B�7� �U�ҳj�y��H��l��O�T�90�b��ԈO�� +�2�B3oj� �+��ɏS�8��3+��T�� +�� WO&�i�'l�:E�dc%��7�N�T�9��?��׺N�Ĩ,1�*�K�<0SQ�$k�d��R�N�IJ~IA��T�Sq��i��Ӟ��*M��M��eB��l:e--WEB]��y��I��P�I��$�yڔ%DUb�sд:�&��Ξ�ɞ�@S��)��m��8��2��v��޶��fT�p��Q��1��A��#V��iXB¥z��bH��L�K2>�C"l�m�u �6 ��l�\��ԇ��>��:��'d6�{hT�>qI�T�,�䘠�t-�!BYKh�F�w��G��t +-�$P�`U��X��*۵��5��*��)�IRN��ZLj=4��~��B_|��zWAH��U2�J�0t��(a#�qj�ΐ'y5%<�j�m�N�gy55��v��F�G�C��C#|�� ]Gǳ\��ծk�'�Ք�vG��#�\Ȫ��#4 4�UsZ��q�Y2��~��d��(Ʈ��C��S��HP��q�kv�5��S��P5h�53G��H��16�:��'E�]X~?]�E��!Qk<�b�fq��D�ؕK��s��ܺ�c�8i0N �n�[�W��ccj�TvД�l�vi�~�M�n�i��1�+x*��LXٍ��1�M K{(A��t��>(�i�iP�5G5a}(��Nn�W�R4I�U�L��P[P��i�`�Zɡ +�M�.��IS�: f?�F��3�q&�NM��5�dv%T�a�9Aک�}��ݫ�@h��|<1�*ad�j�:<Ӱ�K��j��g�j)��)a�8�W�M��ϔk��Xg>��0��i�晬i�t�>v��~��Ϯ�>�n��tx>��E4�'C��Fcg�X��)��xi�g��p^�X��A�QA�Y3禆�P�K{$��s��[4 :嬐kI��Ks�t�Χ0ƙ��+|J��%o#�>��35~2N��WhT0�q��fSn�D��g�S\�fT��30Nj��@P3Y��a��Jva��*R�"S�Y$�5�r��P��E��*�B�'\vq�t�p$G�wn� �3��K+� ��;4<��"��B84�Lɮ�8��$ܡ�sC6{h��x�C��$��͐]]q�O�C��4�=�][�LX��wj�˄:!�W��+�)VYD�e�{�2j��Y4+�M��ʸ��qHx�d�VA�3Yqf�M �)v �I��J��dJxby�o��M%,�eq�/��y琮�XYb �9e2�vA�WǞ��)��#%�AU�Z�4%<7�]81�t��r�4��}��lUC�@�ꐈi��.n@oN��&C��b��G}=%��I�CԄ M��(Lq%�QH#jjxJ5;S#�kLV�p�s0啽dx4|�R^��k��$��xXxZB��`�eU3��l�c�!��;5u��&�m�0�"�v�J0K�sTǐu�=�2\�,��Ngɜ�Aeݎ� *�8�(.M U��4*�e��qT��^��ؠz�Mn�AW�~i�\v׍�7�� i�5y?^��}x��0�~�l��O��o��~�=�\��`��+x��Q~~�5�-�5�-D��xK\��lk��~[�k�%��nk��nk��nk�%:z�mM�xz�mM�Dz�mM�dz�mM�Tz�mM�tv�.��=�&[=�&[=�&[ +=�&[=�&[:;n�͞w[�-�w[�-��y�|�{��W_�y��wЀ_��ߞo�#[Wn��r5e?^}h��秷/Oo^^ 4n/?��5�_��?��;|��-8^�7~�+|��a�ҍB��>��Ͽ��姧o^�Y��;��y�v��r=�8��S��𺿌��k%^?@>��3�<�t&Ȑ��WKr�0� +�N&�~^ez�n��`�3�!�:��*L��-�W��&��q�g7{��1z�{��9D�8W4UC��=�#��I�.��!i��),Ν&�D2��m��o!�w ^G��jm��VNw��\"�??$��L��^?�ƫ}�"A��a�$Q%�p��:��EP��@pt��2@�;��^�؜�c�q��<ȋ� { +[D��ŗD7�U�bT��<��@�Z��e]��Ec��A��%��~x�_��߫,/v��jYW��o�S��-�{�I�&m1�֍��MB�V�â�88/<��d"�ʪ,W��ɠ��c4F4uߕN:�z��P��w�#�YU��s�ki-�K��Yv��n��?4��u��t��1��@p��tZ�5�r��9;y��ȇ� +endstream endobj 297 0 obj<> endobj 298 0 obj<> endobj 299 0 obj<> endobj 300 0 obj<> endobj 301 0 obj<> endobj 302 0 obj<> endobj 303 0 obj<> endobj 304 0 obj<> endobj 305 0 obj<> endobj 306 0 obj<> endobj 307 0 obj<> endobj 308 0 obj<> endobj 309 0 obj<> endobj 310 0 obj<> endobj 311 0 obj<> endobj 312 0 obj<> endobj 313 0 obj<> endobj 314 0 obj<> endobj 315 0 obj<> endobj 316 0 obj<> endobj 317 0 obj<> endobj 318 0 obj<> endobj 319 0 obj<> endobj 320 0 obj<> endobj 321 0 obj<> endobj 322 0 obj<> endobj 323 0 obj<> endobj 324 0 obj<> endobj 325 0 obj<> endobj 326 0 obj<> endobj 327 0 obj<> endobj 328 0 obj<> endobj 329 0 obj<> endobj 330 0 obj<> endobj 331 0 obj<> endobj 332 0 obj<> endobj 333 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 8/PresSteps 341 0 R>> endobj 334 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�ˤG��u� ��:hd�4 �� Y$�a�zGVWg�ZS#Bl�L�~��/�Ȩ�)��t�,ҡѡ,�� !�i��0��j� +UyQ$<��o�!�t"3C�x��zC�c6o��#�xh�S3o��c��F��ء�S@��4vhx3�#<�4�S��.v +�@�J�^��Aǣ��45{JFAM3 ��: ��uȊ��@Vp��؅ +��3\EE'�Cƭ�]Cp9{5��tw�D�Ty��ՅMo��L"�{Z=��)��y5��!R��9��)�!\͙+Fw��W��.�c�>v�p��G#uEƳ��%�r�95ܧ�Y�a�B>�Mx��c�32&�a�9��˄�I��ut�/0�j�,�uj�Zդl�I�� l��8R��|�̰��`ͮ$д-�M�a X�ix�w;)�7خ��l7%"��E��n��x-B4vL��S$ ��S��Q��75<��y1HN��s2{$�KL��0:�>AI�M�o�:�X:�f �*�,/<�\��RN:GF�M��D�3�`�-Z�D��adn�ӏ� +!C��t��wװ��~l}�H�q0��԰F��)n#H�UE�)�(s��q��@��:ocø�܈;�U�=��0��a�]E-�[^��_�?|�w��o�~�f �8~��Oۆۯr�" �>�m@A�w�:��8ۄSu:'�:S"�ک��A�8Qޜg0g�:��$Ω�+�g�T$faMI�� )�,��NY�ß��h�p�V�kϪ�2"��f�|��΁1�ԧF|S��7�A�xS�o�p^�]�G��L~�Ʃ0u�Y�-�"$LPH��Hx�z2�L�{¶�S$L6V�O��P��C5�#�I�S��xZ~��I��3f*J�d��tZ��0IBɯ)(�t��a*N�9�%�鴧`�J�'A;%��:��'d6�{hT�>qM�T�,�䘠�t-�!B�Kh�F�w��G��)4��Z�I�&��&]��1�?�U�k��j0��U��S��i3��zh��I�4Y+*��J��]!�V�l��&��)��J�l|h��Ʃ�:C��Ք�ɶi:A��7t�rqjx.V��Tp|&��U�#8z�a�BVm��i�q(��Ӳ'��|Ȓf �CƟ'X�G1v=L �� NM��[#A�C�i�ٕ:�`lf0N o�vR@ՠ��#��V��wau��t-�V:0�D��`�A�ա��cW,��Ǟ��`��v��z�K��]#|�`L��2��.��/��b� <�v6fwO�Ք +��@s�P3�)ai%h�x�N��t��%5�2 ��&lE��t�ɭ��P�&i� +�� jJ�9m,R+9T��e��a\4;ijX'��|�ȸCStr�9�Dک�0��"��̮��3='H;U��`UP�{�h�_��'FS%��P-Y�gvq �Z�e��ٹZʋ�yJ�5�mS��3�کa�֙��@kj��c��L�4 @�`;MK��>�n��tx>��E4�'C֢Fcg�X��)��xi�g��p^�X��A�QA�Y3禆�P�K{$��s��[5 :嬐kI��Ks�t�Χ0ƙ��+|J��%o#�>��35~2Np�� Ш` +��ܦ��'@�X��ͨ��g`��$��f� +� 5ø�� *'T��E��H�kN��!��i�/��U��pK��d�4�H��4f^g�6>' whx`'E��=�C +�ЈB0Q$��3>$� ��Y�C3��s�p��'i6o��c��;dŮ!8iV�\ V�L O�"/�=��2��E�,n��7p� +��5R�S�S��м,�n�`�\�ϓX�٥��)v��%1��o�q�:��<*��jWrh��e\éϦ&��$��j��G�]��ձ��p�aC*E�j3� �HI�hPU�V>M ύjNL)�C��4 o�fߥ1D�P%�:$b�k򂫫Лl��ɐj��Q_O��rDc�5aBӃ>��@<1; +S��(�55<��)��%&�@8�9��^2<�j)�M �f�T��q< �<-�V�v0��q6��h�򊝚:�CxX�_d�cw�J0K�sTǐu�{�e�>�Y��Β9�ʺG9T�qQ\�6�dg�hT,�Λ<5��㨸��A�C��P��z�Ҁ��^=�q��i�5y?^��[�<��~��/�m�w�_~��_ny�<�\��`��x_W�[��k�[nk�Z�l��0�m��d��:>��dK��dK��dK��xKt��l��l��l��l��l��]79z�}M�z�}M�$z�}M�z�}M�4z�}M�tv�.��==�&["=��[�<��׫�/_}��B~��7ߎ��q��e[��G��Ք�x��y��o�^==Ah,O��ގ_��.O�B��+��>K<�l]�>>��ϟ�,�>/��ay��o�w�{��߿[>�w��ӏ��x�/�{C~{�Ƒ�m�6��?��?�g��r��0|�9�Jl��/�J��{7u�!�Kơ��I �T�կ��?�1�s�_�(�<�ͱ�q��(N8�?��鈘Nz� +�" +>��]/b��+��;%DTqA�RE}_�h/Y�pYG�S�JF��@�}kw��n�K��ҟ:��V��0h��cӉ{ۍ�1�s�kS�c�#h +�pG�f�R�!�l��R.�OYr�<�^�%}��"��P=�'��!�J��poJ��%�z��͉#V�p��(�ĩ� +�Qr��͍��E�_џp��ۼIEFTNM��:�yicR~�$��h�6X��&�q݊؇�+5(��>ɳ�?�<��,0�S�ƥ\�i��k%b� �I�5ɤ��7��{k��m�\��/U�9��QU ��ӌ��Be�]��2��U��E��?*r�Rp�^�o�H;��i�]��̥~+//y�q�ԃ5糧[��C��78�2��m}�Ӽs@��)� ��`�E�& +endstream endobj 335 0 obj<> endobj 336 0 obj<> endobj 337 0 obj<> endobj 338 0 obj<> endobj 339 0 obj<> endobj 340 0 obj<> endobj 341 0 obj<> endobj 342 0 obj<> endobj 343 0 obj<> endobj 344 0 obj<> endobj 345 0 obj<> endobj 346 0 obj<> endobj 347 0 obj<> endobj 348 0 obj<> endobj 349 0 obj<> endobj 350 0 obj<> endobj 351 0 obj<> endobj 352 0 obj<> endobj 353 0 obj<> endobj 354 0 obj<> endobj 355 0 obj<> endobj 356 0 obj<> endobj 357 0 obj<> endobj 358 0 obj<> endobj 359 0 obj<> endobj 360 0 obj<> endobj 361 0 obj<> endobj 362 0 obj<> endobj 363 0 obj<> endobj 364 0 obj<> endobj 365 0 obj<> endobj 366 0 obj<> endobj 367 0 obj<> endobj 368 0 obj<> endobj 369 0 obj<> endobj 370 0 obj<> endobj 371 0 obj<> endobj 372 0 obj<> endobj 373 0 obj<> endobj 374 0 obj<> endobj 375 0 obj<> endobj 376 0 obj<> endobj 377 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 9/PresSteps 382 0 R>> endobj 378 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�ˤ7��w��>�T��:5��I`� 3��ʊ��(B�I2_vu��/2*l +e��7]6�tht(˨�<@D��s�xZe�x"��B�D^ ��,��{(F�;��P<$>��>z��H*��~h.�-��`�|��4vh�P��D}-��@�L��TvNf��$t� �S�@F'�'� ��MY�KG��a[E�%��C+:{@�I��i�K�M1�&ڢEOt�xF�8�h�24~;Mg��]p��h�c�+D��1ȜĦ�5*�4�Oq+A*>� +(:M GA�3̌+�H��G�yƝ�F܁��~F~��z��y=�{?��\��hb��H�C��7rP��]�΁d$�%0%�A=�6�T��ĉ�Δ�v��u$N�7��Y�NA�@�$�s�� + +��,,))�>��:E�ar:c�)k s�}��M�ܪy�Y�¼PF��h�Qȧo��C1K}j�'u^�}AD��75�� u~�I��ǩa� +Sי�R*B��d�k��ǫ'�ȴ��U�"a��|JěJ�o��YLʟ +�A��k]�HbT�MϥX��(y��S�]\��i�w��$ %��H��i�C��8y�K��iO�n�&O�vJxЦ�H�Xс2�N��Dt6��"��b�v��Y.N ��j�5� +��j�x^U;��F.d��9-{�ʇ,`�P?d�~��exc��ީ ��vK$(|�8�5�R��)��N +�4ך�#S�s��rW��"�.,��Ţ[K搨5L1H�8�wS"}�J��x�M n�Ʊ|��4��V��-�+\�|v��1�`*;h�x6v��S��U7��٘�<VS&��F�B͘˦��=��m�u:M�i�Դ�4��>��i'�J�+@)��*H��pr�-(��B�H��P��&�O�u +�q�줩a�� #�M��8i��f2��0�� T��UA��UR� �~M>�M�0�C�d�i��%dj��ӳs��W��k�+ۦ^�gʵS��3��@kj��c��L�4 @�`;MK��g�p�`7Qu:<��L��"È�� KQ�1��h�Q�p +Wz��̅�S8/ �,U~N�� Ш 㬙sS�[(إ=�_͹�I�-��rVȵ��i�ȥ�9h�L �S��w��>�Omj�S��<�� ~�o�C��g��Ϸ~�n?ܞ|�� +�0��+x��Q��k�[�k�Z�l��0�m��d��:>��dK��dK��dK��xKt��ۚl��ۚl��ۚl��ۚl��ۚl��]79z�mM�z�mM�$z�mM�z�mM�4z�mM�tv�.��==�&["=��[><�{�|�{��ۗ_�y��Ѐ_}��=��G��#��j�~��|��ǧ�/Oo^^ 4n/?��5�_��?��;|��|��?��ஆ ��:`X�t#��_~y�û��ϟ��?��/o��Ȟ9lG��/y�/��#�F?� iȯ�˘N��Z?@>�<��6�<�c��/��a��TM��5v�� "A2%�o?�T�2:vAX��?��Ne��j1�G-hw!��&hsC��4p% \��$q��fX��(��iِ�QƢ�E�V�%�BC,��vWy�Ζ�d!�t��"�MWP1Q�&'��;��<��Њ�_��`p6 +endstream endobj 379 0 obj<> endobj 380 0 obj<> endobj 381 0 obj<> endobj 382 0 obj<> endobj 383 0 obj<> endobj 384 0 obj<> endobj 385 0 obj<> endobj 386 0 obj<> endobj 387 0 obj<> endobj 388 0 obj<> endobj 389 0 obj<> endobj 390 0 obj<> endobj 391 0 obj<> endobj 392 0 obj<> endobj 393 0 obj<> endobj 394 0 obj<> endobj 395 0 obj<> endobj 396 0 obj<> endobj 397 0 obj<> endobj 398 0 obj<> endobj 399 0 obj<> endobj 400 0 obj<> endobj 401 0 obj<> endobj 402 0 obj<> endobj 403 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 10/PresSteps 410 0 R>> endobj 404 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�˥G��$r��,/ܞHHS�� m�X�H��ʊ��j[�v�ĩ�y��fT�� o�l��P�Q�y��f�DU^5i�*��(WY^��P��w:��xH<��>z��T<4⩙��c��F��[i��)�؅ ��V;4��4�S��.v +�@�J�^��Aǣ��45{JFAM3 ��: ��uȊ��@Vp��؅ +��3\EE'�Cƭ�]Cp9{5��tw�D�Ty��ՅM��L"�=�u��x�I�ۼ�W��Hۜ��W��.�7c�>v�p6BA�FꊌgW�K<�{ȑf��p�fg1hTr�� +�T6�2N��Ștr��&V8&� +_x��-��īQԲHש�kU��&e�6T��H�'g�]3�rJx�5��@Ӷ�6�%`��L�`��J��ݔ��vL��5&��S��1��N�4TO�nGev��$��͛Arj�� ]b��0�� :H�lrxS�!��E0k�VQg��px�y��r�92�l��$¦�AmѢ':T<#s[�~4T��3dt�.��h�c�D��1ȜĦ�5*�4�OqA*>� +(:M GA�3̌+�H��G�yƝ�F܁��#��7��*ja��z��}~z�>9W?�4�~�f �8��-�� ׿� �$8�7�E� +��K�9��7��&��1�8�ԙ��N��ĉ��08�9��)��$qN�Z��b�� 1kJ��O�NQgA��Xp��^��F�#�j^{V�0/��x4�(��/�t��>5�M��z�@PQhf�M �a�yv�aR3�qj��uf��0A!��"��$2�n ۪N�0�XI>%�K��/�t�,&�OQ��Z�)��5CS�s)�W�T�<��)�/��Ի�a��_SP$��4�!�T�4�FF��T�!O�jJx^�d�4� ��jj��v��F�G�C��C#|�� ]Gǳ\��ծk�'_ɫ)�yU��^'G��U۲Ghhʫ��*�d�YC��Q�]Sï�fw|�S��HP��q�kv�5��S��P5h�53G��H��16�:��'E�]X�>]�E��!Qk<�b�fu��D�ؕK�j��65��u;��85<��]n��å��>v0�v��2��.��`SŪx��l��<VS&��F��1�M K{(A��t��>(�i�iP�5G5a{(��Nn�7�R4I�U�L��P[P��i�`�Zɡ +�M�.��IS�: f?�F��3�q&�NM��5�dv%T�a�9Aک�}��ݛ�@h��|<1�*ad�j�:<Ӱ�k��&O��R��S®q��l�z?S��>0`��=�O��a�ɚ�H�c�i��3�i�O��:��|�F�x�a��Є��Y4֨w +8�+=^��Y�B�)�ׄ��*?�`whT�q�̹��-��ɯ�\�$�VMF9+�Z��4D�R�4]��)�q�i�>�Omj�S��>��35��(E�+��C��!��=4�j<��whx�f�fȮn8��O�C��4�=�]ۈLX��wju� +tB:�f�7(S��ʲ��e�D��!�hV��ݩ��q��q��$�v��Fg��ޛ�S��f�ɕ`eɔ��*��*�JX$��_|�y���@�s�x��E�U ��+�u0�t|a2!Ŏ�1�%��2U�0��Ge��!�V�Jmu��k8��4��q�dq[��B#�zC��@� �_�:��|N1lH��Pmfd)� ��ʧ)�Q�i�#�t� ��4f��BT�DLsM^puuzs��p4R C]=��)�@W�hLB�&LhzЇv�'fGa��4��T�s0�1��d;S��K�G�W-嵩��J��:�'�5 ��%�*�&XV5;�&<�M]^�SSGpk��,r�Y fIu��`v�ׇ:Kv��,�s0��q��Ae�ť�a�Jvv�FŲ��SÚ8;;��x�KT;T�� 5��/ �ˮ��Ǐny��Q�F�0��j��Fs��e*A��K�A:�"��O:�C��&�i��3ʺ�Wz��׮�Z$詢N��v�v +9�z��l��S�ư�1\^/ƛ�4/��P��(�[ń�~��'��]o[V� +J\��r��+��_YQ͘�72��F��ڎP#�e�[�%'�%�ȩ��+�(T*7�N��mޞ��_��]� +endstream endobj 405 0 obj<> endobj 406 0 obj<> endobj 407 0 obj<> endobj 408 0 obj<> endobj 409 0 obj<> endobj 410 0 obj<> endobj 411 0 obj<> endobj 412 0 obj<> endobj 413 0 obj<> endobj 414 0 obj<> endobj 415 0 obj<> endobj 416 0 obj<> endobj 417 0 obj<> endobj 418 0 obj<> endobj 419 0 obj<> endobj 420 0 obj<> endobj 421 0 obj<> endobj 422 0 obj<> endobj 423 0 obj<> endobj 424 0 obj<> endobj 425 0 obj<> endobj 426 0 obj<> endobj 427 0 obj<> endobj 428 0 obj<> endobj 429 0 obj<> endobj 430 0 obj<> endobj 431 0 obj<> endobj 432 0 obj<> endobj 433 0 obj<> endobj 434 0 obj<> endobj 435 0 obj<> endobj 436 0 obj<> endobj 437 0 obj<> endobj 438 0 obj<> endobj 439 0 obj<> endobj 440 0 obj<> endobj 441 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 11/PresSteps 448 0 R>> endobj 442 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�˥G��)�$��~H�Y��\`5�H��6�A#��ʊ�Ϭ��'2��ͨ�)��t�,ҡѡ,�� !�i��0��j� +UyQ$<��!�t"3C�x��zC�c6o��#�գ.�1�p�W�j"�i��x��՜�bt�0#�فtߔ��v�E�jl<4��F(��H]��v��9Ҍ��,�J�0[!��&�C�ɱ��N�0��Z�d� +Ǥ_�G��L�E-�t��V5)�iR�naC%��<�Tzr�53,��'X�+ 4m�An�qX�h��N�� $(�M��n�djkQc��19%�A��ICE`�q��vTf�HR��_��Iм$�F��9��%&HN I�l��$�&g�7eR,]��mu��G ��)'�#�˦�M"l�t0�-z�C��02��GC��4�!��v��4��`F�_[� Rd�A�$65�Qa�1}��R�AU@�iJ8 +af\�W�5�?��0�<7�tnO�3�{-�?�@�oWQ��˧��7��' ?D3s?��o��m��G9A��~rP��]�΁d$�%0%�A=�6�T��ĉ�Δ�v��u$N�7��Y�NA�@�$�s��cw�L�Ab֔ O��O��΂09��5�9�\��F�#�j^{V�0/��x4�(�Ӈj:�P�R��I��z�@PQhf�M �a�yv�aR3�qj��uf��0A!��"��$2�n ۪N�0�XI>%��C5�#�I�S�?�xZ~��I��3%O�vJ��p:-��t�$��p:Mu�0'Ϝn�t�S�[�ɓ��4�!Vt�L�#4�M��H�� X>�S0�|��-yR�9�|�6e Q��4�� 3��s�g,�Ԯ�e2�}Jx��n�4�x3N帆L?��[eo[uz3��B�J��(|�M`� V�+��4�!�R=d|^�f1$�t&�%�!6Ŷ�:�R��q6k.�{J�H�C��f�xF�M�C�2��=4*�R��B*a�xDsrLPO:��l�%4p�Ղ��p�#�n��t +-�$P�`U��X��*۵��5��*��)�IRN��ZLj=4��~��B_}�� d6�u�@a�YQ%P6>4�FF��T�!O�jJx^�d�4� ��jj��S�r� +�ԇ�ŇF��g�85<�]��O*8�ɫ)�yU��^'G��U۲Ghhʫ��*�d�YC��d��(Ʈ��C��~�S��HP��q�kv�5��S��P5h�53G��H��16�:��'E�]X~��Ţ�J搨5L1H�:�wS"}�J��x�M o�Ʊ��4��V��-�G��4��S�ASƳ�ۥ��l�XuO��=��jʄ��H�y�P3�)ai%h�x�N��t��%5�2 ��&l/E��t�ɭ��P�&i� +�� jJ�9m,R+9T��e��a\4;ijX'��|�ȸCStr�9�Dک�0��"��̮��3='H;U��`UP�{�h�_��'FS%��P-Y�gvq �Z�d��ٹZʫ�yJ�5�mS�gʵS��3��@kj��c��晬i�t�>v��?3��gj��F�MX��!�Ec�:p��S��㥍�e.��yM(`��s +v�Fag͜��B�.��j�L�n�$�`��B�%�OCD.e�A�eJ8��g>Mˏ�)}kS��Z��4�) ��05�>�kl�F��\��5%E�&I�cm��B[¹�"�(��xN 1NI�'��WhT0�q��fSn�D��w�S\�fT��30Nj��@P3Y��a��Jva��*R�"S�Y$�5�r��P��E��*�B�&\vq�t�p$G�wn� ?��٥�Ƅ�I��D��cw|H!Q&�dW6��C��!��=4�j<��whx�f�fȮn8vǇ'�!�a��ʮmD&�`�;��Ʉ:!�W��)VYD�e�{�2j��Y4+�M��ʸ��qHx�d�VA�3Yqf�M �)v �I��J��dJxby��M%,�eq�/~�y��@�s�x��E�U ��+�>��]:��bG��^S�~��2U�0��Ge��!�V�Jmu��k8��4��q�dq[��B#�zC��@� �_�:��|N1lH��Pmfd)� ��ʧ)�Q�i�#�t� ��4f��BT�DLsM^puuzs��p4R C]=��)�@W�hLB�&LhzЗv�7fGa��݉BQS�S��9��^c� ��)o�%ã᫖��p^kvH%Y]Ǔ��j�n,��g� ��.�ة�#8��5��A9v��:GuYW0��X��C�%;��,�s0��q��Ae�ť�a�Jvv�FŲ��SÚ8;;��x�KT;T�� 5��/ �ˮ��<X��p�6�&��W b��z�o��ȶ��◟��?,��[>.O>�qz��~��+x��Q��k�[�k�Z�l��0�m��d��:>�\�dK��dK��dK��xKt��l��l��l��l��l��]79z�}M�z�}M�$z�}M�z�}M�4z�}M�tv�.��==�&["=��[><�{��\�z��7o�?��;h�o�}��ޟ��_��[}��\Mُ�>4/_xyz{yzs�@h,�a'dq��0��]��/�k��W��|��ʆ<��e��B��~}��ey��/_��/��>}�[p�W�7��]�c��٣��Q��?�ܯG�?�U��&�|ŕɌ� �ƯK�"�k�� 4�D��Tf��N��Q�nh��i �5}y�R�k�U-Y5E�%fe��?Qj�֟��v|20��m� �$��c:��r��UL��$��|��U�jݙj�,��N۸yb��/��g��%��.��[��2ņ⮄�vsQQ�c��(>�� OM�וAޮԙ�D�tB��b��Ů�$�q�ɦ%��!2��ԨE�۫?o�C��S��>u�(�+�e��d��8��ص�>lP��1�Q�W�>��=�7��jr�;J&��:*Z9n��v5qP��K�QO@c��N��ݗ +c��Fb�@��?"�H�{-\�`�&S� +endstream endobj 443 0 obj<> endobj 444 0 obj<> endobj 445 0 obj<> endobj 446 0 obj<> endobj 447 0 obj<> endobj 448 0 obj<> endobj 449 0 obj<> endobj 450 0 obj<> endobj 451 0 obj<> endobj 452 0 obj<> endobj 453 0 obj<> endobj 454 0 obj<> endobj 455 0 obj<> endobj 456 0 obj<> endobj 457 0 obj<> endobj 458 0 obj<> endobj 459 0 obj<> endobj 460 0 obj<> endobj 461 0 obj<> endobj 462 0 obj<> endobj 463 0 obj<> endobj 464 0 obj<> endobj 465 0 obj<> endobj 466 0 obj<> endobj 467 0 obj<> endobj 468 0 obj<> endobj 469 0 obj<> endobj 470 0 obj<> endobj 471 0 obj<> endobj 472 0 obj<> endobj 473 0 obj<> endobj 474 0 obj<> endobj 475 0 obj<> endobj 476 0 obj<> endobj 477 0 obj<> endobj 478 0 obj<> endobj 479 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 12/PresSteps 486 0 R>> endobj 480 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�˥G��)�$��~H�Y��\`5�H��6�A#��ʊ�Ϭ��'2��ͨ�)��t�,ҡѡ,�� !�i��0��j� +UyQ$<��!�t"3C�x��zC�c6o��#�գ.�1�p�W�j"�i��x��՜�bt�0#�فtߔ��v�E�jl<4��F(��H]��v��9Ҍ��,�J�0[!��&�C�ɱ��N�0��Z�d� +Ǥ_�G��L�E-�t��V5)�iR�naC%��<�Tzr�53,��'X�+ 4m�An�qX�h��N�� $(�M��n�djkQc��19%�A��ICE`�q��vTf�HR��_��Iм$�F��9��%&HN I�l��$�&g�7eR,]��mu��G ��)'�#�˦�M"l�t0�-z�C��02��GC��4�!��v��4��`F�_[� Rd�A�$65�Qa�1}��R�AU@�iJ8 +af\�W�5�?��0�<7�tnO�3�{-�?�@�oWQ��˧��7��' ?D3s?��o��m��G9A��~rP��]�΁d$�%0%�A=�6�T��ĉ�Δ�v��u$N�7��Y�NA�@�$�s��cw�L�Ab֔ O��O��΂09��5�9�\��F�#�j^{V�0/��x4�(�Ӈj:�P�R��I��z�@PQhf�M �a�yv�aR3�qj��uf��0A!��"��$2�n ۪N�0�XI>%��C5�#�I�S�?�xZ~��I��3%O�vJ��p:-��t�$��p:Mu�0'Ϝn�t�S�[�ɓ��4�!Vt�L�#4�M��H�� X>�S0�|��-yR�9�|�6e Q��4�� 3��s�g,�Ԯ�e2�}Jx��n�4�x3N帆L?��[eo[uz3��B�J��(|�M`� V�+��4�!�R=d|^�f1$�t&�%�!6Ŷ�:�R��q6k.�{J�H�C��f�xF�M�C�2��=4*�R��B*a�xDsrLPO:��l�%4p�Ղ��p�#�n��t +-�$P�`U��X��*۵��5��*��)�IRN��ZLj=4��~��B_}�� d6�u�@a�YQ%P6>4�FF��T�!O�jJx^�d�4� ��jj��S�r� +�ԇ�ŇF��g�85<�]��O*8�ɫ)�yU��^'G��U۲Ghhʫ��*�d�YC��d��(Ʈ��C��~�S��HP��q�kv�5��S��P5h�53G��H��16�:��'E�]X~��Ţ�J搨5L1H�:�wS"}�J��x�M o�Ʊ��4��V��-�G��4��S�ASƳ�ۥ��l�XuO��=��jʄ��H�y�P3�)ai%h�x�N��t��%5�2 ��&l/E��t�ɭ��P�&i� +�� jJ�9m,R+9T��e��a\4;ijX'��|�ȸCStr�9�Dک�0��"��̮��3='H;U��`UP�{�h�_��'FS%��P-Y�gvq �Z�d��ٹZʫ�yJ�5�mS�gʵS��3��@kj��c��晬i�t�>v��?3��gj��F�MX��!�Ec�:p��S��㥍�e.��yM(`��s +v�Fag͜��B�.��j�L�n�$�`��B�%�OCD.e�A�eJ8��g>Mˏ�)}kS��Z��4�) ��05�>�kl�F��\��5%E�&I�cm��B[¹�"�(��xN 1NI�'��WhT0�q��fSn�D��w�S\�fT��30Nj��@P3Y��a��Jva��*R�"S�Y$�5�r��P��E��*�B�&\vq�t�p$G�wn� ?��٥�Ƅ�I��D��cw|H!Q&�dW6��C��!��=4�j<��whx�f�fȮn8vǇ'�!�a��ʮmD&�`�;��Ʉ:!�W��)VYD�e�{�2j��Y4+�M��ʸ��qHx�d�VA�3Yqf�M �)v �I��J��dJxby��M%,�eq�/~�y��@�s�x��E�U ��+�>��]:��bG��^S�~��2U�0��Ge��!�V�Jmu��k8��4��q�dq[��B#�zC��@� �_�:��|N1lH��Pmfd)� ��ʧ)�Q�i�#�t� ��4f��BT�DLsM^puuzs��p4R C]=��)�@W�hLB�&LhzЗv�7fGa��݉BQS�S��9��^c� ��)o�%ã᫖��p^kvH%Y]Ǔ��j�n,��g� ��.�ة�#8��5��A9v��:GuYW0��X��C�%;��,�s0��q��Ae�ť�a�Jvv�FŲ��SÚ8;;��x�KT;T�� 5��/ �ˮ��<X��p�6�&��W b��z�o��ȶ��◟��?,��[>.O>�qz��~��+x��Q��k�[�k�Z�l��0�m��d��:>�\�dK��dK��dK��xKt��l��l��l��l��l��]79z�}M�z�}M�$z�}M�z�}M�4z�}M�tv�.��==�&["=��[><�{��\�z��7o�?��;h�o�}��ޟ��_��[}��\Mُ�>4/_xyz{yzs�@h,�a'dq��0��]��/�k��W��|��ʆ<��e��B��~}��ey��/_��/��>}�[p�W�7��]�c��٣��Q��_~=m�?�U��6�}�%�JsN��^�8�HI5��#Q�39�c{�X,.�i��fM[-}��Z��;��O�js��;�!ؘ��>��v��4^m��Y�-ȡ�M�DT�7� C��O�FM��L�V��Uz�]��R��k��<*h�1��V(=t�Np�N�N`� �� {O�o�@��H��<��hx4M�@?�Ώ��q�>˦a�_U�+3cx=�#�yzF9�6��H��E��.�Y��0��2�6�e&huM��v�<��,ڞ��Zѧ��7�-D��Ph��$Z:W�d�&��X4fEdA�TVea�~QAq92��~Ȋ7� _��#��D'��*� �H�:ک�[�ץ��UӽS�h=��;;��*Uݒ��d�o��[-Z�/�\��x.2��#��z�[��e�Ӭ��'� ��$:�FC�Js2�z_j؜��$t� ��#�� R��T��v� +endstream endobj 481 0 obj<> endobj 482 0 obj<> endobj 483 0 obj<> endobj 484 0 obj<> endobj 485 0 obj<> endobj 486 0 obj<> endobj 487 0 obj<> endobj 488 0 obj<> endobj 489 0 obj<> endobj 490 0 obj<> endobj 491 0 obj<> endobj 492 0 obj<> endobj 493 0 obj<> endobj 494 0 obj<> endobj 495 0 obj<> endobj 496 0 obj<> endobj 497 0 obj<> endobj 498 0 obj<> endobj 499 0 obj<> endobj 500 0 obj<> endobj 501 0 obj<> endobj 502 0 obj<> endobj 503 0 obj<> endobj 504 0 obj<> endobj 505 0 obj<> endobj 506 0 obj<> endobj 507 0 obj<> endobj 508 0 obj<> endobj 509 0 obj<> endobj 510 0 obj<> endobj 511 0 obj<> endobj 512 0 obj<> endobj 513 0 obj<> endobj 514 0 obj<> endobj 515 0 obj<> endobj 516 0 obj<> endobj 517 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 13/PresSteps 524 0 R>> endobj 518 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�˥G��)�$��~H�Y��\`5�H��6�A#��ʊ�Ϭ��'2��ͨ�)��t�,ҡѡ,�� !�i��0��j� +UyQ$<��!�t"3C�x��zC�c6o��#�գ.�1�p�W�j"�i��x��՜�bt�0#�فtߔ��v�E�jl<4��F(��H]��v��9Ҍ��,�J�0[!��&�C�ɱ��N�0��Z�d� +Ǥ_�G��L�E-�t��V5)�iR�naC%��<�Tzr�53,��'X�+ 4m�An�qX�h��N�� $(�M��n�djkQc��19%�A��ICE`�q��vTf�HR��_��Iм$�F��9��%&HN I�l��$�&g�7eR,]��mu��G ��)'�#�˦�M"l�t0�-z�C��02��GC��4�!��v��4��`F�_[� Rd�A�$65�Qa�1}��R�AU@�iJ8 +af\�W�5�?��0�<7�tnO�3�{-�?�@�oWQ��˧��7��' ?D3s?��o��m��G9A��~rP��]�΁d$�%0%�A=�6�T��ĉ�Δ�v��u$N�7��Y�NA�@�$�s��cw�L�Ab֔ O��O��΂09��5�9�\��F�#�j^{V�0/��x4�(�Ӈj:�P�R��I��z�@PQhf�M �a�yv�aR3�qj��uf��0A!��"��$2�n ۪N�0�XI>%��C5�#�I�S�?�xZ~��I��3%O�vJ��p:-��t�$��p:Mu�0'Ϝn�t�S�[�ɓ��4�!Vt�L�#4�M��H�� X>�S0�|��-yR�9�|�6e Q��4�� 3��s�g,�Ԯ�e2�}Jx��n�4�x3N帆L?��[eo[uz3��B�J��(|�M`� V�+��4�!�R=d|^�f1$�t&�%�!6Ŷ�:�R��q6k.�{J�H�C��f�xF�M�C�2��=4*�R��B*a�xDsrLPO:��l�%4p�Ղ��p�#�n��t +-�$P�`U��X��*۵��5��*��)�IRN��ZLj=4��~��B_}�� d6�u�@a�YQ%P6>4�FF��T�!O�jJx^�d�4� ��jj��S�r� +�ԇ�ŇF��g�85<�]��O*8�ɫ)�yU��^'G��U۲Ghhʫ��*�d�YC��d��(Ʈ��C��~�S��HP��q�kv�5��S��P5h�53G��H��16�:��'E�]X~��Ţ�J搨5L1H�:�wS"}�J��x�M o�Ʊ��4��V��-�G��4��S�ASƳ�ۥ��l�XuO��=��jʄ��H�y�P3�)ai%h�x�N��t��%5�2 ��&l/E��t�ɭ��P�&i� +�� jJ�9m,R+9T��e��a\4;ijX'��|�ȸCStr�9�Dک�0��"��̮��3='H;U��`UP�{�h�_��'FS%��P-Y�gvq �Z�d��ٹZʫ�yJ�5�mS�gʵS��3��@kj��c��晬i�t�>v��?3��gj��F�MX��!�Ec�:p��S��㥍�e.��yM(`��s +v�Fag͜��B�.��j�L�n�$�`��B�%�OCD.e�A�eJ8��g>Mˏ�)}kS��Z��4�) ��05�>�kl�F��\��5%E�&I�cm��B[¹�"�(��xN 1NI�'��WhT0�q��fSn�D��w�S\�fT��30Nj��@P3Y��a��Jva��*R�"S�Y$�5�r��P��E��*�B�&\vq�t�p$G�wn� ?��٥�Ƅ�I��D��cw|H!Q&�dW6��C��!��=4�j<��whx�f�fȮn8vǇ'�!�a��ʮmD&�`�;��Ʉ:!�W��)VYD�e�{�2j��Y4+�M��ʸ��qHx�d�VA�3Yqf�M �)v �I��J��dJxby��M%,�eq�/~�y��@�s�x��E�U ��+�>��]:��bG��^S�~��2U�0��Ge��!�V�Jmu��k8��4��q�dq[��B#�zC��@� �_�:��|N1lH��Pmfd)� ��ʧ)�Q�i�#�t� ��4f��BT�DLsM^puuzs��p4R C]=��)�@W�hLB�&LhzЗv�7fGa��݉BQS�S��9��^c� ��)o�%ã᫖��p^kvH%Y]Ǔ��j�n,��g� ��.�ة�#8��5��A9v��:GuYW0��X��C�%;��,�s0��q��Ae�ť�a�Jvv�FŲ��SÚ8;;��x�KT;T�� 5��/ �ˮ��<X��p�6�&��W b��z�o��ȶ��◟��?,��[>.O>�qz��~��+x��Q��k�[�k�Z�l��0�m��d��:>�\�dK��dK��dK��xKt��l��l��l��l��l��]79z�}M�z�}M�$z�}M�z�}M�4z�}M�tv�.��==�&["=��[><�{��\�z��7o�?��;h�o�}��ޟ��_��[}��\Mُ�>4/_xyz{yzs�@h,�a'dq��0��]��/�k��W��|��ʆ<��e��B��~}��ey��/_��/��>}�[p�W�7��]�c��٣��Q��_~=m�?�e��&��{�B�f6�{g:ә^{��"��1e��eH�dOq��i�q�|M]-~��Z��3^q�Ş�y�:��4��_��ᝂ?��/�(�֏e&��GV�C0փA��^sx�oJ��7�#Qү��:�֬d�.؉h�!� ��$�FFCs�<"��Z}�(S�w� (��^��d��6�$�M��`l/��|pNG~r�{W��(��X��EW�o�r�<��Bc]p��6=?��ΐ�˲䋅�k��EZ��W]t��iG��nw�-��ݟ�JG��g�7"�N��MU��'��M��nʣིd�z�G�c^DU��_%_�d�ז�"xJ��С2ڜ�N]��Z7��sr(s�3��X��|V�a��@ͧȊ�6��k�o�7MC��pf�) i4��o��6�YC@G;�5�#�,3y��]?�-�J�V.��MG��C�T&?�Bk͉��{ڜ|EZ.HgO�Muk�(��.7ّ�T�4C��,�<� ��` d��k�V��9E�G��V7�*�7&x��q��>�L +endstream endobj 519 0 obj<> endobj 520 0 obj<> endobj 521 0 obj<> endobj 522 0 obj<> endobj 523 0 obj<> endobj 524 0 obj<> endobj 525 0 obj<> endobj 526 0 obj<> endobj 527 0 obj<> endobj 528 0 obj<> endobj 529 0 obj<> endobj 530 0 obj<> endobj 531 0 obj<> endobj 532 0 obj<> endobj 533 0 obj<> endobj 534 0 obj<> endobj 535 0 obj<> endobj 536 0 obj<> endobj 537 0 obj<> endobj 538 0 obj<> endobj 539 0 obj<> endobj 540 0 obj<> endobj 541 0 obj<> endobj 542 0 obj<> endobj 543 0 obj<> endobj 544 0 obj<> endobj 545 0 obj<> endobj 546 0 obj<> endobj 547 0 obj<> endobj 548 0 obj<> endobj 549 0 obj<> endobj 550 0 obj<> endobj 551 0 obj<> endobj 552 0 obj<> endobj 553 0 obj<> endobj 554 0 obj<> endobj 555 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 14/PresSteps 562 0 R>> endobj 556 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�˥G��)�$��~H�Y��\`5�H��6�A#��ʊ�Ϭ��'2��ͨ�)��t�,ҡѡ,�� !�i��0��j� +UyQ$<��!�t"3C�x��zC�c6o��#�գ.�1�p�W�j"�i��x��՜�bt�0#�فtߔ��v�E�jl<4��F(��H]��v��9Ҍ��,�J�0[!��&�C�ɱ��N�0��Z�d� +Ǥ_�G��L�E-�t��V5)�iR�naC%��<�Tzr�53,��'X�+ 4m�An�qX�h��N�� $(�M��n�djkQc��19%�A��ICE`�q��vTf�HR��_��Iм$�F��9��%&HN I�l��$�&g�7eR,]��mu��G ��)'�#�˦�M"l�t0�-z�C��02��GC��4�!��v��4��`F�_[� Rd�A�$65�Qa�1}��R�AU@�iJ8 +af\�W�5�?��0�<7�tnO�3�{-�?�@�oWQ��˧��7��' ?D3s?��o��m��G9A��~rP��]�΁d$�%0%�A=�6�T��ĉ�Δ�v��u$N�7��Y�NA�@�$�s��cw�L�Ab֔ O��O��΂09��5�9�\��F�#�j^{V�0/��x4�(�Ӈj:�P�R��I��z�@PQhf�M �a�yv�aR3�qj��uf��0A!��"��$2�n ۪N�0�XI>%��C5�#�I�S�?�xZ~��I��3%O�vJ��p:-��t�$��p:Mu�0'Ϝn�t�S�[�ɓ��4�!Vt�L�#4�M��H�� X>�S0�|��-yR�9�|�6e Q��4�� 3��s�g,�Ԯ�e2�}Jx��n�4�x3N帆L?��[eo[uz3��B�J��(|�M`� V�+��4�!�R=d|^�f1$�t&�%�!6Ŷ�:�R��q6k.�{J�H�C��f�xF�M�C�2��=4*�R��B*a�xDsrLPO:��l�%4p�Ղ��p�#�n��t +-�$P�`U��X��*۵��5��*��)�IRN��ZLj=4��~��B_}�� d6�u�@a�YQ%P6>4�FF��T�!O�jJx^�d�4� ��jj��S�r� +�ԇ�ŇF��g�85<�]��O*8�ɫ)�yU��^'G��U۲Ghhʫ��*�d�YC��d��(Ʈ��C��~�S��HP��q�kv�5��S��P5h�53G��H��16�:��'E�]X~��Ţ�J搨5L1H�:�wS"}�J��x�M o�Ʊ��4��V��-�G��4��S�ASƳ�ۥ��l�XuO��=��jʄ��H�y�P3�)ai%h�x�N��t��%5�2 ��&l/E��t�ɭ��P�&i� +�� jJ�9m,R+9T��e��a\4;ijX'��|�ȸCStr�9�Dک�0��"��̮��3='H;U��`UP�{�h�_��'FS%��P-Y�gvq �Z�d��ٹZʫ�yJ�5�mS�gʵS��3��@kj��c��晬i�t�>v��?3��gj��F�MX��!�Ec�:p��S��㥍�e.��yM(`��s +v�Fag͜��B�.��j�L�n�$�`��B�%�OCD.e�A�eJ8��g>Mˏ�)}kS��Z��4�) ��05�>�kl�F��\��5%E�&I�cm��B[¹�"�(��xN 1NI�'��WhT0�q��fSn�D��w�S\�fT��30Nj��@P3Y��a��Jva��*R�"S�Y$�5�r��P��E��*�B�&\vq�t�p$G�wn� ?��٥�Ƅ�I��D��cw|H!Q&�dW6��C��!��=4�j<��whx�f�fȮn8vǇ'�!�a��ʮmD&�`�;��Ʉ:!�W��)VYD�e�{�2j��Y4+�M��ʸ��qHx�d�VA�3Yqf�M �)v �I��J��dJxby��M%,�eq�/~�y��@�s�x��E�U ��+�>��]:��bG��^S�~��2U�0��Ge��!�V�Jmu��k8��4��q�dq[��B#�zC��@� �_�:��|N1lH��Pmfd)� ��ʧ)�Q�i�#�t� ��4f��BT�DLsM^puuzs��p4R C]=��)�@W�hLB�&LhzЗv�7fGa��݉BQS�S��9��^c� ��)o�%ã᫖��p^kvH%Y]Ǔ��j�n,��g� ��.�ة�#8��5��A9v��:GuYW0��X��C�%;��,�s0��q��Ae�ť�a�Jvv�FŲ��SÚ8;;��x�KT;T�� 5��/ �ˮ��<X��p�6�&��W b��z�o��ȶ��◟��?,��[>.O>�qz��~��+x��Q��k�[�k�Z�l��0�m��d��:>�\�dK��dK��dK��xKt��l��l��l��l��l��]79z�}M�z�}M�$z�}M�z�}M�4z�}M�tv�.��==�&["=��[><�{��\�z��7o�?��;h�o�}��ޟ��_��[}��\Mُ�>4/_xyz{yzs�@h,�a'dq��0��]��/�k��W��|��ʆ<��e��B��~}��ey��/_��/��>}�[p�W�7��]�c��٣��Q��_~=m�?�Uӣ6D��1&��aUi{�RU��b<�vv��;{"X��|�yy��)n�פj�w��h�3v�+~�P��u�!ؘ��?޴k��{��jZ��(lA��*��&�x�a�?�nZ)&�/�E�n=�S-�)s��M�~��ol�Q5`,t��Nځ��:@�5��2ږ��8�'r�)rN��lC�OL7geg��V��l�C��?3��;t�3��m�+@2�D1�Xl쥜 ZP�*SFN+��(��qh�h>`�Y�%�0�Ue� f�U��:�tw��=��b#0��/`8�"�[��'"��\�:SR+�4��TU�wW0<�s8y�Lϳ N� ��N�j��P��Bo��a2-n�`��r�gy��ٗ�>�%mTDh��cB��8��1S�P�ds��;�� +*�\E��?B,�Tߎ��L�ݶ.�wרȠn��9x�Z�C�?��{-(H��7�=V��[�E�Ӧ&��^($��rR��c�H��Ps6��Z�eR�W��b��e��~A^s\�8R��-ғ��-6�M]�0�� +endstream endobj 557 0 obj<> endobj 558 0 obj<> endobj 559 0 obj<> endobj 560 0 obj<> endobj 561 0 obj<> endobj 562 0 obj<> endobj 563 0 obj<> endobj 564 0 obj<> endobj 565 0 obj<> endobj 566 0 obj<> endobj 567 0 obj<> endobj 568 0 obj<> endobj 569 0 obj<> endobj 570 0 obj<> endobj 571 0 obj<> endobj 572 0 obj<> endobj 573 0 obj<> endobj 574 0 obj<> endobj 575 0 obj<> endobj 576 0 obj<> endobj 577 0 obj<> endobj 578 0 obj<> endobj 579 0 obj<> endobj 580 0 obj<> endobj 581 0 obj<> endobj 582 0 obj<> endobj 583 0 obj<> endobj 584 0 obj<> endobj 585 0 obj<> endobj 586 0 obj<> endobj 587 0 obj<> endobj 588 0 obj<> endobj 589 0 obj<> endobj 590 0 obj<> endobj 591 0 obj<> endobj 592 0 obj<> endobj 593 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 15/PresSteps 600 0 R>> endobj 594 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�˥G��)�$��~H�Y��\`5�H��6�A#��ʊ�Ϭ��'2��ͨ�)��t�,ҡѡ,�� !�i��0��j� +UyQ$<��!�t"3C�x��zC�c6o��#�գ.�1�p�W�j"�i��x��՜�bt�0#�فtߔ��v�E�jl<4��F(��H]��v��9Ҍ��,�J�0[!��&�C�ɱ��N�0��Z�d� +Ǥ_�G��L�E-�t��V5)�iR�naC%��<�Tzr�53,��'X�+ 4m�An�qX�h��N�� $(�M��n�djkQc��19%�A��ICE`�q��vTf�HR��_��Iм$�F��9��%&HN I�l��$�&g�7eR,]��mu��G ��)'�#�˦�M"l�t0�-z�C��02��GC��4�!��v��4��`F�_[� Rd�A�$65�Qa�1}��R�AU@�iJ8 +af\�W�5�?��0�<7�tnO�3�{-�?�@�oWQ��˧��7��' ?D3s?��o��m��G9A��~rP��]�΁d$�%0%�A=�6�T��ĉ�Δ�v��u$N�7��Y�NA�@�$�s��cw�L�Ab֔ O��O��΂09��5�9�\��F�#�j^{V�0/��x4�(�Ӈj:�P�R��I��z�@PQhf�M �a�yv�aR3�qj��uf��0A!��"��$2�n ۪N�0�XI>%��C5�#�I�S�?�xZ~��I��3%O�vJ��p:-��t�$��p:Mu�0'Ϝn�t�S�[�ɓ��4�!Vt�L�#4�M��H�� X>�S0�|��-yR�9�|�6e Q��4�� 3��s�g,�Ԯ�e2�}Jx��n�4�x3N帆L?��[eo[uz3��B�J��(|�M`� V�+��4�!�R=d|^�f1$�t&�%�!6Ŷ�:�R��q6k.�{J�H�C��f�xF�M�C�2��=4*�R��B*a�xDsrLPO:��l�%4p�Ղ��p�#�n��t +-�$P�`U��X��*۵��5��*��)�IRN��ZLj=4��~��B_}�� d6�u�@a�YQ%P6>4�FF��T�!O�jJx^�d�4� ��jj��S�r� +�ԇ�ŇF��g�85<�]��O*8�ɫ)�yU��^'G��U۲Ghhʫ��*�d�YC��d��(Ʈ��C��~�S��HP��q�kv�5��S��P5h�53G��H��16�:��'E�]X~��Ţ�J搨5L1H�:�wS"}�J��x�M o�Ʊ��4��V��-�G��4��S�ASƳ�ۥ��l�XuO��=��jʄ��H�y�P3�)ai%h�x�N��t��%5�2 ��&l/E��t�ɭ��P�&i� +�� jJ�9m,R+9T��e��a\4;ijX'��|�ȸCStr�9�Dک�0��"��̮��3='H;U��`UP�{�h�_��'FS%��P-Y�gvq �Z�d��ٹZʫ�yJ�5�mS�gʵS��3��@kj��c��晬i�t�>v��?3��gj��F�MX��!�Ec�:p��S��㥍�e.��yM(`��s +v�Fag͜��B�.��j�L�n�$�`��B�%�OCD.e�A�eJ8��g>Mˏ�)}kS��Z��4�) ��05�>�kl�F��\��5%E�&I�cm��B[¹�"�(��xN 1NI�'��WhT0�q��fSn�D��w�S\�fT��30Nj��@P3Y��a��Jva��*R�"S�Y$�5�r��P��E��*�B�&\vq�t�p$G�wn� ?��٥�Ƅ�I��D��cw|H!Q&�dW6��C��!��=4�j<��whx�f�fȮn8vǇ'�!�a��ʮmD&�`�;��Ʉ:!�W��)VYD�e�{�2j��Y4+�M��ʸ��qHx�d�VA�3Yqf�M �)v �I��J��dJxby��M%,�eq�/~�y��@�s�x��E�U ��+�>��]:��bG��^S�~��2U�0��Ge��!�V�Jmu��k8��4��q�dq[��B#�zC��@� �_�:��|N1lH��Pmfd)� ��ʧ)�Q�i�#�t� ��4f��BT�DLsM^puuzs��p4R C]=��)�@W�hLB�&LhzЗv�7fGa��݉BQS�S��9��^c� ��)o�%ã᫖��p^kvH%Y]Ǔ��j�n,��g� ��.�ة�#8��5��A9v��:GuYW0��X��C�%;��,�s0��q��Ae�ť�a�Jvv�FŲ��SÚ8;;��x�KT;T�� 5��/ �ˮ��<X��p�6�&��W b��z�o��ȶ��◟��?,��[>.O>�qz��~��+x��Q��k�[�k�Z�l��0�m��d��:>�\�dK��dK��dK��xKt��l��l��l��l��l��]79z�}M�z�}M�$z�}M�z�}M�4z�}M�tv�.��==�&["=��[><�{��\�z��7o�?��;h�o�}��ޟ��_��[}��\Mُ�>4/_xyz{yzs�@h,�a'dq��0��]��/�k��W��|��ʆ<��e��B��~}��ey��/_��/��>}�[p�W�7��]�c��٣��Q��G�N��Oy�47 +�� +��L�CH�=�q/��`��`3�n��hC Й��0~z~O�4EL�5ֳ�B� +4n�S��OtGA��O�c�Y"$��C��0:��<� +l��UTБ&Ѐ)?��W�$�XqϦ�J��V�T�"�M��/��iQ��ЩRCpzeo�j�f�?�hN8�_�1w�:�C7��ôs*�� ږ��Մ�n��p ��С$��.�؆, +j��\A��X��6e��ndz*V��A��u&�@��~f8��4��WH�36��5�RY��4q��b�k��x��v��dz��R��T�R�ـ��䞑҄s>َ��<��^Z��S�C�_�.(�ϙ��ʬ(��&D��!VhDyP#��n��;m�E��n��&��a��T`I%�@P�W�T��֛͠��#4�ukS�۷��fB��tXE@5��#��?)(W +�vK��_� KR"'��2��c +TN{��-&·��ts�>��˾Sp��qI�� J�>|m,p�> endobj 596 0 obj<> endobj 597 0 obj<> endobj 598 0 obj<> endobj 599 0 obj<> endobj 600 0 obj<> endobj 601 0 obj<> endobj 602 0 obj<> endobj 603 0 obj<> endobj 604 0 obj<> endobj 605 0 obj<> endobj 606 0 obj<> endobj 607 0 obj<> endobj 608 0 obj<> endobj 609 0 obj<> endobj 610 0 obj<> endobj 611 0 obj<> endobj 612 0 obj<> endobj 613 0 obj<> endobj 614 0 obj<> endobj 615 0 obj<> endobj 616 0 obj<> endobj 617 0 obj<> endobj 618 0 obj<> endobj 619 0 obj<> endobj 620 0 obj<> endobj 621 0 obj<> endobj 622 0 obj<> endobj 623 0 obj<> endobj 624 0 obj<> endobj 625 0 obj<> endobj 626 0 obj<> endobj 627 0 obj<> endobj 628 0 obj<> endobj 629 0 obj<> endobj 630 0 obj<> endobj 631 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 16/PresSteps 638 0 R>> endobj 632 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�˥G��)�$��~H�Y��\`5�H��6�A#��ʊ�Ϭ��'2��ͨ�)��t�,ҡѡ,�� !�i��0��j� +UyQ$<��!�t"3C�x��zC�c6o��#�գ.�1�p�W�j"�i��x��՜�bt�0#�فtߔ��v�E�jl<4��F(��H]��v��9Ҍ��,�J�0[!��&�C�ɱ��N�0��Z�d� +Ǥ_�G��L�E-�t��V5)�iR�naC%��<�Tzr�53,��'X�+ 4m�An�qX�h��N�� $(�M��n�djkQc��19%�A��ICE`�q��vTf�HR��_��Iм$�F��9��%&HN I�l��$�&g�7eR,]��mu��G ��)'�#�˦�M"l�t0�-z�C��02��GC��4�!��v��4��`F�_[� Rd�A�$65�Qa�1}��R�AU@�iJ8 +af\�W�5�?��0�<7�tnO�3�{-�?�@�oWQ��˧��7��' ?D3s?��o��m��G9A��~rP��]�΁d$�%0%�A=�6�T��ĉ�Δ�v��u$N�7��Y�NA�@�$�s��cw�L�Ab֔ O��O��΂09��5�9�\��F�#�j^{V�0/��x4�(�Ӈj:�P�R��I��z�@PQhf�M �a�yv�aR3�qj��uf��0A!��"��$2�n ۪N�0�XI>%��C5�#�I�S�?�xZ~��I��3%O�vJ��p:-��t�$��p:Mu�0'Ϝn�t�S�[�ɓ��4�!Vt�L�#4�M��H�� X>�S0�|��-yR�9�|�6e Q��4�� 3��s�g,�Ԯ�e2�}Jx��n�4�x3N帆L?��[eo[uz3��B�J��(|�M`� V�+��4�!�R=d|^�f1$�t&�%�!6Ŷ�:�R��q6k.�{J�H�C��f�xF�M�C�2��=4*�R��B*a�xDsrLPO:��l�%4p�Ղ��p�#�n��t +-�$P�`U��X��*۵��5��*��)�IRN��ZLj=4��~��B_}�� d6�u�@a�YQ%P6>4�FF��T�!O�jJx^�d�4� ��jj��S�r� +�ԇ�ŇF��g�85<�]��O*8�ɫ)�yU��^'G��U۲Ghhʫ��*�d�YC��d��(Ʈ��C��~�S��HP��q�kv�5��S��P5h�53G��H��16�:��'E�]X~��Ţ�J搨5L1H�:�wS"}�J��x�M o�Ʊ��4��V��-�G��4��S�ASƳ�ۥ��l�XuO��=��jʄ��H�y�P3�)ai%h�x�N��t��%5�2 ��&l/E��t�ɭ��P�&i� +�� jJ�9m,R+9T��e��a\4;ijX'��|�ȸCStr�9�Dک�0��"��̮��3='H;U��`UP�{�h�_��'FS%��P-Y�gvq �Z�d��ٹZʫ�yJ�5�mS�gʵS��3��@kj��c��晬i�t�>v��?3��gj��F�MX��!�Ec�:p��S��㥍�e.��yM(`��s +v�Fag͜��B�.��j�L�n�$�`��B�%�OCD.e�A�eJ8��g>Mˏ�)}kS��Z��4�) ��05�>�kl�F��\��5%E�&I�cm��B[¹�"�(��xN 1NI�'��WhT0�q��fSn�D��w�S\�fT��30Nj��@P3Y��a��Jva��*R�"S�Y$�5�r��P��E��*�B�&\vq�t�p$G�wn� ?��٥�Ƅ�I��D��cw|H!Q&�dW6��C��!��=4�j<��whx�f�fȮn8vǇ'�!�a��ʮmD&�`�;��Ʉ:!�W��)VYD�e�{�2j��Y4+�M��ʸ��qHx�d�VA�3Yqf�M �)v �I��J��dJxby��M%,�eq�/~�y��@�s�x��E�U ��+�>��]:��bG��^S�~��2U�0��Ge��!�V�Jmu��k8��4��q�dq[��B#�zC��@� �_�:��|N1lH��Pmfd)� ��ʧ)�Q�i�#�t� ��4f��BT�DLsM^puuzs��p4R C]=��)�@W�hLB�&LhzЗv�7fGa��݉BQS�S��9��^c� ��)o�%ã᫖��p^kvH%Y]Ǔ��j�n,��g� ��.�ة�#8��5��A9v��:GuYW0��X��C�%;��,�s0��q��Ae�ť�a�Jvv�FŲ��SÚ8;;��x�KT;T�� 5��/ �ˮ��<X��p�6�&��W b��z�o��ȶ��◟��?,��[>.O>�qz��~��+x��Q��k�[�k�Z�l��0�m��d��:>�\�dK��dK��dK��xKt��l��l��l��l��l��]79z�}M�z�}M�$z�}M�z�}M�4z�}M�tv�.��==�&["=��[><�{��\�z��7o�?��;h�o�}��ޟ��_��[}��\Mُ�>4/_xyz{yzs�@h,�a'dq��0��]��/�k��W��|��ʆ<��e��B��~}��ey��/_��/��>}�[p�W�7��]�c��٣��Q��ܣ��A x�W�[+�I��|A�{!2F�,�$�}f Y��{�1==�CJ�^�/=�r:��L ��(��@��9v��H"��릻��{�D9�D��I��C��o��f�N��[fK�L:`��A��s�~h�`��b(|��_瑰�q�H��8�^�d`��[��T��i=3e#Z�(�9��"Ͷ+c�$�a��~��@m>��WB�mS<�J��"'��,{��F�2#X�/��o�'~�m��[cO��;��<�kT��}��^8o��(��kY��n�o3�q�Zn�V��֋w�lz%��^��$�~fM�g��zvյ\��`8��W��J��d�9��:�ɱ��5Zm�aW�f k +endstream endobj 633 0 obj<> endobj 634 0 obj<> endobj 635 0 obj<> endobj 636 0 obj<> endobj 637 0 obj<> endobj 638 0 obj<> endobj 639 0 obj<> endobj 640 0 obj<> endobj 641 0 obj<> endobj 642 0 obj<> endobj 643 0 obj<> endobj 644 0 obj<> endobj 645 0 obj<> endobj 646 0 obj<> endobj 647 0 obj<> endobj 648 0 obj<> endobj 649 0 obj<> endobj 650 0 obj<> endobj 651 0 obj<> endobj 652 0 obj<> endobj 653 0 obj<> endobj 654 0 obj<> endobj 655 0 obj<> endobj 656 0 obj<> endobj 657 0 obj<> endobj 658 0 obj<> endobj 659 0 obj<> endobj 660 0 obj<> endobj 661 0 obj<> endobj 662 0 obj<> endobj 663 0 obj<> endobj 664 0 obj<> endobj 665 0 obj<> endobj 666 0 obj<> endobj 667 0 obj<> endobj 668 0 obj<> endobj 669 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 17/PresSteps 677 0 R>> endobj 670 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�˭Ŏ��+��Ҫ��<�CJ:��O�1��C~}v��S��==֨fjͪ��ogT�� o�l��P�Q�y��f�DU^5i�*��(WY^��P��w:��xH<|R��1��}�� +��T<4⭙��\�1ZS# +�L��ki��)�؅ ��Z;4��^��x�x��U ��Y%L/�F��sh��=%�� WP�xȺd�� +��pH�tR�B��!�Vծ!��X�d�;T"N��&�_��L"�sZ=��)��y5��!R��9��)�!\͙+Fw��W��.�Wc�>v�p��G#uEƳ��%�r�95ܧ�Y�a�B>�Mx��c�32&�a�9�zȄ�I��:z��x5�Z�:5u�jR�Ӥ��J6�y��kfXN O�fWhږ�ܦ�,�4�߻��l�QIP��_ݎ��֢�d7crJ��!;&S�)��)�� +HR��_��Iм$�F��9��%&HN I�l��$�&g�7eR,]��mu��G ��)'�#�˦�C"l�t0�-z�C��02��GC��4�!��v�=4��`F�_[� Rd�A�$65�Qa�1}��R�AU@�iJ8 +af\�W�5�?��0�<7�tnO�3�{-�?�@�oWQ��󇧟�7��' ?D3s?��m��G9A��~rP��]�΁d$�%0%�A=�6�T��ĉ�Δ�v��u$N�7��Y�NA�@�$�s��c�� +��,�))�>��:E�ar:c�)k s��P��&Gnռ��Za^(#R�h�Qȧ�t��>5�:/��~Æ� +�:?¤f��0N�� +n)!a�B��5E��ՓIdZ=�U�"a��|J�C�Ӈj:G��P��Z�)��5CS�s)��T�<��)�/��Ի�a��_SP$��4�!�T�4! O�l�ШJ}� +��)�Y��1A=� ZPC��%��V*�ñ�|{Ph�)�ē@MP�UM�b�c�+~��l�b��`PC��'I9��fPk1��i�VT}�� +B�'��X�M��SdE�@��SSu�<ɫ)�yU�m�t�<˫��O�� 4*)��t-�V:0�D��`�A�ա��cW,��`˗��v��z�K��C#|�`L��2��.��/��b� <�v6fwO�Ք +��@s�P3�)ai%h�x�N��t��%5�2 ��&l/E��t�ɭ��P�&i� +�� jJ�9m,R+9T��e��a\4;ijX'��|�ȸCStr�9�Dک�0��"��̮��3='H;U��`UP�{�h�_��'FS%��P-Y�gvq �Z=d��ٹZʫ�yJ�5�mS��3�کa�֙��@kj��c��L�4 @�`;MK��>�n��tx>��E4�'C֢Fcg�X��)��xi�g��p^�X��A�QA�Y3禆�P�K{$��s��[5 :嬐kI��Ks�t�Χ0ƙO��>�omj�[��<��l�Pvm#2a#ߩ�C&�� 鼚]ߠL��"�v(��P�QP�xȢX�n +�w��V�͇ǍC$۵ +�Ɋ3{oj�O�kN��7W��%S��KO,�h*a�,��~��8�t��)�)R�[xh^Uw�p�K��I,��u�x�;z�@��o�q�:��<*��jWrh��e\éϦ&��$��j��G�C��ձ��p�aC*E�j3� �HI�hPU�V>M ύjNL)�C��4 o�fߥ1D�P%�:$b�k򂫫Лl��ɐj��Q_O��rDc�5aBӃ��@�1; +S��(�55<��)��5&�@8�9��^2<�j)�M �f�T��q< �<-�V�v0��q6��h�򊝚:�CxX�_d�c�J0K�sTǐu�G�e�>�Y��Β9�ʺG9T�qQ\�6�dg�hT,�Λ<5��㨸��A�C��P��z�Ҁ��<�w��i�k�~|"��oy��'�f^~�l[�-~��sW��Ǻ�g��_O�f��&[��Ȗ�o ��і}M�ܯ�c�}M�$z�}M�z�}M�4z�}��DG��O��ɖH��ɖL��ɖJ��ɖΎ��q��dK��dK��dK��dK��dKg��k�%��k��w��7[��ͻ�?} ��7�~7��m�V�?�/WS��˗?>}{{zs�Ah,��a'dq��8��]��o�k��}��ʆ<��e��B�_��}��r��qy��_��O��Ϗ��?�˿��?�g��X��Ѐ}phw?��j��1�@�m��x��8 C�=_q� ��y4a�$l�qwjMj�b�e��sm�r�t��R�>�9��ؗ<� +�q��H�gX,��#�w�A�y��魢�V�경�VNư� ��\�-� +A�nu��H0�c�3 "e�}��a��?G��Z�U�ڑ,�S�׊B��RaP�X:Z�_��[|8��m(+NZG�hC��2^��cy�i�F��:<�١{%i��$�"n�R�l��"I��J�c� m��S��7��`��]��;{�n+/��n�,O��Wn^z�i�?qJ`s P� * +��d�~�/�h'�gm��2��7�q�t\�Vر��z��5\Y��֏��mM`�Vv@�({/��=WuH�]�@^�-�+��s �u��(��]�蛧֎}R��'ل� V�he%�$��ǽ��-�'�LH�n�w��Z��5YU� +��Ă��_�ͮ��K��'f�܍��h��Eh�|��a�"�԰�6n�[�l��u�qJ/J� �^ �Ύ>��jTK��O�綑� +endstream endobj 671 0 obj<> endobj 672 0 obj<> endobj 673 0 obj<> endobj 674 0 obj<> endobj 675 0 obj<> endobj 676 0 obj<> endobj 677 0 obj<> endobj 678 0 obj<> endobj 679 0 obj<> endobj 680 0 obj<> endobj 681 0 obj<> endobj 682 0 obj<> endobj 683 0 obj<> endobj 684 0 obj<> endobj 685 0 obj<> endobj 686 0 obj<> endobj 687 0 obj<> endobj 688 0 obj<> endobj 689 0 obj<> endobj 690 0 obj<> endobj 691 0 obj<> endobj 692 0 obj<> endobj 693 0 obj<> endobj 694 0 obj<> endobj 695 0 obj<> endobj 696 0 obj<> endobj 697 0 obj<> endobj 698 0 obj<> endobj 699 0 obj<> endobj 700 0 obj<> endobj 701 0 obj<> endobj 702 0 obj<> endobj 703 0 obj<> endobj 704 0 obj<> endobj 705 0 obj<> endobj 706 0 obj<> endobj 707 0 obj<> endobj 708 0 obj<> endobj 709 0 obj<> endobj 710 0 obj<> endobj 711 0 obj<> endobj 712 0 obj<> endobj 713 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 18/PresSteps 720 0 R>> endobj 714 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�˥G��)�$��~H�Y��\`5�H��6�A#��ʊ�Ϭ��'2��ͨ�)��t�,ҡѡ,�� !�i��0��j� +UyQ$<��!�t"3C�x��zC�c6o��#�գ.�1�p�W�j"�i��x��՜�bt�0#�فtߔ��v�E�jl<4��F(��H]��v��9Ҍ��,�J�0[!��&�C�ɱ��N�0��Z�d� +Ǥ_�G��L�E-�t��V5)�iR�naC%��<�Tzr�53,��'X�+ 4m�An�qX�h��N�� $(�M��n�djkQc��19%�A��ICE`�q��vTf�HR��_��Iм$�F��9��%&HN I�l��$�&g�7eR,]��mu��G ��)'�#�˦�M"l�t0�-z�C��02��GC��4�!��v��4��`F�_[� Rd�A�$65�Qa�1}��R�AU@�iJ8 +af\�W�5�?��0�<7�tnO�3�{-�?�@�oWQ��˧��7��' ?D3s?��o��m��G9A��~rP��]�΁d$�%0%�A=�6�T��ĉ�Δ�v��u$N�7��Y�NA�@�$�s��cw�L�Ab֔ O��O��΂09��5�9�\��F�#�j^{V�0/��x4�(�Ӈj:�P�R��I��z�@PQhf�M �a�yv�aR3�qj��uf��0A!��"��$2�n ۪N�0�XI>%��C5�#�I�S�?�xZ~��I��3%O�vJ��p:-��t�$��p:Mu�0'Ϝn�t�S�[�ɓ��4�!Vt�L�#4�M��H�� X>�S0�|��-yR�9�|�6e Q��4�� 3��s�g,�Ԯ�e2�}Jx��n�4�x3N帆L?��[eo[uz3��B�J��(|�M`� V�+��4�!�R=d|^�f1$�t&�%�!6Ŷ�:�R��q6k.�{J�H�C��f�xF�M�C�2��=4*�R��B*a�xDsrLPO:��l�%4p�Ղ��p�#�n��t +-�$P�`U��X��*۵��5��*��)�IRN��ZLj=4��~��B_}�� d6�u�@a�YQ%P6>4�FF��T�!O�jJx^�d�4� ��jj��S�r� +�ԇ�ŇF��g�85<�]��O*8�ɫ)�yU��^'G��U۲Ghhʫ��*�d�YC��d��(Ʈ��C��~�S��HP��q�kv�5��S��P5h�53G��H��16�:��'E�]X~��Ţ�J搨5L1H�:�wS"}�J��x�M o�Ʊ��4��V��-�G��4��S�ASƳ�ۥ��l�XuO��=��jʄ��H�y�P3�)ai%h�x�N��t��%5�2 ��&l/E��t�ɭ��P�&i� +�� jJ�9m,R+9T��e��a\4;ijX'��|�ȸCStr�9�Dک�0��"��̮��3='H;U��`UP�{�h�_��'FS%��P-Y�gvq �Z�d��ٹZʫ�yJ�5�mS�gʵS��3��@kj��c��晬i�t�>v��?3��gj��F�MX��!�Ec�:p��S��㥍�e.��yM(`��s +v�Fag͜��B�.��j�L�n�$�`��B�%�OCD.e�A�eJ8��g>Mˏ�)}kS��Z��4�) ��05�>�kl�F��\��5%E�&I�cm��B[¹�"�(��xN 1NI�'��WhT0�q��fSn�D��w�S\�fT��30Nj��@P3Y��a��Jva��*R�"S�Y$�5�r��P��E��*�B�&\vq�t�p$G�wn� ?��٥�Ƅ�I��D��cw|H!Q&�dW6��C��!��=4�j<��whx�f�fȮn8vǇ'�!�a��ʮmD&�`�;��Ʉ:!�W��)VYD�e�{�2j��Y4+�M��ʸ��qHx�d�VA�3Yqf�M �)v �I��J��dJxby��M%,�eq�/~�y��@�s�x��E�U ��+�>��]:��bG��^S�~��2U�0��Ge��!�V�Jmu��k8��4��q�dq[��B#�zC��@� �_�:��|N1lH��Pmfd)� ��ʧ)�Q�i�#�t� ��4f��BT�DLsM^puuzs��p4R C]=��)�@W�hLB�&LhzЗv�7fGa��݉BQS�S��9��^c� ��)o�%ã᫖��p^kvH%Y]Ǔ��j�n,��g� ��.�ة�#8��5��A9v��:GuYW0��X��C�%;��,�s0��q��Ae�ť�a�Jvv�FŲ��SÚ8;;��x�KT;T�� 5��/ �ˮ��<X��p�6�&��W b��z�o��ȶ��◟��?,��[>.O>�qz��~��+x��Q��k�[�k�Z�l��0�m��d��:>�\�dK��dK��dK��xKt��l��l��l��l��l��]79z�}M�z�}M�$z�}M�z�}M�4z�}M�tv�.��==�&["=��[><�{��\�z��7o�?��;h�o�}��ޟ��_��[}��\Mُ�>4/_xyz{yzs�@h,�a'dq��0��]��/�k��W��|��ʆ<��e��B��~}��ey��/_��/��>}�[p�W�7��]�c��٣��Q��_~=m�?�U��6C�� +7�b�i��7`��e@/�L�Zː��=�kl+iwH '��#E��$n��_��Q`��d��H��J��B�إ]?ٸ��;��t�">�<�� *�=j.r�2��x�2-SB�1��i�_r�n}�4�A�@��81ӽR�|Ud��#6X��N�w��LӅ&l�*o%��`�a�n�VG��LFXҫ;Rhݱ�I�'y�Y��=��8�U9�V��dx�!�G�]�N\�p��i��7 ��͌�h��,�]�[��oFM&*��Z?�B��.��\布�}$�г哰��6֐>֖{�;�4�$�3��}��,�D�;vC��d��ǐ��eUL/3�WW�_��[4�j]�6��KdX��Dj�؊"'|>3�j�Sՙ��ɧ��4aDH.��)9.�cD��h��(��ZU��]��Z)�գgr]h5ǎ�u'��e�4,� El�H��q��jJ�%4^��WL��g+��9�0�P��1��?�vU/��[�6>�tG��(Pѯ�-��IH�i�8?�e��_��!zm"��P��+� -�� +endstream endobj 715 0 obj<> endobj 716 0 obj<> endobj 717 0 obj<> endobj 718 0 obj<> endobj 719 0 obj<> endobj 720 0 obj<> endobj 721 0 obj<> endobj 722 0 obj<> endobj 723 0 obj<> endobj 724 0 obj<> endobj 725 0 obj<> endobj 726 0 obj<> endobj 727 0 obj<> endobj 728 0 obj<> endobj 729 0 obj<> endobj 730 0 obj<> endobj 731 0 obj<> endobj 732 0 obj<> endobj 733 0 obj<> endobj 734 0 obj<> endobj 735 0 obj<> endobj 736 0 obj<> endobj 737 0 obj<> endobj 738 0 obj<> endobj 739 0 obj<> endobj 740 0 obj<> endobj 741 0 obj<> endobj 742 0 obj<> endobj 743 0 obj<> endobj 744 0 obj<> endobj 745 0 obj<> endobj 746 0 obj<> endobj 747 0 obj<> endobj 748 0 obj<> endobj 749 0 obj<> endobj 750 0 obj<> endobj 751 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 19/PresSteps 759 0 R>> endobj 752 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�˭Ŏ��+�1 �U�C<�Cv�Cȃ"��0R0ɯϮ>}��ՙZg��ogT�� o�l��P�Q�y��f�DU^5i�*��(WY^��P��w:��xH<|S��1��}�� +��T<4⩙��\�1ZS# +�L��ki��)�؅ ��Z;4��x�x��U ��Y%L/�F��sh��=%�� WP�xȺd�� +��pH�tR�B��!�Vծ!��X�d�;T"N��&��&�=�u��x�I�ۼ�W��Hۜ��pg�}Sƃ��e��;�c��PУ��"��U�� +9Ҍ��,�J�0[!��&�C�ɱ��N�0��Z�e� +Ǥ_��:��x5�Z�:5u�jR�Ӥ��J6�y��kfXN O�fWhږ�ܦ�,�4�߻��l�QIP��_ݎ��֢�d7crJ��!;&S�)��)�� +HR��_��Iм$�F��9��%&HN I�l��$�&g�7eR,]��mu��G ��)'�#�˦�]"l�t0�-z�C��02��GC��i:CF�킻kX��f?��A��8��IljX��Hc��⃪��Ӕp�9�̸�Dk �D��a�yn��*ܞ�w��Zb��߮��-��?`�󇇟��s�󛆯��9��˧m��W9A��~�6��t��H@K`Jx�z�m©:�M�)N�� H�(oN �3��N��΁�I�T��f*��Hx��u��υ�o4A8r��g� +�B�z�G��B>�PM��Y�S#��R�� +��7l8��#Lj&?N �T��ά��R&($[]S$<^=�D��=a[�)&+ɧD|�t��ΑŤ��_T<-��u�$Fe��T�\��3%O�vJ��p:-��t�$��p:Mu�0'Ϝ��t�S�[�ɓ��4�!Vt�L�#4�M��H�� X>�S0�|��-yR�9�|�6e Q��4�� 3��s�g,�Ԯ�e2�}Jx��n�4�x3N帆L��[eo[uz3��B�J��(|�M`� V�+��4�!�R=d|^�f1$�t&�%_�!6Ŷ�:�R��q6k.�{J�H�C��f�xF�M�C�2��=4*�R��B*a�xDsrLPO:��l�%4p�Ղ��p�#��t +-�$P�`U��X��*۵��5��*��)�IRN��ZLj=4��~��B_}�� d6�u�@a�YQ%P6>4�FF��T�!O�jJx^�d�4� ��jj��S�r� +�ԇ�ŇF��g�85<�]��O*8��WS��N�0r!��e��4�8�W�i��U>d��!�ϓ ,ã��_��x{�&\Ïۭ��!�4��Jj063��7^;)�j�\kf�L ϑfcl�u\ �O�Ļ�:�y��n+�C��x0� ��P�M��+��c�mj�s�v0��K�qjx`u��R�¥��>v0�vLeM��n�v��T��v;��jʄ��H��@�sٔ��m�N��a:�c��V��^sT��`:��V�q(E��[�TN�%ڜ6��p��N�0.��45��`�C>adܡ):9�g"��t]�q�LfWB��^�*(ٽI +4�֯��Fv��3 ��L��2��\-��<%��ʶ�W��r�԰��e~f�55�~`�}�~�y&k� ]��_�g�s�p�`7Qu:<��L��"È�� kQ�1��h�Q�p +Wz��̅�S8� �,U~N�� Ш 㬙sS�[(إ=�_͹�Iح��rVȵ��i�ȥ�9h�L �S��O��>�Omj�S��<��l�Pvm#2a#ߩ�]&�� 鼚]ߠL��"�v(��P�QP�xȢX�n +�w��V�͇ǍC$۵ +�Ɋ3{oj�O�kN��'W��%S��KO,�h*a�,��~�� C�ce��)�-<4��@8�%W��$`v��:d��/��7ȶ��/��Y��O��<�\��`��x_W�[��L��d��r[�-ap=ڲ�ɖ�u|l��ɖD��ɖB��ɖF��q�5��q�5��q�5ْ�q�5�R�q�5��q�8nr��l ��lI��l)��li��l��]7{z�}M�Dz�}��}��כ�-_}��? ��|;^��/�e[��G��Ք��3@��o��<=Ah,O?AO?��X�o�<}�?��g�+��p}�x��>}|�ݟ>��|��>��3��w�Ͽ>�K�,�>}��?��0�v׸�r�q �zj�K{�|��?��Q��9��V=UꅋIX��w��8d�EZd��x~~o�mY�Rė�ch~~3��O�î��{0־$Fŭ'�1*YR�+1=RR=**��P��K�w�2z]�VV��ٖ�J�+J��ZY��e�ڤTZ��>�ܖ��;(!8��V��՟�3R�5ާ��f=:��SC6s� +��(��aa%h�B�ҞB�|'�T`4�}u� �p2��i��QEC��$r�E��%�r <�x$��'Ai/��k؋R�r��)s�!CO��N�Il�4�ɿg�� ca*�3y�,H5�o�v*�Z+J��Y�RTҮ|��/�/�0 S��a�VVj�j�^��8��Ј�?�>��8 +�Ob��4p��a@��ߠ�ߑ��< q�H�'/FaEyD��}yͩ�_}n��#<�!á8�x��2�:m�-�=�P��5(o'+]i��N�*��Y�8�Ĩ�!㷟?�(yeY��y��X/�]?.'�\ �뺍0��4͒�]��<��\ +�˸w��|��.4J=�^X42_Y|_>�/2��"N.a1%�W��YżWI�QUH�DwGs�+��'d��NU�.��3�<�5[��Ͻy(6EN�#��"9�@BF[=#�g��}E�Is��У�(�zՠ,� �gj�+d��G�`E��X�CwHh.�O��ܛ��F�1dA��9e��GY��+Jnj��)�#1 +endstream endobj 753 0 obj<> endobj 754 0 obj<> endobj 755 0 obj<> endobj 756 0 obj<> endobj 757 0 obj<> endobj 758 0 obj<> endobj 759 0 obj<> endobj 760 0 obj<> endobj 761 0 obj<> endobj 762 0 obj<> endobj 763 0 obj<> endobj 764 0 obj<> endobj 765 0 obj<> endobj 766 0 obj<> endobj 767 0 obj<> endobj 768 0 obj<> endobj 769 0 obj<> endobj 770 0 obj<> endobj 771 0 obj<> endobj 772 0 obj<> endobj 773 0 obj<> endobj 774 0 obj<> endobj 775 0 obj<> endobj 776 0 obj<> endobj 777 0 obj<> endobj 778 0 obj<> endobj 779 0 obj<> endobj 780 0 obj<> endobj 781 0 obj<> endobj 782 0 obj<> endobj 783 0 obj<> endobj 784 0 obj<> endobj 785 0 obj<> endobj 786 0 obj<> endobj 787 0 obj<> endobj 788 0 obj<> endobj 789 0 obj<> endobj 790 0 obj<> endobj 791 0 obj<> endobj 792 0 obj<> endobj 793 0 obj<> endobj 794 0 obj<> endobj 795 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 20/PresSteps 802 0 R>> endobj 796 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�˥�n��\&L��^h�E�8 �ddaزa'VI��T��6��X�F=�gN��W6��CÛ.�E:4:�eTp "��9d<��Y<FՁWMZ�J"/��U��=#�Ndf(�To�v��m��c$�xk�m?4`��ԈB0S>�g�Z;4| +(va�� o�b�|�W��x*2��!@a�`�@V �ˡ>�xt��fO�"��(��i��A��!�nY�;"8� +�<<6>��P�6t�� tȸU�k.g�&V=��*/񺺰��hm��V��|J<Ĥ�m^ͫu�Tf�mN�qJxWs��]aF��)��U��xh��1\�F(��H]��v��x�iFN �iv�F%G��Oe�!��쌌I'G�{N�2a�cү��&^��E�NM]��4)c7��pG*=9��S��ٕ�� 8,K4 ��n'e��uT��D�W�c2��1�͘��E�Ǝ��u��"�x�8�u;*��W�s4�ɩ>vNf��$t� �S�@F'�'� ��MY�KG��a[E�%��C+:{@�I��i��bL�E��P�0��mu��P!dh<�8Mg��]p +8��7�c�9�M kTiL��6�T|PPt��0g�W��h ��>6�;ύ�]��S��^�O,��U�º��G�͹��I��O��ö��H�C?��(��.]�@2��ޠg�p�N�@�DSgJ�S;u�:'ʛS��S��s v��9U��{�L�Ab֔ O��O��΂09��5�9�\��F�#�j^{V�0/��x4�(�Ӈj:�P�R��I��z�@PQhf�M �a�yv�aR3�qj��uf��0A!��"��$2� ۪N�0�XI>%��C5�#�I�S�?�xZ~��I��f*J�d��tZ��0IBɯ)(�t��a*N�9=$�鴧`�J�'A;%�C"l�m�u �6 ��l�\��ԇ��>��:��'d6�{hT�>qM�T�,�䘠�t-�!B�Kh�F�w��G�=(4��Z�I�&��&]��1�?�U�k��j0��U��S��i3��zh��I�4Y+*��J��C!�V�l��&��)��J�l|h��Ʃ�:C��Ք�ɶi:A��ϧ��7t�rqjx.V��Tp|!��U�#8z�a�BVm��i�q(��Ӳ'��|Ȓf �C�ϓ ,ã��_��NM��_�[#A�C�i�ٕ:�`lf0N o�vR@ՠ��#��V��wau�y��n+�C��x0� ��P�M��+��c�mj�{�v0��K�qjx`u��R�¥��>v0�vLeM��n�v��T��v;��jʄ��H��@�sٔ��m�N��a:�c��V��^sT��`:��V�q(E��[�TN�%ڜ6��p��N�0.��45��`�C>adܡ):9�g"��t]�q�LfWB��^�*(ٽI +4�֯��Fv��3 ��L�2��\-��<%��ʶ�W��r�԰��e~f�55�~`�}�~�y&k� ]��_�g��p�`7Qu:<��L��"È�� kQ�1��h�Q�p +Wz��̅�S8� �,U~N�� Ш 㬙sS�[(إ=�_͹�Iح��rVȵ��i�ȥ�9h�L �S�̧i� +�ҷ65��E��Hs�Oi�L �()W��)�cs=��> kJ��M�0��>� +�B[¹�"�(��/�-�l+��A��_}h^>}��~��[߾_�y��o�[��'h��}��Ɔ8��@�p��;��~}��}~��?��77|�?��O�� G��qL��q�|?�8��˘U �6��,Gax��>泝]�f�%��B��#�4��d��B�_��$��O ��(nC}lE�mb!��Q�H��W��P[$&@�O��H:҉�G��:t��l��)�'Ld>��V0m��y��=Z� �V��f��`�>��4O}�`��ӗ=�ׯo�*'��qIz��휛�I��7K�5 �g�Κ�*l6 ʞ*Tl��.�]\�� +�&�� 佊�l*��,Ng1��}�A�ޚ�P��h�듯N�&t�UeZ�t�ʓ��F՟��A��8@�RӖ8�z�ܛG��Ek@YI�i'�Զ�Z��u��= �u��l@�� _ �(E�3�雿1��u38��>�G�A�{ꉫ� �B-@�$q؎ޜ��q��Z9+�[G��$��pR�[^��^��<�W�� %$~��"z�x��ڍ��}��+/��!��H�oY��܀?��/|&[�6_�ꎸw�r�0/�Q +endstream endobj 797 0 obj<> endobj 798 0 obj<> endobj 799 0 obj<> endobj 800 0 obj<> endobj 801 0 obj<> endobj 802 0 obj<> endobj 803 0 obj<> endobj 804 0 obj<> endobj 805 0 obj<> endobj 806 0 obj<> endobj 807 0 obj<> endobj 808 0 obj<> endobj 809 0 obj<> endobj 810 0 obj<> endobj 811 0 obj<> endobj 812 0 obj<> endobj 813 0 obj<> endobj 814 0 obj<> endobj 815 0 obj<> endobj 816 0 obj<> endobj 817 0 obj<> endobj 818 0 obj<> endobj 819 0 obj<> endobj 820 0 obj<> endobj 821 0 obj<> endobj 822 0 obj<> endobj 823 0 obj<> endobj 824 0 obj<> endobj 825 0 obj<> endobj 826 0 obj<> endobj 827 0 obj<> endobj 828 0 obj<> endobj 829 0 obj<> endobj 830 0 obj<> endobj 831 0 obj<> endobj 832 0 obj<> endobj 833 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 21>> endobj 834 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�˥E�� Ή��ji!��4xa�F�ea�Č��7�2#�3+[��&��ԉʺ��w3*l +e��7]6�tht(˨�<@D��s�xZe�x"��B�D^ ��,��{(F�;��P<$��>z��H*��~h.�-��`�|��4vh�P��D}-��@�L��T5�:/��~Æ� +�:?¤f��0N�� +n)!a�B��5E��ՓIdZ� ۪N�0�XI>%�C��t�,&�O��i��S$1*K��R,�T�<��)�.��Ի�a��_RP$��4�!�T�m�&�r�ꈕ�t��p�2>/d��k:��/��b�tH� �C��8�5��=�E$�!i|h�O�}��ć&�� C�O\R!�0e<"�99&�'Dj�P��jA�]p8��o;�&`1�BK< �5Xդ+V8&��v-v?@ 5�J��aJx��:m��Z��9��f kE��_��U�=a��ƺ�(�"+��ƇF��h��3�I^M ϫ�l��Y^M �~�]n�Q��;C��,��b��I�y5e<��#�jk�M�Cy՜�=q\�C�0k�2�<��2<��aj��n��wj�5�� +2Nsͮ��c3�qJx㵓�͵f��iv0�f�\Ǖ��H��ß�k��ҁ9$j�S�,�ݔH�`�6{nSß[��q,�/ Ʃ��rK� +�2�]#|�`L��2��.��/��b� <�v6fwO�Ք +��@s�P3�)ai%h�x�N��t��%5�2 ��&�E��t�ɭ�� +P�&i� +�� jJ�9�,R+9T��e��a\4;ijX'��|�ȸCStr�9�Dک�0��"��̮��3='H;U�/�`UP�{�h�_��'FS%��P-Y�gvq �Z�2��\-��<%��ʶ�W��r�԰��1?3КF?0��>�?�<�5M�.��N�ү�3��5�'�MT��g>S#|��0�dh�R�h�,kԁ;�/��,s!��KB�K��S�;4*�8k�� +vi��Ws.`v�&a@��r-�|"r)c�.S��8�Ӵ|�O�S��Ԣ�m��4x��O�w�Ԕ+|ڗ��%%E�&I�cm�vN�-��^�q��vA|\�SB�S��O� v� +� +�0.��l`�m�H�8z4��u��݌*�x�IM�j&��P30��@�.l ��pBE*Yd�8�$��TN�*��H8]%Y��.�@�N��!��-��Oc�uvi�1�s�p��v2Q$��R�F��"ٕ��;4|n�fa͸�}(��ټ��+�=�� w�x�f��k+� +�N�v��B'��jv}�2�*��ڠ,{o@YFMt@��!�`e�)�ޝ�Z77 O�l�*ht&+�콩�>Ů!8iV�\ V�L O�"/�-��2��E�,��7p� +��%R�S�S��м.��`�\�I,��u�x�;z�@��Z��8T�`�B�MK��CX�+9��2��g�T� �ŭ5��F��𑁶�_�:��|N1�H��Pmfd)� ��ʧ)�Q�i�#�t� ��4f��BT�DLsM^puqzs��p4R C]<��)�@W�hLB�&LhzЇv�'fGa�+�=�BQS�S��9��^c� +��)��%ã᫖��p^kvH%Y]Ǔ��j�n,��g� ��.�ة�#8��5�n��A9�CV�YR��:��;��!��PgɎ�v:K�*�v�hPY�qDqijب��G�Q�,;ov�԰&�Ύ��V^��nrC ��K治n�y��ݞ��p�6^��.�ۻ�>��wȶ�ۛ��w�'��=�Lu?�� +�am��<��l�"[�k�%�G[�5�r��-�5ْ�q�5�R�q�5��q�5�=�&[<=�&["=�&[2=�&[*=�&[:;n�M�w[�-�w[�-�w[�-�w[�-�w[�-��fO��ɖH��wO߼}��z��~��7o��/��~��?�*WS��g@��zyz��Qq{�v� ��ad�3�^~��|��'��i�?�_@ � +n�mM�~{��~��/��o��o�?��ק��/?=}��G�m {]Ry��w��x&{d�w�mGNp��]�� @��^ �!��HMlT`Y�}g!M�(��F�3��*cs��^��ґ|8"~v��ʰ�K�!y�V��a��T��,j2NS�U"��]��dͷ��+��!ş��d�Tc8$ L/��ȭ�hFMD��`��m�A�ͯ� +�6�f:��?9�P;=��8`}؉�>��!Vҋv)��#�B�K��5rƾ�/��;��ZY o��A�bRW�0Ui��ƝD-z +� � ��^��qEBp ��u�t��C9�<��F��ʴ"�E_�C��q��{�x�"��7�U��$?�[��|bȤ[YKg�_�B[0 �U��#�Z��~ +endstream endobj 835 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 22/PresSteps 838 0 R>> endobj 836 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�˥G��-9'� ��a��HSˋ��ɿ�7��"�3��q�Df�{�6��CÛ.�E:4:�eTp "��9d<��Y<FՁWMZ�J"/��U��5#�Ndf(�To�v��m��H*��~h�1ZS# +�L��ki��)�؅ ��Z;4��^�4�S��.v +�@�J�^��Aǣ��45{JFAM3 ��: ��uȊ��@Vp��؅ +��3\EE'�Cƭ�]Cp9{5��tw�D�Ty��ՅM�Fk�D��z��S�!&n�j^�C�2#ms�S�C��3W��f�;;��2\�.��_��F��A��=�+2�]�.�!G��S�}��ŠQ�f+�Sلw�896;#c��S��LX��+<�[|��W��e��SSת&e;M�� l�d3�ǑJO��f��kv%��m9�m:K�M��I��v�e�)��Lm-jLv3&��;h��c2u�"i�,�2xݎ��I +xS��9 �W��;'��A��ɩa ��t��C��`ְ��Β��!a��= �sdtٴ�I�M1�&ڢEOt�xF�:�h�24~��3dt�.��h�k�D��1ȜĦ�5*�4�OqA*>� +(:M GA�3̌+�J��G�yƝ�F܁��~F~��'��*ja��z��#��\��hb��e��m��('(��O�A� +��K�9��7��&��1�8�ԙ��N��ĉ��08�9��)��$qN��0H��"��S�Y&�3��2��h�p�V�kϪ�2"��f�|�PM��Y�S#>��R�� +��7l8��#Lj&?N �T��ά��R&($[]S$<^=�D��-a[�)&+ɧD'��Ӧ�!�{��5�p�w��pN��ڵ�Lf�O ��m�foƩא�t�q��m�n@oF�P�O��O� �Ī:be<��5$\��,��Τ�$�C:$¦�6]RjC��<�fͥqOiI}H��l��(� yhxBf��F�P��TH%L��bN� �Iт"�-��n�ZPq�}�ۍB��N�%�j��j�+\�^e��Z�X�0%�N;�Ur�J�$�VA2��CmA�6� �Ej%�*69|��S0��f'M �$��Owh�N�0ǙH;5F�p\D0�ٕPut��i�� +Jvo�M��k��h��%��L�.�!S��L�=;WKy5?O ��a��m�#�L�vjX��u�s��hM �{l��?�<�5M�.��N��g�s�p�`7Qu:<��L��"È�� kQ�1��h�Q�p +Wz��̅�S8� �,U~N�� Ш 㬙sS�[(إ=�_͹�Iح��rVȵ��i�ȥ�9h�L �S�̧i�>�omj�[��>��357d��f\��>� O�l�� ;dŮ!8iV�\ V�L O�"/�=��2��E�,��7o��#0V�H}NO�b��C󺨺 +��]bp��'��K��!�R��kbj�/~�]ơ��lZ:dª]ɡ�N�q �>�F��:n�,n��[h��Boh7A��WǞ��)� ��#%�AU�Z�4%<7�]81mt��r�4��}��lUC�@�ꐈi��n@oN��&C��b��G}=%��I�CԄ M�Ү��(Lq#�;QH#jjxJ5;S#�kL6�p�s0卽dx4|�R^��k��$��xXxZB��`�eU3��l�c�!��;5u��&�}�0�"�n��`�T税!� +f��p}��dG;�%s�u;�r4��8��45lT�Ύ#ШX��7;xjXgg�Qq/{� +`��q7��]��s�U7��y~��w�� +D��oO�ͼ�ٶ�]��3��:�@0S�/�uϰ5��|�t�uM�@�-�5�ף-��l�^�ǖ�lI��l)��li��o��w_�-�w_�-�w_�-�w_�-�w_�-��&G��ɖ@��ɖD��ɖB��ɖF��ɖΎ��q��dK��x˻��>/׿�|�|�ٛ��_ ��_|9��m�V�?�/WS��˧��<��\ 4�ˏ�s��ˇ��ϼ\��ߺ\>A#��/��7�f_.�<��돿��?��_?.߾y��ϗ��'x��~�O�oO1�G��[>�e��H�7��9�Ow�� п�c�E-��+0�E��H�-e�H��j�$plY|�{E�ڦ��͞ɫ��3>��Ϩ��9 .�h*�h* +��j��*�Dկf��y��_�P��j"]|jM� �O_��?6�}� +A��&�m�p�с��F��W��"��Y� �� A�9�_��t��}]\��;�0�`ud��t�X� 5UsD��FxA9w +� w�N��:�~� |��g�"W�|��mf��ݒ�ݳ��u~VYL�}�_?q��~�þG��M��Rx:��I��c�j��4RM��2G�m�[i��Y��|_P�L��{�.9��Б�K�N{�L�{S�JPW6u.��[|� �w�HB�[��J�|IT�?�I/��woI*E�`��#[�n$�7.a�2��=�Y�w��(�ϓ�Zn�\U2)9��dGz�S,ml�q��G3x��N��&��̑jb��Uk�ڀ4l��c?��r�'�57lU]�Z��bmh٢��eN7�T�DξCٝy�t�.��7k}� +endstream endobj 837 0 obj<> endobj 838 0 obj<> endobj 839 0 obj<> endobj 840 0 obj<> endobj 841 0 obj<> endobj 842 0 obj<> endobj 843 0 obj<> endobj 844 0 obj<> endobj 845 0 obj<> endobj 846 0 obj<> endobj 847 0 obj<> endobj 848 0 obj<> endobj 849 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 23/PresSteps 852 0 R>> endobj 850 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�˥G��-9'� ��a��HSˋ��ɿ�7��"�3��q�Df�{�6��CÛ.�E:4:�eTp "��9d<��Y<FՁWMZ�J"/��U��5#�Ndf(�To�v��m��H*��~h�1ZS# +�L��ki��)�؅ ��Z;4��^�4�S��.v +�@�J�^��Aǣ��45{JFAM3 ��: ��uȊ��@Vp��؅ +��3\EE'�Cƭ�]Cp9{5��tw�D�Ty��ՅM�Fk�D��z��S�!&n�j^�C�2#ms�S�C��3W��f�;;��2\�.��_��F��A��=�+2�]�.�!G��S�}��ŠQ�f+�Sلw�896;#c��S��LX��+<�[|��W��e��SSת&e;M�� l�d3�ǑJO��f��kv%��m9�m:K�M��I��v�e�)��Lm-jLv3&��;h��c2u�"i�,�2xݎ��I +xS��9 �W��;'��A��ɩa ��t��C��`ְ��Β��!a��= �sdtٴ�I�M1�&ڢEOt�xF�:�h�24~��3dt�.��h�k�D��1ȜĦ�5*�4�OqA*>� +(:M GA�3̌+�J��G�yƝ�F܁��~F~��'��*ja��z��#��\��hb��e��m��('(��O�A� +��K�9��7��&��1�8�ԙ��N��ĉ��08�9��)��$qN��0H��"��S�Y&�3��2��h�p�V�kϪ�2"��f�|�PM��Y�S#>��R�� +��7l8��#Lj&?N �T��ά��R&($[]S$<^=�D��-a[�)&+ɧD'��Ӧ�!�{��5�p�w��pN��ڵ�Lf�O ��m�foƩא�t�q��m�n@oF�P�O��O� �Ī:be<��5$\��,��Τ�$�C:$¦�6]RjC��<�fͥqOiI}H��l��(� yhxBf��F�P��TH%L��bN� �Iт"�-��n�ZPq�}�ۍB��N�%�j��j�+\�^e��Z�X�0%�N;�Ur�J�$�VA2��CmA�6� �Ej%�*69|��S0��f'M �$��Owh�N�0ǙH;5F�p\D0�ٕPut��i�� +Jvo�M��k��h��%��L�.�!S��L�=;WKy5?O ��a��m�#�L�vjX��u�s��hM �{l��?�<�5M�.��N��g�s�p�`7Qu:<��L��"È�� kQ�1��h�Q�p +Wz��̅�S8� �,U~N�� Ш 㬙sS�[(إ=�_͹�Iح��rVȵ��i�ȥ�9h�L �S�̧i�>�omj�[��>��357d��f\��>� O�l�� ;dŮ!8iV�\ V�L O�"/�=��2��E�,��7o��#0V�H}NO�b��C󺨺 +��]bp��'��K��!�R��kbj�/~�]ơ��lZ:dª]ɡ�N�q �>�F��:n�,n��[h��Boh7A��WǞ��)� ��#%�AU�Z�4%<7�]81mt��r�4��}��lUC�@�ꐈi��n@oN��&C��b��G}=%��I�CԄ M�Ү��(Lq#�;QH#jjxJ5;S#�kL6�p�s0卽dx4|�R^��k��$��xXxZB��`�eU3��l�c�!��;5u��&�}�0�"�n��`�T税!� +f��p}��dG;�%s�u;�r4��8��45lT�Ύ#ШX��7;xjXgg�Qq/{� +`��q7��]��s�U7��y~��w�� +D��oO�ͼ�ٶ�]��3��:�@0S�/�uϰ5��|�t�uM�@�-�5�ף-��l�^�ǖ�lI��l)��li��o��w_�-�w_�-�w_�-�w_�-�w_�-��&G��ɖ@��ɖD��ɖB��ɖF��ɖΎ��q��dK��x˻��>/׿�|�|�ٛ��_ ��_|9��m�V�?�/WS��˧��<��\ 4�ˏ�s��ˇ��ϼ\��ߺ\>A#��/��7�f_.�<��돿��?��_?.߾y��ϗ��'x��~�O�oO1�G��[>�e��i�r��}4yZ�Ow�� B��~@dp�xo{�R/��K/��cT�X��w�|��\��<�A?�5�/ޅffP �Z��.��~'C��X�܊�$9�m�a-/��e�ʥm�/��)T +endstream endobj 851 0 obj<> endobj 852 0 obj<> endobj 853 0 obj<> endobj 854 0 obj<> endobj 855 0 obj<> endobj 856 0 obj<> endobj 857 0 obj<> endobj 858 0 obj<> endobj 859 0 obj<> endobj 860 0 obj<> endobj 861 0 obj<> endobj 862 0 obj<> endobj 863 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 24/PresSteps 866 0 R>> endobj 864 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�ˤG��u� ��:p��0 �`�@h� X�@��;��:+֚b3{��ˮ�x�EF�M��f��e��H�xO�lO�Qu�U�V��ȋ"�q��u�y��7��y�G��I�C#��y��05�̔�� ]؀�o��C��)��_�1��p�{PX2(�U��rh�:=��S�ȸ0 +bh�qu�q~��@V��;� ��O'�.T� ��**:12nU��٫�UO��C%��K��.l"�5Z[�$��գ.�1�p�W�j"�i��x��՜�btW��@�o�xpU��36�cq� +z4RWd<��]�1^!G��S�}��ŠQ�f+�Sلw�896;#c��S��LX��+|�QG��FQ�"]��UM�v��@�P�f8�#��w��)� ��JM�r��t��%��{��2y��:* �vS"��1��ZԘ�fLN�w�"Dc�d�:E�PX6�;ύ�]��S��^�O,��U�º��ӏ�s�󛆯��9��O˯ۆ��r�" �>�}@A�w�:��8ۄSu:'�:S"�ک��A�8Qޜg0g�:��$Ω�+�W�T$faMI�� )�,��NY�ß��h�p�V�kϪ�2"��f�|��΁1�ԧF|S��7�A�xS�o�p^�]�G��L~�Ʃ0u�Y�-�"$LPH��Hx�z2�L�G¶�S$L6V�O��P��C5�#�I�S��xZ~��I��+f*J�d��tZ��0IBɯ)(�t��a*N�9=$�鴧`�J�'A;%��:��'d6�{hT�>qM�T�,�䘠�t-�!B�Kh�F�w��G�=(4��Z�I�&��&]��1�?�U�k��j0��U��S��i3��zh��I�4Y+*��J��C!�V�l��&��)��J�l|h��Ʃ�:C��Ք�ɶi:A��7t�rqjx.V��Tp|%��U�#8z�a�BVm��i�q(��Ӳ'��|Ȓf �CƟ'X�G1v=L �� 띚p ?n�F��\�+u8��`��x��As��925)��Z,��t`�Z��4�C}7%�ǮX�W�=��ϭ��8��/ Ʃ��rK� +�2��F��0�4e<�]ک_`SŪx��l�� +� +�)Vv#��f�eS��Jж�:�&�鴏JjZe�{�QM��鴓[%� �M�n$SA89��hs� X�Vr��a�ç�:øhv�԰N��q��s��S�at �E3�] UGgzN�v�z_{��d�&)�Z�&O��J١Z��4��2�z��ӳs��7��k�+ۦ^�gʵS��3��0��i�晬i�t�>v��~��C�}��D��|�35�ǋh#N�&�E��΢�F�S�)\��2N�&�?�T�9��@��0��f�M o�`��H~5�&a�jt0�Y!גΧ!"�2��2%�Oa�3?M�W��>��O-J�F�|J�gjx�Dy7LM�§}�M��].�iXSR4l��9��iW��G�h7��8%�8%}�3N𰱯 Ш` +��ܦ��'@�X��ͨ��g`��$��f� +� 5ø�� *'T��E��H�kN��!��i�/��U��pO��d�4�H��4f^g�6>' whx`'E��C +�ЈB0Q$��+>$� ��Y�C3��s�p��'i6o��c��;dŮ!8iV�\ V�L O�"/�=��2��E�,��7o��+k�>��H1o�y[T��.1��_'��K��!�R��[bj�/~�]ơ��lZ:dª]ɡ�N�q �>�F��:n�,n��Gh��BohA��WǞ��)� ��#%�AU�Z�4%<7�]81mt��r�4��}��lUC�@�ꐈi��n@oN��&C��b��G}=%��I�CԄ M��(Lq#�W��F��jv�4F��l�`�`�{��h��65�ךRIVW��$�0�Z��˪f`�لǶC£��+vj��aM��~a�E�= +�,��QC��!��PgɎ�v:K�*�v�hPY�qDqijب��G�Q�,;ov�԰&�Ύ��6^��nrC ��K治n�y��i�5y?^��[^��o��wȶ��◟��o��-�/O>�qz��~y�+x��Q��k�[�k�Z�l��0�m��d��:>��dK��dK��dK��xKt��l��l��l��l��l��]79z�}M�z�}M�$z�}M�z�}M�4z�}M�tv�.��==�&["=��[^��z��><��+h�/�|��x�o��m�V�?�/WS��g��w�>��=�� 4�ۏ�s��w#=��y��u�}�F��76��/��7@��y��.�~��_�>��?�~~��O��~��8�n��q�o y}[s��fk��޽j��S^m�i�@�W�(�|��~Bӗ> +{��ؒG�}w%|�Q�vF�9G�g/��i<#�j<>e��?�-h��,Ⅷ�n<Ŝ��Ej�=��U�$�'5#k��b�>s(yh�L�,��x�A<�m�( +S��e4��)��l��F:�@��38/=v��V�n�� +�� Ċ*ckBt��#@�n}��gi��\�5��֠��X�(k��\L�"_�� M��H�z��R��Z~8��IH��3\B�s�C�H�d�b��i�e��)��ѕE�P��xq�|-d�[!�{��B��˻�ߥS|�!��+R��IVop��g�O�H��H�m�� +��V/��BR�<-^9�}H��(��$��8Jw��Jn |�[��WFoҕ�1%�S�� i��M�V͙�+h��B:��8�f��U��b�pB|^|��p��M7H]sg߀��nƾQ�_��|�P��+R�#��*�qw\��9��׈{�5�Okh�р�sX�U�[��!��а�B��u��D��(��3�z�O��&��3ｐ��b'� *ȊRSb�P7��qfQ�H��O��KH +M��0��f�0��_x�}Ư"��ƲbXsEK�ȿ�_�(�t%F�Kc��P�:��]��??B +endstream endobj 865 0 obj<> endobj 866 0 obj<> endobj 867 0 obj<> endobj 868 0 obj<> endobj 869 0 obj<> endobj 870 0 obj<> endobj 871 0 obj<> endobj 872 0 obj<> endobj 873 0 obj<> endobj 874 0 obj<> endobj 875 0 obj<> endobj 876 0 obj<> endobj 877 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 25/PresSteps 880 0 R>> endobj 878 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�˥E�� Ή��jia�؂i��x1�%�B��7�2#�3+[��&��ԉʺ��w3*l +e��7]6�tht(˨�<@D��s�xZe�x"��B�D^ ��,��{(F�;��P<$��>z��H*��~h.�-��`�|��4vh�P��D}-��@�L��T5�:/��~Æ� +�:?¤f��0N�� +n)!a�B��5E��ՓIdZ� ۪N�0�XI>%�C��t�,&�O��i��S$1*K��R,�T�<��)�.��Ի�a��_RP$��4�!�T�m�&�r�ꈕ�t��p�2>/d��k:��/��b�tH� �C��8�5��=�E$�!i|h�O�}��ć&�� C�O\R!�0e<"�99&�'Dj�P��jA�]p8��o;�&`1�BK< �5Xդ+V8&��v-v?@ 5�J��aJx��:m��Z��9��f kE��_��U�=a��ƺ�(�"+��ƇF��h��3�I^M ϫ�l��Y^M �~�]n�Q��;C��,��b��I�y5e<��#�jk�M�Cy՜�=q\�C�0k�2�<��2<��aj��n��wj�5�� +2Nsͮ��c3�qJx㵓�͵f��iv0�f�\Ǖ��H��ß�k��ҁ9$j�S�,�ݔH�`�6{nSß[��q,�/ Ʃ��rK� +�2�]#|�`L��2��.��/��b� <�v6fwO�Ք +��@s�P3�)ai%h�x�N��t��%5�2 ��&�E��t�ɭ�� +P�&i� +�� jJ�9�,R+9T��e��a\4;ijX'��|�ȸCStr�9�Dک�0��"��̮��3='H;U�/�`UP�{�h�_��'FS%��P-Y�gvq �Z�2��\-��<%��ʶ�W��r�԰��1?3КF?0��>�?�<�5M�.��N�ү�3��5�'�MT��g>S#|��0�dh�R�h�,kԁ;�/��,s!��KB�K��S�;4*�8k�� +vi��Ws.`v�&a@��r-�|"r)c�.S��8�Ӵ|�O�S��Ԣ�m��4x��O�w�Ԕ+|ڗ��%%E�&I�cm�vN�-��^�q��vA|\�SB�S��O� v� +� +�0.��l`�m�H�8z4��u��݌*�x�IM�j&��P30��@�.l ��pBE*Yd�8�$��TN�*��H8]%Y��.�@�N��!��-��Oc�uvi�1�s�p��v2Q$��R�F��"ٕ��;4|n�fa͸�}(��ټ��+�=�� w�x�f��k+� +�N�v��B'��jv}�2�*��ڠ,{o@YFMt@��!�`e�)�ޝ�Z77 O�l�*ht&+�콩�>Ů!8iV�\ V�L O�"/�-��2��E�,��7p� +��%R�S�S��м.��`�\�I,��u�x�;z�@��Z��8T�`�B�MK��CX�+9��2��g�T� �ŭ5��F��𑁶�_�:��|N1�H��Pmfd)� ��ʧ)�Q�i�#�t� ��4f��BT�DLsM^puqzs��p4R C]<��)�@W�hLB�&LhzЇv�'fGa�+�=�BQS�S��9��^c� +��)��%ã᫖��p^kvH%Y]Ǔ��j�n,��g� ��.�ة�#8��5�n��A9�CV�YR��:��;��!��PgɎ�v:K�*�v�hPY�qDqijب��G�Q�,;ov�԰&�Ύ��V^��nrC ��K治n�y��ݞ��p�6^��.�ۻ�>��wȶ�ۛ��{��?��_�'��=�Lu�� +�am��|�t�}M�@�-�5�ף-ۚl�_�ǖ��lI��ۚl)��ۚli��o��w[�-�w[�-�w[�-�w[�-�w[�-��&G��ɖ@��ɖD��ɖB��ɖF��ɖΎ��q��dK��x˻�o�>��o�>��kh�/��=��G��#��j�~��|��W/Oo^^ 4n/?��5�_>��𮷗�Ј��z�� q��d�� 5A��~�ϯo߽��ߟ��_~z��ݏ��~��q$ +{]�!�ƙ��ƃn�y��a��Y��Ks�0��]��4�u��n"�LH��P��&�sϹ_y�i��W ��r�4��'��b��0;<��ǒt��?� ��΁07h�2��v-ئ�Ny�]��P��}��Z4��q!f[s��8�+��ji�Ϙ'�#��K��d�d�skE�ZZ^�~�;�K�Q¨Z�$��}> endobj 880 0 obj<> endobj 881 0 obj<> endobj 882 0 obj<> endobj 883 0 obj<> endobj 884 0 obj<> endobj 885 0 obj<> endobj 886 0 obj<> endobj 887 0 obj<> endobj 888 0 obj<> endobj 889 0 obj<> endobj 890 0 obj<> endobj 891 0 obj<>/Font<>/XObject<>/ProcSet[/PDF/Text/ImageC/ImageI]/ExtGState<>/Properties<>>>/Trans<>/StructParents 26/PresSteps 899 0 R>> endobj 892 0 obj<>stream +H��K��W�:�x�~�K� �2��0��YHcY�H�}��}��b{&HGr<��?��wן��ۇ��~|��]߾}��*��!EeÖB��/��?��ן�~��v��~}�~��?��.׿�ۮy��Xksq{�ޒ��.��r}�O�=|۴��;�s��b�GX��oo7�op~K��u>^� +�h��p!�xc]�E�/益�g�FeX?��y��(�6�5�^\�צ��9��V\1��^�V�ʶ.��6��>_��W�{��6�W��< �2�`�2��h�2F�U��w�N:w�>1"��Ĕ��>/�D��R��Y�p +, ,��\,��i|�8��Dqm��]�m��n�p�2I\��8Q�H�+�9��u�*c��x4v^� +�@�'�;k�D��6��bR:qNԸTC6]�L�D�(G��˾�*N�3i��Fߙ, +��Tm�q{��;'��x�jZߢ=N��T�(S��M��贜鬓��3St�;��Բx{�OF��_/nmʊY9-��(VA��09�)�n��GJ� q_��0M,#��8]]��lj�,�p��d��[,��[��P �3٪�%��3��!q3�+�%��p�7!�M�L�請��p�)\Dc��^�]�yAV��v��qV^�]��V_��Af�˾(��h�c��s��hg�2 ol{��7~�]�7�U7�~�"�_N�[P�f=�k��=��_B�OAy��6��/��<�x��}��-\��,�Zځ�C=� +��pB��:hO%m�)�2r�N�N�#L��l�u,*_�wf{��(�q)��MgB�.u�~�e溬t8n�B��O�2�a�Ǯ�r��K��"绕�!Be��v[��*��$oZ�� +�$H�(!��1"�ڠ-�O��G,g�� [��9"��>��9��3��te�8��׉U�u�$��jY�v�d�D+��ܠJĔ��"e�d5�� ?�S4]�^Z��x�H��F�{�R N�uj��"��R��p%X*�ܜ�,��LJKχ��iʎ��~�,c��ʘ�Ȥ��[��'��agrV�c'-\!nb�]�b��+㧀�kŒa�P�f}k��jY�t�e;C�d��`1�;C��?N��F +� $��e�і�:�}�zm��]�#�۪d��JS��ّ靖�g�y%-�j �t��2�LJ�d촔��=��m�!��e2��Z��n9OY��84Op��2`핳S��X��p̶vz��[;��- +0��Q�yg��9��rb�u�F�[�>5_Q]��#�Yy�{��TX6��،E�yY�3�!��l�X��b�hN�� ϻ�}�9�y_��+�D�Ȥ��y�/�D�LJ�<�+:��Uw�Կ��.��)�I�uѣ�r�ǛX� 5"6F@su4��a��Rs�@[o_ljKS��X�d<#P,b\+��d��=�!�:^h�! +# t�C�e��H�A�l7(�4��H#��Pd*��L:��!¶D��W�;��V=0��l��lM��I�dd/@�q��郡��J�m��խ?��ײq�r�`hN�ģ:�/��M��%1��o� �{j��Jc8�JM�{C��a��*V� ��eJ��eG�P,�?0|�LA��X��i��A��a4�uѶ��0Q��92�q�L��a�A�1P��XI��tF6Ct�v�R�a�Н*��e��nH2��m�B�S��ɶ>|�v��zVV`�y��xG�9�YƵJ|"3��Ф�r�&�*�I��I��tVN�Z%>��3""b9Qa��yr�;3�)��h��rz�Z��M-��(��r8Tᆋ��Mk�#Q��nι��#d��v��ku��Ц�f��#_]�T:B��a +׮��u�~4?��ؗ��~��θ��1w:3��}}Ao��CnxpQZ鼠4B�/��'��/(�7��.��Ӱ��(�D�2�$�߯<��Z�#8��d�� 08m�R�?�˥7��{��Y&Ԫ�0��E�( "Y��ĖFd��H~}n�tW�gOӄ9*�OMϽ��QaS(;4��Y�C�CYF�".�C��*��aTxդ� �Hx\ey��C1B��Df��!��I��n�l��+AI�M�o�:�X:�f �*�,/<�\��RN:GF�M�]"l�t0�-z�C��02��GC��i:CF��5��`F�[_!Rd�A�$65�Qa�1}�[ R�AU@�iJ8 +af\�G�5�?��0�<7�tnO�3�[-��X �׫��e��O�s�󓆏��9��/��r�" ��}@A�w�:��8ۄSu:'�:S"�ک��A�8Qޜg0g�:��$Ω�o+�f*��Hx��u��o4A8r��g� +�B��ǣ�F!��s`�,��y��e�f��6�W`��&5��q*L]gVpK� �-�)��L"�jO�Vu��J�)o*��sd1)*�O˯u�"�QYb 04U<�by`��I�N�vqN��ޝ�$��"N��.N�=�U�< �)�A�N#bEʄ:1B��t�ZZ +�� +��15��=l�ȓ��I��)K��iu|M8��=5<��=c��v-.��S��=uۦěq*�%d��2n��m� �ͨ +��)Л��ic4��XUG��Ӱ��K��y!�Ő\әԗd|H�D�ۦ�@Jm��٬�4�)-"�I�C#|��u%>4! O�l�ШJ}� +��)�Y��1A=� ZPC��&��V*�ñ�|�)4��Z�I�&��&]��1�?�U�k��j0��U��S��i3��zh��I�4Y+*��J�� d6�e�@a�YQ%P6>4�FF��T�!O�jJx^�d�4� ��jj��S�r� +�ԇ�ŇF��g�85<�]��O*8>ȫ)�yU��^'G��U[�Ghhʫ��*�d�YC��Q�]Sï�fw�y�&\Ï�-��!�4��Jj063��7^;)�j�\kf�L ϑfcl�u\ �O�Ļ�8�~��n-�C��x0� ��P�M��+��a��65��u;��q��`�X�.�T�p)��5��.��)��N��*V��ngcvW�TXM�� 45c.��P��49L�}PR�*Ӡ�k�j��PL��*9��h�v� � +�ɡ��D�� +�"��C�>]�)�E��u�~�'��;4E'g��L��k8."��J�::��s��S��V%�WI�&��5�xb4U��Ւux�a��.O��R^��S®q��l�z��)�N K��|�3��a�c��.�3�3Y�4�}�4-� +?3�]�}��D��|�35�ǋh#N�&,E��΢�F�S�)\��2N�$�?�T�9��@��0��f�M o�`��H~5�&a�ht0�Y!גΧ!"�2��2%�Oa�3�M�W��>��O-J�F�|J�gjx�Dy7LM�§}�M��].�iXRR4l��9��iW��G�h7��8%�8%}�d�`�� Ш` +��ܦ��'@�X��ͨ��g`��$��f� +� 5ø�� *'T��E��H�kN��!��i�/��U��pO�� +d�4�H��~7f^g�V>' whx`'E��+�=�!�phD!�(�]Yq�I�C��l�Ќ��܇"ܡ�I�͛!��p��i6{(��j� +tB:�f�W(S�� ʲ��e�D��!�hV��ݩ��q��q��$�v��Fg��ޛ�S��f�ɕ`eɔ��*��*�JX$��_�}��!]��@�s�x��E�] ��+�1��]:��bG��^S�~��2U�0��Ge��!�V�Jmq��k8�Y5��q�dqk��!|d��=5��S+R)2T�GJ:F��j��iJxnT�pbZ�H9*��ixc7�.�� *��!�\�\]܀ޜ`+M�TC�P��zJ<Е#�� Q��J`��F��jv�4F�ט��`�`�+{��h��65�ךRIVW��$�0�Z��˪f`�لǶC£��+vj��aM��~a�E�퐕`�T税!�f{�e�>�Y��Β9�ʺG9T�qQ\�6�dg�hT,�Λ<5��㨸��A�A��P��z�Ҁ��_o�?��k�~��w��'�a��v�p��?߾��x{�+��T�y_W�k��L��d��r_�-ap=ڲ�ɖ�u|l��ɖD��ɖB��ɖF��q�5��q�5��q�5ْ�q�5�R�q�5��q�8nr��ۚl ��ۚlI��ۚl)��ۚli��ۚl��]7{z�mM�Dz�m��||��o_��C~�ͻ��g{��m]��G��Ք�x��<�{yz��q{� vBG��^~�o��ߠ��Ə��7��! ��_@�n��Q�o/��O�?}��/|��|ڿ��{vⰝ�6��ެ��|o�Y�1HC|�_�p��2�(?��}y�L�!�!��*x�h��Fa z�9�n��m� +{&`�ؑmJ�}ǘ�,�)?�2�{o>\�RQ�j��d�� X��/��n +�2-��$��,*^��bd�!��yx�b@7/xE�t��v��E:�g�P�؀4 ��n�Ǝ�J�W9i��'�I)��+�o�j��P�+�a�� R=��#ۈ#O��Q�"��y9PDY��xY�E�3=<��XR�]�uD:�nξk�aT��N�(��1�r�y��~�'�/��/9#��L�xU��ϬiA%�:[cL�Z��7��C��Vt�JZI�K�*�+�� O\Y�2ZK�=J*��!tZ��2C��%h4m�]��R�zKmh�^��GCHS�/�n��-N�8Nٌ�_�z�`��q��j��5�\�� իSh�?�u�."B��Q��t�H�!$��0��UQ�> endobj 894 0 obj<> endobj 895 0 obj<> endobj 896 0 obj<> endobj 897 0 obj<> endobj 898 0 obj<> endobj 899 0 obj<> endobj 900 0 obj<> endobj 901 0 obj<> endobj 902 0 obj<> endobj 903 0 obj<> endobj 904 0 obj<> endobj 905 0 obj<> endobj 906 0 obj<> endobj 907 0 obj<> endobj 908 0 obj<> endobj 909 0 obj<> endobj 910 0 obj<> endobj 911 0 obj<> endobj 912 0 obj<> endobj 913 0 obj<> endobj 914 0 obj<> endobj 915 0 obj<> endobj 916 0 obj<> endobj 917 0 obj<> endobj 918 0 obj<> endobj 919 0 obj<> endobj 920 0 obj<> endobj 921 0 obj<> endobj 922 0 obj<> endobj 923 0 obj<> endobj 924 0 obj<> endobj 925 0 obj<> endobj 926 0 obj<> endobj 927 0 obj<> endobj 928 0 obj<> endobj 929 0 obj<> endobj 930 0 obj<> endobj 931 0 obj<> endobj 932 0 obj<> endobj 933 0 obj<> endobj 934 0 obj<> endobj 935 0 obj<> endobj 936 0 obj<> endobj 937 0 obj<> endobj 938 0 obj<> endobj 939 0 obj<>stream +xڼY�n��@��x'�`eda�F��Ƀ�Ş8�j5�D��SŮ��Lzlg&y��tw�X<�:${ZbG�%��=�]. ��r��fW��7�&Yp��*�6�b�yy��4�A� ��s( +�AR{v:�c�#��\A"Я$׻��~�~R +T��l��O�ߞ_�?�І7��U�`tq��?oI5N;Q��U�U:A]�%U;AY�NPQ�TU�4F��^'Ke�O� +� ҧ0��)L* +� +�¢��*�2�ѓ +�BV)T�U�C�QQaRIT�EQV�ނ��o��n�q��ÇO��'d��J��e��y�_^\H:��o�{�5?�`9;R`�� G��Y�f��7ko�ެ-�EG��=��Ձ'3"Y�d�R�s��kd� L�j�r�D�R�qZ��)h�,��l.�T��1ېD��7��\�I",Z�9h3� ��0`U��ޙ q0Oa��fm��6Kl��~�6R��ڄȀY��5ӛf�d�4YW��fm�E6_�&k��)�b�լ�dm5OV�d�NV�T&k+�� +��l��l.��l.)YUɺ��˪8�p��'�:XV��|�U�I��dm.��v�p�@�Y]��ݭl��2Rv�x��Fw��4^{t�׎�W��?o>=�>��y.�}�>��|��;m:��_�$��TN�8jP�PJ[��f ]=�f�>�YN�m�2�̇#��Ƚ��n��V�/�1��R,�^!MP�95�K#� v3��b�8��1O��y��+��Jiio��_�v�s+�V��[�n�ǭb�1��y7�W߶я�{�`�kbd�4��i��Oޮ��z�l�5�?V�ަ��Yi�X�N�q:�5�tpzh��C9Gh� E�z� JXw1̄"�Pă��ʟBY�=��u��#� /b�;c\��+0LTʗ��|ֈKX��L��n��޾m�T5�m�=x|�bG:��@�&ōq�K��I�^�=��߂�N�Hw�/��i1��Y�f��߯�m��'��g��;��\�3��9��[oi�_�+��_*��s�R��$�3�ċ)��X�-��Qn� +endstream endobj 940 0 obj<>stream +xڼ��o�6��KDQ?(`(p��+��졻f��d��?��8��"Q��&)�Y�6��lp&z��|�"��w"��"z*�T��Dd��Fk(˩|@�J�M�Α(v�*6A�ȗag�8�͢�$8U|r�&N7YU|yU#S�9i+�*��Y;��q�ŉ��V�%k\��/ƥ$W�3.�� Ib�7h�Ġ�J=w�%UɠK�� fU٠G��-w�4�f0�@�3H^��1�x��0k�f��E�S�c� E��ʬ��Z�U��E9=R�Ɠ'Q�'�*�]�1rP�L̢�TeVX��e��τ �p�� +Y%��9��%H=CdE��:�RT�*��"FU`b��b��kyb��<+.eV�I\nL$P�L�VϒIrHY$�b�&�U&y��q]�*4)zi%zVRk��NI�`dTq��z� ! ]G2d��8`��!#t��Ʈ �_RG��lf��vw��Ǐ߿n��ݗ��U�Q��۾<�8�v7�p�q��%Z�}��g�w��_�tT��~��u�!_O�5Mŭ�2d/�[��VtA2�e2�W�:�B��ܹ/��$�P� ��)g��G-�f�Tf�2Ӆ2�E[S^�)��df�Ô��W��x�F��M��9푕k\�|۝��4�6��)d&��_�z,��J='�9a��SYtaf�9�C=?�m��Q��->stream +x��Z�n7��@w$��CP]��F��.�-��}y��xvǙ��j�(��P� !�]��.T��*BtP��!f�Ì"AgU�J`!yW0�\�:"ʂ,�d + <�) +c��1�GPp�5(��߲%Q�d��2��E�e��IԋY��< |2;�_��@б��{��TtFu� �ϒ��NAT[ +�Tt� s��+�F�u�-(j+E�u�8�@��1�F� �7�� G�ä�UbI��M�B��(6�ר��5��{ ��,�J��@XG(�Q�o�Q:Y��O,��,eGI�_X�-�F�2�ͥe�8�TD�5�l��#�).r�xK.b}!��8�RR��Dx+�UXʂp�� +��'R�U��6��:#��tFv A��M��N\�@�R�xZ�Ap�LJ��k}�ֱ貭��Xv��J�%�I��Rɻ�fv&��J�r��D,I�o^ rĀCW��]]�S�U>w�('�qo�t��m�l{��.n�u�;�A��E#��ۏ��ÿ��/�/�w�=��vv_�o��rw�]�o��~/��r�ĺɱ��r�� f�Y��o��'�;ߙe�kx̸�j/0��#�_��OΙڭq�}+&\>|}Z& �h�U��\w׻��Q�� &7g ��i�7��a��a�(fѷ?i�g�ߦ�Y��qb�9�� |�Ų�� +_Hp��Ye�C�2 + ��(RTJ�(-'�g7n��P�s�h�}�꧖}�ڶV��x�J��2F��l��+~�<�q��FO�F��m�u��X��n,�i��3��Qc�gԼ� k�\�Gb^�,[�� )j�zxG��C0��8�B��)�!�e*-��Py U�QɈ�.��#��+k$iA>��7Fh��g�,b�yč5��tdX'��p��Y��r�(cn�e�17�2ʘ�2ʘ�2Jߐ4?�e-süS�~��1BcF?{e��٠koz�37��;�R��UU�"�mP��j��^�󿶫wܴ� G�F��cld�Fۢ��kgݘ&w��%hr �׃⎳;�d��"�h{��t��P�ɗU��i� œi��d��>��i޽l��9v/��' OOz�g��@�p��[�ѠS��`M�9xE?��~6ќe}Z�ؤ��J� +endstream endobj 942 0 obj<>stream +xڴV�j1��F��!ԥ&�!��P��l!��ά�u�m�U?,#�̜��Ҍ � �I`�#�g�� V�d�(Y ��5x'`o�G{�8��N��T�Q"ϑ�q�I�� ́�< .0[4�\ ��Ӏs�hE�C��ZCpu�>��4ܪ�{^V��W��v��~�oZ�8�}Sw�c�c��Ԫ_��6m׳�� R��v�~�ZЍI� C�#�Ս�;�%��+��͊�T]��x�6��s��p� +u�L��M��!�-�rr0o�'y�=��%3&Lal��)�˘4��Çn�3�LaB��Q��W��<ܗ�d4~tw�p�i4c|�cU[��^8-s�Ȼ"j�<S�eK�D��U�Џ��Z�v�-v�U�87��05U��(55��w&�:�՟<��Gϫ�� s�o��瓇��q�Φ�e��3B��!�7�G�Д�� 2�ΜkM4Ȁ�~0�� +endstream endobj 943 0 obj<> endobj 944 0 obj<> endobj 945 0 obj<> endobj 946 0 obj<> endobj 947 0 obj<> endobj 948 0 obj<> endobj 949 0 obj<>stream + + + + + + + + +How to do the BasicsMark Souverville + + + + + + + + + + + + + + + + + + + + + + + + +endstream endobj 950 0 obj<> endobj xref +0 1277 +0000000000 65535 f +0000239981 00000 n +0000248347 00000 n +0000254829 00000 n +0000254872 00000 n +0000254915 00000 n +0000254958 00000 n +0000255001 00000 n +0000255044 00000 n +0000255087 00000 n +0000255130 00000 n +0000255174 00000 n +0000255238 00000 n +0000255290 00000 n +0000255342 00000 n +0000255476 00000 n +0000255611 00000 n +0000255675 00000 n +0000255730 00000 n +0000255769 00000 n +0000255833 00000 n +0000255888 00000 n +0000255954 00000 n +0000255995 00000 n +0000256023 00000 n +0000256087 00000 n +0000256142 00000 n +0000256208 00000 n +0000256249 00000 n +0000256277 00000 n +0000256341 00000 n +0000256396 00000 n +0000256462 00000 n +0000256503 00000 n +0000256531 00000 n +0000256595 00000 n +0000256650 00000 n +0000256716 00000 n +0000256757 00000 n +0000256785 00000 n +0000256849 00000 n +0000256904 00000 n +0000256970 00000 n +0000257011 00000 n +0000257039 00000 n +0000257103 00000 n +0000257158 00000 n +0000257224 00000 n +0000257265 00000 n +0000257293 00000 n +0000257348 00000 n +0000257414 00000 n +0000257455 00000 n +0000257483 00000 n +0000257522 00000 n +0000257588 00000 n +0000257629 00000 n +0000257657 00000 n +0000266032 00000 n +0000272712 00000 n +0000272756 00000 n +0000272800 00000 n +0000272844 00000 n +0000272888 00000 n +0000272932 00000 n +0000272976 00000 n +0000273020 00000 n +0000273064 00000 n +0000273128 00000 n +0000273183 00000 n +0000273235 00000 n +0000273369 00000 n +0000273504 00000 n +0000273568 00000 n +0000273623 00000 n +0000273662 00000 n +0000273726 00000 n +0000273781 00000 n +0000273847 00000 n +0000273888 00000 n +0000273916 00000 n +0000273980 00000 n +0000274035 00000 n +0000274101 00000 n +0000274142 00000 n +0000274170 00000 n +0000274234 00000 n +0000274289 00000 n +0000274355 00000 n +0000274396 00000 n +0000274424 00000 n +0000274488 00000 n +0000274543 00000 n +0000274609 00000 n +0000274650 00000 n +0000274678 00000 n +0000274745 00000 n +0000274800 00000 n +0000274866 00000 n +0000274907 00000 n +0000274935 00000 n +0000275002 00000 n +0000275058 00000 n +0000275126 00000 n +0000275169 00000 n +0000275198 00000 n +0000275254 00000 n +0000275322 00000 n +0000275365 00000 n +0000275394 00000 n +0000275434 00000 n +0000275502 00000 n +0000275545 00000 n +0000275574 00000 n +0000283948 00000 n +0000290851 00000 n +0000290896 00000 n +0000290941 00000 n +0000290986 00000 n +0000291031 00000 n +0000291076 00000 n +0000291121 00000 n +0000291166 00000 n +0000291235 00000 n +0000291291 00000 n +0000291347 00000 n +0000291473 00000 n +0000291600 00000 n +0000291669 00000 n +0000291725 00000 n +0000291765 00000 n +0000291834 00000 n +0000291890 00000 n +0000291958 00000 n +0000292001 00000 n +0000292030 00000 n +0000292099 00000 n +0000292155 00000 n +0000292223 00000 n +0000292266 00000 n +0000292295 00000 n +0000292364 00000 n +0000292420 00000 n +0000292488 00000 n +0000292531 00000 n +0000292560 00000 n +0000292629 00000 n +0000292685 00000 n +0000292753 00000 n +0000292796 00000 n +0000292825 00000 n +0000292894 00000 n +0000292950 00000 n +0000293018 00000 n +0000293061 00000 n +0000293090 00000 n +0000293146 00000 n +0000293214 00000 n +0000293257 00000 n +0000293286 00000 n +0000293326 00000 n +0000293394 00000 n +0000293437 00000 n +0000293466 00000 n +0000301840 00000 n +0000309188 00000 n +0000309233 00000 n +0000309278 00000 n +0000309323 00000 n +0000309368 00000 n +0000309413 00000 n +0000309458 00000 n +0000309503 00000 n +0000309572 00000 n +0000309628 00000 n +0000309684 00000 n +0000309810 00000 n +0000309937 00000 n +0000310006 00000 n +0000310062 00000 n +0000310102 00000 n +0000310171 00000 n +0000310227 00000 n +0000310295 00000 n +0000310338 00000 n +0000310367 00000 n +0000310436 00000 n +0000310492 00000 n +0000310560 00000 n +0000310603 00000 n +0000310632 00000 n +0000310701 00000 n +0000310757 00000 n +0000310825 00000 n +0000310868 00000 n +0000310897 00000 n +0000310966 00000 n +0000311022 00000 n +0000311090 00000 n +0000311133 00000 n +0000311162 00000 n +0000311231 00000 n +0000311287 00000 n +0000311355 00000 n +0000311398 00000 n +0000311427 00000 n +0000311483 00000 n +0000311551 00000 n +0000311594 00000 n +0000311623 00000 n +0000311663 00000 n +0000311731 00000 n +0000311774 00000 n +0000311803 00000 n +0000320153 00000 n +0000327308 00000 n +0000327353 00000 n +0000327398 00000 n +0000327443 00000 n +0000327488 00000 n +0000327533 00000 n +0000327602 00000 n +0000327658 00000 n +0000327714 00000 n +0000327822 00000 n +0000327931 00000 n +0000328000 00000 n +0000328056 00000 n +0000328096 00000 n +0000328165 00000 n +0000328221 00000 n +0000328289 00000 n +0000328332 00000 n +0000328361 00000 n +0000328430 00000 n +0000328486 00000 n +0000328554 00000 n +0000328597 00000 n +0000328626 00000 n +0000328695 00000 n +0000328751 00000 n +0000328819 00000 n +0000328862 00000 n +0000328891 00000 n +0000328947 00000 n +0000329015 00000 n +0000329058 00000 n +0000329087 00000 n +0000329127 00000 n +0000329195 00000 n +0000329238 00000 n +0000329267 00000 n +0000337629 00000 n +0000344307 00000 n +0000344352 00000 n +0000344397 00000 n +0000344442 00000 n +0000344487 00000 n +0000344532 00000 n +0000344577 00000 n +0000344646 00000 n +0000344702 00000 n +0000344758 00000 n +0000344875 00000 n +0000344993 00000 n +0000345062 00000 n +0000345118 00000 n +0000345158 00000 n +0000345227 00000 n +0000345283 00000 n +0000345351 00000 n +0000345394 00000 n +0000345423 00000 n +0000345492 00000 n +0000345548 00000 n +0000345616 00000 n +0000345659 00000 n +0000345688 00000 n +0000345757 00000 n +0000345813 00000 n +0000345881 00000 n +0000345924 00000 n +0000345953 00000 n +0000346022 00000 n +0000346078 00000 n +0000346146 00000 n +0000346189 00000 n +0000346218 00000 n +0000346274 00000 n +0000346342 00000 n +0000346385 00000 n +0000346414 00000 n +0000346454 00000 n +0000346522 00000 n +0000346565 00000 n +0000346594 00000 n +0000354944 00000 n +0000361593 00000 n +0000361638 00000 n +0000361683 00000 n +0000361728 00000 n +0000361773 00000 n +0000361818 00000 n +0000361887 00000 n +0000361943 00000 n +0000361999 00000 n +0000362107 00000 n +0000362216 00000 n +0000362285 00000 n +0000362341 00000 n +0000362381 00000 n +0000362450 00000 n +0000362506 00000 n +0000362574 00000 n +0000362617 00000 n +0000362646 00000 n +0000362715 00000 n +0000362771 00000 n +0000362839 00000 n +0000362882 00000 n +0000362911 00000 n +0000362980 00000 n +0000363036 00000 n +0000363104 00000 n +0000363147 00000 n +0000363176 00000 n +0000363232 00000 n +0000363300 00000 n +0000363343 00000 n +0000363372 00000 n +0000363412 00000 n +0000363480 00000 n +0000363523 00000 n +0000363552 00000 n +0000371914 00000 n +0000378586 00000 n +0000378631 00000 n +0000378676 00000 n +0000378721 00000 n +0000378766 00000 n +0000378811 00000 n +0000378856 00000 n +0000378925 00000 n +0000378981 00000 n +0000379037 00000 n +0000379154 00000 n +0000379272 00000 n +0000379341 00000 n +0000379397 00000 n +0000379437 00000 n +0000379506 00000 n +0000379562 00000 n +0000379630 00000 n +0000379673 00000 n +0000379702 00000 n +0000379771 00000 n +0000379827 00000 n +0000379895 00000 n +0000379938 00000 n +0000379967 00000 n +0000380036 00000 n +0000380092 00000 n +0000380160 00000 n +0000380203 00000 n +0000380232 00000 n +0000380301 00000 n +0000380357 00000 n +0000380425 00000 n +0000380468 00000 n +0000380497 00000 n +0000380553 00000 n +0000380621 00000 n +0000380664 00000 n +0000380693 00000 n +0000380733 00000 n +0000380801 00000 n +0000380844 00000 n +0000380873 00000 n +0000389199 00000 n +0000395761 00000 n +0000395806 00000 n +0000395851 00000 n +0000395896 00000 n +0000395965 00000 n +0000396021 00000 n +0000396077 00000 n +0000396167 00000 n +0000396258 00000 n +0000396327 00000 n +0000396383 00000 n +0000396423 00000 n +0000396492 00000 n +0000396548 00000 n +0000396616 00000 n +0000396659 00000 n +0000396688 00000 n +0000396744 00000 n +0000396812 00000 n +0000396855 00000 n +0000396884 00000 n +0000396924 00000 n +0000396992 00000 n +0000397035 00000 n +0000397064 00000 n +0000405439 00000 n +0000412220 00000 n +0000412265 00000 n +0000412310 00000 n +0000412355 00000 n +0000412400 00000 n +0000412445 00000 n +0000412514 00000 n +0000412570 00000 n +0000412626 00000 n +0000412734 00000 n +0000412843 00000 n +0000412912 00000 n +0000412968 00000 n +0000413008 00000 n +0000413077 00000 n +0000413133 00000 n +0000413201 00000 n +0000413244 00000 n +0000413273 00000 n +0000413342 00000 n +0000413398 00000 n +0000413466 00000 n +0000413509 00000 n +0000413538 00000 n +0000413607 00000 n +0000413663 00000 n +0000413731 00000 n +0000413774 00000 n +0000413803 00000 n +0000413859 00000 n +0000413927 00000 n +0000413970 00000 n +0000413999 00000 n +0000414039 00000 n +0000414107 00000 n +0000414150 00000 n +0000414179 00000 n +0000422554 00000 n +0000429225 00000 n +0000429270 00000 n +0000429315 00000 n +0000429360 00000 n +0000429405 00000 n +0000429450 00000 n +0000429519 00000 n +0000429575 00000 n +0000429631 00000 n +0000429739 00000 n +0000429848 00000 n +0000429917 00000 n +0000429973 00000 n +0000430013 00000 n +0000430082 00000 n +0000430138 00000 n +0000430206 00000 n +0000430249 00000 n +0000430278 00000 n +0000430347 00000 n +0000430403 00000 n +0000430471 00000 n +0000430514 00000 n +0000430543 00000 n +0000430612 00000 n +0000430668 00000 n +0000430736 00000 n +0000430779 00000 n +0000430808 00000 n +0000430864 00000 n +0000430932 00000 n +0000430975 00000 n +0000431004 00000 n +0000431044 00000 n +0000431112 00000 n +0000431155 00000 n +0000431184 00000 n +0000439559 00000 n +0000446244 00000 n +0000446289 00000 n +0000446334 00000 n +0000446379 00000 n +0000446424 00000 n +0000446469 00000 n +0000446538 00000 n +0000446594 00000 n +0000446650 00000 n +0000446758 00000 n +0000446867 00000 n +0000446936 00000 n +0000446992 00000 n +0000447032 00000 n +0000447101 00000 n +0000447157 00000 n +0000447225 00000 n +0000447268 00000 n +0000447297 00000 n +0000447366 00000 n +0000447422 00000 n +0000447490 00000 n +0000447533 00000 n +0000447562 00000 n +0000447631 00000 n +0000447687 00000 n +0000447755 00000 n +0000447798 00000 n +0000447827 00000 n +0000447883 00000 n +0000447951 00000 n +0000447994 00000 n +0000448023 00000 n +0000448063 00000 n +0000448131 00000 n +0000448174 00000 n +0000448203 00000 n +0000456578 00000 n +0000463347 00000 n +0000463392 00000 n +0000463437 00000 n +0000463482 00000 n +0000463527 00000 n +0000463572 00000 n +0000463641 00000 n +0000463697 00000 n +0000463753 00000 n +0000463861 00000 n +0000463970 00000 n +0000464039 00000 n +0000464095 00000 n +0000464135 00000 n +0000464204 00000 n +0000464260 00000 n +0000464328 00000 n +0000464371 00000 n +0000464400 00000 n +0000464469 00000 n +0000464525 00000 n +0000464593 00000 n +0000464636 00000 n +0000464665 00000 n +0000464734 00000 n +0000464790 00000 n +0000464858 00000 n +0000464901 00000 n +0000464930 00000 n +0000464986 00000 n +0000465054 00000 n +0000465097 00000 n +0000465126 00000 n +0000465166 00000 n +0000465234 00000 n +0000465277 00000 n +0000465306 00000 n +0000473681 00000 n +0000480432 00000 n +0000480477 00000 n +0000480522 00000 n +0000480567 00000 n +0000480612 00000 n +0000480657 00000 n +0000480726 00000 n +0000480782 00000 n +0000480838 00000 n +0000480946 00000 n +0000481055 00000 n +0000481124 00000 n +0000481180 00000 n +0000481220 00000 n +0000481289 00000 n +0000481345 00000 n +0000481413 00000 n +0000481456 00000 n +0000481485 00000 n +0000481554 00000 n +0000481610 00000 n +0000481678 00000 n +0000481721 00000 n +0000481750 00000 n +0000481819 00000 n +0000481875 00000 n +0000481943 00000 n +0000481986 00000 n +0000482015 00000 n +0000482071 00000 n +0000482139 00000 n +0000482182 00000 n +0000482211 00000 n +0000482251 00000 n +0000482319 00000 n +0000482362 00000 n +0000482391 00000 n +0000490766 00000 n +0000497474 00000 n +0000497519 00000 n +0000497564 00000 n +0000497609 00000 n +0000497654 00000 n +0000497699 00000 n +0000497768 00000 n +0000497824 00000 n +0000497880 00000 n +0000497988 00000 n +0000498097 00000 n +0000498166 00000 n +0000498222 00000 n +0000498262 00000 n +0000498331 00000 n +0000498387 00000 n +0000498455 00000 n +0000498498 00000 n +0000498527 00000 n +0000498596 00000 n +0000498652 00000 n +0000498720 00000 n +0000498763 00000 n +0000498792 00000 n +0000498861 00000 n +0000498917 00000 n +0000498985 00000 n +0000499028 00000 n +0000499057 00000 n +0000499113 00000 n +0000499181 00000 n +0000499224 00000 n +0000499253 00000 n +0000499293 00000 n +0000499361 00000 n +0000499404 00000 n +0000499433 00000 n +0000507808 00000 n +0000514378 00000 n +0000514423 00000 n +0000514468 00000 n +0000514513 00000 n +0000514558 00000 n +0000514603 00000 n +0000514672 00000 n +0000514728 00000 n +0000514784 00000 n +0000514892 00000 n +0000515001 00000 n +0000515070 00000 n +0000515126 00000 n +0000515166 00000 n +0000515235 00000 n +0000515291 00000 n +0000515359 00000 n +0000515402 00000 n +0000515431 00000 n +0000515500 00000 n +0000515556 00000 n +0000515624 00000 n +0000515667 00000 n +0000515696 00000 n +0000515765 00000 n +0000515821 00000 n +0000515889 00000 n +0000515932 00000 n +0000515961 00000 n +0000516017 00000 n +0000516085 00000 n +0000516128 00000 n +0000516157 00000 n +0000516197 00000 n +0000516265 00000 n +0000516308 00000 n +0000516337 00000 n +0000524724 00000 n +0000531457 00000 n +0000531502 00000 n +0000531547 00000 n +0000531592 00000 n +0000531637 00000 n +0000531682 00000 n +0000531727 00000 n +0000531796 00000 n +0000531852 00000 n +0000531908 00000 n +0000532025 00000 n +0000532143 00000 n +0000532212 00000 n +0000532268 00000 n +0000532308 00000 n +0000532377 00000 n +0000532433 00000 n +0000532501 00000 n +0000532544 00000 n +0000532573 00000 n +0000532642 00000 n +0000532698 00000 n +0000532766 00000 n +0000532809 00000 n +0000532838 00000 n +0000532907 00000 n +0000532963 00000 n +0000533031 00000 n +0000533074 00000 n +0000533103 00000 n +0000533172 00000 n +0000533228 00000 n +0000533296 00000 n +0000533339 00000 n +0000533368 00000 n +0000533424 00000 n +0000533492 00000 n +0000533535 00000 n +0000533564 00000 n +0000533604 00000 n +0000533672 00000 n +0000533715 00000 n +0000533744 00000 n +0000542119 00000 n +0000548908 00000 n +0000548953 00000 n +0000548998 00000 n +0000549043 00000 n +0000549088 00000 n +0000549133 00000 n +0000549202 00000 n +0000549258 00000 n +0000549314 00000 n +0000549422 00000 n +0000549531 00000 n +0000549600 00000 n +0000549656 00000 n +0000549696 00000 n +0000549765 00000 n +0000549821 00000 n +0000549889 00000 n +0000549932 00000 n +0000549961 00000 n +0000550030 00000 n +0000550086 00000 n +0000550154 00000 n +0000550197 00000 n +0000550226 00000 n +0000550295 00000 n +0000550351 00000 n +0000550419 00000 n +0000550462 00000 n +0000550491 00000 n +0000550547 00000 n +0000550615 00000 n +0000550658 00000 n +0000550687 00000 n +0000550727 00000 n +0000550795 00000 n +0000550838 00000 n +0000550867 00000 n +0000559254 00000 n +0000566181 00000 n +0000566226 00000 n +0000566271 00000 n +0000566316 00000 n +0000566361 00000 n +0000566406 00000 n +0000566451 00000 n +0000566520 00000 n +0000566576 00000 n +0000566632 00000 n +0000566749 00000 n +0000566867 00000 n +0000566936 00000 n +0000566992 00000 n +0000567032 00000 n +0000567101 00000 n +0000567157 00000 n +0000567225 00000 n +0000567268 00000 n +0000567297 00000 n +0000567366 00000 n +0000567422 00000 n +0000567490 00000 n +0000567533 00000 n +0000567562 00000 n +0000567631 00000 n +0000567687 00000 n +0000567755 00000 n +0000567798 00000 n +0000567827 00000 n +0000567896 00000 n +0000567952 00000 n +0000568020 00000 n +0000568063 00000 n +0000568092 00000 n +0000568148 00000 n +0000568216 00000 n +0000568259 00000 n +0000568288 00000 n +0000568328 00000 n +0000568396 00000 n +0000568439 00000 n +0000568468 00000 n +0000576819 00000 n +0000583521 00000 n +0000583566 00000 n +0000583611 00000 n +0000583656 00000 n +0000583701 00000 n +0000583746 00000 n +0000583815 00000 n +0000583871 00000 n +0000583927 00000 n +0000584035 00000 n +0000584144 00000 n +0000584213 00000 n +0000584269 00000 n +0000584309 00000 n +0000584378 00000 n +0000584434 00000 n +0000584502 00000 n +0000584545 00000 n +0000584574 00000 n +0000584643 00000 n +0000584699 00000 n +0000584767 00000 n +0000584810 00000 n +0000584839 00000 n +0000584908 00000 n +0000584964 00000 n +0000585032 00000 n +0000585075 00000 n +0000585104 00000 n +0000585160 00000 n +0000585228 00000 n +0000585271 00000 n +0000585300 00000 n +0000585340 00000 n +0000585408 00000 n +0000585451 00000 n +0000585480 00000 n +0000593765 00000 n +0000600297 00000 n +0000608612 00000 n +0000615388 00000 n +0000615433 00000 n +0000615502 00000 n +0000615558 00000 n +0000615614 00000 n +0000615686 00000 n +0000615759 00000 n +0000615815 00000 n +0000615855 00000 n +0000615895 00000 n +0000615963 00000 n +0000616006 00000 n +0000616035 00000 n +0000624350 00000 n +0000631036 00000 n +0000631081 00000 n +0000631150 00000 n +0000631206 00000 n +0000631262 00000 n +0000631334 00000 n +0000631407 00000 n +0000631463 00000 n +0000631503 00000 n +0000631543 00000 n +0000631611 00000 n +0000631654 00000 n +0000631683 00000 n +0000639998 00000 n +0000646891 00000 n +0000646936 00000 n +0000647005 00000 n +0000647061 00000 n +0000647117 00000 n +0000647189 00000 n +0000647262 00000 n +0000647318 00000 n +0000647358 00000 n +0000647398 00000 n +0000647466 00000 n +0000647509 00000 n +0000647538 00000 n +0000655853 00000 n +0000662482 00000 n +0000662527 00000 n +0000662596 00000 n +0000662652 00000 n +0000662708 00000 n +0000662780 00000 n +0000662853 00000 n +0000662909 00000 n +0000662949 00000 n +0000662989 00000 n +0000663057 00000 n +0000663100 00000 n +0000663129 00000 n +0000671492 00000 n +0000678213 00000 n +0000678258 00000 n +0000678303 00000 n +0000678348 00000 n +0000678393 00000 n +0000678438 00000 n +0000678483 00000 n +0000678552 00000 n +0000678608 00000 n +0000678664 00000 n +0000678781 00000 n +0000678899 00000 n +0000678968 00000 n +0000679024 00000 n +0000679064 00000 n +0000679133 00000 n +0000679189 00000 n +0000679257 00000 n +0000679300 00000 n +0000679329 00000 n +0000679398 00000 n +0000679454 00000 n +0000679522 00000 n +0000679565 00000 n +0000679594 00000 n +0000679663 00000 n +0000679719 00000 n +0000679787 00000 n +0000679830 00000 n +0000679859 00000 n +0000679928 00000 n +0000679984 00000 n +0000680052 00000 n +0000680095 00000 n +0000680124 00000 n +0000680180 00000 n +0000680248 00000 n +0000680291 00000 n +0000680320 00000 n +0000680360 00000 n +0000680428 00000 n +0000680471 00000 n +0000680500 00000 n +0000680996 00000 n +0000681242 00000 n +0000681498 00000 n +0000682169 00000 n +0000684373 00000 n +0000686074 00000 n +0000687760 00000 n +0000688412 00000 n +0000688449 00000 n +0000688474 00000 n +0000688545 00000 n +0000688685 00000 n +0000688827 00000 n +0000688944 00000 n +0000692566 00000 n +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +0000000000 65535 f +trailer +<> +startxref +116 +%%EOF diff --git a/tests/gnarly_pdfs/large_prompt_hint2.pdf b/tests/gnarly_pdfs/large_prompt_hint2.pdf new file mode 100644 index 0000000..7e0f584 Binary files /dev/null and b/tests/gnarly_pdfs/large_prompt_hint2.pdf differ diff --git a/tests/gnarly_pdfs/large_prompt_hint3.pdf b/tests/gnarly_pdfs/large_prompt_hint3.pdf new file mode 100644 index 0000000..9387a61 Binary files /dev/null and b/tests/gnarly_pdfs/large_prompt_hint3.pdf differ diff --git a/tests/gnarly_pdfs/load_v_error.pdf b/tests/gnarly_pdfs/load_v_error.pdf new file mode 100644 index 0000000..342f13d Binary files /dev/null and b/tests/gnarly_pdfs/load_v_error.pdf differ diff --git a/tests/gnarly_pdfs/lots_of_chem_tables.pdf b/tests/gnarly_pdfs/lots_of_chem_tables.pdf new file mode 100644 index 0000000..0ec64ee Binary files /dev/null and b/tests/gnarly_pdfs/lots_of_chem_tables.pdf differ diff --git a/tests/gnarly_pdfs/lots_of_sci_tables.pdf b/tests/gnarly_pdfs/lots_of_sci_tables.pdf new file mode 100644 index 0000000..cb41032 Binary files /dev/null and b/tests/gnarly_pdfs/lots_of_sci_tables.pdf differ diff --git a/tests/gnarly_pdfs/map1.pdf b/tests/gnarly_pdfs/map1.pdf new file mode 100644 index 0000000..bd8bd55 Binary files /dev/null and b/tests/gnarly_pdfs/map1.pdf differ diff --git a/tests/gnarly_pdfs/most_content_in_image_form.pdf b/tests/gnarly_pdfs/most_content_in_image_form.pdf new file mode 100644 index 0000000..fa7dd8e Binary files /dev/null and b/tests/gnarly_pdfs/most_content_in_image_form.pdf differ diff --git a/tests/gnarly_pdfs/newspaper.pdf b/tests/gnarly_pdfs/newspaper.pdf new file mode 100644 index 0000000..275b6ee Binary files /dev/null and b/tests/gnarly_pdfs/newspaper.pdf differ diff --git a/tests/gnarly_pdfs/not_parsing.pdf b/tests/gnarly_pdfs/not_parsing.pdf new file mode 100644 index 0000000..252e00b Binary files /dev/null and b/tests/gnarly_pdfs/not_parsing.pdf differ diff --git a/tests/gnarly_pdfs/not_parsing2.pdf b/tests/gnarly_pdfs/not_parsing2.pdf new file mode 100644 index 0000000..b103501 Binary files /dev/null and b/tests/gnarly_pdfs/not_parsing2.pdf differ diff --git a/tests/gnarly_pdfs/olmo-page-1.pdf b/tests/gnarly_pdfs/olmo-page-1.pdf new file mode 100644 index 0000000..0ab910e Binary files /dev/null and b/tests/gnarly_pdfs/olmo-page-1.pdf differ diff --git a/tests/gnarly_pdfs/overrun_on_pg8.pdf b/tests/gnarly_pdfs/overrun_on_pg8.pdf new file mode 100644 index 0000000..1042afe Binary files /dev/null and b/tests/gnarly_pdfs/overrun_on_pg8.pdf differ diff --git a/tests/gnarly_pdfs/pdftotext_two_column_issue.pdf b/tests/gnarly_pdfs/pdftotext_two_column_issue.pdf new file mode 100644 index 0000000..38405f1 Binary files /dev/null and b/tests/gnarly_pdfs/pdftotext_two_column_issue.pdf differ diff --git a/tests/gnarly_pdfs/repeating_references_on_pg9_pg10.pdf b/tests/gnarly_pdfs/repeating_references_on_pg9_pg10.pdf new file mode 100644 index 0000000..f523ec0 Binary files /dev/null and b/tests/gnarly_pdfs/repeating_references_on_pg9_pg10.pdf differ diff --git a/tests/gnarly_pdfs/skinnypage.pdf b/tests/gnarly_pdfs/skinnypage.pdf new file mode 100644 index 0000000..50d4eba Binary files /dev/null and b/tests/gnarly_pdfs/skinnypage.pdf differ diff --git a/tests/gnarly_pdfs/slideshow_mostly_good_some_pages_should_get_filtered.pdf b/tests/gnarly_pdfs/slideshow_mostly_good_some_pages_should_get_filtered.pdf new file mode 100644 index 0000000..b903ad5 Binary files /dev/null and b/tests/gnarly_pdfs/slideshow_mostly_good_some_pages_should_get_filtered.pdf differ diff --git a/tests/gnarly_pdfs/slideshow_mostly_images.pdf b/tests/gnarly_pdfs/slideshow_mostly_images.pdf new file mode 100644 index 0000000..40da519 Binary files /dev/null and b/tests/gnarly_pdfs/slideshow_mostly_images.pdf differ diff --git a/tests/gnarly_pdfs/small_page_size.pdf b/tests/gnarly_pdfs/small_page_size.pdf new file mode 100644 index 0000000..c9f8806 Binary files /dev/null and b/tests/gnarly_pdfs/small_page_size.pdf differ diff --git a/tests/gnarly_pdfs/some_ocr1.pdf b/tests/gnarly_pdfs/some_ocr1.pdf new file mode 100644 index 0000000..8098eda Binary files /dev/null and b/tests/gnarly_pdfs/some_ocr1.pdf differ diff --git a/tests/gnarly_pdfs/ti89_guidebook_programming.pdf b/tests/gnarly_pdfs/ti89_guidebook_programming.pdf new file mode 100644 index 0000000..6dde295 Binary files /dev/null and b/tests/gnarly_pdfs/ti89_guidebook_programming.pdf differ diff --git a/tests/gnarly_pdfs/tobacco_missed_tokens_pg1.pdf b/tests/gnarly_pdfs/tobacco_missed_tokens_pg1.pdf new file mode 100644 index 0000000..b02206b --- /dev/null +++ b/tests/gnarly_pdfs/tobacco_missed_tokens_pg1.pdf @@ -0,0 +1,2002 @@ +%PDF-1.4 %�� +55 0 obj +<> +endobj +53 0 obj +<>stream + + + + + Acrobat Distiller 5.0.5 (Windows) + + + 2004-11-15T14:58:55Z + 2017-07-12T05:22:36-07:00 + 2017-07-12T05:22:36-07:00 + + + uuid:769b8e27-1dd2-11b2-0a00-720927bda400 + uuid:769b8e2c-1dd2-11b2-0a00-5c0000000000 + + + application/pdf + + + + +endstream +endobj +51 0 obj +<> +endobj +56 0 obj +<>/ExtGState<>/Font<>/ProcSet[/PDF/Text]>>/Rotate 0/TrimBox[57.25999 58.11 666.34 848.69299]/Type/Page>> +endobj +1 0 obj +<>/ExtGState<>/Font<>/ProcSet[/PDF/Text/ImageB]/XObject<>>>/Rotate 0/TrimBox[57.25999 58.11 666.34 848.69299]/Type/Page>> +endobj +5 0 obj +<>/ExtGState<>/Font<>/ProcSet[/PDF/Text]>>/Rotate 0/TrimBox[57.25999 58.11 666.34 848.69299]/Type/Page>> +endobj +8 0 obj +<>/ExtGState<>/Font<>/ProcSet[/PDF/Text]>>/Rotate 0/TrimBox[57.25999 58.11 666.34 848.69299]/Type/Page>> +endobj +11 0 obj +<>/ExtGState<>/Font<>/ProcSet[/PDF/Text]>>/Rotate 0/TrimBox[57.25999 58.11 666.34 848.69299]/Type/Page>> +endobj +14 0 obj +<>/ExtGState<>/Font<>/ProcSet[/PDF/Text]>>/Rotate 0/TrimBox[57.25999 58.11 666.34 848.69299]/Type/Page>> +endobj +17 0 obj +<>/ExtGState<>/Font<>/ProcSet[/PDF/Text]>>/Rotate 0/TrimBox[57.25999 58.11 666.34 848.69299]/Type/Page>> +endobj +20 0 obj +<>/ExtGState<>/Font<>/ProcSet[/PDF/Text]>>/Rotate 0/TrimBox[57.25999 58.11 666.34 848.69299]/Type/Page>> +endobj +23 0 obj +<>/ExtGState<>/Font<>/ProcSet[/PDF/Text]>>/Rotate 0/TrimBox[57.25999 58.11 666.34 848.69299]/Type/Page>> +endobj +105 0 obj +<>/ProcSet[/PDF/Text/ImageC]/XObject<>>>/Type/Page>> +endobj +134 0 obj +[139 0 R 140 0 R 141 0 R 142 0 R 143 0 R 144 0 R 145 0 R] +endobj +135 0 obj +<>stream +q +61.5 0 0 38.75 103.25 713.25 cm +/Im0 Do +Q +BT +/T1_0 1 Tf +15 0 0 15 184 721.99997 Tm +(industry: hope or hype?)Tj +0 1 TD +(Corporate social responsibility and the tobacco)Tj +ET +BT +/T1_1 1 Tf +10 0 0 10 184 699 Tm +(N Hirschhorn)Tj +ET +BT +/T1_1 1 Tf +12 0 0 12 184 662.99997 Tm +(doi: 10.1136/tc.2003.006676)Tj +10 0 0 10 239.0199 675 Tm +(2004 13: 447-453 )Tj +/T1_2 1 Tf +-5.50199 0 Td +(Tob Control\240)Tj +ET +84 649 m +528 649 l +0 0 m +S +BT +ET +BT +/T1_1 1 Tf +10 0 0 10 408.48874 614.99997 Tm +( )Tj +0 0 1 rg +/T1_0 1 Tf +-22.44887 0 Td +(http://tobaccocontrol.bmj.com/content/13/4/447)Tj +0 g +/T1_1 1 Tf +T* +(Updated information and services can be found at: )Tj +ET +84 601 m +528 601 l +0 0 m +S +BT +ET +BT +/T1_2 1 Tf +10 0 0 10 184 573 Tm +(These include:)Tj +ET +BT +/T1_0 1 Tf +12 0 0 12 99.30817 553 Tm +(References)Tj +ET +BT +/T1_1 1 Tf +10 0 0 10 435.72888 544.99997 Tm +( )Tj +0 0 1 rg +-2.72401 0 Td +(#BIBL)Tj +/T1_0 1 Tf +-22.44888 0 Td +(http://tobaccocontrol.bmj.com/content/13/4/447)Tj +0 g +/T1_1 1 Tf +T* +(This article cites 4 articles, 1 of which you can access for free at: )Tj +ET +BT +/T1_0 1 Tf +12 0 0 12 119.29999 512.99997 Tm +(service)Tj +-2.83299 1 Td +(Email alerting)Tj +ET +BT +/T1_1 1 Tf +10 0 0 10 184 516.99997 Tm +(box at the top right corner of the online article. )Tj +T* +(Receive free email alerts when new articles cite this article. Sign up i\ n the)Tj +ET +84 490 m +528 490 l +0 0 m +S +BT +ET +BT +/T1_0 1 Tf +12 0 0 12 84 464 Tm +(Notes)Tj +ET +84 182 m +528 182 l +0 0 m +S +BT +ET +BT +0 0 1 rg +/T1_0 1 Tf +10 0 0 10 84 143.99997 Tm +(http://group.bmj.com/group/rights-licensing/permissions)Tj +0 g +/T1_1 1 Tf +T* +(To request permissions go to:)Tj +ET +BT +0 0 1 rg +/T1_0 1 Tf +10 0 0 10 84 109.99997 Tm +(http://journals.bmj.com/cgi/reprintform)Tj +0 g +/T1_1 1 Tf +T* +(To order reprints go to:)Tj +ET +BT +0 0 1 rg +/T1_0 1 Tf +10 0 0 10 84 75.99997 Tm +(http://group.bmj.com/subscribe/)Tj +0 g +/T1_1 1 Tf +T* +(To subscribe to BMJ go to:)Tj +ET +BT +0 0 1 rg +/T1_1 1 Tf +8 0 0 8 423.39191 779 Tm +(group.bmj.com)Tj +0 g +-14.508 0 Td +( on July 12, 2017 - Published by )Tj +0 0 1 rg +-13.33897 0 Td +(http://tobaccocontrol.bmj.com/)Tj +0 g +-8.11398 0 Td +(Downloaded from )Tj +ET + +endstream +endobj +138 0 obj +<>/Filter/FlateDecode/Height 155/Length 4254/Name/X/Subtype/Image/Type/XObject/Width 246>>stream +H�� PSw��VEQ��@ +�U�x�*�(�"��Z� ^��h��U)� +��Hut-**Z+��D��VE�.#�n ��$o_��]_�K�$��gƙ��~y_�AQ��jkk)(p� +��n��A�۠p�mP8�6(p� +��n��A�۠p�mP8�6(p� +��n��A�۠p�mP8�6(p� +��n��A�۠p�mP8�6(p� +��;�^WU��Ueɋ��*Ajt�^�H�bW��[{/W�6r7�H��mfT�R�@�<^s�5��$��a9ړ�??�w�A~��*�D$�tZ�go~)��/@"��ŭ�w��˯�|�N�>.uiޣ�<��7��y�'Zs�JE��+�7� +*�54(�G U��X��+�:�#��^��7>��x�zX��@��9=��*;LረK��7d�e�.�|��y��lU�sC��8[��]�AZ�ˣ��Љ�!��Z��P�Ɓ +�5��\�*\��ղ�1�~�Z�z��GU��ɐ7.�%]�j��~�N�vX;�ېE��ܮ��GH�l�E��I��L��l2�n@�6df��_��t(?��UѦ[3 �:�wL��z�� j>P� �S��eQi�w��F2��-�i��D(90mKt`xMlrP� �.<�� W$qn�}��iӶ�}�ԣ�nC� +�^^4f��ڪf��ё�?,��I��ـ-A�6Ģp��d`R��ڗH,m�}1 ՞r��#]=;5�6l +�!S�W��=��=$vj�}1�!�2�{��%M�ۄ�p2Ux��E?��d��Pz��h ?��S>�Ƿ�]�A�>b��,܌]��m蝅'^>�<�+$u��tϤ��B��^g��2�^_]WS�x�T��OH B�^��:1_�w$��%��KT�ǉSn�\�ϩ几/qBsN��JS�m��9a�̷?)xV<(i�J�A�yǤ��4H�Z��7��_��^�r��I-��I��kO�df�*+4_(6��P�ЫK��Yv�?T��<+6�|��+�+n�&Q�@ ��/1��Q-;��?�gnNq��p !Kb��/ ��o��Z-y�v��3�VT�#��9ST��fܶ#r�w�A��@b��z��T�x��6��طq��lLZ��7#B��>��R��z��ug +�#��wH¿�aB��]jQ�g��˾��kNy��V�b�C2Q��Uj��ZM�߾��v6�[�C��#�E��;��ϗ\�g6��KW��R��;��\��NC��ASӦ|m񳱊T�޻0��'$��x�Ҩ��C6��k�8��3�M�* �/)�.�6��CD��Y�v�̺��J�ݶ��Z5Ғ}�.2+*p�5�� +��6��{)"e>b��A��xv΋?�F��u �47��觠�޾_��h`47�߀٦�ȴ�̂,$aY8�#5��0�ߘ|q?��Y��m��b��psn�k��٠��U�=�F�u��_B�!��g�܆k��O*��@B��[ �?lև^]0N�+P83��CqG�!��j��~�2�[\h�x��Db� ߐ �� +g��p=�N�Z��p�<�{��FuD}�w+��3��|�wG�3�]�Ӛ�mGh�gE (*7�pp�X��dP83܅g]�w>�l�j5I��U��=��2�%��.<�Ԏ��H��j5�\?6aݨ�XF�pP83��)�o��[e��hJ/��Vfvo�� +g�� V�V#��j�گ}��U'|��k�%��{n��i��k&�\:|�� +7k�犯|�w.�G��]Q}FZ? (��o��NnA2��ţ��`�\��pc��؆��H��f��ԥt��8��I� �n�Bg�Kud7��5�]$N�� +7k�#R�](��DK�?��_�pL�g�N +X�a-|��!��R�*�{��e� +A�̰�q┛O~F"s� �_\�j�P� +g��p��u��$�W�룂>?��ʡ� (��µzm�mK*JYN��:0m��C�P8��+|VhT�ά��R&�ˈ�a�P��f��'�X�i��Y��k�v��}x �e��Dp��pfX�2�BQ�8�^��V��H �~.@P�1X�J_�q�$�8�^��z�z�ý��r�w�Y} �� ]xkG{ܦ��ԍ�M�jΉ�ܺx� �dP8?��d�ִ3�j�Zs��B�12L� ܯ��\�+Ӑ=�1�o�o�p~B�m�:�sp=Riܽ�j�x6�8�#�6��C��b>��f��͏�� +�'B�MƖq�1I(�.pr�_]��9(��h�c�Ζ'��j�$�]�XB��o�Lyw��!��ܹ�ힱ��DRi/��歟��O�±�E�*��+g�E1<:u�t}Blh$�7��gz�ε��c��nlv-$q��F��w��Rp�?�ȅ_h��TaJc��\�kG,�3,2nxԨ�#�4C��nKg{��o�U��k�h ��3c&�ec� +�'r��s�:׬��'�J��x�}�w0<��w��ʎ_�Œ��%��p,m��( X�%�I��J��8��5��.��>�{pL�(_oEHP8?��6w�ٞQw��p�{��R��j�X�� (�� +�ںn'g�6� ��<ɳ��[��/�fb��y�p��Nz��*\��ųov +I��2�/��n�mgro�Q�Q2$SxgQd�*d��ԍ��/��uP8?��?�=�I��kH��RW��1� +�޽l�s�SE\�נp~>*��ys��ҺvƂ�j$�s�o�h/_�pd�0��O�p~�+�G�� '��7�vՈ7L��c � ��t��P��L(Lq��X� +nί�Ў_Uu�v_mYycu��\&��T��7J�J�t��W�,xM��ڨ�k��&c��2׫\��~˚�� +_��w��Ś�ֆ7�8�k��S��B�<ǹ�C�W|hK��Q�fO��T��`�o\<�z��|�FkG�Վ6��f9��$ %�ʤ� �&J1xD�6|��OF�"��ȷ��~�m�fh��%$��UrE�20НIl`��٠p@6(� +d��٠p@6(� +d��٠p@6(� +d��٠p@6(� +d��٠p@6(� +d��٠p@6(� +d��٠p@6(� +d��٠p@6(� +d��٠p@6(� +d��٠p@6(� ��L�X +endstream +endobj +136 0 obj +<> +endobj +109 0 obj +<> +endobj +137 0 obj +<> +endobj +139 0 obj +<> +endobj +140 0 obj +<> +endobj +141 0 obj +<> +endobj +142 0 obj +<> +endobj +143 0 obj +<> +endobj +144 0 obj +<> +endobj +145 0 obj +<> +endobj +152 0 obj +<> +endobj +151 0 obj +<> +endobj +150 0 obj +<> +endobj +149 0 obj +<> +endobj +148 0 obj +<> +endobj +147 0 obj +<> +endobj +146 0 obj +<> +endobj +131 0 obj +[153 0 R 154 0 R] +endobj +132 0 obj +<>stream +H�ԗYo��)�T��Y��(I�"��ȃ\��b#�I��Yf�h�o�A��Y~��9��_��Y5�A5�O&��Cu2��I�2��XxJ�`��K�XnV'�S��e�z��j}z�'-!x��`��Ю�{: ~8f��c��y�`�>��_?]�wO��/��T�5��%,6�A�hF-��t��K��lf��[�.��C�|��dzB&�`,<�CAK�W_<��ơ�*�O�d�;֟7��dûsڥ�=��ҽ�ћ��u]��y��#��nr�&��3��-ʙ��'��*�4�o`�q[�#�zu�o ��"��2��Y�{^�L\�v��c��l�H��LO�ūE��l1�Q�e'��9��C/�D�>t?ƤH_k2��U�=�r+B/��(*�ꓷ^͋Y�͊��N�J�ڧ7� +M�;>A�rh��l�g��^8mf��? +�o&� ΐHx� ��D��Cp�8g�(��l�lWx�צ�V�"��ɬ��Kb�E��b[+;Q�e��e�� 9�Ӷ"qi!|!��ዸ�k��p�m�q=��"�a��"��h��`��X��$V�� Php3��xw��V"��!�!��: +�t�O݂U�|�Z��K!�Äm��`G F�6��g��"��o�;��r�x��#�~��B��>̤��6ww��?��z�(��p��.�~� +��"¿�^�Au��ȱN��{�更|��w��e�vla �b�a�=��v��s5�d��DP�|�_��Ҋ��HmK��C�tL[��((-۴$֥�%7Ds��v�� I��Q�}M��o�]��Zn�J�vx��y��I�T�W�R��\�a֩m�%�֔�x��ŷB�պȫ쎬��~4�8�-��Ѭ�L��Rs�_Rw�ӟ��>��J�b-��R&��hi��E@�$�{!y�:�|j�� Ӵ�-��}p��I��å��ď٬�� +��ؕ�ڇ൨ٙ� �8K�ni�>��n�hq�_� �H�L�>i<��̊|��ezO+�-�JQ2��T��<V�8�S�n��a��9wLL�b3�ʓ��xL��n�o��G�XL$�mI5�#n.��>�ج`p��n�;N��1�:~M��c=Už��6�8{��Q�[?��h#�?��S+��@�E��|)� +��̉�� j�l9�sQ� �Bv��A��[o2��U=�E�gޙ��=��F2÷�{]��]� )�y��m�K�rר�DNz�`kA&�)1?�#�G3��'u��G(�:��Cs�yi1��e3 &..��4��0��MY��v��*�Z� �n��ûm� �|U��;\R_�}V�.�� tx�qԕ�v�y��ۓ[��CT�p"U( 5j��Q�߲/5lsu�9��!�4R�O_�ݘN7�{o� ��U�Z�I��f&GD�"y��*Oo��o��kAM~�_M��ޜ�1uX��=<]��=��b��w+��^,խ�p;�Hv�dV��g1ցL��'�@��XO�(�G�&h�4�@��$q�ay�sۍt�)6G0��'��!��+��"�fj]>�t�V��P��igB�'�r�mD՗i�FQ`G��R� ��e\�9bj� ��$�$Ѩ�>@�Nb5�}�0�t,�!�2��x㸃��!��-��%��Y�M�*�/?Q��Y��)ሲ� + ��s�l�E�х�D�G&u�d�� +M�|6+ޞ�ɚ�� +8��:��m�f�۹K��bhI�LV�-X�r�5S��֜�i>��5��E�wQ ɫ��ljd(��L��C�]D�&,|�Y>F|��mI��2MJ��IY8M�M��z�+oϡ�@���Et:&t~�>x@�l/�$��C��hpv��1�tО��.��~� S:�q̅nϱ�a��ǲ3|`��/P�J�m�̴�6�Kn�=�`G��6&��f�fAJ�ƥy4K�$�_��6�<}0%��&%�I�j��q-ƾ��}2s��_)Dܳ7k"�5 ��"��WD~ZU��u�Sg��<��K�,�%�K0bVl��8}j�CZ.�Eެ�k�\B��m� \��v՟5Rf�(a��j�/��P�8��J��Y�Η�iu�h��2}� +6Z��|3�y;��t�W�l=�2�- +D�<g��"�x��W>Д�0�K^4��V&��Dyp]�� 3�h~.6�OIɪjC]�l��F.L��{�[U��9�3�S�?e��Ϭ#�D�o��G�W��.0b>9ZmJ�+��d��Q�l��;SY� ��8�|)�A�t]�D��_/� ��\X�1'`��N�7��~4��8�<V��Lr"i��(w7��Xr�#��[S2u��-B��f�'S��d:m��f�5e�!�_�D�C�� (��ǉ��X�D$͸� E6�r;�$��A�Z�4A3hr�n��`:��E�H��I�h��;h"��9��(N�F5�B��M#�D�om��łF�Q%�ݹ*M�ҹ�i�R��/��.]�r�L_-��"��mW��~L[�J��"�t�M7{�:��I�pO�Q<]fU6�:��Ξ�o� +l��m6�Ǩ�g��R1?]钪�s�h]7YQfd#�R,~zWN��8�oZ[}��)� 6�֜L4��Ρ�Hcེ^��W�B<^��\O�B�h�D]o�-�P �O��h��cVz�Q��.k��H��]e�:�0$A��-$X�~�p�֗~&X��;�Y%�/��!�/ȩ��NP�%L+ ��HK��+¼AX0�@�E��=� +�(y��e��u]��ؖ�'�0ғ�B@|Lf��k͸�Fս�+�0�sLz{S˼�:��z��AF�E'�t�$�k��悜��CQ��bx�i��i`�ہ�0��,��᳃�9r��[>��v�x�Q��2Է��T4ADb�2��v��g"��\e�b��{)��w(ٲh��e*h��+��"�o�]��I?AcԞ��%�S��m��ʰ��Xdq�޼iVg�!`�$ ~�2�N�@�9�}�֋��J�P�G]nY��>n��Ǽz,�/��]�ı�П88﯂g��{��%8��疞W�y�c-,��g�1F5��!�C ��c��Ȇ��d��K�"V�e�/d�E��l׫#O��Ʃ��y�p�DmJ��j��9%s�@/sRW;��|ۀ�XoG'�zK��cݞ~K&��[�曎\�\��P��{y9�)-o��(��I +�2#T�u9ЪG�=�_��we�We�`�T�؛K�`�@��`��" #c�$�`�i��!�=��[��ө9�K�R��[�yG��v]5$��Nq�Cߡ僦��ؔZ��P{�$d˲!c�׫�$m�W�ʐ@ǃ_��EwkB|��TR=I��'U��vEl�7�5qn��.��A�v��:�>oy0)IhX�D�i�`@L�fܬ�� {�G ��pK��&�G�yAb��`�A�Єd�)�ﲨs��N��.��I��G3đ,�T�#��-�{��G�ͱ�,��4�s��mH[��x6ŭ!'�I��s�_},��$$�t�}..�^�kW6��(v`��^B�c�0��Dv�)��ņ��M��v�x�ב8c�q(��aĆb /2�w��L'ތ'B��p��R WP2��!��"�W��ix�,�ȥ��QL2*5|-��|�M6� +��z]N3*8�8;,��z4�a͆�׋�}�%��捕��˯$F�Ub�3��{p�<[Q=<��7qL��6��m�>��tx�#�� 3݁�i�E��7=~�u�&�Ij��w�m@�ý�h s��K��\�l +�}��N��'�5�' +endstream +endobj +58 0 obj +<> +endobj +27 0 obj +<> +endobj +71 0 obj +<> +endobj +72 0 obj +<> +endobj +60 0 obj +<> +endobj +62 0 obj +<> +endobj +61 0 obj +<>stream +H�T�Mo�0�� +;��S$��M�aZ��tH#D��$�ې O�ؼ��^�3�W;��ZY�Ƌ�'<��T/�e�r��i�a��:bot8�� +��֏�:��b�^�auL�?(p��bhP�E�}�Y��x5��'K��V�3B��j.@��EUP�:�)l2�8mb�@� �ٖ8/<�B� �qrr���3-�<0o"�e�Z��Z��,�!q��ȃ4�@�2��J�Q��B�=��P�; �\�Zʇ��7��b-��O�[��5�~3祻�o1�� + +endstream +endobj +63 0 obj +<>stream +H��UP��G�-�w!`j0A��*�"��> ��X�Ø5#SL1� +((�)��Q�"��Ԩ`la�mƞ�,i��8��iggg�{�=�|�w�9W#99H�fdDHȼ�p��ً�}B��ƪl��qN��@Gm7��8仓�FzI�FH�$i�$M�$oI�,IS$��A�.I!Z�M"�i�FzU$��$E��v��2��M��a�C�C��Sǯ��:u�Y�H�H'饁�HO;�tr69�t��v��O�Ͱ%��!v�+�u�k��:��5��6��,��آ��V�!l��{�q�z��1��Ŗ3h!�)h!�� +VS��|*��J��Q�2�2- +R��;_e�=]ޤ��,��)Fa,��WhK��XP9�� *�b��[U�0�1 )�I��p�Y�ͪ�6^ �dɛ�i�g��;̰p�o�9�]��p��:Ǩ��.�6z~O�,-��h�/lEƵ`T��p��Nq�`1�=_@�_��S�#�k�c-V�l2�Ј�Qb�e�<:�d7��p�A�i��ݼK5YuS�hU��Zf+&t�ՂOlp��녲�ba� +$W�f +��[��ݘ�P��!x-}84��\�ћ��r7�(�X~��$Sg浘�ƷB��cH�u>�� @�$ 6k��P��VO��5M�8<�\��qmi��Y��-)�3�� +�F�d3c[n��/m��y�j�.v��C�ծ��/��m��k0,��fu��G��%��D|��zV ��(D�f��P#�{��8J=�Ͷ3D~�6eR��f~�� S��ȏ�=� ϟ(Sh�l��~��.�q�8]r8��u��+��}g��ZB��w꿇�MQJ5]u��Jig��5\v̓!�X��5��ڥf�)T6��b�Z��=Qѡ�z��~ώ}VV}��g�*y�a1�{��k��}�ܔ��1�vnv8�fUE� [�3��$[IO��oH�{�RHSJ�xM8� ��:J�)`�Ͼ��<�;�s��ĮX�2��'vu��rM�ɫ��j�UU~k��MCm��w��%I�pi�l.B�VՅ�HZ~��H��{�� Wڹ��`� +��/&��G��Sё�Xr�N9�B�5�V�a0�ML��b@��n��z��e_��<��'b��> �'�rzѕ�,L_�F��Ld\�)N�-E��n-PR�i)��Ż�Y�'��R�o�ao*�7�q�H@�&��0=��TC(��ޅ8��(��0y}��\��G��<�g��&/x��ĭ�v;M��?�Y�Q��z +��2�G�E�^��ZY��/�'�!A�s�� `y�f+��K��.��{�d�P� +��ܯ`z�/�]�%��k�}��f=-7j D3<�ZL�A�H�G��o�UW�y�-b��,�`�S�Z�7S9P�z*��cX`��"�^TcƱ��e�Um��&�ju�*�"�v_!��B"w\\��}q��M7�kůj[��Բ��2��P�B��E�o��~t/D�nL�� +'��|\�D�!�}�_�?��q\��!�Ih"AE�D +�6�w�x��\?N(��r�]�!(F��o��"棿�,�K��t�b> +endobj +65 0 obj +<> +endobj +68 0 obj +<>stream +H�T�Mo�0��>v�!!��J��mR��N��F�=��NX�!A޼~l[nw�;�M _�`�8A۹&�8��E8�s�ih:;ͻ��}�AR��2N��\;�1B�Qp��w�Y>fKu�%4:w��!{� c��{t(�*h�r�T��G��C��G�q��}m1��`��Q�k��D�2��H�R��IoYc��3]��RF��M"K��&_E�4L��d0q��^�h�f��H5H�f�HĚ��"e;��h��ӑ��<�k�9jn�[l��sx�>stream +H�tU{PT��+{��E��]�X��QkXQPJS�d$�hlր�U+f�qb|`u@� ��HUu�H�ѨM�q�ww�e�Y�4��t�?�|��~�;*N=�S�T��NO*^��a�u�t��]��W�.ۿȀ�ϓ`��v��͓j��r*��87��<8ΓpS�8n�%{pKx.��9�[gP�qF.��:%y,��Q�oWqɬN�}ĵq��'�+�lU��ϪG&Mșp�I��S��U��:Q]�^��A��BRHv�� /y��7�m�S�`�w��h\�\�]z\�z��v��6�-��Onܮ��.�g�W�oto�?�+^𦸪�b)�-+:�w멣m�(?�S��Y{��b�{�+k��5,�:�5�8*ツPБ+'z/k�� +��DT/ڿ�H�TXN�+Rc��4��_;~�O�`� �&��b�~̿�i��M �@�C+��|�m�؋>��-�kД�i$��~ȇ8-DW@� �$t!�*z��A�#�L1��8/��=O��f��h��Z(j! +��P��2z�_�`vB��@�ۭ�=��l��`[��yG4�� Dr�n�08iA�}��Ǣ7HA/ln�ѯ�;b~+�pJN��@�m�MY��ƃ��:��ei+W��Fҿ�cA�6Ӓ��w�� p��)m�ٯ[��M�0jP��ZI�VV�VQمZ��I*�ǩ �Ի��c��(�b(5� +�M�R��C|�ݗ|A>��n��9nNμ��=��r�L�_�]�P�`"�A4�M��,�l�<bD�|$cI�%c�"K��H�e�^1��()R�>�^l�|>�n��뿦1t'��(S��K �>��Ւ�eeK��z�:��:��S.�nu��o>�B�/0��N��t��e�_�H�[��}ӔQ��d�� 9)�p��S�7�� _�V0��_��+��.��Ր�f�q��m� R��v0�>`�Jp�L��PC�>z��7�r��*�z +�F��0\�h��򙶽}�eE_ 2�1��s�X֘�uTءXm��Y�g��c��b��rٺ�K�&>E��4N{��h��۾Uˎ�B�y �P(�P��a�`��:�y�a��!oc�\M_%s�/v�X��e��R�Sn�]�;p�]>q��w@+��|q\�%��F +��@E� +��`*��$��YL�`��{��q��b��)�DO��ba�8{%9K!A�t�bY��F��ٿ /mM�ԭ[��O��nۦ"�F$2TjЀ��@�&��-=��U�m�9��1YL��=�`(��IX��wD�`|p"!!�[�)c�8L6w�~z��А��**��(\��9��Z� +�� B�@ ��+� 6�;��W��k��T�Rĳƿ�ס�1=��z�OE��7j�h��]��}�C)�H"�og��%L�a;v瀗��3�G�K�o �� +c��Eos�_*�vA��iwP�B(��f[��2��y�F��2zn�)O$%�y��A�cl�pڶ��L�oط�-T`��6�5�G�_�s�)�^7~�ak"h�h��0Bi!�C�-�C��S��@es��M�{��'`-c�c/�B<M��P�&R��`�c._�L�DD��v��.ա_,�l�O��~ဳ�-�mrc!H�M�o��~�Ʃy��&�,�vi��t#-:xnM��l�F��%e��<4��$�^��R�bg� ��ZC��p�s7��ĉ�MsI�,2�'KfdOB%[�G��jj�E��FRHI��>z�X��i�&F!�s�Y�C��lǢs-��3��{�;�g�B��N8U�Ō��6yIB�s��ޗ͵�v�RZm<��~�P̮��Yz�E��t��w��rf��K)s(��8��^A�R؞k�H�#�u��J��ntɝu �R*� �5 �A��0�5�ꮬ8�W��X��9�#�lB_g&��m��"U}�7[�X�3�gd4,��?�t�j� ��6�ֻ�G��R��_�uB��x�QzL��DiQ�"Jq7=��~�C�x��E�� +��6��V�R�1.\�J��TJI��䤆��ֆ��멍I�_��|��vn� b(+Ax^ r�X�0.��\�WN4_��l��?%@s�xRjQ��bǑ�g��S\F��k�􏉵[v)�4�� r�ű=S��F_T��I>��x^;F�`��9h�� >��0 ��*N�v��\��呛&�w� ��c��e�f�N�}�t,W�a�P�o��r��-�i�-Q�l]}:^��6z��v{�Cmw��H߶�|m��v��c�'��7��U��j��+�5��u�b+�j�"`1�q�u7�<� k�xE�D�]0��B��tF�͋��ʻߥ�A=L�H��"�� <�z��_�>��;��r9�Ȃ� ��K�f�.�!��)�rF^��$�� 9\��y0yD�L�T"+�ϭx<��c8 ��N�p-E�W�VF�Ie�܂{��?^�e�6 �aJ�� + +endstream +endobj +43 0 obj +<> +endobj +31 0 obj +<> +endobj +44 0 obj +<>stream +H�TP�j�0��+tl��I{�%FG!�uc�vwm93,�Q�C�~��u�@��$��SG.�|e�{�`��Ϭ�88�z��v��Q��/Sı#�i�|K�y��c��aWmA��Av4��R��A?��#R� +�Z!��*�Ո ��W��FV4 4�� +��0�V*�ͪ:�Z�v�if��>�̜,�닅,�� +>d��[��ln + +endstream +endobj +32 0 obj +<>stream +GBBJGP+AdvP2A83��0�y�|��$ AdvP2A83 G��r?0�?e\qc�moA��~q�� [\�|]�\�_�!' + +endstream +endobj +70 0 obj +<> +endobj +64 0 obj +<> +endobj +59 0 obj +<>stream +H�T�Mo�0��>2qH�R(RTi!q؇��UZ�(-��c[�4O��ױ�f��v��{}��C�጗�B.��z�N�;�S��6��mӃ��s� f��<{�� ^`v�Od8\��5l�<+��:�'��u�9��TFoppJ�W�� Ų��j@k��:e��>stream +H�tUPT��]{A��*�Jx� �<�y�@!>�%��"��3��V+ƉcꃇQЪ5V�(P@E\-�VE �ю��n�b��>&��ܽs�s��|߷*��#*��5)>>91�߰�t�!�j��ǖ�n�\!pԎ��v�m;��KTDC��q��'$��B��$e��pD��G&��%��h籀,"�d ��%١ڭ�'dF�d39J�F��j��C�~��Uٝ��Ϸ��9�.u��R}��d0��/��9ce��Yl6�){�}��|�i�zi��9ڏ��^��C�C��a�A� ��96:�u��)�)�i�S��s��_:��M��m�ֵ�̜�%q��Jk�pR��0��L�I�L��ˇ��_[Ǭ�-�mΝ��U�Y�yET[+xв(B� ��)�8�2g{/?��h �7{܌��N�/-^��Ҳ�tn��_�,qa��aN0�f��̄1% ~��cK�l��;p�A�B?�J0r&��z3�^�=�&��"�2��7ŧ��vX�\��c�>��T��rMM��=Q2, +o }=�F��|��S�xȼ�AIPN�e%a9a�u,�L�2_ b#.7.sQD1r_�l�e ��#�� +��d��r?7� .1�N�tI��G�W�d6I�l��;�o��S��/.̓�u"�l4�Ԃ��yu��nL=�[.��pA��>��Yy�pL�<^��+��[茮Q)�Z& �W*�:�%޼��\��9ed��5wd�h�n��wYLpǕ,,�Nf;�)U�fl֠��cOm��Z|z/b�_d�or̹�91��vfhѐA� +��O��W�p��e7��-�dNތX=A� ��jP��Рk��O��n�l�bp��x>�`zx��ͻ��I᱉�r +EcF��Y��T2֓a�1��=+2֘b0��w��ɸ�� d*��<�;�=�5R��[s\��8��!�}�]��`�R�w��R�f�L�j3��t�9��,��g̙��+7�DtG/TSZI�<�jU]��[110�舮s�i�):2bF��S�S��ockh��_l��M�E��?��a�.�ƺ�޾q0��J�X��"��kn��l��r|�Ҝ/��5P7�?�Q۩��zyG�Kj?=��\��dӁ[��%��Z��}�W`4]%��q�p� 9��)�K�8M�X��o��p�Qff��T�S>�L��S^�3/X�ZjGj��h3�Z�B,!|5��T�$�,D^�Ʃ[��@��x� ��l@8 ,m��f=�+o�\8t��Y�p�D�̲�,K�cs��0��*�N�\p�w_}��VDj*��"��7�F�E) r:�!��V0��~��F��0�(�O�b\�z#f�+Y��L4c'4B3�C��:��)�6YYU#� +��sXNg֣|��a��A�_��:��0 B �P�D>Yh�(_�`��ږPDo2�>��mS��Gӗ�Gh�BQ2y�@�׀�z�R��5n�p꡶ ��y�ݺN�ݐ��[*eG2��!�|�WV�ъ�@��b��84�K�a�2a�A��g�2j#^JUP�RIM呵�y�b �g�'�� +��@O��*�ǌ�*�zQc�P�y�(!6+)��K��H�(XEW"�Z �8��#�մm�D��d��(��Z=�z� �u��J=K2��&��!�u;��"�j��M<:��v�y��/ӝ��ۥ+)Ά&�K�ͬ�Nߜ-�,cQhz��%�Ãj�i�`ʿ1�A�O3nK��t��hb�e�NTo@~� ��R�x�'w��'T1ȍ��ܴ��R|e�^�JO��S��m��n�7!��<�Rr�E�z�� M��u�޻hj�2��@�&�J�.J�le5�hT� +��e~$(�(~��񱦃��T�si:k��i��wϳ�}��<�[5�BH/O�9��mn��K�(:7x��ԭ��h�d� +p�)��8�n��0f+�9MK!�fQ�H��c�Bԙ�O"t�B�@��0_;A��F�(Mf�2Tv�v�Z��H��h��Ѯ��]7�K��N��t6�y�pH�DP!��AT^�1@��]��=˶P�I��ܪ~vVj+=v�]�P��XJ�d��䑶o��i��#m�"U,\��~��>b��3_T��~�Bxw��9E��`g��x1+��bb8��/Òz ϫ�S� ��C��;0��9 +V��K��'X�W;�"��x� �&�4��<��X[��_{�ᱫ2��(c��7i��ԐI�0��F��&��p�ˏ �~&]��`�_�b��'ӗ+�1=�Jn��=.Y�r?|%��ӏc^��_��rl�+5��U�9k�j L��Gr�c�-��m�0�`>��a0��w~�yƂ�yAGo��K}�p|��?��6 s�x�:e �zШ�e��v+ޫ®��a[Q��r�GHGs3j,r��8�9��t��U�Nn��u��W(T�l��s��=�Z�.�ծ�VWT��{C�U��W�f~)��,��z}3 ɄI/�6��$a��"�cȢ*�ݠW`� �Ą��fsak�e��KyS��m&S�Y�L/A�=��ݴ@3m-(��ޣ�i�(N�+$DW�VB��a�A�`��h�`�%$�|��S�# J#��;��u^��<�f8"��x�D��\/hKba7�vM/�<ipw?�yB��bC��t�T��V��!% ��iT +��Y�A)a��P��9=ϔ��7 + +endstream +endobj +101 0 obj +<> +endobj +133 0 obj +<> +endobj +73 0 obj +[/Separation/Black 74 0 R 102 0 R] +endobj +74 0 obj +[/CalRGB<>] +endobj +102 0 obj +<>stream +H��Ŀ�þ�½��~��}�}~�|}�{}z|z{~y{}xz|wy|wx{vxzuwyuvytuxsuwrtvrsuqruprtoqsoprnormnqlnplmoklnjknijmhjlhikghjfgjegiefhdegcdfbcfbceabd`ac_`b^_a^^a]^`\]_[\^Z[]YZ\YY[XXZWXZVWYUVXTUWSTVSSURRTQQSPPROOQNNPMNOMMNLLNKKMJJLIIKHHJGGIGFHFEGEDFDDECCDBBCA@B@?A?>@>=>=<=<:<;9:9898787576455344233122/01./0-..,--++,)*+()*'()&')%&u�� + +endstream +endobj +153 0 obj +<> +endobj +154 0 obj +<> +endobj +156 0 obj +<> +endobj +155 0 obj +<> +endobj +128 0 obj +[157 0 R 158 0 R] +endobj +129 0 obj +<>stream +H�lWے��|߯�#��`0DO��J�*�)/��RXpv \��|}N�@Rf��4��\��qw�Ç�8��(��j�~��|>w?��2�{��,�m˿y�ml]�i�M��UǻmL�7�i�nm�ه��}�L�u�f<��G�>��&ܶ�6��l�F��C;��8ȔO�y�9�_��o�hs��z8��{��]�'�ĥ6��$�pI;r]b`S��0.��>�j��-�Z\R��e��&��U�tϯo�Sʖ�'�L��(<��㴯ð}��|g�y;F�nM�"�ZS$Qn�6��"�[`-B��u&��8Y�[c�a�1�>9Q��(��m ��?��} �CM��A��;uc4 �N=�cY��{�]�s��ӱ<��wI懿��9�}��q�a�?�Mَ��N��tV��@��N��?L��_�>I:9/A�=Xl U�vǺ�Ju6�:��X��\mbbH=`��n�q:�� Pص�+>R�p�0�A�� C��$��SU�A�)?z�� (�8ە�U��๯G��&cf��%��1��a��{#b/�� IkPC�z��0�aL��w�5ˡ��#�c�m]��K�<+�i�sp�c��`��H��E� ��P�w�b6��z�|1��VG7�s�*�]`��$��rڹ,�B B�8+��J��껒��(2��sbZ��J�f�Z��\�&JP�)"��,�s�~lq[s��,[qc@h�[A -9Ȍ��;�s��J��c�k��VМͩwZ��uJ�ȮQ&҃�{m��ӡ��8�� H�/� 1LOO5U�7�C�%�'9\�Aj�Q��z�B�𒤌ҟ�o�ά!m�:�F.�9�%��𴂛�I#�l��Rk��~>�<�}/�j�oI�Y[U��xj��0#��J�W��p`�w��z�p k�(q�k�LB��V��v��1^�㍠t/VAu(�!��X!��/z�2%��-�<��d�]��-�硇ꧺ|d1a<m8ޜ��LEJM�{&R`��8W��X�qvC��n�:�$�N}�V�I2~#O�/� �2�Ȇ;K^��q�-�t7�yL��D� Ķ�!#�B�FnJ�M��zK�LU6KĞ�I�&g=��#�l��8"Mu]��9R��[��кј�i��܎z��n�8�\�gR_^�[5F�#�hn;[59�B�< ��Qw��\��]��˒� �!2��~��4�6W�a�T��A��6��4�˱_��]�:il��SL.O�r?5#��E�c�*��N�� +��X�>�\eg��J^u��-'�/|�)_�ƛmrn[~��U��ჾ��x +�2�v�Ŷ�k�yf�v·S��0�>��ٹD��Kn6Ќ��<��J��Ƴ�b�3��9C��[ߞ��Er�;s�\��GZB@%��n��fǊ$kg8��Pc7��N5�'�I#C�sD-k��VM�� +i`�k��?_��5�R5��`vmjB Xzj�<`��Z�*ٰo�3#�NN�#qy��`�!ߞ��C�`��V�"4`=[}Щv��r��>��n;!�,x��w�j��=��푽��$Ctt� +�T��a�\�h5Ա�u4"�;�z��78C'@{�]b��}�cy��V�^Cf�'��n� +=,d�kbCv�:�f�;��f�k�:0)?);�4��,}��ݻt�>ΆD�P&��'b��u0i��弖� +n'��ۭ��p��e��:��K��A`�q��Y|��{3�l2�D".�Ц@g�L��c2:�I�t�2c�~y:�� Vb��V��>ãj��O�g��l�-(�C.c�c#ID]A�n;�#1�?ǃ�TjZ�3K��-�]��n5�8��Cܪ�%{�*N�2=�KC7��r%]D\c�B|i��bq��-�Wo��؁9�q�j�/Cyc��bS#�9�HÁ��L/�\e��e��M_�xjK�� +[��g,��Y��)�jW�<�?�}��?ֽ��Qw��)7=��H(�?�6lJ��#�s��Ȩ��o+ �of:;U�wZ��!�ɰ�G��:��Qm ��!2��yL �WXȾ�bmf�[&��1a�6��ˌ�Ŕ�* =�y��dr��4�O~|�h�*me��ֿ��m��x2� �]p{�S��Z�Y�̌�M(!��F�c�$1� $j01fH:��d�>ӈ�_nj�E�d�-�'�n��0�a�-S}�{�Ħ�,ٜfcW�B�ZyS��^�(�,�Q�Ȭ��̹(�E��'��kXp烑,��2'[�v��O +5,�{�_��]�~�`��,M\��Df� ��ږ��Ѷ�5��]~@�ͮ�Tf��gag�\��.�O_��'��w-�@�Ša��B�M}O�OFLZ��`&B��q!� ��\�%m��~�+r�oV�Sw�xRC?��0�g��LQ(<��Bd�\}�;�Ν��d�=�'��ۼ�Z��ĒBx�y�I�~�7��&�&ƮHܒ��a:KąF3&��)��\7b��H'�D �%��x�e9b +��8m#��V�m0�4��h#�_�F��Ș�e�b��\V��'f��+`x;o��` Q +'ʹ)��3��h-��gN6�#�9y��uW7j��ԃti��Tr V뮗8��-�a�)&�B�֡�F^��^W�= n4�aK@��fK�icrRfH>��R��1�"'졵m��%��3��сes9��i��.We��ŀ�d��7��>�u��YtlJ��P��"r��j��B2A+'��.��.�$��l�e+��x0�}�\��!�N^.i<��b��4�׭��v�U�l�9]�X�4��<'/{�v��q��khn��|� +��%=,�*��$�"i��m��;Z�$�#]'�z�a��,�ԟ+Al��w�F��+عbJ]0�+��P�_}3�1'��;�-y��V�`¥C49�"�$ ��՟��?�?]�v#�,�Ea9Y |��p9�$��7�_��R]L�f]��O�ou�Lq1��#G�܀W<�Ӱ��_�[��/��RčT�r��,mq��2�W�Hw��L�R>�;oʻ�XEA��(N01�z��,�>#�^>�E�:��#v��61\?\:�e�3'��&D]ة�v˟� ��_5�iX�F��}�Qpk� ��9�|8/��c?�T.�0�rG�u� �@��(��6H�oڝ�a�Pq$ �nI��a��b�,&ގ��@��IQ��[�Jއ-�j�U)��y;A� ]ԝ;��0_aU7�V��n^8ۙXth<�vB�]ۡ��Џ�v�O�{�.�--��v)FR}RXF��d�e��\��l�~`cH4�%��9�sj��O՟+�l�=|��s��,��\eyc��@I +3�v��.|�\�q�; +�CK��h[p��a��lj8�S��6�C{Xt�l��& ��CcIJ�܈��C�6�,A�5[/�g��~Fd��kG��o�.��K��)� JJ&��x��԰��1 �=Ds%��T��L��h0E�hp�DÛ��K|-��"*��OO�)�4td6,;L B��J�i�� Dk�*,%��QBT��~�c��4L�q��km�|��;COtS�� W0C��H��?�S��N6��cT��C��\�h��R�1 ��]dJ,ˮS�ź��^=<<��b C3��I�p4�h��Y��R7]�~�֣l��ݬlw +d��U5�^�s��S)Ī�k�i|HB��Ot��j�g��gC��%?=Ǚ�l�?櫵�q��|*�¦��AH��gw��v�"K#ib>>�u}�cfD�բ-�~0,��>�=�ܳ8��ϵ.wJ��!�b��Vuy� ��B��Bf�i�q�.|�&-3� �0R��C�mk��3��8�m%q��[�;�X˷��"�Z�Z�4��#�=F��H' ��#tgl�Br�Ϻ�zu�H}-�ְ(��g�lGBS�0�(��5x�8?0g��(�(��c"W��|��.xz�^KݴW�GD#x6�4 ��Ӌ z��^�dk�3�S�;L$*L��{�S��q^HE��pܜ&)��$�}�͑uӵ��m�⤣�#U�� +-_d�8��I�*�%6N�8+�g+��!$��Ӻ%��/@ ��4s.5e`��%�UΎ>��Aql�&�gd�� 7n�#CfQ0��l�嶮��C��6=�l[M��P��\�?NFb�@�]�v��(��+�V��ݱ��5�:9�Ǝg/�?)�'|��D�sY,�BZ��Qr��$^�*6�뛊��ʀ�br�mY�k��3��p�؅��{�~x�8pz9!.?�8��ۺ(�U��|eǇk�Ϩ�D��1��(��u�)��f�U��I7W�0�� 4@��#YxF��|j톽mj=(�� ;'��;� _)��$�H�uc +z�xƶo;i�WW��S�G+��"�>��S�h��c� ��_Y-u[S�@�3��]�T+p��|A�C��X�D�q�X-��^�;�*�'Ɉu�Y8�y�d��h�<��Cy��R�[�n{��`~§�?_9՜�͓r�~w�tq>�B�G�(�!RY�| ��!��bp +z�KS]b�<��M2��l-�T��^��h]�ՍV-'��|ȶ~�%J�_ђ�1��,�۝�I�dP��F}t{\�%��br��2vq� ��F1>k�Cd��(�U�/q��~?��nh��R��H�2k�{YѶw�7��j&9�ۑe��֮t1ɥ�.��آ�)�:,��M�S>��uY��b9��&� ��7��8��m!>�nd�^�j�H+�y��y�eƍ9��4&�7�ѻ�5~�;�Xi�Ȭr�!�!��B��b��ՍZv��-u�z�1��U��EA��+wE�� 9��'RZd&?�{��EF��+��Q�8 ~�@��d�0;�)�E��t�MF@9X�rb�G��~��2;�w��m��5ld_1 +�aە�w��rI��۶�UA@��`�K� ��S��ſ��E�?��G-WB2p��I<��"�y8�v-�0l�@��]:��ݼiPoĔ��#��C'yd�@�� l|�5�Kz<�)be��p��>Jf��;ם��ڭe^�;�8M��A��Xŋsk��mW�9Ѐ$7��#�S�M�� ^��|� �ª��B-��a1��"4S`Y��a�[��s��^ɰ��;!M�Z.�#��Z�̀&�H6�T�+w�-�}��~��G��--�lM�^y�f6j��d��ď��$�L,��*S�Y 9�G��u\�Qg��$H�Qv;�F�,]�]�#��I%�:��X㺟7�g��{�c�'�CdD'9�,NF�B9�a[��7��[Y��r',��t��de =��x�(��nTV�%�%)��_p/)��`Ǎ�X�d�:��=�@O�x&L�Ĩ�fDm�$�F��1�,w�)�`�p��r�{%��:i:Z7��&K��Ɋ�#��d�c�7�� #�[�˲y$��`ui8@&�&�84� p2�� ϖ�"��Y[6 � ZY��X�#@�2��B'!"sR4�M��O �܆�g�Wu�7�w�V�1��N>�m]��ȃ��`XJB�{�S�+j��%�]E�l�9G��DgQ(Ɯ>װ�V��w��N��w[<��v��9��s��B�i[�;��|+.�IS2� M¡hy�`��E~��M�,O�SX.w�Z�J;��v}';]W7�3U-��l��z�!��]�z�N�␪p��F=9�Q��~�i"9��8#��u��{w�g~�r�,H'A��Q�!Xv��U��;�$mT'�B��^��Eێ�m�X6v>?ئ +�|�0 m��!~��E�+�wz��O�?��ŋ��e��n��lx��L��~nH�c�P?�p�*�S��%�^�?�w8d�E�!��m�m#��~ +^Va��m/��N�u�XrM�k�5N��R��C��7��7��RzĹ��G�q�#R��Y��<6�F�*�և��F�qi��XV/� we5m��5�7|�D7��8��{T!S-&=�n�:�(�7v�� Q�M� ��3��\*�$,Bv`4<̭5��e�U:a]|��I�A��)5*�[�5"��z 0�� @��h��l��V~֛t�&��;&��G ��b�t~�uc@�9I��mV�imYR䪺(~�Vig?0:e󥢳H�v��yEe��&��ĸ�� ,1p�ݳ�7r�p��K�?��j ��8Ϫ%��I#��)�i��?k�LF��=�|�RHaZ��𻏍��VCk~�>?� +ڷ6}�)R�ESVh�%��~�Y��u��j� ӧ�$k�I|��*޴��ʦ�#߲ �w�':ۯךN�?�uo�ˎ�vm/��(vDdt��ʨYC(]V�lMͳ��U��8�T��9ܖ�H��I�&��W�pB��>� (t�Po��똟��jm��Έ��O��ݤ��\�ML�)�D�Y��?�@��Y�i�ԫ�Zo�h�>��-�}¤��vb[��0|�m��c`1�4-��;0��0_�~7��Q�N�NS� �.@��i/�*��ȓ"�ۮM�!(Ce��\�s �>�8��-��7�*�F�,��O��GKye��U��쳌��NL�u��!��%�ej��q�/�|��n9��TJ }s�3DFdr��$��X�*�,ʬ�� +��տ0A��s�&/��_�/��Om��d�L҆q ��<��4%IH]RJ�M�G�B�s��b��8W!�F�,XB\�57F� �k� ��M�gk�:�"4/qĜ?�<{��#"XUD�2�p>3ڭ��nP�F{�eF��gf�M6��<��_s(_��6��J/!(�.op#$O�&�p��!��>�$eRMUf��pN�[^�,/p��S�/��Y��,7�J +N:!��(_�� +c�\We��)APc��{�P�YMq��fo�Ϯ�a�:�*K�~rVo�y|o��T�en�f��b\��@��@�%ܕ_��{�U��!��Pt� +endstream +endobj +78 0 obj +<> +endobj +80 0 obj +<> +endobj +81 0 obj +<> +endobj +77 0 obj +<>stream +H�T�Mo�0��>v�!!@K%��>�v��F�=��/NP�!O^��`� ��'h;k<��k�3^:��i>ŧ��q��`��'��c��b)��z��X�� }c�vM[Ʒ�ʽ�� nA�s6��Ni��^j)�7��SŹ�_ʳ�)Dx�:o#��9r�Dl�Wy`)䊸��e�a��{�Ժ��L�JV$l�P��穷��" +��2\��#r��2�v�y�S�~�m�>`}�>�>�5Ζ��Y�o� ��H7�`�r�� + +endstream +endobj +79 0 obj +<>stream +H�l�iXgǻaz�U�QH#tk7"F"�(��c �<�� +�Q�� :�5��)2� +��!r#�Ơ��1�j\��j4DS�/�<�$��X��T�[?`8��lZ��u��K�a� �,�x�d�c�̐=��q q_&�a��,Nm2)O��k�1��)8f�ac�ì$��ư`�Z,�\0�İHK�01�K�0k�_��;c./"�vY$a:<��|s�ۈ�� ܠ��w{A�`��w"��V�.�+Z$*'}��S��Z0��#ܨ�8ԸU�H|W�r��3�ft�,4�0��۴��l�ٕ��o�U�m+-Zi+�(�άg��gݚ�uv��SYցuf7O��Q�˱�T%D�M��EE)�d*M�F�V�(֚�.gkt�t�B%��e՟�x��:�U^�&��I��_�L?��D��E7ӪT��aQ��܊p!\�i��o;��P�N8�V2�1E�J.�qL��U� a>I�;��ᾴ��xO�V�jz��[�P��zh��ۀ�� ٻ��k��}��`;'NS �1��"D�l'�x��ˡ�?ly�7��Rǵ�sR{��>~�W_�bK(AaT�� lh��V`�"9x��ۋ�NL7�0XXo�� K� +��&Y�!;V��N�U�tp"�Nf��!�q{'��ԁ��Q��e��2>��P�� +�Q� +B:%��d��Bd��H��\N�j�(��m�P��^ŉ��%��2*��m�I��)��)Y�*�L�M� 3h��&Y`�J�6��L#��锤N�݊�4 +W"ô�l�b_�8R|�,��Ks�y�45u�˄E�q�bGbtr +�<�'ݍ �N�8sO�~��M~s��1=�2�H��-lY� mK�7��(��G䂉F*��Y��T5��p��'�Ȕ��z�GtUݔ^�&;�I�-H2A��il�X!k$A.��~�+ ��X��NV��'��y+H�M=�I@R1�� g�\��z�`),+�15��\.W�yK��y3��t�'�2 �R�}�y�� ?Z��cs��RHn��{\J#�m��ڙ�F��o��G��4."��oɺ�s9�tVj�*��O��z��?��FZ��{�5�r^�.�^Cݿ�A�FֈRR|$\��u�C��1�l��ͫ�0�#�nxN!��At��U��|g�� ّ��[��E��q�φ�';هM�c��.��z�7��6��̻�3Wz��5�ׇo�^��Ċˠ�cH�$`4�۽E��;v�{C�zҖ\��FGi]-�$��j �c�ԃ~gD8H��9��81��-�q��ǭ�ǭ��-F�N`��AP��Wn�&��㹕Or��?ʔ�e��,#v��2XR��[��Ȋʓ�W��N�:8�<�ES�E5O�B]YN3z�o�*�L��uh�NZv9T��G��W:�.(��u�f�1� �I-�^��WD��Խ`A+l��} +w+e��e62�<�<��'#�*��r)��B�hB�Ic�E%��2�.i��R�d�釁_G6]G�)N'��z߷�ON}h)��\��E��cdv��e�DV�(��)��}�y�ͪ(Ϩ`��\��?�[)��rX��{��3r¹�� ſ�]�QZXAH�uRf��#��ͧ>Vh*��F��w� +[��Be��`�l�Rʠ ��!��9�`o�޼0~µ�9��/�"�l�u?�y:�KjW��y4|��L�gs2�h�T4��?��uF�\:sMVil�<�I�FkU,| 5�L��00 �S%b��s��{��<F�*��a(��J��ZFkbM�5�C�鏞1i��V�VW��S�/�:�=�7B*'g��!� m��3��{� à�k��F8R�d�� ]�!ʙ8M �P �xZ��^��*��+m�VeUF��I��~�^8j줊$�c +rJ�)QpA`�>%\1tQ��1p��B[��5�C�A�I�B��)�DU*�R6�-t��vy�ǫceu�KWI�;u�3d�V��B; +ٶ|��4�M�Qñ,M�a�w�+�iJ��[��-�� Ξ��)��d��̥�4�p��m��Dc"��xgr=��Yh29��o/��5h��~oyO�.�L��"�R&:Q�2D�*��Ň�8�V ��L�RXU��Mh��l{_ +Cn��_ÆЯ��H<1�?a��1��5��&��p��ʄ�j��3i��R��k�M +F�zك��/a�9��״! ?��E�ͭ�5�K7x�<#�kk��i�sf��~瑔jk�:PWǱn��c�nB��>-��d��}�SN��I��b�Zw~#Q��u��Nt^m?XZ��D�H��}'9�7Y�Z>}�P +�M�*��Y�T�TS%m{�Q ��N|!��GKy�ь�?�R��s-�L�Ԏt�*�H��̖��V�h�]�t��b9|`��o&�N4��5j�؀ v�f� �Aw><�~�<$��"{��y�"v�u��2�Suo��dє�/�4�2�F�$Z�q!z1"��;��p�"a��َ��p��J�x"bu�@��d��0�e�~�=�%� +S��ZrzF��m�xW�J+ǻ1�`􆼮º�br�z��Ph�l��4��}�&�X��J�)�\B��KP�-t��x��p�5�x � ��nD.�l��X�x�O�Y�d��^��W,*&�h�\TÀ�V��̻D�d˴7��@�� i�ύ��71�˕ ��}|!�VɄ�V�~`��h�4ܖd� +ܢ�)�!a��,��j�.)��Iza5- ?�"�D4�^f��F��+��B?R�"!%@_�Q��$�vP��.鈺��UMAe$bt�F�G-�<\M�� ?�V��M�Ș̦��g:!A��d]G;��`��Io��Uܜ(_rEw�G�=NwS|�~O�G_�Q�^7_*pl��&[vX� +<�\ϝ^��l��,O�|\�>�F6:�W5562 >��fl�(#I��U�(��0@�/��h>Q2��j�j��« +I�3�^;��^+c�U��̈�U�_��4 ^ބBx�u��$�`B@Hxyc@� +��s�h�϶��a��s"�Ugٵ�1��η��>�D�x�ň�a�5<`mĳ��]�|ݵ}�ދ X�Y� �h��D}��h�\�>Fs�.}6v�e��^��u�ͩf�6n|yv��ST(��=z�a&Q�y��AYc�.��hC�<&/ߚ�O�'\�h��|�Z��19�Iνj�U�5j�F��'��oH��y;�4��B��Q��U\�)JŒs��5i*Y3�G`Ը?q�铫��t��©��4�eh�HčG�h�e鷸��"��3��XT�ϓp�\��(q�#l"�܄Cc��Is�}Hy��o�֘�{׻F��/��R�O��%'��kcۏ�Է �R�5F�� +�-%�<�DS��>�Ҳ󸈎�g$��tDF�%4&\��ה��`;h��$�P��o9O��W�ʭf��QU��n_��5�7Ѷ��J��~ܛx�y�{@I;�P2[��6Ȍ��P��J�aE�{Хk��_�nf��g��a��oO��p��򆱽N�v�)�掲=6;�Z�"��^٨̣ߴW �v{��8 �z7l��?݅K�̡�n}�)ZB�n �E��f-�#��g��Bo��}a��[�fU� ��?w��E�Ӓ�V��_�3e�r�^ kw��z|�.��M�%$�g�y�� 3'Ӄ�B��j��hR5F��A�n��̑�:rK��R6j@3�q괒\*%?qOiA��A7�� 2?O�j�&��m:8��ů�[y ^U��}��v��f�ٔ$��IT�&�G>��J��|��'n8��z�W[Z��i��j��J�KH��A7�Cb�.�Dw�I�Ȭ�C;��%�\b{��n�\̥��#�jS8I8�҉��[��;#�G@_�s�lU�:e��E�l[!�&)e��s>Xb��R��X�Y�I��p*=�=Ξe/RpZ>иE�ԭ��,0�1��%w�\�� +�+��y��C[�)*|�r +(]c��v��f�υ��w��4^zГ��;�/�=έB��{*�_��D�f +��Ȩ��V}��z9�9^��p��Rlʵ�ǐ�cd�Y8��s~�O��!R��x=��x��$@=o) +��FPe/;b�T��^�Ͷ F��O@�:�B�Y� +�m8�}�}��Mb�_�� +m�>��T[S��R�<��8l�Żue�Ӟ�Y ��` +��ڌ��0j�,\ET�@��"�|�AYcU&�!�sȈ��s6k�du�ZًO��\�Tv�\�O��$��p��-�qi{��*Eh�q��Кc�� +��p��8&%�I'�E!�*m�"P��_�� +��:PL��ɉ�f��d�J��;��#��Y��JFq��@;�I�v��図��0�l�bw�j+��*݉��r�{��!��:^��w��w��Oˤf/��i9�}ҁ*�4�j�Z17��o�ǿ�b�r+��K��q�m߿��b��u⒙�O9��O뷳�o;�j��zoH�9�ձO~�w��ؚ��6��6w邜��t��\�Ӟ|W8~2��>ҾA�"��,�E�k��+V-��).Α��PS��J��L��۔��Z�zvÊ��Ytf�w�怜�gM�>[jvմ��9`O#��GǾ?l w��5�}��^p�wPb�ȏ��@��W + +endstream +endobj +130 0 obj +<> +endobj +157 0 obj +<> +endobj +158 0 obj +<> +endobj +160 0 obj +<> +endobj +159 0 obj +<> +endobj +125 0 obj +[161 0 R 162 0 R] +endobj +126 0 obj +<>stream +H�|Wێ��}��У�d^E�yڝ5b'��1#�FТZ=�(�h2��s�)i�Y{`��dwUW��s��͛�V��`6��q�߼Y��er��ɖI��3_��0� �gi�-�<��;��H�x��mӗ��*�gL��y�c��}�2�e��!��ln6��^��8f(��Џ�/��E��j��EL��r1'�2��x��_�cq��6��'��؃9�mG�뿈��`o��2�WKw6��Q/�ټh�zlʂ��GQL;��߲6{��zWY�U�x��ι��G8�|��W�+�t� II�z+��}?Z��\J*��D*[-�Dxӎ�~%�Ǯݎ�@kW@¥D�n=g2^��y��MD4\��p�jIvߔ��E��U�smSخ��E/�U�^��6�e�gr��V�ϥ��`�9�q�<��dic��*r�}��獱� w̵��6�9�}Y%;��q�c�#�m��`Bc��S��9R`�D��r��[��!��c~�"�]ᾲ��O��7�[�Y=A��l��N��1��<ư�0�G{h��E�q�6{�pm�{cG�K��k��w ��[e��x�g})��uSx��VXH� ��+�F V�v�� +��%b�'�瀀*7ނ�,7�(g��`��|}��i�-�� @ex��A䩬*�hc{�#5�|�k�G��%�X�Wm&""��-4��#l��k��k$�F,��Ո�2U#��c��)�xt��Պ�<��J�!u�0\ˈ��W��6F�uk�� l{��t�l*L��.�K?J��lBR|FR�9b�^�>�0��6��_H��k��ڣ[��6�0�m +�~�!ò%F�NwXD_�v��S��nx\�]P��M��a +�ל}��5�Z��^��H�,��z��ڑ��o�j�j�r0�+�h��ѵOo`�8�ݨ��2 ��ՔF�^6{�m1��&�h/E��m�|z�9��x��M̚��:x��(Jĳ��F��`JGA��j�G2�Dڃ�r*� �8�k�FO�G�:��S�FQ��t�m�4��^�%m!ɢY(2K�:��| ��k��i�&��lB%a(�(�Z˓d��m�Nk.��tk��⠎r$RVI��Y*?�ja�׸��Y%C�<��b�r�� )M�CE�l��G��bo��JݲA��N��B�/�Z/�Z��x�}�+;��H`(YN�ɮ�w�D�.K�g�&�B�شO����l!�).��n��f�}��l N@.�eE��{[�4��{t�jEi[�� H:%F��7�ٕ��G�0��{�'o�0��j�k�g�f��wb��kҳp�Q^F@�f�&��#��(#�[�i�:YV�`6��Y��ܩ?�_�7&?N�ߝ{r��\��I0�#��]��|��+8 +m�4:��Ƹ��aꁪ ��F�(�r�{kz�1��H36�{�+p/#)�e�[��^��JA�;l��v�~V��~��)�V�=^��L�=��}� �w��n�Ӡ��K�H&}6��=c�]�;�;/ ��^@�s�~�g/�ƺ��KJ��4��}=e�K��@��_QΤ~�])e1o��69E�Ϟ��7Rav�J�0�� '� �tH�|��쒴��W �4�Ne�6B��q��b�A��4͠��O�a�V�"��K~:\��޷�Bi�#��K��;e�w��/�d��tb��p0��iG��5��AH��"�Q��3D>�%ʔ��k�1��b�@�WP�?UC��k�QÄgn�cI�W��iB*%��ة�vc +\� +��4�=ovYq��@hÄ�#uɮ��] &��]��V.1��?�}�~��r� �Q�+�=_�cz nk�$��svpć�Yin�K��A�m�}��>�j�n�gִ�0f[5�0��w��#@�5��6��7mU��3� �Rê�qeE9��/,�rj�[��%�;�ݮ�f��'y/�í.�ׂ��X��Vɷ(�`�.<��Y�O�L��p3Mӎ"vV��%��Kd�0��:�'��^��8��-*��%�B��j�mɢ��eH��d�7�q^tfAI�ń"��?��[$%[ɬ$��ȅ4<.gl��M_ �C��h�"je8 +ip-��X�Ww��\�C�6�� +3��V�VG v ��#+Y3��m��l�� +��6aR��W��mR��L�w!*�?>lq�"��j�d%U�IIgN��&{ˠ�t�CP��}��`)4$c��T��[63�%ڮ�b�2��T` +��a4��ֆ+�޵}P�¤[��l�\e(�(W��̯̎��ŵ��^�"!�T��>�*��g�2��+R��=5EJ�]E�a�EuN�U�G�-�&�[�4B��VT�(C۰�6�(�y��b*�� >��Ȁ�|�j;�&��#��*C�d$��Qu�۪9��X�Hot�Y[!�V{��G�Y��.�� P�i�Y��7�o��8a3�u=O�K�4��t��alv�C�Or��VZh��B��_,��n"O!w�bؽđy!ֈ�`�y�<�ǖ5G��y�e��{��x^r��[>2 ɠ��:�8�V�ʸ��AC�˫\DJD��.wۦ]��ҙ�3�W�֩��1��ҍ�Rb(�JI�;� �R(9�?�I�T��{@�YG:y��/��s�Ź� �-�9��ZOo�d�s�{��s�[P�cwdY�_\3_,��*�a��e�4��-�\0r�F[Z��V�XB�w��`��R7��+�I?u�k�d':/YG��$,�5��/�'��H�Ȱ��ǁ�`��"�)�jv�b*��k��$ ��z��e�|��{}*�ʵ�'�{A��$��E�foU�KY܎�#�W�C3l�@�i[�$)��wpl�;V�q2��i��Ѡ�EM��:m��X�I<�=r�f��53��4D�Ϣ�36��(ܕ��!�2b�]�U��B�^7T��6��z�r��A)]Y�k�8U��(J�\��fZ(D�B�v�]��6�Y�;U��0V1'UA�0��:��$��N�AT)�@�q��U��2��U۬%Q��h,��^hb�KH�BE�P��qHr�X�%��X��D0x��sE�ldc*�KܤlG��U@��Pם �E��VŞ��j��h��TW��(/�S�[��(`��r6�o(�H��q�Ic�c�&�W#�:F=��_ +�0�{�O� ��\,z��X1r&!��a��Hc��R�٭j}��gh��7�-�W�GP�,6_s�*U��kQї-��jTt�'��#"T�㤝�� +{@1�D�D�g��jĘ#4�Q��s��"�7��L:ϛ�:��(�xb#�$�Ed��F�r��o��g�;3�� @7C��9��$�E�\�%�Ί�-��>ρ$4�c��ڈ@��%,0�-H��C o ��V��>�'H�No�:��cGvR�&P�(o�?��n +�R2�NJ��C�C�%�4�t�N��S�]�A�G�'A M��~�R��D�y��U�O��)�q2~G�9�� z��^�ަ[�� 6q6?��ԻAe�C�[�{S�`Tg��$�l��|�3�b�ۋEq�aIH̒�$㌲�q��ws��gLv�*'ǥ7 ��fD3��ʘ@��9�`:>)��LVR�X�2��_��@��v�9N^6��-�� .#=��q�(ei��jP�O��m^��h?\�y,�b^ +��a�*X�ٿ@0+%��"(��k��<��_��Ӫ��Z��8]:�lx��N?t�>#��Gn��*~�h�}��O��q��l*�:A��\�1��üBi��]�I�2InW��J(��F��l:�4vMXd�YC�T� +Z��|F��BF��~&z-��s�J膀Ƚ��R�~�� R��z�J��: c�\D+(T)�z�W +t�f.�"yZo��v�6��zi �WqL1��6�%�4�;�s��Kt�`+K�ë�pD��S�T��S�K��套�3I� k� +ߗ�� GE₅�zVY�"��;z5��V�'�T�\��\�[��Y �Zxز�m�f�ު�fGQ�;ƐS��us��01t�i�N��g�)��&(&!]��l�~��L��r��h��"�|��VO8�ǱR��,��W��d�Pf�s�9F��̖WحF�{��}; +�hj�� +CL3��%ݾ�j�EqZ;)T!,-��頛��M�G��K��N� ��7��'}�U�b��ɻ4E� 0!��\>.�ϳgQ�Y䓨{w�8��06�EEof�'oh�n�7�2��n �/lj y��:��n�$�䚬;��%��C ?5��O��㻦}VȎ�I ^�:&q��B{�>p}�o}��>��^ ��ރ��v��T:�G�S�ނ�_K�f��t=�Cۯ��d��i�3��J�/�r%� Y�GX~�Vo(UV�7�B�$B7oHs��(��+��*h-5ERtV3� �`L`'�xCQT�6%*$՝��s��&6`��oݺ�<.�f�v��l¸�fl�[�aG�j.�x%��J��'�Ѷ;z��r3Z�<�}��.�' ̤�{��?�ͱ��S�ƑM`��_��4��ںn��Y��s֬� �1�V��@��#�p˃�"��w�'��2Ia�� '&X;o�u��a_��Q\Q��c��?n_#� +^Y�]��??�@��ri��v�̬Fڅx�4��<�Z�ΊӲ3��mr��Y�1�(v�U��;��~�_eoy��QO��#��L��W��E��Ĝܦ��#rqŤ��5�5_G\1׬�[yg��J�1�U��(o�1Вkr�1>��2l�H2e�-l��M�g�XWq;�ս"*��4�hr2��iP��ʹw��[I6�g�D�jb26�S>uK[�#��W��E\��6Us+�ސIk�&�l�~�n�ت< ��3CYRB��7�4��_��+�YB��&�D��# ê��p몦��[�s�VI��{6��Hc!~�8�+^Da-�r��&f y�{>*Ӆ��.8j;�V��l� OoyO=�)�s%�g��5e*>��Z��g�5Q��t�X+V٘�8ӌmGfiO�|ȍ{��Q�Ӈ�d�bW��(��ۆZC��5�vc ��wq�o��-򭥩��7�l��fg%��6;&)!\��0�VF��eq�Rs��Y8�/��R�G��d$uG�qa��y덍�0�|��n�'A�H�)�n�vvy��w��㇏QP ��R��Fq�iqbm��)��Fh�4�m�p� ˌ{G.տ��P� ��t�M_��s�W�t:��ǁ��J�#�l�y +x7r�b��O�q�,��)�ד~�t��%6т��JI��L�KN?��ɔ��M��A�Z;ֽ=��7t�SR>��(@�K��ߊ��8I^��j��|cE��F޵$0��P� �.��Ba ^��:E(=�[ +W#)1z�ڌf�]��Y�ׇ��^b�B�IH�0`w�KB�LOBn鄂�K��O_E�N�9vk��o��Ԇ�A�Ev��?G(��I�"~➩��ݩ��A褧3i�3�`~6-U��0z��]Z��[<|uuƬ�M��{0\e�o��1�͡2{��#�Ws�,9��I�I�9��m\7W*��˻%�X��*<2]��z��ɀ��p"�t�/�x��Y�M�Ys$>�7]u�b: �%�Cm��T(��%�NF��TVY�(�w�D�&�+��0�>�M�^KCIe�^��x��9��|�Y*]�u'P�۲#�4cf)��⎯E~��Q�)k:��a�%*��9�F�p�f��h�~V��˶~0��JX5Y�LmO㩯��Q�m��8L�� VX"�ZMS�`��3%}�e �ylW{A�L�i�1��Sdf��MSH�"[շ孌|F��2��K�-o<��1��e�:3'Lߖ�ݩ7�Mc_@�x̔�k��tEq"[g$��u�I��&*iy)�^C��|�Ȅ��a�.��wy�[NT~W��$4�"`��W&��j��]��!F��zl*qdbp�ٸu�r�ki�"7�Jr0!��ZA��ي� +c柉[<�� +Ѿ��W�ؙr�v��<�IR��5��{�FT��T,��)��$U�3��D�Ǿk��N�~��>< +��ǿB\> ��I��ZѰe��q��Ş{��8��N�k�AF7��:��M�~ l�wQ,h9��vٍ��㣗Ӂ|�o"�Gt��v%�U��O�l.j� +endstream +endobj +30 0 obj +<> +endobj +49 0 obj +<> +endobj +39 0 obj +<> +endobj +50 0 obj +<>stream +H�TP�N�0��+|�!i��JhhR��Y�Hԉ��п'��K�e??��=w�"�W��#�8�� GGP�`��[U��t��:G�:<4��o �#��{*�?Vj�@��Ev4��R��F��Rm!'�zB��e u��M��8m�5�M��0��,n�J�$��ͬ|�]�Y��r��;��yWv�-��l + +endstream +endobj +40 0 obj +<>stream +GBBMDJ+AdvP497E2�8��|�b�� +AdvP497E2��b�R�fy��vjVlC[RY�L�z��P�?.CC/W�`�S�$> +endobj +161 0 obj +<> +endobj +162 0 obj +<> +endobj +164 0 obj +<> +endobj +163 0 obj +<> +endobj +122 0 obj +[165 0 R 166 0 R] +endobj +123 0 obj +<>stream +H�tW�n�}߯�c\�^I�(�u��^4��- +Yflueɐ��;�%��AR��̙3g�<}��G�%Q4�ݧ��d�?}~J��ǧ�O�&Kr��$�4Qn�M;�u��ϻצ��uZ&�,�; ~lv��%�$�]��0.6et��w~�Z� v?��n��jm�ު�vS�Q�pH�5�>ׇj؇Mt׿��W�z�˧8Z'�ĕE��U웪��jm�{^��&<�Q~"cCǾ�tA&�?��w�*�m��$1l�z-��l[��p�PM�d�<��&9��۟`̕�&S#��d��軝x��b�yl�=��[% w��~�qJ׏�O*|4��1��S��AGqn��2��S� +��&��ҟ��W⺺ט;�ʌ��W��wbC3�g?~Oc�tu{�� ]SOM��M�л�q�İ�v�݈�� b�$/�%��~��<�a�-r�[ӶXb��u��8�Db'6��q5��g̝"�1v�8 ��_�/POx� L�^��T�7M~�軪m�?��=b��3��+b�%��"Q�c�B��ݸ��Vc��L�6x��0}�k��6�U��Tŏ.gJ��S�~L�2�8�]e��ݵXތ�lA=b�H�[��.��f��ؑ9�Y%��ʃm^:睉WO�Q�-/,[Fg�"r`ٲ̕e�%�M%�iS��``{�_��Ӣ? �j��4gn�H8�p�6��+��և�x��Ə�32˲�� +�J��a��;��y��yؘ)�Kf��D]�eo~��w��|��}',8�8w�G��x;�#�`�2p��P�M��M�s -��Q�I��j��㩝��$k�>�� J�@��u +śTq��-�u��-�ޏ�͛Q��nc��/��.)H�I��b~��śD�G`�ҥ��O��NӍ Ɇw;}�i�O�|9�"G��eb��Q*��I�?�4�/m(��*!�B��o'ͻ�d�L�^�2ٷ��)&��O�EL��y��Xd�%Υ��Dn��4�x�l��t�ۂ�a�_}۟�!�X��P��L(`� �O��)N5��骼�] +�3��x�0��RY�J.h�3%��X ?��(/��,f�,a��v�~��{|��ߝ� sK��tԻ2�J�o�� d��O�,T_cҹ��i��F�th/�1�ΛߎR�6�&�RM��e��(sB��Hu~��_FOu�s +�z6�d��i<�� )�W�`XE��P!�M�H��`+�y��'ó�� ЧkM��>c��X�ѯ�+p�^�P��]�l�ͣ��ln��vU��Ƈ��{�oػ�V$-�|��!��_]1��%�懄��~��Q1�+��"T��kC��oM-L�T�W$�q��>��6*K�8m|��/kl<2��C؆YJ`{�An_u:x�L�&%�Ϯ��m��]��75M�.}m.��7�ZƖMv��)K��N��u��ɔ��i\FVdy,�W�)]�M�`N��8Ej5q��.�d!-ڙ�Sn��Pd]u%��D9�Fu��N��=T�(`��V'��@'�f42�T�*��V��Nʛ��`�P#ĉ%��MC?쐈��f��'m=��(%m}eTj�c��A��#�v0�p��e��0b��lf3BDhD��g��Nu,_��q�|�� AC#�<��Oq��,��J)�4��x$͓@M�*�J�6��4��)g�՟��h3L��N�?�0G�� b��?��X�X��4wH� 9T�*�Y-�� DYw�}&c��-��q�F�%�:��<�Y��@�Yz^�3�E3�|F�ͻ��;�2��[�� ;)�ޥ�䨗[�Fs��H<��g�5H-I;�5�3.��"��"�r�S� ��ǚ�1g%G�� x��b�a1+*LK5[앆��Q�Q��x(M�.�-!�(y ��,j�,�<�r�#�46�Z��>��[!0�!��6WP� +P�>\�V�R&&t�;VTY�R��t�Is�F�� +�f�j�"�:��ք�9�@��hca��{Y�R��Ǜ1��kh�f�� +�$$,5�� s�)��G��f��}stS�A�,�5eo� 9��7�6rJ�b�4e��x��l��ˈϴ��Eb %��̽c�!��t�`�ނe��p��Ǽ6�%�s��*�Y�2��SRQ��Ӗr��+�y��Q|狫�� vWrp��T��f�6�w�j��O'W٭�9ˀ�0��q��G�/9MG��6�AK +�a^�� .iR��HZ1c�E�X|V��j��tcMOsn��'��^V�;�Z{�}�S�'r�djC�T�ϔɉ�1{TM��Ҧ��{x��/�&�BK��n�]˺��IX~�X�jF4:��&��=.|L"�G��Ƽ��M�NF;OQ1�r~��7�7��v�[�%��HDa�Z�y�2�2��؊��.#USM��E�ի0��;�Vg��y��jm9o�<� n�KQ�]��@�+��k��R1k�P=��2��x�A"�(6Cw�~��GG�&3��a��B˿?��N�9N��0L��.��d�;�Q��Q��$��g)dO��]q�rn�\�Wӡ?L��[\�VOo9��M��\]�Q��m��A �6Hj��܂�r��i9n�V:�޴&_',g�i.6w��/�as��9��?�jw�(� +�,��(��J�I4[ߣ3\E�&�Y��J/&��~PS�>��K?��R'�n;�Jӳ��*O�(��C��q�4��f��|\O��{=��IN"f�f�>D�x�5j��V�Khb��D��Gn��Bd"�uf�\ڗ&��8��2e��3t.�. \)C� [RH��-��4��cD�p��/Hj�xk�س'��S{X$��T��x�Ru�y`�:��7c7��B�{��4��o��]{i&��`��+ �u2B_Fй*!>�pݱ_�,%�5/f;��JB��x�R��B��(K+*rv��c(�� (2ţ��D�HIg8�s��E�hIY��m�{"eX��&�2�@��??��8`&� �u��x�!�N��5TR�:��0t�$S��H6��J%��j�I��`fc��犎y}��`6�v�^֧��۩CL*��F ��I1�*��skL|�I��-v�~��[1��QO^�i_<\��j��z>%�Ub�}�^غW(��T��,"�44�t\��X�ӱ��,��ioL_&K�]�2�U�q09��<�}+�f��o�^N��8��a��_%�d�E�Kΐ`�e�"�M�}�`��hu�E�k��#�o�2�E��0?��a��"�iGW��6�i�'��`�/��9�NZK��q�`M��Ur1H��{ R��Ò�6�g� +�Y]$L�&Ƴ�NM�iѿ��\2+ꛢ�wׅU��|�� g�U��<��͂�1��W$G,U܅*Y��By�M,m��ی�� S`�8W_��E +ކ��$a)��P@ȗY�I�5�U船KK�6�b�G��z��m�}��=�A6<�v�KS2��?��# +��'�qRG��j#��󪄋�@��ǡ�,5�S�Y)/��S+�>~�H4+'��R�F>�3��v��_�GY�÷$_Y��[_��}w@�՘h�'1�Q��3FaFn/,�͏X:5eyS��l]8m��2�X�p�i�[��`�;掇�LF1��q!��qA��bE��x��i�o� +{@�[74LwFaTF��^_a��7��=|[��g��0|�i-j$��1��0��~�(�u��,<��a� �;TܟUjp�#�O��zi�⅏1��sm�·�T�T|��Y$h*B�� g��b�)��&O�f��HH]8��E��Y�O��׸\��E��!r�|�P�-��;��m��l�>|�Z�Bά�e�i�87�ir�܂�n�ˌ�.�&G�[ʌ��Ȕ�޽�g�q��&�;��\9�M��g��V��UH�Z��ʯ�9��\"�k̲όHj&�?�g�"��|C��8LhW�G4�Q:;�s�Z�S\��/��N��?��L8�9\��&7�Zѕ*͡9�6++.��ĭ�ʵK�e��";9��Wf�Yߗ%��܌�a��s j�_�~:��X/�+��j^H�E��f~-�?��%�)�b�<��Jk��dC/��Xkux�U+��(��Ŧ��*��*��T��3��[շR��]/�)x��5�+w�2MB�;IEI :m� +W�SLvr��L�>�ۂ��0��=��q��?��C�V��D&�� x�>WK��ʜ�aK9T�_9~�Y�Z�ܗE2�"��t��N��c�)� +4TRVy�Ÿ)ᐩ8b��1~/�:�� y8t2>w��#C�Q��8l��¥��X��)�� .{/��I� +H��^�ƤfG�RVtԈ��LS� �H��sl �5��z�ӐB�,��.��1*��gX�E�ISo��W��aPCL��>iX�-�.'��12j�)㉿(J��i��b�؛ʦ��<� +(��b2��dX��5��w� Ð�.B +�yM�f�Q/R�1��Kv�vt��G�W@�a�n ��5|��ر��h)�� q#zׯ�q��$>Hpn'��Vew{k��%Qה��I@��=�%6� ��߻[��F��Lev�[-�ȥ�Ʌ½~=t��޼�r.�Opa4߅��l5ę[y��VfBC9�tc��P&T�/0��dy{�O�-|��4F�!)��ױl�e��G �1w�(�%�׷XG��Au�%N�� +��뭦w�J4�0��K��a��rd��'��ve#E$��$C�k�L�ۑ;��1�}MA�Җrl��c��g䡡#+VL|*��g1�m2�L��,6�֤tB�݂��>�&��4�t6��,�"��p ��'��Y��#��޿Jyv1zނw��w��왍�o�]U�:�He'�,h��"o;��5��u/��H��x8�J��}''�X�ktV��TӉś�Ρ'�* "챮ڜ�+!7=��i�S��Cs��L��T[�S4�c"ıS��;Ecα�C�0��[E��'dU�^w2�ux{��oU��k�I��:٘6_S$/��]��&x巁��mm=��/'��1��T!��}c�ܰ�QN��>��L]j��NUȬ7eRN��6��X��d�,�r�eL1.�B��ᵦh��ㄥL�ײ��/�p�eD�a`�UQ��H�3�r�0<�� a��Ph�HI� +�hk^8��l��~;\L ��d�nUj+㥜Ƌ�ǋ�7n��I��RZ��b�X|p��{��u�|C�PP��>�.��"7�ڹ|TG�M4U��Ѭ�;�!!�e��{'��8��0�� 3s��n�S,�|�e�1�ݯD�feΑ4�j�R��N"��A��;��=Z34 +��'P�E��h`aN�b�JK@QH��w��2�M��Qi��(�Ɩ��P��h�I�*��w�ߥ�6��[�ɣY�y�j�pŤ��dk.��k̯ptQ�'�l�/��y��E��@j�e� B�bZ&}��OY��P��o,�/��G�S��;�~J�S"p��](�"X��@A��s��([��3�1k�p=z�:t�΃��]�(�"X��t�l�=��,�T�fv�2�P��<"y��="<�bm��@�٠��Gd��&�3�Cѕ�z�L�%9Dx��$��y8�D�pY:��fn{��׻q/pluO{`.��q+c��x�'�pC*�%��)��rY�pz�2>�K��K\G�4�;T�>K��Ï`q/g�H��is��*j��j�;��S}�ɼ�=�c�D>$�!��ܾ'��2�>�r^�Ͱ�7��=�ׯNq��1�]v��I�2^��g4'��t�C�U|o�w{�"n�fS6 +endstream +endobj +29 0 obj +<> +endobj +47 0 obj +<> +endobj +37 0 obj +<> +endobj +48 0 obj +<>stream +H�TPMk�0��W��҃��T��1ȡ,�d�E6�sȿ��&��Г�ym��v��%�8��5� {KPn�X�*G=(2��y84�9�*!�#8�a��m��k�g6ȖzX]ˏ��h'�q@ +P@]��N��Q��/t�=�6��"��Ȋz�jW��d�cƭ�_��}�(bqv�&V:�CO�Qb�>KH�-��A��+��`m�k� + +endstream +endobj +38 0 obj +<>stream +GBBLDE+AdvP40798�w�n��b�� +AdvP40798��HL�.�Pb]k{U�V�g�n�~\oZ\�R�^�h�{�lݒ�C�w�y~|WV2�R�B�u� �ԧ��~�}�\�b��@�W�n�b��v�tb��kzno|}�}�{�k�e��\ + +endstream +endobj +124 0 obj +<> +endobj +165 0 obj +<> +endobj +166 0 obj +<> +endobj +168 0 obj +<> +endobj +167 0 obj +<> +endobj +119 0 obj +[169 0 R 170 0 R] +endobj +120 0 obj +<>stream +H��W�n�V}�W�$l�M'Nf��.v6�Ѻ��:��NE,�ծ�� tq��f�"�M��4Ѫm�m�8sh7yQ�JsE[wys�Y�hU��%n�lD6|Y�r�D�i��,Z��Z��C|��J�)Kg,9�cc��D +pJ��&):�zH�tڹ��p��Τ�,ؠtf9k�P)��fY2�/��S�[�_��E�� j7V�P�w�)��Z�o�c~(��I01�}ח��H�ft��l��$β,�(9�x� ��.�Y6�q8��pD��Xʮ:��#�e��eh@yn��r(�'ߟp��R��0 ��p�-��-؁κ�`t겪ʶp ��e�&V�>��ɫyn�Oعm�� eS�`�( +��8�ŁM�Ҡ;):֍��,��Lu��K+�XUH��*�'8��%��1�pl3��2��طG2�X�ۻ�*e�^�P<�J��h�">*EG0��y�R��O9�J*!;s��N��.��m�(�IJ�� [HS(�fWn}S�s�_�a��F��J�e��[r� g��Cn�Li��lnO8�gp��QU��v�]ٔ�^�4�-@ �"uC#p"L@VN�DR��nN<��V�T�O��?�d;�뜙Lsb8�6}��H��)�$�i��m� +"(ܒJ�Rid�Z��MHd6�0u�3�@9.�62��t�q��z��h!,h\�`n��`*�^ߵ�� &��Ĝ��H�o��<�Hf0�8"�(<��k�P�Xb��c"2b��7w�G�5��:dOdd�;��泛��󱂎}� ��jIUдx�̈/��fE�3=X*�U�k�E�?��C��,`�طm`�r3�O: zDEu�k�@ �3�H��IA�]ju� +�أ��F��Vo�'L��n�� +�`� ��[(��ؾ�q��;�Ցӱ�4�"o�1��n/�?X0N>�$��?�ǐ��j��'n*��Z��#�)%'�:O��%�w�H�+C�**�J� �Ұ�S��S0DR�y�{_|��p�`��.?��%R�7��+-��2qi�-�*��J��D� ��<�ӧ�\|�[0r(��X�ށ�ǲ�µ[,� +f�g��m��%�Q2T!��@$�障�zV��,¾�o +U�wˌ1�=��(F��\�-��^��᷼\�p�Ρ)�rWf��=�}~�'A�Z:s��?�?��zJ��5�|tg8 p�=E�� +sA��>䫴�D +*��y�=��c�հn�!��{DF�I��F6�p��>��o�1�ã�R�,=��y��"��В��Q#�jn4�z��.�p�Λ�:��oj~)��"�}Hݸ0�j>�A;�o,��/��~��OB��?�z q��4?��E�3�eP`?�MX1uBM��M��{oi��h6t߾�؋��$�Tz�w�c8�e䫶y�G�U�R �!6�S��K2��P��xG/�9��,#�,<�H�`��9BN��Z�$)�LL"y:uh��`��b��f��Q� �a��c�]܉3�:�H�N�Yj�Y�@%�ܦ(;�]:��>3r3e7txbpg��a�_�0�&a��&��^b�ɐ�(�Yȳ�5�� 5��D�{.�+��z��;�)Yr��'ZQ�[�R�[a! s� j��49�I�3it)tN; L��hc��"�x��g��8SV^��V//��m�e�e��!��[lK3E�=�1]I��6|��7��A��|�]��tY�H�V�S@i��&=9�G��ǲ�JN�Bh "�b)� � +�0Z��Ӽ��r�2��3D�߿�od�O��7|Z�O�H�ߜ��>]$�-)��5�}��]֐p�� XCO +<��Qf,��ΤH�B�.m��)��⬡B@F��ƚ�oA�\y��a�C�m��!>_w?Q��!�U:ET� v��m��i�N�1W:�+}��հd�k�c��,�Q�+�h��z\��&Z4i�m��+4�"m�+�W1?��|qU +�O�S%��J%�p��P�ew~�׎@��m9�H|Yd��|��,8~7b@��)��2Ǿ��7'C+r�z 4KÚ�#�7�{�|��_�5��AU��E��>�;�j��}4+��S?^Ĉ7��쟘W+<��ܬovUz�FZz�:��M&�8�D��m�㮚��cl#��<zjԵA��.)�&��5��X��)j��Y،-�a��f��S+��hVX�0� +��Mx��_UR��Uˊ��^�N��{u��(b$J0��&��O��H�Z +��<�C�.2�>6�:��V��@��$w�� m�Y)��Y�O(r�c��z��rLm��]@n�2. ��6��"��>�5_E��Ԉ��rY�� P�4_��q�7a�'�E��i��Du-h^s��$Wr��Ga�t�q��U#ۗ��:�<��Kz鵞0�|��:�vz�-WN��{=�f� +��{�N�&K30Q�N��Ҳ�^��J,�֥�|Z��K��зc�#��W��0�M�Y4v5͢rF̿�̛E.0��e6�S�X��P*_��U��r��Im��A-*�nr�J�^�*u�{ͣ(�+��5��q��J;}�L��ni�� |ԏ�dJId�^�:=q x#x`\��GR��,#h��G�"�2/�E��i�YE��n��\��L^��b�ݷ�>��7t�\R�O9B�[��-�,�Ԃ�~��^��o).�zM�bu�9�?ذO�vRj�_�� i7&�@`�}L��)�%ܨ6\��$'�֧~L�R�.4�p֥Vϗŷ᫮F&^��ی2R��a��Hf�ө)��2�T��?u�i�Ǥ�]�a��9#?v�fׁp�P�X�&�0�4;�vJ^�6�ܧ;�S-��&_d=�ۄ�d�|^N��9 a-�o�=��=u�y��V��"5�A |�k»4R��*�Q/V(Ov�#��G�9H��t%��[��et��J��G��Ha��}�Xxmo�j�6�-ύSd$�/럱F�y��O$�E�چG�u�6g�Ȃ��4�L�6j�׽�v��Ź�-~?;Ѣ�� +�a�_X.y]r��D��IX��2��$!�T��`S�B^��"-ɒ�双dp�f!I�ˣqg�'h^�B�w��A�ds0��yC��x�YrL.��A��8'$z`��D��"�=<㚗 +�3�c��N�lR�|J�pW�p7J�BF.ϼ+T2��r��p ��c=��d��J�Wq%//��s<�$��olc��λ�)��Ϳ��r�iS��X�z��}�;��y�P��F�^y��v�oy{/o@��~�'�O�f��l��i��&�~��D<,䝭�p(e��هǄysJ�r�6�QO#��?�K9�?[�T|��(X%�V�ԅTT�m��Ё�pmN2��.��Fo��p��zRjq��z`/}6�t�n� ,��P��Q�\��LR5�x��#05�{�)��Yx�VT�2�W��D;Į�';a�7�M�BՂ �U��o +�mhNÝ��&X~�7�C��p��"�^\��;��?��!�k b,�af�{+��*V+�)G�Ea��ϸibW�>��⵸&�5Y��L�^��vF&��d~=IG�Ąu�< ֪5��;ug،�n&�Ft�3\(�me�P��f˪�z釭PVQ��=8�M�X��He��T��zR��x��A/b^�/��\�_2gQ�9�I��!EeO8�I��*cx�h��5G�M��&J��Q�Q� 2&�� K��J�G�6P��"�1j�+��H�#�Ɏ�v��qP -<��u�2�(bO�� .1eB��@Lے9Ւ��EA�d��gX�R��D��Th��(;��Pl�v+B��6�I�PQ&�+��8�T�6|M�'�J�2�U�U�F ��l�.=jp�Ķ�53�ɓܺW �s�-S*_��ё��<-��)�){W��+5[��QAʯ[1��؆K��iu�S�?-��) +��2��~�d��vTi�$[��D+�Z.9m��~7F}$��f�~F>�IӜ��N��Ly��Z O��X�"k�&V�<�_��n"؃��yC��R�/�е;3��f5X�q)�� [�9i�8ie\�mU�0� zˑ��7��P[K��Nh�$&e��/�+�� +*��4l��f) ��JU��;�Ir��\�C�O��g��U~\��C�)N�O�~>��7�6$�f�xg�sǚAsf9q*��:�Yv�~j4��$}ZZ!��%x�f")a +�6+g_e5j@ItV3�w�|2��=�ܜ{mF&��"#�g<��P�M��A��e�(%C�y��"y��r�8qn��w^V/P��ϕ^��o�J�Ȼ�*�;�>a`X:=gI�}��⼄bnpL��O9$�0�|�I�2� +�+��f��^u��#G�]�B�k��5�򴷹;l�E.��{�-�w)R!%+��S�=��HX�j֐Þ�*��w%�Q0��*EAN�@��/��HN�i��1�դd�'1g̥nU�ߦIϬ6u��ha��L��Æ�K��{��5��>��x��H��w}�!=�o�x�a9�EoZ(��C��O��Ƹ�I%�+�9��n��Tt4p�o��䅻�q+�x��[�q*��t:|�A�k{��\t��ܛ��u��;��;F�`n��;tk� +�b�"a��NG��p�W�u+�ltr��V�� eR��*��~+��+'��a��>�0��{��2˭j�S��$��:�&�<��t +��ks�Q2��r1��w�Mޢ��A��&�b�y�DX៟a%w� �hg�ɻ@��H��X��Ф��\�f jW�T�Z*2��* �WB��;0y0�5ڑ/C�;7��\�k�V�r&ڙ2_ +�"��Z�ݵ�)z+��RSA�V�c�zY7L�M<6��ĽT2��<��j��3͎�vL� TN�V 퓇r�|�Ŏ��89�C�fNQ��_��9�e�+&��uF�m��K��Oì`�Um�Wɓ��.�wR��i3�� +e�M��i��z>C[֢��c�U�B�x��T�%��|��7�l�C��h5��s�_��Y��a��%�ǿ�;n�2��\�թl��_�}�c��xSw�~��|��>:�p߉K-�<�V�^3��"��&��ÏP9�qBIvAx� (`2^�"�T�pi:��99��}��f}S6��ڢ,t�MߞdR ��R�{C��x�R�c'%F|ұ>�Z� ��Z +�f�K�|��떽@T�� u�o��]L��\Mw�2�ahs}�y$�x�j�S�G�CXq=��Uv[��[r]�'BrsW%BM��2[��~(�C�Y�0��nY��%c�W�1��I�_��n8��\��$;��B,w��%Ԡz��E��S�q�-;��M��%�;��hl��p�2d��W~n�٠yMt�N>+~uu��6͓x ��.�6�m��6B[`��;:�;��d��f�s��\��J�'n��գnz1�Acߒ�;��D��G�s|:{��Z��[��y��^_�J��PǏ�LDK +�x��b�B�RWYkbG}u*�Uj�ơnt��e��^�]�}w��P��T+O��6�pY��R�o��y/�?�� /Lg�_��]y�$xƓ$K��X|C(�R2(*�8"H��K�]Q��j9�J1��`��(W��.쀃kN��ώI�T��iq��]��fa��z�Q��<�Z�tK�Mn�� +��h]��iA�NV/�~�r�q�/�n��X��H��m��?�� 8��+�cP��'�v�u(u�SH�~_��@$�d��}��=ah�3#�z��p-�-Ե|��Jp.�� cD��i�`#�m|u��}/��(B~!��,��;)�G=sP��m��Bʢ�"��36�o��V�K�]4:��k枑E�;��p>+��gᤩ�PށĿ��(P��6ϐ�d�(w��n�Y��4�~t-{4��/��A*2Պ��x�s �Eqc��ph��%�f�C�� 5��Js*�t��P]�P_�8#5E!�R��66�v=!�@��6a�:@��&��%� .��ʮ@�Ҋ�s�6}`�m%\ +endstream +endobj +28 0 obj +<> +endobj +45 0 obj +<> +endobj +33 0 obj +<> +endobj +46 0 obj +<>stream +H�T��n� �w��D�U)JU�CڪN�8�H5�3��b�-��?uO�w ��c��yK8��G硑`�I[VW3�<��:'�:?h[�߳8'Zaw��x{�d��awi>>�A��㧳�/zB��_�FY�f�,�Q$�G�Vj�P��ט��|ib�J!�Q� +�},s[E��|��,D�n�D�S�8��f�˻e�%�nv + +endstream +endobj +34 0 obj +<>stream +GBBKBA+AdvT373��n�n�|�\��C192AdvT373�|n��wsib��b��_�w�mƴ��i�s��u�d~]ym{��Dn�{�I��y��q�t�}�v�uF��|�=��X��X�Z3�Z�W<�W��}�=��H�Y��Y��Y��Y�L3�L�X=�X��X<�X�H + +endstream +endobj +121 0 obj +<> +endobj +169 0 obj +<> +endobj +170 0 obj +<> +endobj +172 0 obj +<> +endobj +171 0 obj +<> +endobj +116 0 obj +[173 0 R 174 0 R] +endobj +117 0 obj +<>stream +H�tW�n��}�W��l �^��N6�z��w_� ��1��<ޯO�*R��^��S�N��x��C�4C�%��W��Rs7\}z��R��\٬0IJ��3/WU�ڤ��Jg�I��_燐�o�*Wyr�n?��?�l�f�;j�2�u�o0j�݄x�z��]��ԎSt+,��"��?��.[eEN�~!+>�q�7�Z�V��U�?�áno�68h�cR��߿�^Zd�ou�q��01�V�tr=|+{��k=�a_��0�TV�zػ��F��1L�V�ݾ>�0E��+%Fc��;}}��k��h��q�{ �*Vr��wc_w��-�N�n[��^e�uW�g7�~��4�^6.� :*8��3٪�h��Mm�v>�B�n�ѵ� �Vlr�~D�aK��ud�F+P\"��]��||Y�'8[��{i_��@ɍ�~�~��SU 3��uS��W�𱸡NS8��/� +4�q��X�C�٥D�v��Nɰ1��o\��k��E +oc��(p��"oRx�?㴜m^�׀�P$�,�'�M��FGy�S'�O�y�a��s8�YZֆa@#gW��}ߴ�M�A�ӣ�� }�|s+��9y`�!:��E[��JR_F2��O�݁��9N�d�I2�G�2\i= �w��)��g�q<��t�~��g�(@/~��Z��&�#�/c�Ӻ� �]J�Tv�U��T��G��B|�Lho�8a��]@�E�i�jKl'��a�;(�+O ��-pE@�I�L��14�X��e��ξX�i��s�5�9�iu�ۼ�}iN�a_G��Ld��̦��9f�\D+]�U3�㴎0[3�hߓ� +�o2��at�!�_��t��HDF�#fg�̃O��~z�Y��'ܞ� �Ivr2F�2]2�$�q��p��bi��0L��a$��1��k�5O�k��:R �WYf�;Y�{렏�,;^��)��e1��I��BP �, �wg'r��eރ��s�ۊ�U�y��r�EYW9_��a�&�� +ׁ��B%ʖ3>�j�-W�{��L�ي��q�(/N*��h��w�8E�o��3_Bl7�/�&XL�²�Kʨ��Iڌz��eNuK��|m�W�fr�90��~��%��F�� +�s��`��z/�,�\��e:��m�&��3r��Oj`�PP��9x �2�rI�r]��=, ++�ٞN�!K3��D�iUhR�4*yT��8�Rf�Յ 1��tG|�ʪx��,̉�E�� //Y �EuW��|��8/�*|\ŝ�M��m�=�[Υ��/�/jP��>�8�R�X[�#��j ��w=@�) Pw��ΊA��H-`��$�YU��bk�DQ��Cm�"iR��õAH�}Q��UBɨĦŪ(��F�� Ti+᥎`��|~��oaں�9i��_�iо��M��oG� )�?�2�P�@D�k��n �7�Rt3(��GvS��ƕ�>�+7��+��<��9��|�<�Bv ̋4u㇎�{�fZ�'2�*`�H'4�`d�ѡ��Yp�S�O�2갲�Tjj�,! P݅��`q��DiEiEMV�S��xc��3�1�)˅N=�I{˱<��S�,l� +'�ȳE+\��QCܙ�ϬS2��N�mJ��w��..qk^Ě��iJ��[*d��0�4XK�9��:(X-��f�޳t�*I?�2�t��SAO��I�w�2TΌ��|Q��u�>�ѝVϻK8�|�[��:��N2��.��ؔ��$��$7�$7IF��L�"1��(�J��qC�CF*s?��jN�%��Ü��"5��oٽ��2&��a�ۗ��AwB�p��g +yBDϯ#��$:o��/ �[��0��=2j�>z�RR_ �1��P��+NI��c�Bx�n7��$�QLu�D�(��Ω��?*�a@�1�^J}3�7��ԓ`c�Ҝ%$c�2]��W�s(L��E,{�u�� +A��L'0�U�� =��R�*J�R��y�mpc�)�P��d��F�.3��$z�/�K�7A�ވ;�|��S��^�>�a��4 {�Hn"o��4��6�:�$-�T4*f%��h��ׇRp�8nJ��_Z\ /EL�fy�4�4�S-]��G��.a53 +f�a�"�1��&E�p��Y�8�!w�;T ��->9U�t�k�E�Â��6� +m��O�5�5�O�"[��Vg'�$�u#U%� �4��!�l�wX-��Je�&�X-J�5� +��:#V%� [�J4ɛ4׋�a[q��,�H#�e5{��e��l��b9/��y��b�-��t�Ys��?��$'r�� F�*`�$�Z�ް��Av8�n��n&��wf��OkG��]PU�e�2R ꎶ4nvW^b�Ͽ|� +�ͅ��N��o�f�Q�C�#*��Q�8\��{&��r+_�d>M�n�ҼWJ�O��$�h�m|J�sUԛSiL��U��,hQ͏}�Z�y'��m�K�C�س�<��|��k9�]�w� +��JYOլ}~�>\�(q�/v\Y��*inTq��<+¡��zFbx˔��Z�%��`��]f�Ύq��EU�2ȭ(�[k?ߖN��K'�� ĐeR:ŘEN�:��ɞۮ�a��5f��e~�[�6��uc�] +��-I��Ʀɲ�k�zލd��h�g�g�G�O��ԁ� &,��o�B��BJ)8�{��Bm�RH��"��ɟ�i|F��ZƝn��4@�o�D/��:5A>ј��O��c�VNu3z{��Ԟ�vgy��J:�UKx�'}~�Fˇ��B��O��?6�W{��U)|$X��Z?6õE��p;��s��5��yƿ�<g��܃`�#�bq}�+�w��I�v��"o7�b�v�y�l?�F��1[s�_�� s��!6�9��x47�Z�έRYs�V/�5��*V�|;��ҧ�e��ke�v��V�1�^�j��PA�A�Io��Ǥ"]�K�Ɨj�\g�?I�`:��ǿ�V�\�r� �b��C��~��<��ӽ�S}~�{:Nb��.%h��Z��A#v��̟?d��~��t�R٬r�h��S��0ݵ��O�`�Η�ւN;��A�^}��4b��M��L�W}��ږ�=�o9i2��h^�>�1f�Ȧ�:9W�g��rkm�[��vcN[�⮛��z|؆�B�3BT��|��hf��]�詨�w ��`�?#f�i��l~s��ٴ�4[��d�.��RSQ!ǥ��J��+j��y�Z��y�C̓��}�=Ǿ3�5g��1s<&�j��5R��x�:VY8d�"��y�Խ*�h|z08��p_�~�Z�<^IL��^޿��kQv>�f�/4�;A��w��1[��x�i�d��AA�x�n��/؝�V}�,=�G��3�u��.��H�-��n��G۶#�Q�t��:p�/=U��B��9N]<��닸p�B$߫*��X��͹L�tџ��snF��C��AQ�$�{MNv��I,�vx�`bP[e��%2�M�f�vV��U9ϼkD��m�D��~��D2�1��~m}MEt��|�h� +�/�p��O �4 5V�ΐ�ˀ��6_�KT��+zr�u�~�~L��9�7ht��5�!��t�B��L��:��#5?c��LOƐ�|�Bd�y��h�؎ݖ2�V��g�6�k��^��?�K�s� )��>�al��y�v��\�l��z�� HU.�t�jv�:VJr�7H>��uD�V�Ee�Ɉ��F&Z:�{l��v�je9~��J��u��7�T~�">��Sw��:4��SW��4��W6�4��I��-&�S'�б)�5��9��f��:�/#��1 ��M��EW�u�<1�7�G��-7gݽ��de��3��.풍�|��o=Cb�7��u��Х�>�1v�>"��K+�TNb��/]��ʈ�[nC�}CqeY��H�|�9C'��p�a��՟�:8�S��.�Z�>�,�UŇ��f�H1�vߎ��zs=:��?�>�P�� +�=B�X!�]��I�]��L\Fe��{Cb�s^m��lwe��`��*�tQ��ԓ®}!�٣�'ˣ[�:��Kc�g��&�Pl0`�%�c��4ִ��hScO��M�#6�M��g&�Ѩ��s�Sx<@��̒2i�k��e�1�c|R�9�ӱcO��rs�Xo�I�;oJ�L�ŝk63C.��6��(V�?�]�+e�� 8��P��#�q1�� +K��e~�M��O/�Ad��M�EA�AӇ׳��s�/7[��u콖�V��~�I��/baV^[��c��\�ؘ3v�{ݎ�0�XW�f`V)Sm�q��V��5�̡�xA-�5�Qƾ>4�v� w7��'�A7�ٴv�~p��[U��N��z #� k��6Te�홏��Ԏ`��a�I/��.Ł�co��0��y�ˍ��Ʋ��ۦ�vk=��x,�uԦ��n��7x��۱ч��}��w��1H�_�䰌��Ϋ`) ;�v��/��t��[��c%�� Ul*s3��,c5��9�/�$4��2��@7�)~�'n\��#��,�IK?MC��lU�-��| �>Ʀ��.k��H�t��b�b�q+��S��'��)�N]��Ba]p�ti�c��B˯��¡�ج�}�<ݤPf��6�@_[�Wz-��q��Q +Y"��j4L��W��w9I m��X{R%�I֧y��{8��Ŭ�E�GD��38Z�S��X?�u"�|�@~z��b+�sĖ��=T�Œ��kR�i ��G}�.ĴC}>6֑�§��j��/�>G@��y_n��jYn�Ȣ{�v��xf�qǞJ�eڕE<5��1I�)Y��sι ��B*��}�uV�'�Ju��bh3�Z�W��ι`U�k\��2� \�Y׻$+��C�"4�A|�]��p�8��Oű/̢�z��2(~��G�B:��q*��e��g��堌f�[הb��4�9�}x��-�|�a��X� V�Y]=#�m�&��fu*�!�ܦ�\��]?�_��)�۱�z��A��WF�x�nO֌��ZBlj�a8c$G(�j]SDR[ו`�O�)��|�#^�A��-8��@u��o�K��nas��q\!�9X��twW�A��`�o>�3>e�p��2��n��E��Y��FEV��E��0�Q��rۢ�� g�9�sv+V�B�;3�o��"`V��HL��J`�n�pJj��Dc�\��˷�}y�j^��H��[h��sH�;�.��.K��g,�i|�/��_�Fu��t�mWm��c|��ϑ�G�� ~l Ź޴:S�摵�w/��{�U�x��d��'1F6�@n(�"��x�`G%eS �RTsQ� UmlI}��9+� +��4��D��jV +P�=��CH��LN eh��B~Tr�n>�^#��j��ɟN�FT�Dg>E}�$&)3v�u}W�7��T2bB +��n��n�x��M�ܷ�.��lv�r/��#ޢ��\��k��&�u��D��6��x23�R��'_��k��MP�F�R��]��s5`�%H�FZ��OdqTɖ�e ��VeewA�:Q��(-'P\"� ��H��3Q�C�D��?�~��K�%j�I/E�!d��V��L,c2��8�Tp.:(��<�e�z]/U`y.Սg旪�P�F»�H��nd�<�'��x$�<ӠuF�38��U;�9��5HC&��[��$��dwA�z��Rvm/��Wq�T4��fh2!�<��R!Q�\��q�WT��V��pZ�¯�10#` +��˱��L�T�GF�v��g�#I!1d�� +�z(Y��~�v�f��n�>��7VK��F�LΩe�p�]K�_Z1f�w÷V��* V10�YT�(�Ȍ:H��=��bE@�,3�t�K�%�W)�0a�)�wm�>�}�}ݶ_Q˦�PH�u,�<��ò��#9�~X56[=I��O��k�0��F��3��)�a8UԪ�;pq��6�o#O� +��M�3e�o�0q�{��dD*��B#��fd��|~z~��\!;�{O?��P�ۡ�xg�+��#� +�f� �2W�q��kӤ��m�-%�r�Z!<`��T80��wh7�� +Gr��.��@z��n�b�� տ迎rG|-C�"�L�Y��f��YN��T]f��@<� �!�CL�ܟ��g��6V�H9�\)�p��qY{|��?ӑ�y��.��$��rk�F7��K9��js��Ж��N��yy��R��-Ɗ +Q\�牓7��ҥN��<�ϧ��{�a��gP��v<<&��iq��m[��9%�/�S7��߭)s��u�?J��\.��j�a�H/0��>�R�{co&��*�b�kpl{V�<��4O�!�\��ݛRh�Nje`�,�ֿ��F�!C�� FX�x*Br�5� ��U/�H�eNu�H��,�!��񅲏6�G��g�P*pb� aW��O�@�B�za�{�]ϳ(��MP��\�٢'�~R��}��Ż/�HÉ��]�&�Ȳ́��;ŏU,��vVEӲ��V�쪭4��DW��<��ӧ�?�y�A�M�x��}۬ZN�l��L3�m��o�V�W��+ �d��?��`��&�Rg��B��*йAry'��w|��3�ci��k1�?[K(��.E��/�̊)��M��f .9�bB:"�I+��ݖ0iG�=?��i��q��d��RQ��{�8�ե�,�z6�g�KTwu?.�<��C86S��o��g�.��Zf +W��4��Q;�[��'�UkM'I͜Ja�T��%�&`�K��׮��'��^��l��lY��ۛ�}S?�>�o҉� +W��#�tQ�ͬY�\́��Sǈ��S �6;Y�D�`�>�|�y�!J�[�=s%��z4��M��c ��c!i�Ȱ��()��t�@&��Q휿��ͽ�T�xS?HCf��)��L�tFTe��Q��Q�,��D3�3pt�i 円J't�Ot\E��J2�-Hg\c�}��ŷ�؛3D2�Puh��;'@t��?g�?�Yĳ7�+� +��,�T��}�_��Mk"��9��k��ь��@{��x��g@�d��?�e��&Eѽ��u��!��L��3�~$ӵM��l2O&��Jv�ݙ�d��Æh��B$��X��R*'�H �$T@"�U� +@x,u)�� _�:��]>B�<�?H�T@">*B zB'��hA�U �U�R: ��e]u@�7ʺ�u�IB$`/AV^�"Yp)�Y@��$T@"��$�p3�D�g��A�繍��c�oҷ,��5`?�H`��I$�P��H`m�I�\�A: QÍ�mQ7��:��[YW�@�X+B �X��: �m�R� +H��idiu��V�v^�,6`'+��EB�u��hKE��D[�^�B�³��DSD��@��, �[*B$"�� +@��lu�$��:E��5��0��ʼ"T�/~-~,�n�ú찐�PfU?�DO�ϧ4��-V��/�I��W��|��]ҧ +v��v�h,��37Kt P�>�d��H_[�R�|��a|F��a��A��D��ϙHW�i�S�g��[{�p��44��{{Ln�_bV��V��:��oCR��@ǵ�m��~g�E�H��)�۷�\��i��K$��R�� M� �e��m�77t��{��W��y�;X��[�^:��f�=u�˽0=q��`[��*��ڳ؃��cP1��I�� go� ݀Z }� �t��⋘O�=��Op�4U�s��=��&\,ciK�I�0��b{��{8��y�� +��6O�wf8�/��Ŕ�SYl�+��v��g�XA1v�p��n�?��S~�?t��*]��6(��^*=�� +endstream +endobj +93 0 obj +<> +endobj +95 0 obj +<> +endobj +94 0 obj +<> +endobj +92 0 obj +<>stream +H�T��n�0��<�S�`��$B�h+q菚�w�,��b,cy��4m��3�z,��5ڃxu�:��^��4�N!�� d9tZ�e�j��К~��J�5'�.��q�No@��6gX��/�xH��>͓��r@ܿ��"�a�-c�NV*tҜ�ⶆj��M��q�էtIUf�LS�Q�b�� G�q�b�T�3��yV�QQ��1�G�c�<^��mΠ�j�r�2O��ĩ_CR�s�_��'� ^oώ��7�`}H�� + +endstream +endobj +96 0 obj +<>stream +H��{PS��rC��)�� +�c��խ��[W�vEE�uAY1��( ��$!$��dD��TD��@׶�Xg�Yg��^;�0[��ӝn��w�|?��w�9A��`FPP�;�cb��9��m��Qk�Wç��9�J��_i��5�8�-�;ӣ��b,f, b�0[��A�]FB0#)��̘`�0��AҠ��3�h�aƣ�ܐQl)֍��y��$4:T�-K�V��h��^�� ̋�@�_�&��ୋ#��9�DZ��V�*�}��|��^��h�c��U��a4��6�.��h=�F'es��:o�)E@�Q�U�^o{|�@�\:I�'sӔ�@G�� š�J۔��H�'�O��I��e9%SQ��Bi��6��h��A�eǀD��ϑT�*+˨<@1��z��g^ҫv��N�e�5�)�`��$@�`�F&<�,˼a�vT�Щ��l�#��Fy�C��a�@��6�5�t��䳞u��S��!s��2�H%Ȕ��Y�-��T��y��sp�F��`�e��m��V4�Ꭹ�`��.�X�?O-��S�1��Z] dզzG#�7��3��?��6l%�5gu�E��%y��%*��vG��#\�>F��@�W��UTX$�R��k��3��S�ۆ��CU5��F��^��jy�p�y�80vs��ˉe�IC׺�n�vh��\��=F�G�~rRO *ե5G��j�)Mv��#�B��n�Ym(��R� +rk�I�i��2Qo)��9��ˉ��`��C��h��\�y{w|B�eC�@��i7y�@��{��z��7'�G��#�TNJI*��ݔ�x9��wϿ`�(4��!�E��O��*1U�'�ї�Y1�٨��^)�s��w�e��4��ȉ�W�Z��"�S�7��4S��K��&?y�|��W�T��l�lzԃ¡��J��=��t<=�d�C_E�<7��47<\��#O��p��a��؉�ܷ�]f��B.��Z�e�35FPg��Dɼ�P~l��o$, �x��~ߞ�G�+�T@��N�]o��+I��j� +H�z��m�#��s掆$xѢ, +�K�ɇ?��]��t>��|f~�OZ�w��0?��)��F+��=�_8G|��J[WK7�el�+��q`R��,�x�>��3'z��C�� w>"��u<D��8�yh�=�r:��q��:��8A��4sG��n�IGs�Aa�՛�b�Jo��\��9z �~t��_��$��iM��pɭ�+Ψ��E��j��C�:��D8�ާ$�E��wc��N��Ly\��Lg՘��Ei��K��I�}�y7a�٤B�DPH�O��d��E��ёI� +��[P��U�)$rB��7�^��\#�~qK3��~��\�'ݔ}ӻ4V�,�dӡ뢷��>��le�N`+�u]�\WY��ř��񠆩*��/D�C�16��^Gs��E��>��+(��,wE�H@�l^AϠ7�*$W�p��rը�B .��}ٵ��NLo2i7@��B]��&�� F��j�|�HTQK'i�6�A]�� !��&��u�Ԙ�tH�@��| +��x�� +��B#�w�Kct-d_O�p�lR��8"^Bza�c��:�-CP��|1U J��Eᓑ�K'��RZ��A"�K��8:t%D3��qnQ�'� ƌ'�Z��E�;!��i�� u��_��So��G� + +endstream +endobj +118 0 obj +<> +endobj +173 0 obj +<> +endobj +174 0 obj +<> +endobj +176 0 obj +<> +endobj +175 0 obj +<> +endobj +113 0 obj +[177 0 R 178 0 R] +endobj +114 0 obj +<>stream +H�|W]o�}ϯ�K�Hd�C��E��7��{{q��>�-ӶjYr��4��s��bg� p@ICr8s��$*�HDQ_67�/��oVk��^�w7y��C��R�4�T'�Q��t�Ǣ��z�~y�d�hq�<ֱ�mY�]�x��E��!ٵ��u��D��Ec��m�[��q-�];��J`j��_�_n��N��Y��m�XmG[��h��ƾj��hxˡ��5��?⽎�h�qފ>��|��sW��L�9��iY��r�ʴ�� +,}7uU�c��.�� 9bO��Ę�c��cګi�0�8wݶǰ0��3\��[1�aԉ�ޏվ�mS�[�ʱ��nlY�xΰ]�6C��;B�}lK�m��DK��Og��~�'�ˮ=�� }u��)ٚ2H�l+BN?8�(�BkN��J��h�� R�zX�ğ�t�jx�ߊ�?G�!8��2��};�>g8�e��i��JvԻ �ug}7�o�ߊš��/S��`�v��ΟPx��smK��*��ͫ��{�x�p��v78ʈ)&y��\Y�6�oI_ X��U��{+��f`��T ��)��X�ao=�Vz��2EN�G�-m��Ӕ��i"�@��0v�m�m7�Hބc�=i�mwn;�(&��ݟ�;��p�T5ӣ�LŤ%�n�#�Pkz�� g�)`�s��@ӻ��*��_��F��:�-�ű�H��2d��,x�)2��6l]�q��-1Yj��=�*Kt0H�*~g +��9��;��v��0ź�?��+�YXX�; =�@�23�3�c=�-{�3��ފ�\՘E�R[؎�Ѓ��9D��+�T�\E�'ci��Q��n��\��7V��Xu$��:2ĸy��b�l��V�Ox�珩� +��@�U��庯+�Q��L�A��+�o��Y�%ab{:��I>f*IB v\� ��rϚ!hX��oT"'J klyXp��)�J��Fu\f%��c��hl�r�Χ̙�h��I�*�,E(��)��D��q��&�L�EE�?��1��T�g��h��Hc��%�ș�w�t2�!�}M��<��!|6h�� ;��H��`�Ȣj��Ũ3�L>��'ӌ(%1�zø��U�`z�2$��ݏ.Q�f��]��]Mr�k��r>�L��(d��C��ﳔr��W)�*�?H��ٽ�a?3�uԔ�\ej��d��'g�:P�Mf�I�K��4��d��i��B�W�S�zr��$^�� fw�CD��B��2�|~OM��$hnf!KP�+�3��c_YOM�WS쯐)q�8qI��*�37}5��e1�1��@��9 ��r[Y_e�ǖ�l��Lt黀�|Da�'p68��C��㎁u��k��U�j�ҿ� +5_��7tMP��z��Z��ⶑ�7�Ko�L�Ά�]�/��M�S-��ח��C�;׍�c�~�K_.A�/b�E�V�$�|�Цg7TX+�"� �^u��J5Ǹ��=�+��|8�.F�R�B�˪l��%��:�� ;9�ezI��]�|�%�-��c�Z��!��FLsnC��c�4��U��!]W>��P��GVQ�K��׸G*Ԭ1��]��?�U��6�� 8^~ 9L^�qr�,6vkq�8�PmqW"u$e��믪{8�%w8Ām~gzz��D��e��V�6O�n�iX$��ѝ��F� �K��n�⧫��dѝ@l�$��[�HQ+azbh[rM��-_܏C��9A f��7 �%��?�0��O�f��` ��}W�Tف �v��'`� ��~�FjyJo��;;��ZA��z�W�3�`��+�x��u]w��d(��e��Zo��K�7�L2!��w'��C��"�~5�E��ر�z�L�kDǅ_��pC� ��8��g�wf<��,�N�-�]�,Le&w9L��&�*{��s��Cg��mF1v�2B��ХĢ'�v)�Dӫz?Ű��5��'��$4'\�jS��"O�C��O��L��9 +#��bY��ۇ[�Iy`� :�G��pz '��N�n?�opM��}(�^��f㹯��]�:�k��UW��k��j�1Yp18��e�WS}Y��y��C�Y�P�fg��=k)��Ⱥ�^"�ޫ��Ȍ �JGc-P��,dR��-��WS��墺��[?��1щ�މc�ѶzX�ݹg�u1ҷk�ݶ��Z�:�r�c!��HY?VmSoʺ�PA��fd��sI�-��6�3��U�X��Ж̿&�B�r�l��(��cҙX��z��?U��Phϵ>��޴0N��n�Cp&�)r[��V�g��s�kK@�D� m�E�(�ܭPop�A�~v��_o^��Ar��k5G,�/#fs[��փ�I}��d�,v<(M2��a�dH�v%�UW�S��m�}�f�J4��e��挓��/=ɞB5J�G�)z��0��Yo��BQ��c�L��{��s�X�^�{��̙��_c��WĤƞVp�e�]�h�� {�Tw�E ǎ�]�8h��Ǭ��ἂ�x��UR��K*��|_P �~s~�(�f�\��͝��v�z�P9bjg߼e£� �n�y)��$�͆LW�O>6L?�.��Y��y�R%�Ov��2�!@-�`D�Ӽ/��ƴ��_G��6e�a��X5��U�$$��gP!��,'��nH��"�)x,7sa��!�~��'��E �<CΩf!4�=�y'� �hK��׃E��(��0E�C-a��5b�Oo[��F��ZjwB`��N��@��,h5ͥ��8R>��M�R^��ײ\K2$߲r��D�sd�$m�ڻ��RΛd��_> ݭh��h��U/�fTz��7��ٹ>��9*˽@�Y較v��Q3��׺LCҸ��@$��PK[�{W�"�� B��ˣyi<}�T��tv}��ՙTpJ�2�jEIG�n/�؉+2��ᾟ�)��(��h�b͟sR�E�]G��D�(��SkB��:�7%{�j'/V�Ё�A7r�Q�Ԫ��.�伔�Z��D��vﶼ�E�Ket�J��4<��@��(��M� r.��H*�9��Bɽ3��d�42-��aǸ��;��M�;q�ҽ��i��%��f�n��nW��΍L�/A�I��Gb��A.K�R��}�{o�d0>�mG�sw��rvw�&`@j�Rꩫ�K1h�О^^�H7��Njb�]p)2��/�Q�K�$��e�X��-�� ~��Jf�4%��!vx�0J%-�ܼ%<�?��L��9�]-w2�S��:a�>}�K�k�a;�ʩ�͈}u]F��'d�l��2zj�S��V��H+5��t��d��M36�I�)9z��2�]w��؉`D�d!�2Z��nMY��SY�~�-��$�W�\l"�&k�&��R5�z�Đi`��b�hӠ$=�Q�K! +ic��$`T��b�Z��bե,��Bg`��Y�g��a�۫)��wss7׳w��% +4�C_4� +�T#�1Ӊ�C�ox#$rD4��EbD��R��c�+�$09e-��1ʄ�6�rY�e�t��UHFY$��$-��z�\�29�ǂKA�k�dЦ,��cZV��c!t$L��6,ڰ��m�lץ� Fh��i�d'\DD�B��9�|K|"� #��fSj�H��Yry9 m>�p��P��vMR�J��Q��ڦ� } 5��{FS��K� �Dü_��k�X��r-#ON�E�P�@¦^�:�JwL|��Lr'T‸�-t��a��3.#�I�d��Y��Ձc�<�*�v7�d��I�*ۨ�K��h��XFu3_�p�A+r(��c"�8ϢA�y-D_7��,��R�˝��J��ٗ�!��7��ˍ?(~��[��c[z�1�:!ڶ0MR��i�nG��˛jL(ٛ��19�H�#��34�� >�}K�M�A0��&�RuKl��X��p�'m}�5Y��Z;�P�9�/?�ݞ+R��J�T�z�D9��+�'��V⚒H�j[��6��ۊ�}��9m$��+��@�j�� +O�P�]��8u�dLNH��v��gv%4�Sj��kfg��L3 �! ݦ��s#�s��KF�#�^�I�B`R��R ן/��s�t��R�"/F�f1*��"L�C�O��B�EJ_�0��ku$#�ŢlVrj��N��q�یcn�GX�ƾh�<�\�(��t�q[M�F�v�Ǻ_�K�p��%��-�v�ȇI��O�)��I�Jޕ��t��EYP_�5�\�w%dgI��7P��t%�m�C��=�O��\ٮ�/"�kf�*s[��O��1��s�M�.�L��4�<�-4e��近��x�(Wo��T{M1�ǫFW��+3��o��[-�c&0�;��/��.��;�4� �HÎ�HpqP)�H}�D��B� �A�LR��q2?��8F�A$8�8��9�F��8�8��H�8jV(8)GQ��9C�8�}��}�D�Ak�b �'�-�E"��EZpq`(S#8� 5�� }�%G"��8� ��W"�>@��I�D��c>D��K�D��J�i��8$�e@qP�Sɑ8��3@��./�.bd�"��2�e=�!��Dd��M"��8� +��H��*��G}��6�G"�1�U�#pЪuЏ�!�{$9��lI�3@��P<�#�?J$G"�h�C�8h̑�H��P�V��G�O,��H��0��0ҩu[��)��*`��zN�Rs��N�A�V�o��V�[�2lu��Wa��*�W/��2��Ӎ>��?�VN^��X3K�]}��l��X,�+^#k�E��v��Q�sz*��K�+Ƭ�Y]��x��t��Ұ�妮�ެ[-l)/��k�c�V%��NG��K)�N��{,v��>f�:��v��~G��"qC<<�`9�}��2yO��-�V��쓝�,拉=O|s�Ŕv�m��9وD�x�6�rf� wp�x��>� w�U+;�c�q�]�hZ ��V̺�ޯ�.��{UҰ��Uq>��l�a��P��:��7�\�22��ti<�07��Ը($��]^�O֌�x��Π<� f��b|_7��~e�+��{��qu�x��4�`��C�ڥ�$û�?��'N �?��?��N9� ��롩wM~p�)�3Eh� �#*:,Clw;�۷z�o��Hɖ�8|Z��a��$8�PyS�0�B��h��Ϛb-��m�,C��Y��32gL:Ǳ�q��M]��z��phe> +endobj +177 0 obj +<> +endobj +178 0 obj +<> +endobj +180 0 obj +<> +endobj +179 0 obj +<> +endobj +110 0 obj +[181 0 R 182 0 R] +endobj +111 0 obj +<>stream +H�tWێ��}��8��d��}�]�7H`� +��z��-�"��Sդ$[� �M��ԩS?o>��%��c�FѸ�>��&�OB/7�elҢ��Ǽ��,ёN�8)rm��'��I�$Σǭ�k��O{�犟��nyn��y��J%#�͒��ѷG׹~��A�E�*�B�ʔY��2C�:5f<�w��b��T�ĩ"Som�3��]g��|֓R��E=�qZ�e��q?\"�t\E�mӿa`b�ų��ַ��ٍ�׆ܐi:=z��x��^��o�n�i�A�L��v��Leg�W60�]�#Y�yߴ~�=��tg�fj�^XL�w40 ��b�326��2.�G��xVˌ*��)n�ꇓj��Cdq�N�2��<�e�o&L7Y#n"}//��ỳQ �~rCo1ٶ��;�kC�"��i�l��ƞR��7g��<��֍�s�nT�x��Zޯ'�(��ۚ�p,|�j�Ca��߶s��K��]��J�V�8d��|r�n��v��%0�m+�Nx|��:�Mp�.5��Z��$ជi?�Z2��~�1�$��@9#��)1��s{�}c%�SIA~Ϊ*�I�撖� H<�h�#(��@�3�q݌�ppv�T�w��#CI�P�!�LS��$��qN!��G�$t��ɽ�%1�0��S��*r-��'��0��3h��4�!�m-�-yG/�/;J�-�V��e��XO4�&<2`� +� T�1��?X�%�!��،�q��#>U��ȳxA4N��n�(�� j +A �t�Π]#1h-�(�2�싟'L��~ʌ�>^�7L/��C��^��b��^��S��c�ᇡe�2Z��b%M�Hǃ�G?�{Z}͵Zn� �$E -j��Χ��}"��.��n`JȒ��D)�d,0��$l��!%��)wB1�߄g +��[��S�=�΃�a�Jw�v+�̣��ے;�)��a�<��blz^c%�i�Jn}��ȝ=?�=�R�� C��d��;�:<�l�|��H8�B�¥��B��0P�YYH�.X,H �n;�y�M��z��H�/H�un��{��2@��v��83��B��3?�o�\�1�"��`�X��IP�H*.qಖ�Ҹ0�38J0i�V��R�*� �/�3�*��J��⦉��CȰ�#�9�R$�އ��t=b&攬`��XY�g�찣&�bE��nV�ҫ��%s��^T��U ��]K��G&j�<;��1V2�}b+��ȍ�W��q��vRˍ�s��Lḓ��Zz�58&MChL*=�.IeO뷳PR��ş�{ ��>07A��w@N塁�0�B�2c�~tN�[�j j��e8!;�g�͝�q��܉�b�P�永W��M��@ �cPB�}eX7z�� k�@ޗ-�Ox��8j��]i�4`��0��Y��I6��`*U��d]�e'�kW��ƨ�Ok�m��x��4��y`�9��}<1�2��n�GޭC"�ǡf� ��f�юE)0u��k}^�f��ܵH{�kK��߫�D��7��Ζ"��m��ʩ]p�I,�"�f�#��ҩ�jɯ��CdM�+��b{��f�ǦZL{�ҽd�iv +��|-S��҂s�=��_��U��3�9�b�o|se��~�/��8��G&��A�$m�t8��(��[��e#�7�=�lc��7�ݮ>a8�aB�r��ݩ#[R��7�� 1dR�L��j�Ӯ�"#��ɔ��YR��Xw��vxM��N�pYpI5C� + 8n�2:g`~v�Z�92l�B1!�cQ�w��܁^Q+3Q!��A��:�=��h-��:�fh�S�(�w��ӲE�xK�Y��|�N��ѽ��u%r�O�Z�@��k6�a��; W�� u{h*�&E��2�t�Kэp��# �~"N�:<{��V� ��v3`�wyk��î�s��0-��򕨵b��gR'�u��1�E�k�<�=��K%��X�#�0z�6�״�B;�_��e�@7�Dȷ�Y�>��)�n`k�"�v&��b/S��T�Ü� ��#��J��7+��ԕ5wU��s�Y�zI� +��Y�^�Qa�田8��י�CX��G�_��ߦ�}2ƾk�S��m��}^��a~��'%+��2Zn1 ��4��>R�^ � :��0R^^|N�ҏ�Ky��x�ۺ)z��P4�fl47=��yQIST�ph�� +��fI�m��*E�Ƣ��(X*��O�� +��H��ޏ��n�-jK\��e#Cd��J2�Q_ݼ��Fev��**ۛ��E�?�q��~oV�f��)��_�67��Մ��:iB?��c�#Ů.��:TA 5��g��ZK�"��"��5c��t��5T8�D�� +n�J$ /�.�^��֋C67�*�!#81�L��HF΋V��*W#b��ګ_��R��p�g��T �+/�6_��(�23 l��~ y��Z��تh��ٱ��8��wE��ZcΪ�7��Js�6B�[E��'m��z]"��@W��\E�� @*�n?њ-��8�VX��H�̋��G��s��]�c��"��h�U�h}�d��($�H��ί��÷��|��L�<��;훮��~��#��_lը` +endstream +endobj +4 0 obj +<>stream +H��WPT��,��4&�V��fw�/aw%�,��:��d�I3C��jL4M��E"�Xފ˂�mbjM��H��H��{��ܻ��`j�)��=w�y��QC"ђ��;�4��{ꕗ!�|Ŭ'Fl��X��^�^��+���X+�D��F��ϏG��,+�P�C}�xNv�Dd�b��=�ĺ"��z��Υ��{�o~s�Lm]��Gmm}��l8]Q^]W_��h8m�CeuC}uy��p�q4�mht�v:�44666]��gS㩚j��p6�;��5:�*��O�9��zgM��T]UmC�s�55��r?�!0�9vN��;��!�eX�#��A -��?{��a��K�Sd��k*P��^��C{#離`��d�RhB7�C +iٍ��)��A�m�{�<��Ç��ED�e��u��#�aa��HC�"�>,dQ�S��#�a��#5Z�&,H� S"Ízmhh༧�z]�!t�^m�i j�Π�D��ua:�6t��s��jC��#B�J� ��u=��?�g��ӗq��U�MD�P5T�j��9:4��G��(dv��42Ft�� *��P�deDC�G��5бx%��n/9V`�g�1 ��Z�y��"�� C�&��(0�bKlj�w�^BN��r�r�<�1C#�.�ԥ�gdZWa�%�E��1��˻G�� Y�&7Yn��@&W��ۿ��Vۗ�7�>v��TK�/�~�%�]ƭ�^m�@��G�!Q��|qr�~FF��|QQQ;%+�e�ь��4�`r]nn�G7�d��+Wn�}2��$�^��ki�v�� b��f[7��Rz��{��{\h�w�|�?۵cu i��C��vr�坹w��5� &��W��[��@�d��<:��_ŭq��,�ۧ��Q�c�̘��̙��.��w�/��aT�r��M�x^��My��M +��"�@n|��|��+��ɉ(ƞySOn�� *��@q�N�b�t��߀C�y�c�f�J�5�T��=�q?�� 7� +x��|��{t�,�aeB�l��|��d�!�-�3f��?1��'+�JKkJd��Vp�f� +kǝ`N�,��m&%S��g��9�:�_q�2�Y�mĥ�c��Bv +��y.��d+7~�m�rZF�%��A�}��p�̄�Q\�V�H�(]��hO��Y �r*�i>ǩ8j��Ї;2i��4Η�,%4�=�k� >��H ��^秚,��R*s��Tl�y��_�h/Lo� �#V��﬎�O�=b��x'�g7��S��&�V�bO@`�<�aU��>8-&n��i@7�ۘ ��ձ��Q��Um�x�EP֤��e��R��.9�:}��M@LЁ`��9��D{m�/��WVW�O�@f��a��.��~�T��ǯD��ؘ+�1��Vi��}{6g�ْ��hJH<+ݨ;!?]{��#eԵ6͖e�'R��^D��%CJ�~�p-��r{~�jmF�,�k�,��q~¾�@�x�L�7��4��>��n�� \!��p��WDNT��Wp Y��8Չ��,��M��,͑_Q��b�Y'k 1��?�J�4��D��x�9�9��C�lt"��)��gz|�o�e�_{��>��S�X��p� 6L�brX��@�� +r�= �q�sp�p� +B�S��t-h�&__�W's��s��DeUyIEu��g32!��,�V/[��o��<8�2 �_O�L�S �s9�L��$� &C�A�`q%�� Bp)�,A�]9�D@��F�-�]v��"G��a�}�|=��*��ӕ�f��}��=l˾��2�1��c�W7ج��3y�;��a�ߺ��^�n��o�B־w��m��%6+m��R��'n�ӝg��9�M�DS��R�g.GƄ+��p��k�^��zu��{UQ�2n�N"!��J�u��Lg��Oq�);��,��x!2x625:��6_o9�D,D� +NXxD?f4U�}��@�#�� Q��r@eYK��8Ya�J>�Z�d�2-�5P�z7�Z��P�h��H ��L��փ�@�@�Igj(]a�/�m��_xwU��.I�d0�f��+$�'6��<�"d��B�u��O��MZ�,��-$�F��J��Ѻ��j��%pd�r��؞sK��n`�B�r��e��k��.�V9u��l˵AȯXsϩ{�S_�&�m��4zlYYqv:��w;Rl��w�婎lk%��w�󳶄w|��1�QS:��[#�h�#Y?T�HB�&��n�D�/�8E��pU +HR�:��j�D�8��v��B�Ld��Q�N�9N ��h��p/��,")��7��Yݙy�r�ٺ��B^M�y�)�o�$^��^�l̐5��ʾ6�uK��P��'��Jf(�s��[�"�.�X�nS�P�،OR�*�E?��LW2�HO�c��Pu�bV��ꏶH1� +��P��lu��bI��<�WW�PWs*�� +&�r��q?`O�֕ʁsiS��z��T�i�3g��q[_>A��~D��3s��-/�jM>8퐥-��r�-SoK�=��r�/�K�)��2mQ�s4מU�a�|0֘��M�-��w'�Ԣt�x��h�c/*��ʲt�O �O�H��;醏>Զ~T��-pi��c\ߨ�j7��=K��۩��n��ݢ�S�폣l��5<)U�?��?��)��?T��, J'u��]'�"�UurJ�x��蕝��D��@Ok]� �!�ܣ�#ϾBR1+D�Q��[�b��r��KY$��2��Dl �$:��ɮ 0�c<��DF��`�@�'��yK��yw��?IE`LP��l��5��p�ۡP"t#ԯ�[� �Tm�l�L?j�|�š�U�� gN ��_��ֆÍu'kjN/v�7��5׸��"%��^�H�/46��ra�5��o?�K�{��f�Qo�Y�H[86;o�� ܅�:�I��g��T��c�ԛ��A�1hB��)*ʮ��v[��`��c��_�I˸K��9�~S��ӗ�g��ɶ�P:7-�.pev`�@!�W6�<��hP�"j�B.�{�@B#��f�C,��~��=�Hw$@�㜥]��[ !��|�_3 VeH~ǂ�qa�ɥQ �W�fo"�7%H��pI�8�\VT�|1r�� VD/ Y�� n�V@-~��Y톝�2U�zn�_H�n"8�+��0@ �3j��䠤�d��p^��g�{̙0�5�|d�)oj5�C2�g��R��d��vT�@ �)��xI��YS��K��i��^��I��>(�_9X��r��S�1Bf�۝�l�dWvf�[�UUU�g/x��5�6w��]s�s�3W;3��썂W�k�%��:��>�jAU��e�]��i��2܎�f�9��V�=��qYE��YK�3�U�2'uX!?@)M��$�z>!��]>�>?�]��=�� 7񔕘yK r��.(�YQ|F��w� ��nWe��W��p�,��`�Q?B�Fy�X +�ֲ��̜0x��̗�s��{�&!$H�@A.S-7�bM�m��do�e�݄��a�ʨ�L��N�#��[�Z��cKEM ��P:��H��v��s6�ց?�7?��>��<�^ �}L�d��e�;�F�,}�_��|t�R ��om_%��#<2�J��Hy ��tI��8Ϸ��}�x�?z�F�B�H`�O^��ސǧ� 4��$��~c�,.� ~��bʼ<�5c��Wc��[�k�2��T��dY +��E�ÍTi�"l�d��ݘ�$�j�kC$��>A*�(Xwx�L} �D|�Ѣ:dq�Qvfg_�;��oՓF�ȱ�˵��h[��!Ҽ��~Y��h�m�4�ij��U�GZ�#ծ'�q�hS�7��7��L��hj�v�2�ds��$b�f-��S��x.��7֭lH��[ҝ�}7�"�BPĽ)�<.Ku !�b��x{�7̵jp�� +f�0e�@�\��'�p��=P�c��7�ES��oC�*��$��֥��^1+��c� ��<�Bo��W0/��:��!)�f�㐧}y�]J{��U�|s_[��ȳt�_( r��1B|m�3׹��T��|�#Ȝ3%RZq�;W�{)^IE��%fbKv�p�Hw�� n��\&"��k�G�ɡP +=��RB~�\��'b��OO�+d/�?��OF�uy#y+}4@��QѺ֣��>"r`*c��EB��xa޹7��w��ޙ�d͢G��O7 UˤU�0𓮥�)MM� ]׌�֪k#�ɵ�j��#gjh�ѳZ&k�9��l_��GCC�l�=��U]7͜��A�v��S��qH߶!_��J�og@l�W��r��<�>z��Ԑx��3GOC��~h�T��"3��o��w, +��7{E��I�Qݔ��8��<��x��{`w7r+(%=y��R"ǽa;��`��_��ôL X��X�pv@J�>�v�� +�<��3a +$�S{�X�U��.��|��p��0�铘��gϞ��>��>�| k��J~&VE?�^�'v�LQ�j+�n�S��(�kD"�0��3o��by�["J�z��D�}��]c��]�� +��Z�4�xV7�\4�|$:�D,��I�ޞ�Ӧ�i&S��ި]��LJM��65C�2��;oZ�/Ak�[c�N��P�.�+�Wc�=��Z�<ȭ�?�� t�l�� 3��@��B�͟��݋<%��y^��ki^�:��y�0e�6:ea=�8|��-��W�@N�"Ny'�x�Nf�>��D�]��:G�I9�\o ��o� &~'�`}�E��@!�;\ʦ#>��Vv?(�Xq?@�C��S�s1��u��}��AFn��0�� t��jI*��+v��`��ʀD��?B�i�� +�n�M�"ŮR��IAI~��0�H7L��78)h��A��ww��g}2��4��Y|�H��=�ڜnQ�jZm�2��a��&�d[mKC]�EMG!ߙ֦f�D�ds��5U�MK��bZ��6U՛�k��M24û��GTkWq�S�kk�ì��e�� UhQ��%�C�v{��¼Y�h�ݚx�x�_(Dw�r�R��3�~�S�?L�~�� KPRׄ��9lc�C~��O�ʘ�y�|r��|h[��@D��(�Z��^D�:�qo�� >F釡��D�kl�K�z�� ~�{ָ�� Օ�@��"��/NL�J�+�g�OcE��E��D +K�?��c�L��~��Į`^BW�]��U�&H� wJx��՘˅rB"c�p�#*��տ��j��f��"�Y(��i�]m��}��I5Œ� +U��7��"��$��z[$۹"�3�\��֦�:R�x�H�˲�X��x�5�FE[n�Knްw�W%n��/4B5l�Ӷ�Rxd�Xy�Kr�Q��۟�(>��U�,�e��R�� b��c�g䣼7�;�� S�j��\M>$��I��}�x�w�s�9��~�G�(�X�a�y��+�YV^SSq��4ˁ��#�9�/��/�4�gŌ�y׫�lGAF�ww꜋��c�ޖ6��ſ-�+Οۢ}R��^��tQͫ�UK��e�XR��;r��vT®{:]D;fO��.U5�� Զ��k!��N;\��z=�]ݸgh01�� dch��I_��x�(�ih忑�d.�;�_�X��C��E6~��]�J��[�8F9o��#�0e��D,�յn�"��1P�c��f�$��h��e(�6Z�ULp�lQ�@��U�MB,��e��pE��{Ї�+��ܡʪ��Sk��Ŏ��o^��:��n�Ϧ�W�y�i߱O��ⷢײ�㮶�-�((L��?K�KUyz�[�S��vAI7��آ�=�K��L��֋)��ܰwG�q0�=ԧIGB�6y�� U��fI@�"7P> +r�v��i�� !#�@f��$��8T��j��&�z[��QD �Gꬅm%~M�UVlG�5p�+)��y1��@,ǹ{�P��@��<�u��V��ׁ�1HY��E"��@\�Mb�Z��FVr�i| ��^b��X?�i�~zXhdDd��;H��߽�x'�tI��Pt��L�� %i�;�mSF-�׵2��_��S�?�2߰��Ύ}�=ok�rvr�ޝ��_P�6�[�񸱱m4kZ�;v��Q�b;;��5�\R��:tĸ��8��&�D��=��m��[{��Q��m�4�ڎ:��䍐�鼿�5\1[�xy��~1��/��ǖ��\��D^o��" 1�aN�um0P��#��[��]ljj>�iZW��l�'牐�9N��7~#��Ǽ��-TW�=2Du��M +,O��z&HR!R�`< �Ur��+p3��磉dX��g�Y�Wdʾƌm\�N�M"AW{��<%_C��M��P��Do+D*TϚ��u`<�6��~�l�7߲��R/m�?�@�ܓ��Fa��T�� ׉��Z�W��zUF^A��Q�B� +`|n)�L�ON+��)��/�L�g�oQ�'$�K^v:+��̢�9�׷4=�eǄ��رx��J��ZN��'7�u�ak�m�6��f��~�C�f^��ZH��d2�Q��/on^E�R�7��>�v:p��l��ﾛm��@a�2�AyY߸�b��d��1�n��N�iG��x䱦ʱ�Q��0��/� ��-��]�~�38{�ao'G�S&C&�'J)�|e��,��9��G`��P s�A�P}+�"~ +��`��7�m�0�ĕƯ,6#�o��[ p�沘ӳu�߄ +��_�;��I� 6 ��k��Ҙ�c0Q��IϽi'f�$��H�߇cx��x r�*/��9yfG^n!��Ho!|r^��i)�� /�ZĻ�2��s��lSBzzzBj� +�P�m�UX�(c�y��x\�;�� ,�Qu �h^�@��OU�O�\�}M�8�ߡ�.��7֯t?�8 (\��4�{S�x�3�`EV��`�B�OW�F��6�Ϗ0&��&:��U-��(�1��8mV�R� #��&E�Pf��Sy��k7��V��a�u�c�KU1�"T/|;\�ۨ��)��4��bv7�ͷN3 �H��7$K��}�O�� ӱx��_� �5��;p��X�ߋ?S�,ۉE,�fx��9ˀ��jM��q�~-Ax0B�Rv�^��a�`$�D��ɊIz��l��I��j1�\VYu��ꕩr�"��E�Ù�b~��%K�4��YMl�f�圄��չSޢ��^��d��$3��>��a��nV��/f;i}��}jL�/Tv��[��3�q~}�X?�c?�e��ЯcTD��y��}�!h`�� Nŭ|�%��5N�c_� +ܝ$$�C��H�M [ȣ��EZ��R(&v� +$��*��*��,�eo�#��ӂE��@�'/oK� J��aƎv�hu�ŎE��C��B aK�D�JK[#��5!�{��{�=��{ U�2C��s��w�<�� =XOG�M��l��a�� l��Ӑ�<$��K��4%�� (��u��S�#��z�M]�L��\]�5d�� Ob��O�7�!x[Y6R�(��@5k�+�� j[�˳�9YT~_�t�&D|�3 0��8-P��S�%0d��?��{��*�l4(;I.BG��h +��PF�5]̟κt=��)X�~<�F�p)6��Rb��2B ��p�v��U��H�M��ՀW�p�M��X�U"YP/��+$LQ�̵]��me�k��Js�y�J��_��q��ç�.�X��0�"m�U�J˧�E*��Kf�(.*�,�QY[��jּP��n��H��&i��+�o�+�]]5��hZ�U��P�b��ʲys +�ϝ[]7L��?<�Z��GO��f�n�|��)H��rZ��̀��\5��4��kDRH�$i�#q��K!�\[�M﵆|��v�pȐ�;��352��P��H�W,��-e��INd��@�̩�T��|k1��k}�e��ṹo��g�d��g,6��h��LC��?+��Ii_��wհ�0�R�xR>7�iϔ�_c��l��z�/�L� ��Qj��f�"�-ď03��7w�x��l?��p��7s55�@�Ģ�h�1n��M��P(5E�a�6��b��h4��#�Hk��k��KS�_4ԇ�C��)�ɢ��<��N�E�^�7��[qU�k��S�9~��KgצX�[d߄��K{�GU��d D0��@-�f]X��92�{�Ι B8܊��U��@X��XFr�*��*�.�@Q1�I&ɼ�{�]s��'��G��6�� +�[M�m��&(�ݼs�3��l.96 �ӈ��f�@�=^=�L��[�"��|d2�(p�dP(PA�W$�*�ȇ�h�@�GF6Д�OHY�8]�(ϩ-�Wg��S+qHBfS�DԂGЂEM)�~��aK��jě�OJ ��J`��Y�� +PPvRz��xx�{��[[��(��d1��d�UzD|�/�-�B��Pfi�AӨ7"h� +f�*1"��_��d2�Ez��T`2 t! �G��`B�h�.,�bA�^��\@�Љ�Ĕ�GrR݋_��o�\oP��/�˦=?��<|�%�m� ��_|"g��5��|�**�ɱ�Z��y��粋��4��~��9g�+{�9Dr�l�X��;� ��b"�� ?�맞 �0�u��p(*mBσ��r��e~�y�>�� +ǋު��ʪ�3�s��裩 +a�v�1�6�9p�� a�9��awZq3n�c�Cuf��v��Q�'P�,��6�'0��<Gj��H|�d�̉o�6,e��3�&g �?��[�u��"�~��ȩWK�6��Η�h�ؐ��I��F�|i\�#�v��%�F-��~vbt^�ב�ўWRǾ�� #S�O|�|��d��`R��t��D��J�g��*��Jz +�D�P�8�Ju��OOTA7��D��}P�5E�ʤ'Ϭ,N^g��&(��֥(� ��?� ��@e��jo�ʲ�s-��lԀ��,;�v��C��L��rQK2nO��%Y�òl1�fIT�vQ,�]�#��.�Űn4��f7M3r��C��.��]�ď#mSg�9{D9_7>��اJ-X��?��}w]��A��b��$��)#�"ݚOOZq�i��`��հ1+s�냲�;Z֮=U��͍�<=8�,'��/�D{&Em�zS�k�, �L�K��a�su�A��~h�E&t(l��b��)�Eɜ$I�;�GY��Uy��"r��w��7X�]��:y�[q��waa��l�*I^��C�C9i��p��AR��KϣQ�a]8��,I0F�2�:��+�hԕr�Ő��dH�F��I�9o8�8�� 4��W�u�� ~�{wF�;K�ַ�|3�ٰb]ΐ7��s�t- ��!\��p�𒒴9xh�Ӏ+I��|6eg��]h�h^�uJ�� M��4g��Yn|�xۄ��̨ځ�0��WФ�E�(�;��F�A�(%�+/�f�]�Jz�@W_a� �h��P��DP�*- ~u1�J@W��1��u��w�t�`��C5U5uޚ��L��,�L6b��i��$=�(B2K��XC��4�$ I:H'Ab8M6�&l聤'N�,B6��樎u��Ј��8ۡ�� +W0K�$�l��7p��J#��>x�ߒV�NIX�%>��'�� [6�O�D߯�⇿�'-��G�ы��k#c��}��U��4��!��3|�;��K��k�p�&tB�ҽ`~��>�4�|h8h��N��TC)b�;NV��"�Heٝ�yY��a]"��}8�BZ ��=^Ys��~2%ef �f�Io+��˴�END�$A��(��,i��l��v��qXLV�i*�ϱQE��X1AY�F�IRH^P��s +� �r�``)��Ј��6~< )u�@]�G��ucW<�vV��SF�Į��a]4l�?^�7�M"�g��;�6�4&=qt��>��O�JJX)6��㠡��2n��a��/��{��1�f, #ރ�1�)��L��֙[�=�!�rJ Y�*��V��@He #�\'j��O��EE��<Д��O��5Ue-'@M�Q�v@m9! +9((d�� +��h�'5q��~Ş(�A � �r�V��V�8z��C~�ю�l�X�s�.r��FdJ��3��H�,o�=�n��2͠��e��`��!�,q��Q��f��̐�'7|sO/o��5pcg%ڱ��U��߶�R��`��7� ��?[�S;��n�|��⡖�E�j�^�� +@�8��,��ӎ[l�*�6�~^R\[;� |�qW��1av��Z�|>_K]��dg�9=��,�»-��&&$e�,t�d�3��iδ��쬌ͽL��w~�97� �zJQ�a^Mue�3SR��gf��w��Ĵ�̬ܔtW�ӕ��A�� +��s��lAe13!wX�p��ljqLa�ɠ�Z=��P( �6x ;oh,�B�:: ʄ��p@��*�L�8�-�<~��a-)�O��[(Ev@�iH�C9�S FhP��e0A*A��.tc@@ o@%B��*��0�i��31 >��!�f�P�ݩ�cb��-��$>�$a`Fɮ�z|[bNr%��j�TG�H\�E]�e��Gj��[W:e��I.�c�gܳ'{]I{�7��-��/�Tv�K��s:��E��[$Z�/�'�,�U��K��O��H��}n��2ѿ5õ��5۝;��-+��!�B��( �� 4�%<И,xY�ty�3��,\B�3q$�� +�L�ۖ�D�.�b-�o�|�K`.8Vq<6��r�;f�7�^�{+ԮaY�{CX";�E^Õ �B>��G�5�..��Z٩��P +��c�M�R �6qJK��4#Wp�T�&��|t�~�6� +�D�� +�C��h;ޖD �q��M\r��9v��`4T�a�>�Z�ofHq07f��n��u�-,��{��]H+��54�$ՙ��S3R筥��,��S\��e/N=�-�-��Cٙ��Nq�y��Z 3ƾ�v*��䨑g�]��yE3��[��|�G��y�.q��S�~1��\ 2�x��2Ā��5J1p\�0��K�C�A�A�d��D�,u$�%�O�� I�~2��Ad�[��D9jG9��zfFa�QU�a��`K5�v�ą�q��N-�|g�'�ȹ��R��_��n�� {��L��7�u.�� p�gL-1�t��Tǅh��L{�!N��+�;�ۦ�q�^� ��qI��џ��f��4\=cv5�d��^xo�%��q>��ص�?=��l_~��7�@"��V�.��zi�{�w�L��#�W�z�ĥ׸�*�n +��=��q꙾��DM򴄏�~��o�> �FkCc}��勂�)nı+>� u/��e ��"vdt��Toڢ �.��;�__,*���d~��r�3��2�Y!��77�}�X�?�T�xk��֏��{��ѐy��b��Ƨ�Y�LX��R�d �Ҧ�P�N�`�P��e��0Y��{4��b0��4�� +C�T#��ډj� �WU,Vǚb#y,��X�`'D%�O��!$R%�F� ��@�J��F�*��G�,Z��h!�|��{~�^�{� +^mll��.L��-p�MJ*��mj��9��M�p�\�+��M�Nqf4sZ��N(�?yx��A3+V|��YF;q�coi��j��3Ӫ�K�Ʒ��ɉԺ-DMF�B�1Ρ��jQS��P��(�u ��uAD71�� u�j�ς�0�3��Q��N��į��`�s~&ھ��L�<�5�L`�ɣ�a��GF:3<aO�:T�ٱa��L%;��(�1��[?W��]P@��>��״W��vb��QHIo8y�tA�h��!2��{��~�1��tr�0҂�܊��T�[!��dDτ�ٷkJ��D��B:��4��9|%��Q�+�}*��~�iS6�?��q��9*T �{��0"n+�>��D�ƫ�v1�l��a� �m�}QS�}D�e��݊�� h�`�d�b]}[C#��%��Ĵ4W�KO�l��٥�O��k��{]Y9K�/׻�=Z�ĕb�{��GR�C{�FEu]q��{gPZ�$V �+15��֮��|D��130��JYQ��r�j��F#�D��-�� +�@eu�s_��ދ��̗�0�ܹg�}~��r��34m*㲬6�`�Y�(2_:��z��SWo4Vf�t��x�� Ozz��N9��*(��s��%R:T��ݮ��V�@��~/Nǫ��!�]}� _�W{�=�G��%��b&7�Q%�H��pd.R��'��U�a��Z]Hç��حW�)W��c��z�J��ԋp�O|^uX��} @��X{@��*�U��ZR:��ZB��l�s�PV�i�R�TB7��Ϡ�\Ą��;(� +�PΝ�E�1�� +�⪡�_�a��+�HDZx��T��[��A� |�y��R6��$@z��ԜףB�f��s;G��d��S��nojlilkڦM�� +��gt �O�`��\��m��޳�d��`O~;� +�r\3}��W��j~��oн�d0��VM�Fr��\��'�<��!��UQ�ۇ��Tv*�E�T�^��gN>s�� ~Qs<�s��_��i/��En:R\�=�Ow�RIQ��L��Z��ʶ�.�Sp��y��Z��%wyQ��$2y�pd��u�`�(�ը">�b��L��1B{=�4��+ ��&l��am>�/�{D�:��B�_xԨ��y�:H�� z�d��]�Ӷ�x�web^�q�:��F�(��p!%}e�L��O��/ʯ�s�;�2ۄ�� 0l�j�*�o��Ǐ�I5%��*>O�pGa�_:�-��t+1^Z'�/4��]��Љ�q�S'&�Q ��^yhX��A}QF�s:��| ��f��<�b�}�U��RiQ�4�b'�+�/bJ�D�*��g�a3�#�e��[E�7=��E�^��Y\1,F& L,D��c�`#F�J�\�wO��/Dk�6y��z��߉�m��#v�'@�*��0��J�;��P��ց�W�<�N\��l��F�2\��EYt�� D��KƵ^�.��h�@��4O^�Oň~j��(̬��_#��ް��ۚ��[;�7�ic�ZӖ��+1�IK�iizݪ4��`�ȴ�7/KMM3��d]�>Yo�OJLM�ee&bb��S�u�$m�>�h1&�ЧĦ$�Rlf}\B��dH� ��S#Н_�H䚯?<� ��e��D"(�,4��=G��zW��Z�$:^V�'�>OM�E?U.,&�z��E7)k�+-�p�}weH"�7�W�U{'�fS��&�\:pA��Z��%�`�>D�m@_fP�9E܅��`z��B��t��،^��H�0�q��Pد9+-� t�M��)�IF &� S��l� ��-d~��O��w��P�"?�)�^fȿ5�C�"�d�+�%��@*"C��B^x2�ݳ5QQ;�F�kɧ8� '4��0�m�N��z��B!ć�U�3 ��,D��΋49��JwP�7��+�]�@��<��8pu�66775_�7�ط��$X��t�ѐi�[��5��L[�-ݘ�a�ed��fe[ �,��l5�%i-6]�9͚�n[k��L��4c�Ő�b�[�R�S2t��\c��8~wv��B�$MQP�"E�� +4{wf߻3��1��T R�D�*��P��C�6UR)R�BR��.'&%(�y(�5��ww��s�� d��;w��9��B!#��W'm�7�u �q��Cǔ�rN��s��Tl�(�1�8DT2��$�2�AD>1 !b.�(ւ��!b�I�B��)�E�=ͽSQ��)��_G��3�wų��_v�y�bZP,"4k�GY��J�l��k�50�^��S�C�z�n�?�@w-T��F%H��h.%�P�b��.kE�prN�F��*�0{�ۀ�*��j��U=;�g>�:�M]��Pś|�� +��C�gVyv]�z��D�a�H�c�� +�C��D~�<��Si 5��j��\<>��h�T�bB#��|_�+��#��5�ߎZ�5��ؑ��[�ͭ��H`��h:�Ѝx,�H�b�d��-�C�5�� ]K�F4��Z=�lK�H*WC�v#��b��ƌ��$��\ZH��g��*�cR��M��~̭QKo�Pʯ� �py�w\V�r�/3�VSS7��[��/㿨�xxU]�ī�{��^��Q��;��Pj��1�sJ��G�vyݞ��S_a/O�?�P-*{� +;Y]y¤�/�eh�fʗԴs�t��ƎW!T��ye�j͒�@�x+��dK��m�s�++��]��P4Wz+�K��T=x��RV��!lw=��6TU��!^V��ڭ�~��ͱ�9�U�9=7}ΐ�yE�A[p��;�;� �X]��J>�Q�*��r�-��,~��վ�Gh�>TQ�a)��I,R�G�UZ�:��y�@��_�n\��#��_�~8�iAU��VMM��ԓI��7�:6�_�'�k�i#��cQ5�M��ē��IK�45�L��p4m�#4��d��xg��G� �\Q�>S4X'�'�C�7m�Σ�a�P�AV�#�A�Cv�'N�т4��V�e��ل�v�EW.A�:��0°p^�_��h0z��'ȼȎ� +�(ǩ�,(&mr�&r#��.��yz[��;�x��9�0�O�(F9?{��;��S�s� Ld9��}��pp,7�p8̳)��{Pv� ��t�pN�a�cy��ܦ܃G/p1�N� ÊN�u�F1��]��ٽ��G��5�h]�ՠ�!�@@ )�� = #{83?M5�#a-�Ebj4�oٚi��Ц��SK2�?��Ґ ��ё]'��xBvݏ��]��@*|�MM�M��Ԇ�f�$YR5,s)X�垙鹟aY� ~��g��0V�qg�d+�� LZ��T��+�dDW��Y�cPj��s��a��o�m�ÔN��" ��NQȺqV ��rlq.��r p(��J�S��63�o��M0}ބc��M�@��Z��ڕrnL�R�,��h�L��K�� ӛ��vmVkU�p8� +�x�B�` ��z< �h��k��?��H:��~��Z:�a��#?�-΍��~Fv��짱͟X��p��9p׃|�~7RÓ��%��,�i��!�p��"CL< +z,-�\@��@��al[��"�S� 70!�a�.��ha@��T6��C<�"��9ưZ�"�_�?��ǄYPZ��c8��qXS8�m��Ya�CQSТ��t`[�/��D�rN +6�e��UТQl��A!'�)��f4 ��Ą�ƔwF&�׌�C��Lf_owO��׬ln +�#!P��A"�� s~3��6�CO-ys�K��&�� +��vx��N��m��3t|��s�m� �h�5�* �v �%��a��%ҷ��L�[?~ጨC9��A�E�,d} +�l��J�u;"+ h�bq��U��^t��-dYp1?`.�(swr,w�'�®�6�\�)�L>b��mp��}��-Q�$@�(y��bl�\�wV��q؍Ą�^��+{z�2{3�{��~n\�� +�p8TA`�!M7$Ār*�J��3ذx١��Ñ�lh'>�"��v=��f[x�/��5|�y��(�k��M��-��w�I��_��$l�R�g��fva �y�+�ls��a�4��䕮GJK�>\�*�ssY�&��%� 1uݍ_�r��@�~� W�t�%X.'O0��j��Ɇ�A�lg%�c|b)�Z�TծC�=鹷�B��5rz��3��'��W�7>�6��HHm��dB�A[��r��:��wwAP|�jb�I�<4mc�c2�Xk��>Xv4jE�+�$�MG�FMk�c��h��I�A��ƀ��@t|D�e�ݽ�<��{w��L��ICg8��ܽ�r��iI��`_��Ђ3d�åw,��$�;��6�C^�"k9��Wl]̧l˒%+��ȿ��}��M%T�X��O�5H�%��fE��-��z�Y�n��*�v��!��X(*�j�C3��)d0�XP�DAhI��T�S��թ ��E��Q>8 ��D��e�`�|'�^�)[��v��US�P��d(��a�l*�L�M��R��WN߱�?Zz��ɻ�ϻ0�h1Z9i��\�;8�]�p^ސ��wւg� +��I9xS�ȿdEw��T?�C, ~H�7��:C§$(8>��$8(V͔��E�V&!ʐ$�C�NU�QC��:Ā�s�vK�l0n��x5��7,��YOKMC-Ѕ��:W_{�|�줔�t�^o�N�8,�6;h�j\6��l��@ǯq��ͻ]b��o��`qfG�Zz�R�~g?��X޿��/�G��}b��;�(Y�M�� x�z��y�rAd?��90��b�Os�e�:)��wE��$�10w��\��ȝ�L��dEJ�0��BL=��]?��>��rV�SO�)��Q�4�x��;�F�a4n�<��T8>@�M��ǒ�u� +W?Sw $p��U�f��+��κ��tU��U)q�2�&��`��v�Þk�,��u�� +��%6I]_3��̹hnO_��)=Z��,��O��Y��Q��x��b��k ��}�c?a'��Y?�5��C�� 8�7!�0��D9"J<��S�z"�� <&��I!�R��:C�} +��o�Y��nHć��5�qs��}҂7��u��Y]SY^^��.޹V��f�3[�m�� -��s-�>��n˶:��9+��ɒ��(�]�w��V��|�)u��͞ +�`fw��%1B�e��j)iG��jP�n�h�*��('h,ଯ�n�EH��U1�9�5�~ +e�+Hw2зl�)�&�/�u�(��`��D-�z[j��Fw�|��"z�MC;d��'�j�zR�_��B��t��|G�iE�� R&�� H��"��J,Bdۑn7��{!�)�G��h#o�X��do&��O �2�� e�o�--/�v9++��J�+++��̌��,�^o�5��S��2Ӑ?-z�єc6�Mz��Ŀ �T�1'[�m�N��B�&��D��s�$ͤ��'��h�]mE��ן֡N��R��H��M(��paz�.��0�-�iA��"��W�X��,��!�'ݻ��5�Q.c�p��C�W��u�ڧY��E��9H4��]��ϑn�ǋ��W��C+�o�[��o�(4� -�􀘓��7u(�a��;��5�b��w�]��9"�G�0w�b��)v4u��Ymʴ��!9�;�KK+N��+�O��.+��۷��rq��e�E|��$�Uө�N��faY^&MMj��,$kZM5ck�4Q�}"�E� +��(��X_�mՠ`;F�e_��wq��ݽw~��=��5��M]�l�e�Te�w�[��5<�@�#|>\�x� +��އ�6�sU�K'%En��1��>�� H'ً�5� ��Y\�Ti��6�h?�ٌ�@/%S >�2n��8~��a�ht#I��-n�qt#�O�^bl��'Z��o;ж��e_뾽{�[��7��܌4�ղ<�bY��aY�l�%--}�E�l!�l��t�źB֌��4KzZ�҅��Q�MƶP+�,x�7Ar�N�g��(� ��i&̺!�6�r\E^ej�L�m�($j.�sw��0�'�D�g7t �&�|tA��O;h<ưzq�|aX��r�<6`�/��X2�]��b�ҡ9��Q��N�\8 r��@J�c�L��6f^�~�s��z��&�(@��8��w��<�X ��1�l��?��^�jo�K��kp��S�`��݋��ܹ�U��顃 MM{�6��654�i��Y��l]�~ݻv[^��n�ϳ��_h˳��B��]`��؋ּZ�� +�� +�ȧ�`�� 8�U�^ +!|,�N�V�!ϔ#�6 ��i4PJf��.MD<0�;�msB��U�&8�T�q6�M%Ԓ�uϯ�:��V��rV�._�%�ҳ8��h��D87��\��.}�8�1�D��HDq��D1 Na24q$pc��5(MI�� +gqЁ!2~��E$A�B=�ŕ0_e�-4U�f�$Le��f��{Z��ikinln��k�)W��VUVT�y=.��p�\n��u��>w��W�r9K��v:=>��(.��m:��l�F�h��C�b��§�� +? +R�&\R��~Qo:f�"+Q� ɣxHJ�ͰU�1u��9T�R��*��H0nF��c$��pW9�7��Iв�x 1k J䘘%�6�eD.gZX�� +! � Bѕ�?6��(��E�S#Jta:Mb�C��9��0d#&�!QȖq�VC�g�0��p��H�#r'�ƅ�4b&L��7+�?i�pgMc��uMMUEyyeyEY��Q��x��;<"|^�vw��ص��t�Jw��-Ǉ�<4Ez?B'��w�cL��[��_�9�t8'�7��aA�B�b�� ?�@��"��o ��f��žoǑ��tǗ�g�^��̎H��52��.Q��B��x��5�^��h�G%��~8��q�Fb�S��Q��-�J�W��o �Q "@��E�`�@SI�>l�0o.�y��(j��0j#O� �@ +|��\�,")�-≐2�$�7��Ó�Ƙ&�}y��_�\_W��ƿk��reU��u�K=.!g�^��-��r{��m�n��+/u9}�>��%��Ef�+#��3��6�ȏ��F�ܓ�|K�S�Q��ak"4��I ��6C<��L��d؊L�1��!�{a�� +��ǃ�Hc�5�$/U�}�cH2tEE��;��b��1�J0�6cw�� FCh�N.ЧGՁ} M�t�(�"��o��x�50J�L��%�{2�9U�Q5�5�<��`"� l��w9�F��a��*�B��2}�1��p�hLS7��+K�+*v��J��tv�<�sL�^��UR�ԩBHz��A��{!�ر}��A�u_C-�{�C:5(j�y��ۦn�}Bx��O]bUtL袥��E��:U�`צ?�'x�*WYg' +?��9�/( �cm�WBt� +��r�{�7��:�7��ޮ��q�h+!$j* U_*�}h+��B�e�U�R@EM�cZU�S��7�$�� {׻E�զQ�4��)D�m�4�J˥7@��]�ok��B��3��l�]z_!��?��/��]W�֔:�x"�ٌ*��[��쇖�XY��]c�"�8E ��ћD!��ۓ2;w��?��c?��EY��sP!�~�[ +K��A�T��D� �wO�(S�]�I5H\|"�6�pNS9q�/��N��?;�E�]��m�o�8�]7��h��!� +��34��h�Ȝ��J6~*�3�"��LUЀ��4�Ǵ��Q�� U��1v`��Qm +T�&zZ# +g�Φ��T��4�!��Q�$'��}�i��i�h�F�*c +?T�g%�vm��D$��g��P]N3UaS*Ñ��٪�lX�S�`T+8��)L��TQu,J�dY��eǕ��\�k��<2{��5�~R05=��ߋt�o��hĠ �J��(�Z��$%QDY1�)��Ҋ�VХ��sq�hG��4��o|N-�A+Y��8.�G�ֈ�j`+��ˀ�m�� a�F' �> ցh*�& +S(?9�A2��`z��Ac8Xe�;��כ|홏��\O+*�N`�յ��j|�ֈLu��tT�e��9ݧ�2di�a�|�lBS��4K�Yc�X>g��b�$�ҍ��K@5k:V��L�G�nܺ��ة��Xl v*��?080�t` ��l0��?;|�Z��2�%�j��e�7CA:�ߢ��>(k��V��$�P�X\h�.��ߣBsk�D��Q�wr��\!Qd��2�/5dh<�ы�d�g��DI�_~e��z��%+HC(#��PYG��]��}gH�p�ݧ�I&D��e6V~]ƥr�^ ��Bq��m� ��4��9g�B�Za��P�!I�L��R_��N�g@�<�w}H֨�9��{OA�2a��k��IzR��h�� 2}֩�L:��>��4a &L�0a &L�0a �P�誏�M�'Y��g��ߞ��|�̇'�ͼ?��\�X_e(� SiC7RB�!W]��U��z6��YS%E��9�P�ͱ�"ri�D��"��fc�Y��hv,�2x��Xza-��x�c2Ŋ��W�\*ώ��Ƥ�!�L�\��;? +K�tAA`��l��I��YL�?..�f��!�M3]C�Ј��M YpL��b�Y��Qm�`~y){��e>>��B<��R��|��d�5�o2�6��/�,t�066�D��^|.� �9�BICBG��?80�Yj/�>�Ӄ��o޾�Ήh��l�R��<�?0�ŋϟy�d{��ӱ3ٙ�^%��;w��n��?!��O��? ��n�n)�`(��Y*j��Cm��a�^Gm�s_��J��on[�a��͛�ƍ7-QU�zq��M{|��U-5̾��.G՞�ʊ�ʪU��g��?�ct��;��w瓡��ի�V|\�~��+�( ��4 Pq�v��p �E��T-�h��C�p�02j��m�@�e�^XX�8�f�+o�S��6��ym;?7��/|~��9��p��{��!�yµ��D `,=1!�GǊF�B�Ū��qֻ�Ng}}��Kn��#[n�f�=˞�Zw��ٸ�VfC��܆��eV��j��,/��];�c-��m6�� +G��f-�Ufۉ6O��ʲ��d�"_f 2\��Tw�+�o��ml8�i�z�= �^n��64x��a�&O��x� ��˳��D��~��z_�(A�b�P��f��|"T"f�P�G�U�H-pl��b��A��@��V_@j�|>1��D>��Yh�_�/A��Kd�9��P<�L$�J��p�@|854�H��n�D�l�y2��ur(1�s�s 7I$��=�� ߑ)��s��D�H�HO$� +�#�pw�+� �"�pO$�� +��=��h�+퉄�=ў��QTv��tGz{�Yo(�?�-6��KU��ڼ��(�T�mҋ��~�Dj��Z��z��6U��6�� +^�6`��AA!�� ef�ؘ�̞��^_B#U}��v�̜��|��}�14m��M�{|F��|�m}��6k��nuX�6��i�YN��ᶛ�N��f�ڬN��n�[M6s��as��v��a�9Mζ�}V��d�{۬v��c3��hl7�|� +��L2��tG¡��p�'�� EB}��p0�D"�Ho�N��}}�po�/� �OD��z#�`��}�0�� #�=�� 8�B��Pw$zz"o��/|�1 [�5]�p3L-�5%�h��B�A�I+�RQ+� �U�}7�t��Qլ�ZqYi�mPp �BQK�m�[t��}歮��!I��QI�8A%p�x�� $��d�L�c8M&�4=@2$A�'��w�.��5n](�.��g�z]��vY=/ty�N��ᶀ��k78y<��=�g�q�R��L�X<�1T�?�� b�X<1wPB�'��h2Ѩ�E�x�?M$(�r��x~�>��h4L�c��h<��b��>��&z��1%��U��b>�5E@�� +UFUK ��e�Po�R,��2��[��w�[A��=i��nut��<�n<��T]y��ɨ��y�ciN�I��H��)��<=FP� 0�1�B#A��w�8��%hB�c��H��i�ch��0��x!}��`�V�?�co_*�[�u��u9��f�x��;�~�� Xt��փ6��m�@��6��1��+��k��ww�0�i��v��x��R��YF"8B�)�!X4Ű�_��8V�9��1'��֔�_�A Y�9��Аa�b�f� 78ɱ�9��7 ��0��)y��{� +��ԀV�@yN+h�p��n��SA1tH��dk2(U��z;5�H��瀑��y�s�[嚎�r]Ɏ��~��L�ǳ��*J+r��I<|##a%��.�$��zg�\;' +��N)J"G��]EBL�x �\feYN�X�K,�G.��_��hy _��}U]VJ�A��6v`�e8�_6��i[5�� p�Rqe��E��iU��l\(~J� �R�02��<��aP,��Z�?��9��̟�zja&=2��iF��pN�}X��C��g��ؙ�� F��( e�|��)!>z!E��g��jv�j��mYWw��kb��ǧ��_��r��?��7��o�Z��/�<`�9ZB��m�8�g2ljy�h��_i��ďa��;��v.��XR�L��G +�3S��/���8sfFd��+S�d-^;��J��S)y�b�:~�V}N��;M��g��1~��ɄZ:y��2�{2��L��̘��x��G��4p��Ԝ��K}=\i��0--��A��m��&V�j��5�~�^��^�m� 7 +��r��^jl�BcSx�u�al�>�@6ڑ��o9�I˩4Pi-e��(5Y[[CÉ�*Z�jQ�+7�+huy�_��v�L��V��݌0$��%@BRE�� +*��QiH�6��锐��r�;��2��)`��ՙ��<��4$}�'21�t0��;�@h�E7�|$� !�d��1Ģ��C�Ѵ�CiSJ_��4��Ĳc0'r\�.��p4��>��tp~ x�Y�KW +k�< +ւ��s9�6��9�2��T/ɞiQw&8�!��NJ��K�D��-Ǒ��$ �a�IBO&E-��)��&<<�P��a8�b#�C�EC,%�$DG!>�X��}A�DcN��8C��hNv�`�9xIf�Z'��v��}J��4F�Ҷlk��j�ĭ�5v��!t)�� bX]��ut��C�U쨴YmB&��.�/� �)v��⇜raΞ-)!b�~�{�X�W^�O�Fw�Ijz�04n��;?0��}�r��ށ��;�缹��C�[߆?̪�6|wNn* .�W??�@�۹9� ,�� 4[�FC��[j)wƥ��|�8�,��/$As&�_�h�H8��3��iD��֬h�奫��V��N��B�E�4�T��x�N�2��C��O�rR�H��p��T��D� d��XR@��y�G�V�_�zj�IƲ�e^F��D|,c��{�N�@�eI�y�RQ2.�< )�o� +*d!�(��,��y)��/��"K��"ǦDȣ�!�G@��5F�8�bcq7��~��s�Rs��#��,5M��t��Ъu�lK�<�j9^T�ìJ�!Ѧ1<r%gX�`��`-쵵κ�e�AjW^ڵ�#��wg"Xu��pT�'��Ͻ�v�^r�,��3�߹>�|��P�m��n=��G�63��Ҁ��ɛ��$��<@rNAaA^��ṷ�M��)&5��+��y��)�5��s��"��)��Q�rX��TJSlCH?�%�wR=�E��:��_tz��^Mw�U��P��sx5�"a@ڱ �p$�)9��%�1�7M�[�s�/�~�]H��d��G��Ɓ�ae^��x�C)��r2��\R�>I�s��Q7B�� 'X&N��Q1�%�|^9E>��+ �Xd�I_��ɲ6iNUVQ�,��.iV5 O�J��ɣf>aP��s�ZM�tU�5�F�Tk��5�؊xJ풅�,�QqZ�λ��[:�7�&��]�i��Ͼ �[��8�r��W�೥�H&�o�ߠuk.�֬��u?hǕ��y��W�_ۮA�Y��/��{�Ϙ�S˟� ��(��οO��3@��n��Bm��I.J�̜�8}�T��E�\] + mi�p��B�1l��U;k��ҝm��V�qqY��I��`R-��<�}�tq�]��ō%��F��yd-��f,��Z�,��w�� (��CQO�Բ%��ޟVR�7ޡ�{` �x�R��%�a%��l��"J�| ��R��)e��Ԫ�x8�ڃZV)2�� D)�LŚ�e�҉Aj��L�'OSJ��b�H)}��^�~��5O��fK ONQ0L=��`�3�n��^��+#0�k@��l��8�u5V�M�>��J9�C��0� �#@��5(��(4N�N�Q�S�:�D�Z0Rn�²{v�.�j��L��M;�3Τ�ԎNk��!��{n�ٳ**E�@(��\��r[a��s�~g��g��}��>��`��,��7x�Ҳ +��TńE��S��"I�+��n�[�k�K��=_�;�H��U��'��Κ��h��D �AKwSQCc��p�K��+e9��B�z��]��'�Z�#[Q/��P��ji�l��@��z��ʄ�z�s�.��M��'@�_�6H�D��I +��ϯ��:��F��u{T�D<(��2 ȝ��% � $�ќ�5�@��'��b��Ũ��>Yҷ�O��uO.�%��'-��#��ך��lD v$K��g`�d�O�Uc]1)?Q�<`O3��_��½��:�SP��3$6��e��N��4��Κ�2��S�/~P��{,S��x�x��^P��(�5�7�T#s.��LXZ/��i��~�Y�r�.+�� M�@��X��u��g;>�9��;xwh�޴��s�td��u_�|�F~F��j��w��c��IX�.��#M��B��7�ʮ��*l{Xn�8�ZH��<��H�H�?w��i��.8[�}#�J�>��p��i��vut?|Xj�״W�xk��H�]jN�;�ڊ��V�<��w�\g�e��\qy��=�X[��3�C� �/��hj��(Nꝷ�-z��+Ă�r��Z2G�^:)��^��S�#h��2UU�/<�_��J��@�P�]�@uD�;tT�9�Ҭ_#��h�Z�� +d��?�W�Pȇ�d�1�x=?|F�|$�k݉J�r��4�P��aX+�r +��3ebccgt)G��i�+��<�� ]CaLC�04K�v�X,KѬ��t��x��u��#3W,x(�b֌F��Y:C�δ�ڼ��ԝ�;|�mIF��7$ku)��3,�Ĵ�T¤5��-;2��ܜ�T�65)˘��ۓ��7�Į'�B��MF� 7-�.�Mұ��nr8,��n�ܐ��xD�JѴ��!mg9�b �X��d!8�^x��h��x�v@��rP��)H�$R8�.0�"�SOӐ� +��*�P��F�?��<��7�F��ՀV4�E%v�~�b�G�hE< +��h�C�~?Xѣ��k��}�ɽ!$N�|�l��Np苨<$� 8�VizЂy�U�H@�'bIܢ��e65�ɖ;��F�0��kk�.�H�B��U(�N�AQ��ꨮ ��b�� b0X!�b(Cr��y��v�a��Wv�L�D��v�\^�[$��T�N'��G>m�f�A�7��5�2�t��=�� ]��`��js�Ic�V��'�R�#s ��2Mzmj��`" y3�O*�b4��$�h��N�)8xHg�� %[8pZ�,��4%@�f08Y��q4��d9��"�&'�S�q��m8S Yp� �2J6!�q�A�+�<��,F��3� �� ^#��^+\�w��<��k��s�=jnl��?-�U�K�G%��r͋��R��ȫ�:S��C ��-P~3v��,�R��e�5ȉ�:�̬�ڎ��Z�bq}�� ʳ}C~Nyv�{�٣�_��g��?�?NP5��iS�n|3�k�U��Ɵ�F%�ּ��m�[��{��q��#��ıw׎/M/�� +�*�DR!%P��*E�UBB�B��P��(�׆Ҡ$�w��8.�7�Mh+�*"^ع��ung��gf׻�ڭ�o�̹��6��?��y��Ɂm��/�v��-/��?�e�uà_��/��h�ҢSn3Թ�*_��Pņn}kw ��tef�}n��3��Mů�_�|#q��)�oO�e�"��[Uo� ��.z��8��l:��$D�5�uv|>2�X�c�H�q� `�Cldh乑�C��4:}v�(s��(�Hdtx(r��h�F��͏� +2C�@�ht��S�}��!q��J&��.]`8��Y$��2��.VUM��Q�v�]+��;ħ��g/�Lœ��׺$�8t5/#� +-��0��+�b�N��fѩ��R�2,*:�VѤ9��:�kb�|�H�V�Z�Jiސ�*�{�b��R�RЩ]r�e\�5�eK,C��^P�4u˰]��Kj��"5YU�@1�J�5�1纫+1��?ԣn^%J��ON.\I-��g��ljn�Y�/�I��r�5I�f@\^I��+p��R��7�824:�;@�gQD`4��r0"�sd�B��]�(��A��}-��X��3/,ܚ}3��²fS�tz> ��˖�E��N�;�GCI��̥.'n/�t�ߩGo]�Ud^�U]7*FQ�8Q+��a��Q�Tt�dَmئN,�膣�^��RI��i��N�P*�%-_T +%�Rv��@��mbT�zY/h��u��*C�r�\)�EZ�%U*�+eU��9�K�C��g��~$�y��uT� ��#A�ιß��/o^]�-��[��ť��[�q}�W��뙥5-�ɬ��z#s-�4�Ʒ�x��R~�:��m=_}&]F=AD��2ꅛࠏct��:Tah�#�լ_� +"�g�?ٵi�I�88� ��o��޽{7{'{�΍�[��fog��e��O��"�yf��N5��}��S'�>��MfȢ��9Nx��d��DEd�yQDU�UR($I�dp��J��S4��u��W�6�H��!�DR��i��e�Ty��8��;Id�I +�,�U�]F�*��+��5��+o��~��!�I��f0�`(j4?��PW8EaxV��)jf��IGG[��נ� (ol��Z---�z�ϵz�́��F�m�3�Q��Z�"�m��6��v��ޞn|�uwu��t�aZw�n}��=�}�[��IS��P� +�R��"�EI�?U��yU��bN�y��#t%��$�D�4Y�$ � �i��(�A@�΂q��*`Y��0B �x��0��%&s�$�к:x�zY��ڶL��㔑�6`�bjcP!=5��ܨ��++��0Pc¢�� u��Zq�=�:R��)q��X,CH�f3�|�Gݸxk��!�\��7��@��* +"/s�]o�Hd$1��ؾ��A�s��k��r��YU[+eU`�5��O�M��vgek�>6��3F��>��U-\��E��Y�񠊌�AE@�LUȊ�6��MF'MT��KS�h !y�� CsY�1@�Ȋ"[E]!* K�}w��x��Ko��;��>;��7�޿��z�3� +?��p��ߒ��G�ھ��>Ե��7M_��w��p(�j��N<�=t~��>��GT�|�&��dMG��7�ۤ�cZ?��h\"�R�,�0��thFK��7�i�R]#�=��'N��S-��f2�}��n��_k}��f�c��?n��[��v==��'[�W�?��?��,6��w��56)��RR�$Um]Q�*%t �PEHAR��JT*"��$4l-O��h��̝;3^��Nl��8�6��{֏�{�8IA��H�h��rg��Eoz8��ϯ}��@��8v��!c*Ұ�[��+��]�k��!?|��v��i��=�2r"��.yЏ��TR�$I9��Ս�# Cz|�o��H�w�Dw^��|��C@vo�sV��$1z��A��8A|��v� }k/��}0�h.1dp��}�ºA�_� g�X��]��sGZ|�Ǻd*)�˺e�S�Y��v��v9z��QG~[B5��/�3<93��nq!9T� eɺ9��g%l}e��i#N�C끨 +��-Y�1j��Y6�r�]��"�O��/��ӄ��0�J��{�M��Tǵ�PYKP<5��E[ +�8gT+��c5�Ʉ��v��{�'��L�� +vş'O1��M�.��w�f�[��_��2�_��m��ͣ��wd�o�\�ψ={7�@�w�=�Ľg%�|m䍆��t௏�n.#��{ix�~A �b�}�)��BmY\w��ei��L��b�ѹe3$�E /� ��'��, �ޗ��;r�c��b�8#�w��+M9V��a��ey��Y�ox/�r`�lޞ�3 W��1�񦟨��R��孒�f�W +��W��qǆ��6��v�_>�o{��FW��c��ava�~��8��x�$�o,k��V "��m�ĥ�<��)�~Z��v37�J��ʔH"�GJ�؎&�auIBM!�e��o�Z��P�=4��M�� {�|��)774�Ƨ_Z��d�}�n��ڈ�R_5��q�w��:t_AY�x�JV�j,�ׂ_��A4�N'��P{ʩ�)�HL��^�r�� H��j�rS4%��8)�4P"��VT(�g�8L�7�e@S�`Bڜ��u.�%Ƕ8"��$�.�(�S F�!�Q��2{��+9Y�#��!3 �"a+��w�tt�!3 ��θ��5|��p$�FG��X�T~Ç�x�W�6��[�)^t�yg��~��O��X�̽^i�-y��4�X �l94�x$'��[��2r�?�,�%Y�R��$�R�X��3�� +�C`�2GJ9��a�QN��NH"�q��j@9�$e#��U��~I/q9��5�b�s6w]覛�p�X�+�Y��Q*Wν��(��y�N��ܻz��nX��s�hjhe]=��®��J1�䞑��t4`u�Y!s@�=��F��@�Rj�1�`��݊�TFz�Ԗ+�z�窻�)�N��e��ut_�S�F��h4��L�fg�XwT�ZY2^Z\��z�y�ֱ�>Rgi��)6�c�}�/��(X[{w~t�3wG��䥭�NOLg��t��h�j��눆Z³��Ȳ��sb��X��n�%3�D��@8��j�С�&W��qC�9c��nD�A4�Z0��,��Z +*��D��`�l ��)�Ŭ�58�cZ��@�z�� +��"��#��`Zu�A'�B�.t(�2�¥bmh�ӽLW�=(�FY��7��0��e��T��Z�:Fy&,Om��D�� ʪB:"��Ku&��`3��WkPT��.� PP�VI�M�ڤ�ƨ�xM0vbU��R1Q��JS1�L��I�1�j2ӦiR/��Č�i��mA@XX�wvv�=��=k�N�,�Q��g��w��<�y��?j��9PݘKq8W�@�k\3̴�{;�"#w��*))�q��n�8�yMޚ�ڲ�2d��QWf+��Z��W�m߰u�?�b_��#c�R7�8�@u+��zv�ԥ��U�ۦG�ߥ��Y(�P�X��u��d��4�2�aG�5��@��b*�Mg��=��>J�M�l��r�� +��~oM��{Yo�a�̹�޲"��RVX�}��ho+ol*�]�쬱]+�Z�gb��V"ƭu"WܤRU�VG��e�s��K'��sO��z��{A��@��h$ށ�,��,o񵗴��|B�H��Ȫ��>��0�H��H9:*�. 4�d�-�44�)��&0�`hL�T��kf�2n�U ��텉Qx�Q��/��_Q����rw��bw�6�^+4�h�t��ȂS 6/<��Ɗ7�ae�#h�?�?��E��; +�}'�#�Uc��#K��&h�a�°�X��J�A��%��w0ԦD�#&b��k��Vq��2�6B�� '��@�k*�q4�LJgA��q�W��\x� ��@��JI��-��j��V�[��Ey��wgaK��)2q��r�G5_��ȋ�rӣ�\��9��f��[�� "��d�A�hH�#�;��91s�R��~|߱��Ɛ�0�H�_%{ϦGts�R��+(��%溪�מӹp�!i�X�Hq�aC�Aש��,�.��_�]��j6�y�OwFGd �m�˫��s�h��W��klu�TUW\��E9�?g��9im�)x��.��N gk�=:��Åbv1QgƵ�[��s��+>�x��F� _˲� oK�W5m��pL�Ɔ�̿��I��!B�o1)��3��%�<'�'��fE��e=�+��@ ��VWy��l��~O�D��(�8�Z��cX�1(��{��GC��s ��Wu��5�n��Eӌ +�QzdC�U�1t ӯ�Aq�V@j��={��u�aH4��$~��'C��ݴA��T(fE?g�B��ڤ�OM�|�s��2;|�~��,�yҩ��L%O��B�G,� Q��(�%�2^L�]��$�F6��u�9�X�8)k#��Y�Mq/%�W15��݅�Qyd��.b� +t:[��:�֐V1�B%�D��0QP�c�O� +`��U$e1 ?�R�J�H�:|0M��0�i�'��Nj��i��ī�e�i�j�w��|90�c|!�X�?��^䔟�f¯əG��؄DV��\�1�|��Ę&u��~�[V��]}ʼRX2��sy:��v�/<��Ɓ��x��O��ß�Vk�o�cjL�2uJ\��u��_�s�x��u�j��IfҶHr_Bbă�-d=4M�J� .��1z9RpϢ� ��-�2��(��m�dmw8h��}�\̭@c�I��I�ŴD=c��˫��|m�� l{�Ծ82>��*C�xtp4��b�� mmj�M4�G;�C�ʘ�r�`:,2}¡;b�~HL��# ��3�8-�3�� :��o��AL��m�#Ն�� +�)��>�<�� +h3L��ewK�F�x�0X�f��SH��"�qB��-Dƃ�T�!��P�l@ ��:cT�e��j��ԸQ�5\��+BM7^�1b�E�t|[��Q�rÉ_�,��qg;Iej�q�D~�;��[d;��H��؈� �01؋dٿٯ��*�+>�"�FJ�_�P@��O�{�"�X\��6 1�*�(j(��я��6�P��@ي� +k!%�%��NϽ3�Ő�_5��#!s�Ν�{�w~�w E{o�^Ƨi��NҺ�ډ,)>�,�$,F Gጣ�0�&��!0�W��jfᐣ +��7��S^��f-�(-Nv.��5�%�ȟ��&�.�& ��*;�0�i��6r��(�L��ß'ޛ�6i��ҵ�b�VH��,i��W�G�Ąl6��yw7j��wc�{�;1' ��Z��*�O�=T�2V��%��"f�fx��"`q�Ver�5�]>E�"�/c&jw�-֭�<,�CL��%,]��c�9 ��`��L` ��_7B�_}��d�v�,Ԓ�^�󁅧�:��y�Z�ե#��)� ��g&炢�xdt�C_�K��?�`�F�}��O:ͼ�9�]p�27]��:��L;d=�y54��/�9&�� s��w�*a��䛉,=��-\�n5��Im;7�� +�@}`g�L4��|��pLe4x(+�5!�ɑ��^B-Sr�SY�4 �0C��r�E�r��B�"�up,�71X�'��H�9��}3��G~v],h}��P֎�>5 +�k�%�נb�ʰ��e?��5Mw��/��r��Z1젦;�v>zc�&��Pݬ��J��r�'rRHg(O1y��PU&.!��/T��#4��Z01n��R��h��8Ս_< ��!O�0�\�C�� Z��=Z��k7j67?�� s��(�n'��3��-��-J��H\��G<�d�$"��+�H�x�CY��,H]o��pBI�:�ډ��1b[rމ'��mg�e��cU�>lS��3��Pk��Q��N��%��lyb��_u{��5�?s��$6k~�M�� w��;�Y8��JɌJp�~(QGRL��a�>�$��$��D��P��'H$�ѓI]W4�v"�ۺ�4��'��y�/��Wp�0��r8��&�bg?^�+= {V�KK>9u9�v��V��#/wc��5��ռ��ŮVw~�-�-T'�M?ul[�Ւ�.�ӳƿz�8~�� +��19��T��k��?�M��M�V��$�23R��1��V��+�&�"��_1��O�z�3b��þ��`��Q"��i�?>r�q�{c��[Ϥn��s�j��=��o��\_~�n�5�ɲ��b��c��^��V�D6�8��%�ka�K��͇�`M�6��.2I�V�X L�� 5ϯC�!�H�ni�w{��m��o\��l�q��_1��6d�u?�Ϊ��el�1X7k�/�rF�Q9��e`�N,,+d��W�[��K�pc��*t��qbP6��W|�͒�V�w�:��b�q��v��cG�K�8?�U��+�lV�X D��Ҁؒ,x� ��iG��E9�[q\��+��֛3�(�P�)�Dh& �˜�ӏ��ǧ��I�,��gr�Q�H۾o<��R��Qi��BhGH��/�qΝ>2��C��Nj5й�@��2��A�R1��7��Ba��֦UUYUd� �x��P��z4vš�3;��p�D��]a�ξ�:� }��!��kl|�hc�h�c��b��۩o��رiH�J U��P�E�P +/�*��ڗ&"Ay(�B��/,m2Tm�,N�Yb;N�M��n�6��=۽s��ν3^bG�֥��;��g��|�3��ק�$0]u\��}|6�Í ��޼icww��_xaS��4@�RE��3��g��7�LO�M�S"��ͥ�XEm�dp��{��՗u��B�N��/<#\�ԓ��H��c��V�s��е_>D��F�0�ʁ�A��ǧJ��i\]�lEs�Bc�hP��o0��r3�� M�Mm��|9�㨏��<ʃ>��៦[_��%�k�s-t�c>��y��Sp&��1� � �؍�[X��j +�bO�zB��o��t1W�>>��f�<�-��X�ey�&ӟ*��$�-�T׵ �L��j�nq:R=�pU8��?��1|-EQ&��P�R�^�UY�b�� ?��:��h-�Ky��Q��o.A�Bh�- +�^*��!�3��˦M�qM�( +��W��Ⱦw�j.=_�-;ef��V&��X��'��x��6�o�~w챃��q��,.�}�� 6�{�s�#{z�Xd��D2O �A��O��O�D��Қ�N�l�!(V=@��Ӫ*5�ہ�f1�JZN��D��y��l(��쏟 �A� +�2 %e�]Z�)��p�(t��Ea&��f�b�c^�"b~�w�}��rcU��7 rJ�s�Ş��=��(��Ŕ)`(�3��w4��)�?!��N� y�R�'K�4�Z�;�5)�9�-��JM��\��췈cX^j�� J`�k�.8��7�@` +L�DAմ��^�XoB�617(��6{�AN=̥��6ɀ��,`��]9N�%��h�� +Ê�k�9��[��f!��[��w�<��ڨ̫�cB70�@��U2�c�x ńN��YVUݍqC�;�]yP�2��p)��7��5��z�)3�)f��'�E�"�s�$l�Zt��{]��Q��ȩ�#��hG�1��B>p�㷏0��z��'��wpo�ļñ��>��b��.$�AU�zn +�mP��\=K�e +��|�И;�=޷7U? +}錳o��/ćȑ1�{�{־��KgR�>�!��|8y��(��n|X�gы鳶��;��ƽ�� +Q��b�[7@^�.Z��G��n�S�Y#�I��+��Oy,��Y��.�d�;7GX�.NK�~��Dj��$�sDI�T�Ƶtz2"�jɧ9�󍨪ۼn��~V�ad�L{��j՜��З �ښ'��Z��O��a�K��ʺ�͗��/��3��ܽ,� Q�G� VD��FM��%1��U3I�4ѩSm�I�c4��a�GFt��AA^+�.��{��#M�4C��={�=��/Jr�˨<�l��>�@��j� +S��E�G��Q-�zx��GBll�`��RJd_#�G��M��j{C��2)?z��'I~3�/�H0�?T��LJ_�^�n~��F�H�K �R��h +�rH�{�&+�$��!Q�Zt`Z�Ud@��[Oj��ǟy�w��@��tZ�@�,�v�+_��ոÉ �z�S��"+:4��&��m�Vn�,��F�t��"��=��-�$�W�^] �͇��^��2yz��bj3;�Gv +�c�;�΄YW�G��~m��E"i�cd�%�$��Z1v� a�Xu��5=�^�x�~��$��VE��Y��>�K�a�L�P�;�/�"��%��!��6�L �� +��!�1��ӌRJ�^?m��KA��A -@j)�IC�BSo�A��{��v�G��k��b��3yU&o-��1�;,c^��9�;1b��#Kۑ��-��rl{L�#c��jO��tׇ�v��=��Z��Mţj��P��?�k<� �!܉P*��0FB4U��&�N'Z�K�i�*�' +d��oaZ7 9l*�64'}2N��b0�9��Y/�٨j��I0� +��y��Mbr'2R!�Uҹf�=4��90�֫h��Y`�ˢ��-px��J��v#�}"��^��"��i�s'�d�nĎ��4'�=#�'G[�`��^/)v�&�۰��K>��!|��R�,�V��C�\vk$�e��:�E7�[ކ��q!C�z��+Tq��`��Z��Tgi��4��rTڞWu,]}+��Gs�dmi�Uv��fm�@U]͢ +r�Lo�\G�5Ill� �i��)��`��^��n�w!z��`��5W��Ʉ��#�+S��q`��"�ڠ��S�)�U��#o>vw�� 9r�|�T��X5�.��!}�ү��։��Њ̃ +�l�v�%0B,�J�I�݁�`O@KVy�JZ��`��9��VM��>z+NQ�M�+�7hb�>وۻ�̬7�K��ѓ�� +J��Ϫ�~J<d}��9��;�&hbm +�t��N�V'��~#��^�"ʩ��b�� m��ʊ��BՁ��J�sro��6ל}�vN��E +�jX�_Q��%*�+ʹ�C�²Bs�%Es��s�Y͌;Js�%��N�#�)�ΖWrC�{��|S`�C��|q�ٳ�UM��%�,��7�'| g�3'��Hs_��a�D�d߷��୲��O�� 2<|x�a$<􊠦� ^ڂ�i� ا��g��-C�zO6��d��V��Cn܁.�D��!��>ی��3f�IH�n�>�lU�n�婘h ��_ȣ��u/��FL��n�k��2ڟ�1h��t�ґ5jk�Ąѐ66��ix�4|&�t~��"k<��)b��Xk�a�1᭸p�� UUc��:=��2�!nqr�m��5��~J��2Y�z��P~3�Pⓧm��=�3��ܱe4�0��d(P�"SK��D|��C ��ލ�ۯ�.��ń��$�G�ܩ�O-�k*�Ғ�9� ��(.*�Z�0�A��O��X>k�C$�ƅ ��S��}u��\�Q�eG7�H�]�X�n�i��/��ǥ��Vpm��ؐ�2g?c +>�~_��~׬�N�_��;I+ϯN*�S3.5ĿzDT]ZF�iM*ߝ�X]�&%��[��;�R��m�X9e�W﬙��[�6�&��q�U kV��_��b�zi�q�{�*v�}��1~^�jI��O�Ӽ6��㋒��w��ߢ�%�N�(rk��w�u�iu6 ��f��@�C`�U��Mn�K�Y��Aw��]Ҭ��)Ev�Xk�֚�O�� +߷Ą$%�P`KQCP�L , +2�!!��<�(a��%��@�m�aTjm��]��b��Z��jX�HYBHHB��޿��1q�&��s��9��2/ǆ�Y9� o2��=�;�7��P�X8��1R~T2�^�F�{�o��h�0�l��d~*q�f�u�R�ȳ�e��{�ނ�>��.� y�ԓ�f5��b��q�Sb��i��:H��-9z�a�hVN�`�u�5N1�Z[�Q4a �2�Vm {��<�;\�B��} +*��0f~�d�Zan� U()� ��0FW�L�DeE�eSn�u�{�l��T�V_*�^8�Wr��8Z#�,%�/ض�>.�R��h��Mæ#��B�Lju�^{N�Ȝ򏉏��Q[��GbǽVˣ��%��pt�h��Ӌ��h|�=�_��H{̮S[g��=5K�R-j ��[�Ҳ�z�Z��H� +v�YU.�#T��v( ��4 ��Y��L�rV�2��ȳB�R_ '�;�y��d��:%|�5�#�jEpt��^k +�T`��,kS�]L��)��V�lnQ��@�hE�,tPX��?e~��M�9�^�z#d�J\��S�o#��?zm�:B[W�@ӹ�\*�ɸ�LM�#5�!�f�cBX.��ai�)�0t�-�v/rd@j)��V��fQ,�x}��`Fsں��=mݠ�~�3D�+uo��'��GH��݀D{Ɛ�-h^C��LM�] +��ĸtI��$��a&��m�M�y�7��Xs�a- �Ks�cO�ǁn��>��R`!�&��[��{ ڪ��^�_z��zj��gdն��W�� 3��-��c �fpdWaFu�󕮏�ɤ�D��:��u�z�]է~Y�.��3 ��dKw�s1O��Q��{&��<��ꆜ|8�QJ%s-��Mrve�\�)�Nj.��`��t>*��(�v�/��غA-�q��9_�Z_��)��_��4�yv��$fd_� �)=`7�쌒��d�%��$��w,#%U3�Go�t�o^��h ��\��{� +%�)%뾛"��!cy�>V晙�x��Mkz�􂇆�ʚE�[4��y05grjuZ� +Y��>�W�L_+�?�+�Z��`f��c�VȈ�+��ls��˳s67{^�� *XzL!6��^�� +k^Qh��3)��܂#�k۳s�5�f�ђ[�{[^(�)<'�s��f�׳O��_��nX�ou뚥}-�ovk'��u�M�3B&��#f7�LoV�R� X��?i��Ǜ6��۽�j@�l2-qTX$ǎ:�Cw��"`�rT��0��%H��a챑�+��Q4��&"�� R�E�c�ia|IED�&� ��e�rTI*��6P��b_J\ĕ*��n��Y��|k=�5�Q��~�WU�q��BHX�6#˰4e +TY��ie�x[Y,�R�� -��0�CP� -A +�)%��˻�Y�~�ܛ͐��L��^r�=�l��7ˮ<�xt��Az�[�m�ݶ��t'*�"��,]Q��{6ߧ��4�U�3,�/�S�w�D��s�&%�\NW�3##-c�#cvƒ9�6%ޕ��p%'&��'#iN*Vs&��\�Ԕ�$�ӑ1/=��tb��Ĕ��ii)�rN�3ՙ��]��dW�\�+��p9x�đ2��MKqΚ� +�ok��at��U��,��Ȣ��o��Y��,��˘:H��G�k(f5y�/;�.;{�|r��?�B��YT�3��d��c뒅Y+<~��g��V�=ޜ��mX��y��/ۓ�ug�7˓�ҏ|^��x�YY~�ϝ��y<�>��,�'��Y�_��=X9�/�n�~��*oVv�{�f�8��j�v�|�^��(2y�'\01Y*�6ԡaWTf�Z�H8��P 7>H4�ż�t��n�K�� +q�D5M +2f�ݠ7,oSbn +M��p��:��;-�� A �D1�u�E(g��@{�)�ӘA)EMf\��I��@��(��n7��a�UtdS��O��ez��<��=�T�e��63wg��η��l��5�fH]��f@n�z/��}�c��J݋_.)�̿�~�2��D��.(r�6].Eo��,|�)��u�gj�v��un�J��F�Mr�[��Ƕ�[]漃�C�+�&B�7G��+#V��D��yB�E�H3k�vHR[�[�e�9X��A +��|_W`�RS� �L7��^�$��K@A��$ ��ꈐ��Q5h�A�#Pf�o�a�͸Uh��/�4M�(�!�{o��ND�'Pi�̀�V󺀎��Σdt�rk7,��5��++�Zy��r G�J/](#.~D�/-)�P��@f�A��{gPY�K*D�I PZ��^�t��Lp˄T��9��3� +�'dj��B_��[��k�K� E_��q� +�z�v ᡠ��{dN��hmX}�e�N��+Zp��b��8�4XY��08�c^a�x��Iu1灁p�^��qm"�o��`L�ǽ5zN�q�/W�H��i��,!� vu�O��m��"��*C6A|q��@%�E�!`<��*�0��7�5�o��|�=�X�(\��_�`��V�>U5X�"L�|�X�Z��!�93 ��5ǯ��+�?�N�T��̹�� ]Qd�#��`��8��К:"�PN@�'�v�A�/��7��`˴!Š�4lZٽ�S�4x��q8w]:��$��)EFb��<�j��oM�V��Ւw7��)f�QA'A��o�r#�y��YKY��-�~1�&%b��7_U�3r9U��]��Oey=��.�Q}w�B��, �&�;{��_�/�-��k4<��B�w�d�f��6�Ϧb&�A�Vsj^��f֙��ySS"�?��f0��q�尲Y�8��.Y� \?/؄rYS<�ㆸA�1IH��D�8�:^�br��J�2�b�09uQKzNщa.��ܜ�|�t�r��Ki0�1G�� ZbX��`��{`��o��J�~��rY:�)��ZZ]]Qm>z�ϫ��d.�^�?��?ҫ8��MHxI��@��1`��IP��B� S � !I Pt�<�`Ѵ�t��K�XiEyj�!�N1RQK�juH��{�>Ϋ�9�.�uZ��lv��s��-vb�d�.@�֩'�qgì�y�O��YgY�p��c��Wx�􇍧��_|Q�9�fg�Zg;2�w?��߁��"��" �]L�X��-�Pc�� +��f�S��[D�7�'�ĝ ��d\�ʀW�+��U�UO*�m�lH��ʎ��*`��x7�+�%:ř9��1S�qY$Iˉ��#i�-p�XzL��ŷOe�,b�WE�~�(��+91�tJB��`�X��fn�74�Vz��-��e=��O�~�Kf��8��ﴔ3֖,��d�� U�6�w�Oaf��L� �\g� �!*:��y�b��ɦ�gc�4DD�Vt�\Ÿ�!憿��.pv�2��-��bYh��3VA�1�S!r&CRS��7��r�\Y%��+��e�L��N�~��#�\Ŝp��M\}��E�U[aW��WJߎ^T��åZ+0��#��G�),�t��[yT!��,OD[�bD�O5{{u�2��Hv�w��i(��^��s��!P��4%�m� G��ݬ�b�� 2L�Sk�ϕ@�;]��ݐw�-[��lU�EUK�KVW�.�\WR^��|^UE隊K+V��,^��tٚ��嵵��-]Y_V��b��ŵ�uU�U��V�1܍2 �e��}U�@� Ŧ�6 8R��ny!�� +��MY4��#��$|�Wvӗ�`)�L~�>��TIJ��%�+�ĭ�bʝw��aE�1��P_>�ҏܥ��M�j-��&�9(wH;�"/ͣw��K��E�oS5��\� ��]��L�� +-�Y4��H]�%��b��AQr+��S�W�Z�x��J�6��H��ua��&�ߦ��o��",�.6�mp0��~~Nn (��%��-�� `p|~^ n��p/��B��\��<{��@�Î��[un�`25�1��l;�F��uC�,�u�u��=��A,�:��Q�FXD� b��#�H��Eb��l�8,�`��)��F^ΆML��B��۩�[��A+#��M��5�c��ޖV͞�`��^��8n�x#��2W��n�]C��9&g�`�P_-��R2��(��/R��+��G��p�o��q��oh��F��tPָ��L�sP��_ү��N;�l��F��z��'2�!�m��p0��d�C��3h��MpK��C��p�{&�֏�͞��$�mɼ� ��ze��[��ahgeR='s�@7z��<�8t.��}��|��i�+�ξ��2��P��G��^]��T�:�NG0�!�kOA�)��.�i�LAof��Ѱ�s�} ٱ7��]��찯n�G��)��3��Lϴ/��X�<�iP~�+ɹ�R��z�۶HӔu�k�:�}Ũ�)�IhG��(Ż�x�9"۱~�p�P�=��Inf�l��Ӆ��q��U�&��n��x(�2c|NL��Hh�+^��|��ՓP�&��D��8�c��\��E��m�)\U퀉��ZVf�xF��&��_N�Ԟ�4��׼�#)�+�&�7}��<��3��{w��toq��]��z�t��6űyɧ��[3�K�#��Sh;�Z�^��ܒ�X��2�/��㾌��>]��{��#څ~y�bkt7�Q<�i�4^��mI��ɤ��S6y�y��]n�Ы��'4$�@3�y2!�+�(��#l�� ` �D��г�b{P�Ŝ�|c�7�mՠ>�� dF�;��7d��W��]�mwf%�?�im��~�e_�?.3a�Q.h�o�o4��FZ�L�[Fٗ��d��2� a��KT�oZ��Cn��E]�QjDy�o��h�H\��npUʉ �ƘrTD��Bw�M(��n"�Z�'�2B��]8�5�?�?�WpT��MB��!&<1�@+eJ��G2 "�h�2-��)�2��ҠP� �B"�"� B4��k7ٽ��{��ul��t��3{Ϟ{Ϲ��d��TԐX��*Č�V_��(�Ć��lKbm(��3��0��+��;��մ��Ղ>�&�]��Ro\e�9w�⚐̑W�(��K܄Q8anp�j��H*\>��q�^{��ۉ�{,��l�׵�>��0S��aT��/��]�-8cj/d��t��*w�� ,f��ў�v�9/��f�y� �}��{q�r��o��2��彃��Cʧ�Gq�u�z��]R~KŜ�d6\>�ܠKX�� +��$n�v+�P��E�x��\��GAǗLy^��?lC��:��h��wP��!&�>>��%��Bk�t��&�6��w۲.|�'�ERd?� K��-ZTZZRR\\r�[qIiI�%�O��>�b�*n,��͌G�"A(��(�cZYv��Gb}�:#Sp�Z�+��}��V��AD�D�g�Rs~�S9q +��,Ń�6��JlZ��J�|�)��1�X!�:�.ݕr�� Mm� +�R��7D|�묐U>��Mͨ^�6Y�nkF��0i� ��5/3>8��\&�9�FFeu<�k]��A8��Ɓ&��Q��^�[��Nѳ��v�>� ��hT'd�n��=��~l�?��̪��IW0u�U�2�%yLҖ�$"��n��j��{��2�MY�c��ì?�]�Z{Y(3�wk�z�p�r3�H�3'��e�>\��tª��-{��*z�5N��7��&��1��㑔;-��A��¤if�x'��}��2��-'''7�7g��ٙEhz�6�Ԉ(�{F�H�FYT8�2B�]�Alt#,p��_�:#"�_��y^|ei��x� +�AV�qn�u� ��b�:f��^&z �(!ui�FĊ�#�E�� V��l�,4p�Ꚉ�PEϩq��o�v� +�e��)�� p\b)��kF�� ګ�eq�eɂπ�S`��}��=�}�bӠ ��h�W[!��R��Y�P$��T��^��4:��q��Su�+�g� �-�> �8�&lf��,�$�\� b��T��Vu�)��ܐn�U��`��h��wzQ� �?5N�M�X��l�H�.0��6�xňt��X��c�q|��u�c:��4��XOĤ��y�n+��G$� 2�Is�3��˴1Y��=,�s��Ɠz��s��Ң�I��Ŵw��^�0k�a�A5�MY��L�϶l�lԩSlbI;�-a��Z��C�,G�YA�S[=��-٬��0L� +�78��\�v�5y��a��*�f��i��B� ;4u�aʜږak�#��٧��aZ-��YԲ��K� ��D��M�� ~�/Ba�U��Q�� ek;��f!qŒ��hږf�-$�Ȩ�p;`c�Mn1� jhܐ�+��D�:� ��[=��X�eY�i��_�W �Ue��?��!�@.B�@LAjA �6�$�)#��R��@ ��L�Da� +J3�퀺E�l.�BH��޿s�=��9��B%:3��{��>��m]��X��8�# +X��0 ��mX49k�YR�e��L= +�"��%v��*╪��LyQ�-�M�W�rQ0��y�C�[�o�!N.[-&&6r��e+��Ֆ�5�rզ��Jzt_x#��TDv��+>�ԷF�aIFk(�D?�>Zu�[S�.�@�6�Ⱥ��&�=(QSQ�$� ��L�A�9f�e��)�kAK«Px.N��D��|ʏ��Sg�ޚ� +ʲq_�ϋ��3��}�)Cw�O,�ĺ��HTF�JD<��Ō�n�Ē�?/�Uv= �!#+�:2[��$�*��ZZ񁹆J��U�0�-��n1�u��t��Ӳg�\��Z�I�Kp��*�j(s�t�/�z~*x2�>�oa��[(��s�,��Q�̺F~�� 42(�,��D"�F��+ �t!9�D�"*5V�G&��8��f6`�/��cfc��u�PWp#I:�M]ꂅ�)��X�l̦v��+Ja�B�ZE��4�M��2��T3��*��]8%��"��4U�[�a(�p��R ȋ~��tyK2�-�� +H�Р��k�T��*NX.V�� D��V�(EPq�O�� n<�p&Y9� +6QpT+2݇ �� bJH��͹��I��vQң��A��ƺ��s�N>�>��p��ԲS��HG�6M&0��+�-<2q�!p��T��b�P��,a��N��M��F_��x+CY��;��B ��+:gg��U5��O=�?$�� ?*闍�'��ҳ��,��7h��x�C{B�G2�բ#{��ia�ƭRgW�]2�p��i� U��ܺ&T�}�N}��9�<��(zd��3�� +3JMq +�v{At��i�.-Ԝ�S�NLRb,a♩�K�ϼN��A��22�*�/�+��r��We>�yf^�� X��M�9�'0�*��4Xs��q�V��&�;�Ͷ��kK��4��j��=XS:�ij��9�@� k*�g�7콷�M�ظ�x>� =��͝��h@�p��EϾ�pWp�3jXӾ�hwz�ĽS�9W:}�C�H�-�=f��U��U��v9Q]=`�\�8�ڊ�81��g��=��T��;�S�'�O��]\��yۆ�h �Lh̩�V��N��&�gq��֎r��Ü�0�A��V�V��kC�=��u�=k�c��/^�O�.Э��:�%#I�'X��͵/jk7��K\��!��x�r��6��dV��>��Q�a��Π#�u~e�IJ�6:��pPh�ζ��W��8gœ0�/YM�훠9�]�yΗMx�s;8�ZF�EH�β�q��%�[�.��,�m�^+��ڮ�8��eK�%�k5��]<��W��o�3}$r,�Nz�O��WZ��9U;�fN��3z��R�F͎��R�\%K�� \��a<��(D(�G�C!�`) +�nN�u��9�I��nϱ�!�01r�c�(5`B~�7��l�5�_��~(0��.e��#��d �+��p$be�, +�@��f�ѯ�tp�lF\p�n��C I�)vߒ��ft��3�iC�ĦOB��a�}� {�mv�r��in��h��7}�zjL#`R��>gq5�P_��Ѥ��m�'�^�M��Tg"d+��۩W{'��5cl��N��.ۅ�e)�:\�ל-?�݂�� i܌��;��q��6<�e�8��e�F+!�&+_�w?�w��#��k3�싫(�5َk��k��,�_��ݡ׊Ȥ緑��\�s�H��n�=y̙��'�\ +��l�� {�7��1��螒��t]2ٌ�׳�ai��9g��sd�� V{�z;��/�P�2a^��sg��¤S�[� �G�?�H9��_�/�TP�Ä]�=a�#��h�S��#�gF4|��,-:MySu�hK�燿��д��aa�s\�ﭩ,+��z*T5�d~�� ʦ�"S +�9?Zٿ߰�%5�˫�d��~*:��Q�Mv�Bܥ�!�c��3E��T<?�_��*��c�+�-�U�[�χK��SX�^剙J(Nٽ��;�G��7�U[�Y��(,��d�̦�&h"h�R(-��[�ۖ0l#snF>4��)� ��c��a41��4K�LXt +m�m*n++B�P��=��>��s{�u{�Z$}��}�{��{��8fu�$�_�{ �/<��ApoE��3I�&�N��C!%S��U�p�SIM�鸝��^�o�=�}�N��K<��Z�H��O��W�ܑP��s�x��+��%?�3��o�h��cm�X<kmmk1q��X��d�T�5��~��ю��xK�� ;Ϝ�l=�ki��8s��9��z�D�t�8�q��m�Ύ��S-1��dLǭi�,��5��P��r�s�՚��8��Y�:˲f��` ,+ +B� +�0��P8'�[V� +��9֬��;�� pz�O��P��4�)h�B�P 43|׬P��+x�5ÚmÁP0� +�R�1��U�d��(زw��w�%(�c!�ѐ3��W��4`�K2��HF��È��A\h��5ǶB�*q)�0��I�.�8nG�N �y�@��R�ؔ@R�vmwpp�4�n��E��R�Y$�� u�8��뾒��?Yh9>�]��J�L�RjZ�y�^��?�eƩ��$�pJ��U�9��sJ��!p�pW�&S�KM�S�n��&Ei!�k +��h�&�!��0�jBR[��}$??��)��<(C@��l�c�:%%��qO�/�PD��(�]z��8�4.��7�;�?�@��X��I�q�F:��*�L~�U�� S P� ��I�U��W�~�ii,{�G�2�=��ǭƢ�a��G��P�u�q�Y��`�1u@�L�!�u�iJ��b| #,��U�V"�t�V�IN%ZXP��J_M(N%��!a��R�̞w9�G��:R�vr��L�I;��(o��9��o�H��ck +��Y��OS��{��uk +7>�v_׀"��3SeHrt +��rO�l�6�Nz��ye ��D�2eLM�O�P�^��q�&'��/��*�Kɑ��hi�G��hc�G��(i��4��S��?��q݊��H醒�-��,�~��Q@�iې�0JK�N {G�Af��p|��a�넖�.��1� �g,L��(F��{�~,�|ϡH��A؏�TTD��1��Iz��!�d��ZY��bS�Ƣ��ꉌ�,�W�YZ�xh* +�h�hb��86�(ㆂ�f�A՜��a��\��`�Ι$#O��}�Mb�L" ��/VVF�}�� :G4c��Um*�ih�T��Y�*�X/5F12#��W1�yB +r��`Z�"@�� +*�n�~�uՠ ̥� >(�dL��W.P��"�9�*��L��ue�d<,�$��H[x�4��`��1Y6��.Dʣ ��kl��<��9'� ��ӡȷ*�"uѪ�ƪ��/Q)�!ɉf� +d��l/�u%⏣*�h g��2գ]�5��\�~г�E}#�{�h$��)��i!��" #s��uCʌ��fH��{��w��)�D��k6G�G�*Wo?t ��їI�Y}|��p��歴U�=y� +��-��c^��׶.��y_��>�$/o=��d΂9O�]��ys�zW�{�V�\�7wmI��y�X�w��_[~iP�� )�ݬ�|A�tb�?5'��ǟ^�RN��X& �� +��&��4�:��n�#�eG�C蘐��_�^t}EyEMed{��-5��k7��kŘ�K!2p/Á��˛��_5��A��ݷ\��=�~�z��_u�M�ܢ�7�-k�T�o��?^ +}gYῺvZ��u@Uy�`p1}��]�g,eK�x��$D֫4q�X��5΍EM�u!�'� ��(M�9�R�ƛ`&�Bۂ��S�s&m��yLZ!e¼�8��x/�9Io��}:�,w�o��PS��Vb8k��{jˊˢ��vcUI��"|Z�3�0i��ڳN;V�ʏ�@#v��|�I��}��P�^�s��ߩ⹰�;��ʧ��.��ُ�;��W�?��'W컟�b�)"G�j*Ս��(�>��*Gҫ��;�G(�&c��̘RA�lIp$�]@M��#��"XA�~# ǯ>�@3��o��{��,مHa�ֆƺ��32�2��^��_Qv��~[��jU�w��el�!&D�0F�k�j��I��Z��-�@~Q��fι@�v�Vu�R�$ظ�RXӝM/�RjϪK� �w��ѭvo��g]�?��;��s�f�& +\�U1��B�.�Z�Mߧ��[��a�:k��(�(��[Co%r��HSݍ��*ܭL��ǾI;�=ޛ�ހ\��p�ۯ;m=tG&&��l\�t}6��+��R�D�rTf B��aW/�P�d�4�,�[_V��~C��4M��'J7�n��ml��X�p��s��ͳ$� �q�W��u�}!%��p��e�g�\��8~�� j�Nj�Fcۘ4�1�AdA�}��:��ƪ mԡ��1�&�Tkb�ь�ml�6V��޻�y��;N:�h�axs��s��}��8ʒ�Ŏ��09�H��F�е;�o|� �':}.N ѡi]��5Ṅ��fuf��1��J��}�\�1QYt��{t�� w^}M�Bް�JgQ�~��O�_�*9K��nd8&+�{7?��5bfzw�@��h�&��&��{"��܎aZ�ݏ��̝�D��DĪ�&�\v9�&��.��X�@bT3�/�1ˑ�ܛ/��3��x��l��0Y2_�:^|q^��.먹�}#��}A�,ZP��{j��Q�L� 7^��G^��|^�3?k��$��6��wF��y�#��L�V=Y�[T��mG�Â +'�#��bи�pk��(�:��@zE8�e��]��b8ޯL��z4��X��+�I@[��?w,Svkt�� ] 5Q�Cd�.Y�[P�� +��![�E�ޫ�g��%��5��תS�^�ބ��ta]�` +64p�h,ͳ��1�f�Y��0��s��A��)%��\, ��f�c2�"[�^ִ!g Ȭ,�)Qx^�ˁ��*�]�P#,�|��Ʉ?��eyG��pg��G0�ֲ.�u��|�0ܱ�U��(��(�,/�w(�� +#�~ �Qg�#�$X�^��d��/�7�ؽiI ''�Q��@��<�@�7�˟��匁͒2��!A�W�I��߉�C]�=�v�~�)�5�7� +cʬ��[4�E!�tRR1m[<��OZ�� +�tr��v��QY?�ݨtQ&]ft�e�%�ޠ҃�� x��2{ׂ:��P�ɭWU�㈲��H�~�0'%ȊLr��2� �)S�ŵ�2��6¸ۢ��rX��/ +t��q��5 �CS9�<�4s��M�K�f*$t��Ř��3%��"J�� ݤ��N��)?JC7�v�@��E�i@m)�G}�Z�q�n7ݛԔ�bK�f�!&5��a#�nX�Z��Z��yzŬIPkX�?�_��*��<^��wd�8��3�r +�N��|y��iS�v��I޹�-@$��]��RNї��]gƺ��P�I�F�i��j|h�a��b67w�R!c�ͻ�n +��̸=�xꇓX��8�8<�)��Ǹݠ�"��~�$��в�%�6$L�Z}Ż��)8M>sW�[WނUB�(��$vյ��yYF�� /�W�WV.1�ה��ϫ3��t4�� +f7�P��kE��إ�* �c1��vv)X4O�u�ݨ\D�Rw�~&��Q��қ�e��W�Z՗�\sU��˳N�6�/^&!�V�p=�{�ʯ�;l��U��5��frɵ�pΗ�s�[�G��P�̷�%(��N)w��teܠk��Pu��s+��GɁ鮺ت��W�x�?�XC� +\y#P� +)�!��{�t��x�6��O#�1�p~�x�3sf�43��̓��|g��[V��r��pL/��~=�`�/X�x��"�Ī?�pc�2�Mz��#�? �@p��QB#y + ��P� +qN2�g�Ȏz��eE�2U�i��Z#�� a�r-C��m�`��H��G�$�C�Ӿ��U�l$e��o-b�͈J��3A̶��6$$��[�}(ڋ�I��&cR�t�܉�^#R�c`5"n +I��= +e̵[�R�B�I{N��Q�0��1hSʘ0@8�V��U�m��I��#��jGZ//��Dhr��Ϡ��F��Gm��Z!�z�B��yҲHt�)��!Kк�m��a]�FX�A�V��.$�ي~�N|נ��!��$�{��X�UP��^?Z�]D��_x�Ҥ�yPzS*ژa�|?x�u[`x��T�� rʰh�;�Royu���i�� T��&�7�F��s�M��b�Y&$��S*��ήY��wp��ͽ2wd��[��ܞ` �'��}�A��?�e��ܟ.�G'� +U!�S�� *%m�h�M�]ϝ0DcCTd*��]]��t�G�1�$��Ƒ��P��H!�E�}EB!hO��HD��t�y_d3f��߰҅�0X�4C��{��'�*�� 6�G�?m�?�K��f��"�W�`B�!7d��fw��ْV Gm��z�Eu��~3��(�,�ʺ�j�Zj�]-��EEA�.o7Yc��-ר�0\o�pcyĸ*^�a�r��\��A�hS��9��ׯ��Ý� +0|^��*��KY ��h��ُ@׳:�_�*.��M��}��s��w:��8�Œ<ԡ�d '�&F+�re_o�u�Q�K�W��-��ǩ+�`��#��)6<�G�Xf��&�ш\�dVT2� ��=7T��t �ۧIM`Pg:zͳ��0� �c�%1͏Ok��%ۂ��R�d��zn�ж)%�a�7�é�5@f�o_��o��#�4��2�{9#��0F��Q�z�^��(�i�!**:"l�{�g��a{ �z��d��_"-��TH�q�=^�r{��+܋�˽_��C�*��p�?3u��I�/7�Pֽ�(��lDE��B�!jQ��R3B�(�[,x�(̟��ňU�R ü�W�܃.FplqaZ0z +�#6-+��se:t;�Cx�Ն��G*��Q+)ľf��ϓ��liv��tdan�q��M��H��z�Hme�*+�> ؖq"�kc�\��[�\)��ߍ�� 9��0&��\-��;/��s`\��.;��8��(�\F�K(27�b�j$"��=%2"��lԾ��u�7�'`:@N#z�E9A��C��X�QQm��IY/�zIDx��)�W��̤�ߜ��,�� +|�]�r�/��l�,�W�+�}�jrQx/��Y��L�8$i�B;L��:�QM��S/"u'�@~�(�(��]��C��aǄ�+�h��D@��1��kԊ!H��)WpP�Tǫ��cH��*4@3��3�`��z�'��g�E_��6��j}�Z��ƍ��z5��s�o�NWWu_.�yfL��?�{��y=%��"o(S�(.��~ǫ��~�n�p� �Wc��C{��+T��*�p�ąnc[��]�n�;�]�9՝$n�$.�S�E�H��9�,N�S�>B)%بԋt�3�S8�� @W3��/<}��7�Sv��,�\ Z΅P�Q �i!��U!ѱ:F�Qz]T�A��0D� +#ְ�:o��C�T�`&��X�8��w��0O?5��=��&�K�ьձ`��8��S.�$w�9p8s +�9/��`oV5ii��}��R�w��g��Y6ۓ�sc9u�cI�W��)�⭟�,�=C��2�'��W�&d�t�g��:�m)��\&��k $$�;�Yv�~�᦬kM��.��ـ��n�S��J?��J�m׉/��%d��"uJ{�V��W�=�v�H�t5{gc�1��lK��z6/'��4y��Y�;� +�[��v��c��po��G`WJ��)��2��<�;pظ��ܰ��b��"��S��3��(7��hB1��;��8:�K�2��X�g:B��L�$&<U��i�,�~�Vc��v�C��;� +'��=�E�}��nf��6�yQ��y-�3f��;��y�jkA 6��ݳ(�>��7�kl�[� �W�51 ��G+�P@ �Hv�9WM��r��HՍ݀>7O�ؽK��J%n�V��,$��C�4y�M��?��B��K�]�+�CO+Z�Pp�>�%H@ H+��<�Wm��Ӏ#�2�۾�i<�'O�'��M��Q�W� +�eZy�O�?ޯ�pAA��dYp��I]��Ʋ��I�Rf'��|��/�T� +rB�:sN=�2�x��榖��s��6�Ͼg�i�Ǻ�b}��|��^Ҿ�z�>^S�e쁗��r$FxP53�VQX��XN�B f��p%��ݎ��¬��Ǖ�oUy{��(��bX�'�)n.�1U��o��.�Ը +��F �@��)E:,y��{��v�q�iN��TXs��9��(/>��DB(1ZXXp��rF%�{$��')�4��Ԯ��ꓻ��]��"f��P#�00�$�N�?�7|X�K>��ex�Ԓ��!�8�ަ (U��#۟��w�U2ϼP\=~\�.H�,`��a۵��ɤ�`��uu%�c�ZU]=�]��+ʇ��+�G��S�HY��Ɗ�9jl��͉A��e�P�`��V,�&� ��#�5zGIAQ��V�fY��'7��0oL�:��$��X3��}Y4P��W�i��/��=��IfܐkK��k�ʫ*Q��]��oL��呪�ݒ��-�~k� ��[�0�V8�q�g\i.��L./o�jo�FAq�p�@��?��(*]��aj]hQ�>y{~n��-�[.?�ax/1 ��k�3��5�\e4�W��#ۆ}I�xW,�Ǩ��0��I2��mG�\R��J��X��-=�i��x��z�i��J�%�ݓ�VE�t +=vwr�UF�{<�%+�k�e�u��l㫡�?a�t-{9�j ��Q ��7�F�J�>�QN��W`d3�捌-��λؗ;�D [��b�ڔ��t�DOWM<1=�4�9�鬮1��w��;��m�&��|��Vß��'�Σlͱ��B��w�jO�SQ�Th��.vp�ν��?�E�O�E�ϴ�>(��3��I��쟀��j�9�3�۪�ndڝ��٬"ko�% ��H�o}��^�QQ\w��vXT��|r�Am��!6��@��D�ؤ&$��T1��wrl�>H�1�� C�*�� n +��šq`��xկ#m!A��pUtqK��T7�0䵣wA8ϝ~�ӣ�(�|�t��k2I��{�osץf��z\=��p߶�2MEvq9l��+��\ҙh��!ӈ��Mi�T+dM��i��A��x;�S�Y�v[��nGk名�'sjHT��N̲F0�PL��P�sF�=�+��9^7��B��{�D��t��M��]oXò��H�u4��t>u p? �Q��]'��ׯ��a\�HH|��g5Ϙ&��Q�d�Y��E!�9�� b)7~��v& +�ظJ�*�h��f�3��Z�S_�RT��Q=��M�$��B�}�� m*V��} �ҍ�H��{I��R��:�� |�u��ډP�HVE�~R��;^��Bt��7� �5�+ds_?�&��?(�{}��{��U ׻i�F�W�)`� e��(P��ʵ��X�u�!�N��R��PŘ�j~x��/��8&�뺈i��m��@O�-��>e9r�w�B��ǉF��vj��&�Z�ϵ��5�,��N��24��Z��V�n��M@x��x�� "=�~��]y9��P��*�߇.j��%g2��0��G� ��q#jY��v��Xk��E�q5��sx-�+ֲl&��{=��R[��Řk֡��"��Í� `�峘U�[7��.��G�3�{��/��D"�B7K�g��_��x�&z�� +Y,3��g֕p��p�Qd:Y�W� +�EOmv��poy~j�ұ�ޛu.͎�u�Ku?��Nw��v>��2��i +_a/��< +{Q*��U*]4GGܭ?�mt�u9��n��i'�ͱ1L_�V��g��vH�p9�7�K�o˪(�F�i�_WS��}pL=hlRׄzVF=X>J��F��Y튏(T��0X��c��q��>�Yo�C�o辛e��ŗ��'g�U��k��p!�4��#�W��Q�S��s��F�5ۻ�<��up:�g��0 �"�5�V��6o�R��NR}XWMn ��|�)��Dw�ê��>��Kqj�8Vi��%��\� ��l�Qii�*,��1�:"��Yz'�iWo��-`�p<RZ�/��=��_5[ǡ�L��G�+��A��|N2��X��SAݢ�?6�?��9H7��eA�D/��6��1Ђ� +."!<��7�1�#�53�>��w�>��$�jNZ"��\�{�Z,Q�_ ��޴�6p��K�� la�!u0��o+c��jwؔb0)�a�|�e�9EzPc�p��X�.L�`�d3�\�Q�r`�ᢡ?�U"{C�ߎ��f��@Ѕy&��~F��.5�mx�kX��a��C�E��!vlB[�G��P�?�2vt}&��a��t8\�)h��ϫnR!`�� >��R�H|�\��b>,pa�Dc�GF{��t��,U�K`|�T(��k��W��wjOw��RRP��l6d�&�*+��I�(��w:g��jQӏÑM"�lJ�}�6f,KҀ�.�[2!Mױ��)��.v��On#s��'��?,#�"zX|hC}TD�!(Հ[E�"�>EEf+g@Hd*hN�>g��Wil\W�3�5K��!"�U�)d1I�*j~��Z~ �E�H��W!�R +BJ �B� +�Q��f�,�M�M҆��ȱ��D��Y�2��-w�ܷ��x��N�)�g��=��L.-F�;��b�Uk>��5� �0BD�0!k�ϐ&P��ke�f��@�j�ʕ?W�q�v�r��*V��f*��.q<[�G�Ԧc��`��=��LQT+Z�hzߓO�h��ޞ��V��ޞ�p�;�r��ݽwoﾟ�JY��hB.��m#�a2��_�U�=�>��-,u +Rte+;��0�Й��8��c��GP j4��h�Lo��(�pև� u�u䥏��r��}��$��f�G� +�f��3��ڮ+=�d�:V7�G��1�H03 ��z�`��0H(|8L`ٸt`'�Ô�+PȨ�S +[uU��X0��0��CX�ߩ�#O��l��v��ʗ��!r� +`I�u]��R�cU�9v]^�R�T�<8��1�A�j]l�t�K\��[�w��vmxj.�!jq��[��|:��GüZg翸~cW׆��֮]w�m��u��lg�,kz�Aݎ��60��c?�9��b']��6��˪��Y�u!^<(�=�w��߀�1��ٷ?��=��-p�d��yO�b�(�#/�Ѯ�Y�.��x�;4>��6��R��K�~V��7��/ՃD��w#��"r��=h�1��[|� +Kix��m�T��w��@��]%��6tGP�#�L�rФz�kB��&*ٰ��X�6^��8wWq�❨Ǝ�/!tPN�b&�Q7��=��k��[��!�`4(+�H��e��e��իi��0�\�׃�,2�� ̿vW��E�eh��mҵ��0dk�6)��|zԑzhJp�� Op�]t�Z��/�#�!È&'"��&x�p��<�� Іy�wR�� +�z�T�u��4ݷ��6 Z<�R��%��8=W +�##��wF��sܹ2T�{�=�܋g@[�K��W��uk��D�P�`�Μ�� ǈj�D��Შ_=qA��oI;?|��}��zzXX�F�� I��7�� '�4�o�*!#��|8�\(M9eٝ��7�C��X��3��B�,�|�8��Ǥ.d��!��(Tc��l��} ��uc�4�ehUx׍rM�B��'�Q��P�h��ȕ�)��R�Y��A�#��.�%�@P@�}��5�@�fJ�1ՙ�r�GAUj�h��6��Ln~(��'��g]L�4q��o㈧��q;!�E�n�/�X��7H��O�+�Be�F��F�b�uA�sND��4��j�Q��̘Uv�J��6�>��J�Q +~�K .PCi�N�ppQ�Õ��\MF�M�\��E�jC�g�w�V>P0N_��5QNS@��4~��O.�Z�I�4�j1%�¯*b�`��*��\��L6T��Ht�E *��c=��r�|D(&�]q6�m�Z2ǫ�VNz5��_�fs��B!,�~f2��L��s}�L6��d��B�� w��L��l!_��H!7��p��2�,�r}��\��l�� +��CG��k�n��N�DEXHق�ZށG�8�[+rm'�t��"q�pN��Ǫ�I��gH�<(-L�s�Y��Ђȵ�:=��exS-��{.P�z�p��q�%'IS��Vsu�Vju��w�E��S��}fN�W��mf%�bN�YBi*�a��k~Ū��ȝ��m�o-�]7�e��4o!�V4�+��t�-a,�{�X��F.��甐�N�r�N��XM4]��y��1�X��C�r[�YZI��@��h�(��p��7��[ڸ��y ��,��|��9|Dq�.L�b$�E�P�Rg�y}��o +&�sš�s�T�r�� 3��Τ+�ґT8� +B��8��HB�ɥߠ�[b�SR�1s|��lw �gᣍ��r'��B�P�o��~D�=��1wAx)2U�Gs�Kc��V�LV�Xx:�o��Z5/!T\�h��.�~��\��l��W��~U�;Ǚ@r�?�Y�5��8�FL��Lrɺ��Z 0*��|u��T�g��n��$ܶ� G`�J%��[TIj6a�܁ʑ�d�p�t��ǝn�,��95��Vr��'�<��Ua�v�*��чW�Ċ�S1��7#D�a��y3��c�:GY�W��{c"!&��m�}⧮/��ùIc?gk@T��yOÊ_�Ӥ��%�Wu��Zd*[�]��P�l��<�n�k!z1cr��߭Gyj��;��Oa��Ύ~�mcc]��9�2��A�1��ڐyهiց�.�8)x+��'� 1�R]f�ɋ]2�+�m��f[k��LF +�@V�d,J�� +k ]c�DF� +�#_�B!Lv�m��"v�U\��9HB�8��jTgff�'8��u�q|��p[�z`\BuJ��8$�g�.��g7�D��xnd�s9��G�wq��U�¦�@X��ˤIgr��W��{��4��\�[��y�7 v��e�ۯї�y�� Ǎk��F�a8��p��K�v�Z^*�$0�2�p�Y7e��A�)�INO�Q�+ɜޕ�r�sGui��"�:n�>ZU��V��3��95��IX�==T��i��o��}*��E��'��o�3�.G��m��按�� *�7тu�Aɐ7z�[S��P�l�e��4f��x k��:m$,��6��I��t��3��Q� +y๑�/C�� +�v�kh�@��a�"Þ7�X��`��Ԓ�W��P��y��e�Z��gCi�kjbis��ٵ��Y��p:Ao�(򼏧J"��&��7k��o� X��T��jV��s9�j m�lt��mM5�N��UCgF��,�� k��ܹ�G �<`&QQ��c�4rӊ�Ip��l��ݺk?�� _�]�l�Ap2��>T,�־�fÆ�9o�f}�� +��Q�q��JcԶ��lټnݎ��ş-.��/ a�u |��S�q3�}��j�}��g��}��C��r�#{�lgYA6q��^OO�H�`�i�=LJUX�+\��xQXb)+�0U�TdR��d��RQ�� E��q��w��땦�䜼��bcOSQy��jj H?� ��sg�Ϧ��Oj�nQnՉ,`W�{C3k7��"�E��9�e��f��a}��"o��^n�W��w�P.�[� �^n�؍��?�WIl��A$��)`SI�n��hz0�Kz,P h�"�F� (P �5�1EZ�ۡ@ +��X�(�3Ù!%YK�ƒ%Y��Y�lm��ؖDIg�׾��(Sr�) +��3��U_��A-��,0��0��aXw��> +��)@)�󩌞] �H +�J$`�ɔiƤ@5��\0~�.��Vp��ŉ��`[�m]��{�M��n��jV��P8GB �p�1�F�c�x"��C�5D��D4��H}c�x��x$|4��h}"��<�[��x"?> +�G��h8��;'4��9ޘH��ǣ �P(�� _�� Z.�+EU��h;D��ag��Uu��ԁ��_V��#��>�6�J��ݽf��R��9S��X�V��ř��W>�3�fgV�%w'.-��L��S�skT��.�Fg�� bm��h��?�)�c��Y��hG{��͵�.�U#-�}��j-�n�a� ôt��̂��4�̛�mh�nZ�Y�W�isv>ot�0�V��l�2��%(4��iZ��iڦaj�i挜 ��StӶ4��[F��8��k�m��C�_J3~��Kv�_`��Ox� ~��ˢ�,��Z^3��텱�G��]k5��~��9�2��tQ��]��mՒ�)��@��K�O��\m�Rͩ)W.w�Iw��>#T�ۧ��#��dN�_X]�4R��qklenx0}x��*�H�?�,=�G�9��&Q(�S��0�ZB��d��>��%tP�(3`=섭��A 3�YA ,�J¸/*4�y��'� L��<��``� 5�,j�U_��a��O�9J�/��5K�L�g{��z��:H��E�T�!&T��H�?��7��Z��-��k]�Ͳ�h�^Ϸ�w��ʌ��,�=5T��%9ܔl1m]k]`�pڞ�1 +�@��Sc�]�Ŝftwu��]�/�2��z+U�Q�HQ�W�.Ӈ+ +Q8�� /�1�/A ��'��B ��j)T' �U�x!��Qe +�� !�׋��PqR��1��n�A�a��8�x��o_�w�rK5!p.g*��*v�>�q7�b��:��0�tJ|�'�&�p��CÂ�f�-I}��A��gQV\|�>8�t�3޿��Tw�pQ�1��]�f��ɒ��I��+��O\�Pzo�s#�͇[4s��c;c�� l��L.;�z�6�a�\� +�Ҷحww��즶�hֶ��Ϛ�N��7��t^�g{=0��⻋�k7��5skcm��b��֯O��[ +Bxu4�9��-��1E��q#�d +G"�g`H�t\՟2��*�eB��x8�@a��b'��&��b�.��kNп��>P��v��w7T*Ew�L��?��q�:E�t8��S��H(>�e� +�W�!��'��%�}��τ��9�a��=��gN01�/t��!�� V�̕�2pR��b�D�� +s�$��"K$D��u��=�x��;��"��#3��xZV�t~w��1X��!�B� +%W035�v�)K�:𞭋Ŏ�E��(*5E��z��"�F� �P)Z;�ߟ~^��9��y��X'0z>S+��a[��Σ��+˲��};�x=U�B)��v�T��D�q>�2G� Ħ+}��U2G [� ~!��Kr4�i��NN��f��jZr��bp��OLc�x\74� �ؕ��3�g��|�E��|�5��e� &F��l��h. +ʵ/ή��z��{A� �w�2�}��s�*Hy�+#EU|�� +��d��v H +ɴ�@?dǅ�;� �o�0mG��=��b�\��M�6n�8��[A�>ʮ�< ��{R|�]`��A��W�+U%�*��'�c��tfl�Sg��MS��^�6��D�lc�3�Z=�m�)zYۯ�u�r>�T��.dr�e=ʽ�^��X|{.ah8!��G�p,C��qI��:�KAh!��U%�E&��"�$ +Bj0`*W��>!�7_ +�v�\k�w��0M��}B�d��;f��s&�_h��KP��Ķp��6� �g�5s$�}pD��n!��5TX��`��l�b��|*��o ��#W��lφ�S�u�-P��W��3�v߁ �L��W��" �{�zA�A��+�0F��S��$��楃�d� X��u�L��:A�V8��?J'�;`fC$��O��&n/HwAV��2m �j��q��`À��֢�heMX%U��r�"��2�˰�&��%WH;9!O�u#M�ԕ�s�I�Q1P��f��-�zY�N��}�ڴ�j� �\�:� L��q�V�l�9ho��/dA��Ɍ�3#n��w�uj��z/hg��b93�OSMt�X��}/�4�2ҽw�R�ch[�x��X�}��1�%�u��;c�RX�1z�l�.s��>b>��u�H�<,q�5tb��(0c��#��5N+:�#��1�`��[��G�@�\�QO0�O��& �IT 燆��x�K��~G��ܳУ:�-�L��D��3I3,�a�0��1Ga�f��8�𣟟~&��{�C�0�0b +�RRy�y��\� ki�0<�W��/��h� ��,�J��B�ɛ�(W�[�\#� �%Y)>S�ӻM*U�3Ζ�0��59�W}��ʌ�15 +uN��^,UJsE4��8P�N+�WKƾ\�&�d��]�.ץ�dJoT�{��8㱣��;��/t�\�έ��L(�w��L\�a�Id9��\@�X ��q#�q�4]N��f��;�6:rL\C��/;{~|��=d�:,Y��J&m�%�Z[;<��9_}D~X"y�e��ɣ��"Dܮ�>�~f�� +�>ُ��P.�Vއ1�h��Ev��Lp�7o��6��=L�~}��qs^Y<��C\�F�� Eh��s��ȅ��Y�,!/e�;ڨ�G�D��wr��H�P��B�� M s��񮀾 �Z�v�[��3��l�f�� B��톐B[!p.��V��vm��0l�� l�!7a�B�i�ҥ�~�X'��Z��Y�Q5akzzzI�*�bǮƧ�y�}ڍJmvm�*��r�NzK��kIӫrI�!Le��(+)��/*�(-�ɾUV^TCʇ <r�)d�q0�n��$��bqM��By��fN[��߽z�-�{�~��Sz�^M��Y�j�e�o��8��Z�: �]��Հ�r醩��k�G�k'�ݯ�%�K��~,p��Y�.Ʈ��r��;D$��P8=A�g�@�'��!U_o7��j(�<�Ͷ��R�E�4)b��)�lV��>�u<��֙��Dm�/ֽl7��F��svx(B/�~MWT��F�U(F~ۋ�\��qf�&\�_R\�F~^�8� +�^H5�{�ktC6)D<�t&�o'.8#��~h笶�)�LY�yǬŰ�p��TE+j��K��MxO�CW`� ��ԩk��w��w��s��QȜؖ��*��:��6M��JT�3_�Q]g�͌��18ް��-��Q��t��)T�J��J�TMI��Ͼ{Ƴ��cLmL�F��x��ͼ��{�P)�?�F�yo�ͽ�{��sx��8"]��A�c(�w{�m�a��m��N�n�؈��n�t�:�z��Pt��3�Et� +�[C��#��W�F"�}��@��9(,䧞`!c�)z�^��@+��6�5�[Ľ�'��8Q��?�w*4ɓI ��*�G +0T! �/h�<��G��8��T5.��A�Y�7��,��V�çc��;�y�8I��P�I)��s��K�0��G��P:�fS%=�Z$?E{�)��J��S\Z/&�ix_��v-�e�ܓ��zm�^��D��-��N�X"�2\8 +I��T�c�y��\A��!��?Bᛰ�m��3��K��I1��us��+��Wѣ\��h$�O��G��g9#�K.n��ٯ�`��r��S0�~��]p�u�?��&dX�;��9�/ +y�ŷ@��+��9�33n��v;Zٛ�)�gu\Z-��^�7p'�n�OoZ��t�߃N��p؆��v��i{��"�w�� vvu}Ngp�0J�d!�G2Tiˑ��&}��p=�pSe�NmY�f�]^��kNf� +֖�<�i?:_�g�V��=m��9�Y2^��QX�6�r۸`�~�/5��߀Lm��!�$�*K72�*��n�ӫJ5cc\A��~��(��Y�f{�gs��;Z]�� B�E�."=��ŵE��">�~�N��mԾ�@� +n]��t�7N�ܷK��ZN�ES[ה��ܲ�f&��7�hn˩Y�_>[�TE�؛W��^ކ��})��{��*�i�qպ��<:�F-��К��*�V�Qn�ڼ?��j�l��Ú��9c{�E�w�� +�(�o�8�V�W�ʙ��P�=TR��l=��<78�Wt�l\]R}�._i��wq�\-��+��L��#�h��s��9Ǿ)d�Fm�x��h��i��ɝ]j5��o�~��b�"��5�!%hl5��[[�!��h��mn��ˑ�a��l6��A;J��Q^��_�O>�a��ly?��o�x�i��i�W�4�MK��.2�)�H ��Ύ(�<��d��x��)s�80��A�|>��4 �_��7_Q`�\+��g�|��@ �Ֆ�,{q��]�-��&^!�fE4�d��-M>�vP6ܺ�۽��8Hb�ɉ[*+�,��;k��7>euN�×�-�W�Kq:� +��`��u��A��z \�k�O�R��w4��nR2vK4]�� ֠/4�S>��=0��ٝ��0�s��Q*yZ�F͎z*~e�>0��M�`$�ِB��,=htZ��u�}q��a ژ�/� =��G'/��6ڔ+��rH-h1��O0�,�E��X�Ӷ�L;��Q��:K�Өc��ۉW`)�X�W4.'a0�L�N�8��O�_8yR!Zܜ��hPF��U�x.�1*'�(U/8��HVd|v.��ƃ��Sa�S`;��\�Mi��RRĤm�mTY�(�2h��ž�?\ꑀA�TH��w`m��iB��y̐i�Q�Ɖ*��;��"�ȹ��W�-ҩ1��e*��: +!��ޛ�lT�0�Ѣ^��qJ�ç9EZ��i�B�E!z�B�.�Vo��҉C�P/ +,v?J8��,΂cN�O{ _V0*�q��R��q�v�EJ*��)b![:|A�� +��BR��zXj�ԉT��p�M��se��ꂮZb��J��TRy�킄2�/E�w��TI�v{-��j�1S��uu$yb��Я�L�kM (�EN�pj�Qy�Դ4��Q��4�9��^�.�ܥnX�)��_Y��=�1pu�C�BN�G�_i�xW8Q�s�7X0��=y��'/R�^��jr=;�"�;Ur��I�:�1��S֔�/C�M��&L/��F�>s�s��_$�D��@iib!��&�H�i�:�$1�)�S�rS�h��W�ʲ�.!ф�J.S� ��/B;�K�j�4DH�`¶�I��@H$y��%�yU��[f��]t/O��s��(�h�p��P�K��`e��S��L��E+�,(�bK�B��d�B�q&��I�kiL�K� ��P�t��^�,bd��@,�;�M��K�P%,�(��^��5�u,��5v �EE��9/�Y�[J&��O4��x}Zd��\��mSM֘w��/��e�f�[2��[O��=I3/z�BS6�G;��H�1�ù� � +��8�f��s1�tn�i��}'1��߫��m��a��J�1��zz=C��+h��Ї��{�6��X�T��=�rc��:�;�5[1�G�ٛ��k�#e��}�V9`a�p4�=Ko��u�G��O��h�x�㰌�=ז��~�#��s��|�|F��}��AjH|׸I��R��v��-�ӝ�z�0�N�l9ћ�0�oJ��|{0� �O��>~�Z��]^��u�cm]㹢z�*-D �b�â|�S�Bq�:%�$�|��s�4�K{E�T��d�� :��-��-?fui|��D�1�gN��s��Z׶z��4�`��:� f�1�3��٫\��ep��h_FU�.؋5��Y��R��MB�/�e�䖳/��#�H��6��]��)�䧈��Yy�`�X˪_7�C\�*��@�5��ivg�,2^㺋^�J�E��-8p��m�^��[䓌��@�7�uw��P{|&��Ѽ�,��X-~�nOL��٦��1��a/�J:K f�j��6T@t��ҕW��>�{�G3�M��'��p�O��e�.fw򅰷z#Ç��8� �]0�] �W�)�0��T89m��e�e�Lt ��$7TM�d�6�6t˼��(�O��q��B��{� �.�b��I'��b~6��Pq�=�7�ߠ!�a�[X��:�ɽ��&��wA�x4��^f�R�~��_c��z�q�>B�&��,�p��㮷�s��dVt��z]��P/;^�y�H-��V#.��埿%<��Q-އ��q/i��/a�f_�y�y�=��m��R.��h�Aτ�%��J�)�y�}��LI#��N��S|u��h46-��'� �@B�W�s?�*��^O~��m\�c��qi��-W�0~d��]�3�m +�qgHE��,s�*��Euk��a�� ǪH|��Z~&�ݕW�ߙ� -��Wip�W+�꾐��#0��1��6T8��*l��8p�I@��9�$��G@vbaQa`s��7�vu�v�;Gw�t��H~ؕ��ٙ��{��{J��B4@F��7�sHkF��M�%�z{�%!j��Q�,�5�A.�1�Z>�'��X�=�@i� _�W��|��6��:F�&��f9�s*4��h��:��Fuˆ�m�N�0G�a�D�;��oq�@7$�XA�� g��-�/�dމc{ ��e�,��!=sz�-��/�N3��U�FM�B��Μ�4��V��0q.�O�:�7��{a��tB�:)�5Z�B��;�ñ�U�=��!Mr;��N �(��@ ,�*g��j�#<�:+Z�)��T�X�6v�;��u�־��6��h0UT�xPI9Ո��P��vHbʟ]�4�1��A(�n��ȯ@eO��T��6��y|��s�ڂ:� 8� >]��0Xw��s��º�U@=`��T�Td�,�Eb��U) �A�¶�Kh3ΰ�c�>$��$o�]�km�B +%&4֩|�wҋ6��O�5R�*��H��2�B�Q�{^@�V��ض0K�V��ͺP�&��ѝ�&_�W)�&��h��׹W��W��4/T%J�� Ҹ�Q�1Pn��1V��P�sJ�ȵ�4 tt�f4bt��v� ]�^?P/��x{g$oe��&ЮD��D��hn}Z��Co+��П4w�4ۗnm:h��蕟ם��D�$9hn�s��æa፮��_�(��M=P�ZzAz�=�+[P��f�5A)q *�A��nx�]�'��UT ��z۞��.�2�/c�Rd2e&��I+l;*LOX< ��XIa��oB[%T��9��vM�y�2�53��ɛ2�M@R��A}8 +�z:O.�#r|�� 0㖡��z�<��2d�5Y��/��듂c�wM6�)r�;g^Z%��˿�>M��\��R%�C�y� T��T�H~yA�k�:ڶɜ]�W�� !h�2�\{]��x�m�h�v&�97=��rt��/�?��4�!��E��g��{:sd��Q��;�l�w��>��$&�� +[�Ԕ{�R��T�f�'��e�Dk<tV��e�'5��m�ݖ�x�*�F�ށ�4W��qO[T��9��j��n��f[�s_��q��ll��E�[ȓ=/1y�wlT�.JX:�pd75|��<��O�sD��>�K��Kz-Q5ze��,m2��+Kq�6Ce�a��Y�_��l�5}o#J/��Zira�M�uh��;S�ڢbG�M4/�ggf��P(��r^)~ƚm�W�ȱ�T_rjwG⇬�~��W��up�Ǵ�p;�8|��l��H��"O��?��S�>��P�ЄO!?֪7�ڬ'u%�7O��F@g�kk�CM��z�L��ڞ��1��+�(W`��P=�,9�B^<�P��E��ο�bB,�Y�� ~3�)�dP��y��#��nL�S�ʈ�y�!�8D�@��F(+��0�� b20^4��`w�m;�~[S��b>�0� +��=�ƍ}�� :\Z�ز�pk�_�/��zeU��啕��lٲ�ʊ[ʥVh�6�?�@�l��Ů0Ӫ��nY=�}~��ի�L��FD��O��a4nƑ�^�ܖ��,a��0a�0�',��u,l�Fp ��q k�X�²�#�0�ʔ�j�,�-�σx�H&Nļ�� :ئ)DH�y"�a|@J��G]�h�x� s��߀��9�v'��K�y#�Pd��N��̎Lxr.1i>5S�8�Η��AȀ[�ە?둝��÷��9Yy�Y��ka��;��@,t��w}��&$��$* �8��3�#�"��q��!Pj��a��r'��Zd"@��8dH�҄�3<��9��e�PanBM&�rf�Ę�� <`�g@T�YU�]@n"wiD�x��H���#MDj��e=�m�T�ߑ{�{��36�H�xI^�i��* ��·_w�&��%�"Y��fP��*0H`а8싎�d�r,*d��0��j T��,�B�޻�o�� H'�?p�:�I�;��s��P�o�R�+��gם�k�k\% �}�5xj�X��2��" +#k敛��$$v)�8Q�a��vf�3�f�0��~k8�-��,ʪ;��&m8�M䱘5�eMj#�~�*�� mݳ��ŋ�3v5T�Oqʵ��EɈވYá/��ݦ2��|͵�j�ZթbbV��F��9v�^Bx��H��ȱњ�zP��Dg�Yx��/�/��|=��F��ƎYo��V�~Xp��#@�FI�ҭ�d��ʼv�4<�<d��򅱤(�S1E�;�a��v�2ys@%v&��,��&��&S�0}��֘�R��^G��g&��~m@[3V>N�6iзOB��ˤ�=+kXĊΐ(�"�JU ��O�\X2Nՠ/j׳z�`�-d��8]��=�+XUgj�L#�p�sZ�`(�mmn�p\�C��e�뒦V�X~�lݽԌ��=�U��/�C��"1^��Fd��ΘV��0��n^�nԞ��f�0؅��:�\R�[5��H��/��_� ��^��>��UT��b�`$�ʈS4xY��XA��ݜ�p��V�` +)��h��U�V3��7͆6)�,��b��*~��x�X�E&��H��T䝘yV��z��F)��=N�\�5I�݄T�̃$��LX��C��H�{��LX-3nUY-�`KGz�څ�Xc�U��)�-��D��O�%�jD} e��(�#b̺�_g�y��=KT�I�TJQ��.��=%{<��@�0�q>��Hj`T�ᨋx�3�9��J��~�QK�i X[0ՏG7�~O@��4y�fF\s�YUE� ��l�cG�5o�I��6;��A��C��/h";lv 9�O��~�)ci*��..K�M��j��!�v��F�'��ˡ��v��}��S?��M��0l됌��t��u�}u��A��ήS�r�=�:�V]��Z��I�i��$�p}�5�AoC�L�w�v@9�Kq��M� �([~z�U��V�i��-ܥh4��Y9��r�<�Q�$��@�K��у�`��M�P�P��ib4*NucH�4�Z��}P��1�U�?��dd��+�>��9l!�~�&��:l=E��u�IZ�1��z.��_c��x��^Dy,�(�ƿ��|�ae��֝;�wgs�L�d=�аӒ#�Vý��p � �JZ��Lm�q#�bI�:X��3�5� eg. �t��I؃��#� +�b)��M]Ra�@6qg�>V4�5[@N��s��~8�j��I�1��L��04R��ӿ��{ �;H��A�Дm=�5��ע ɋ6��˻��] +��u�K|J�8��`;|��6i�$�6l��z��~J&Q�lo� ��W��mCM�g��:p\o��8��C��۶o)*\_��mú�/�}��+�W�y�G2`�SH�Kl�}�E@�a +��l� +h�D�A�3�&~N��c-��&��o��˾��Mq|VB��Mx[@KȾ��,]P�n�m �u'-��-��n�@�N��6x/�&��3�ꜣz%�y}��؟s�v�v�l��~>� +6��B�|�2�Ǥ% �C��8U,PY��j�A�O $�ۨ��!�b(��ٯ�ਪ3~��&��$<�hQ�2��G�Δ��v,��3��2tF��ZJ��Kba��`-`��)"� "��߹��q��l��xvu�zy��^��˚��d}��ݸ��V��%��k׽#��V��y�7Gq��-kW-�پ��E,��U�k֬Y�j�8��LV� +�U}�L��7�0�fA ��q��b߯W&j~� �V��nJ��`j߶��WU��t +��70c��⺤d��S�"��s��b�n<��8��Ǆ�B��>�CN��G��%畽��Q�rVu<\�G ��5��n0��=ç�+{y�0M\�<�k�է��C�t�Z��"B�9��x�AӦ�3m�4�)M�*N��[0j�5Lqtݤ�R* Y�ݤ�x��I��c��>Jqh��c��صn��R� +\��CCh��l0]��6�$�!)!-��Z�I�j�4�f +ȭY��Х�*Y�,%R�h�R&��X��K:WQ]�X��S��d�Z3��;��.Bw�1(`��Me��{~��Ȣ�EӬol臱 + :�)A��&,�`_K�g��dN�X�m�Z|�|�A7ù+�� +3.h-i�ʔ�)�e��&�'��n�V�SR��-8�G�)��%�Q8��j�-��*) ��@�L��3(�~iȁ0�t$��Ϝ��Ȑ�1LCl�u�� 䶌�3::��"�h+d2��c��(1�x��tj3(��[�9.>q,�ّ�"c��F��m$>�mH��Ȇ�8Ⱥ��&�X�'��s�L��"4��.9�at! ��7��(eM0SB^� � ��J\��G"Wů�:d�ݶ-�US�� +VN쭚�*�躷Х7��QW��f��Ԡ&�N�l��K��PpuO*��`�L$�Y�kH��_��ї��^��D{ ��NK��=VԔ\S��n"x��N#�1�NQ(z|'��/��/��Z��P��T��V�J��:�^j=й�4=��x�D~-�� +c��,�7��.�5o9m�w��j�6%-p��.o50A��-��ǆw�} +>��q��1͏�e��R��:X=MAȇh�X�H["��<�x#��S��]��P��^�� |�OP2F%N\��G�l�HF��5��[�>��8򤧹}��"��p�?F��{�ӆ�ٙt�ʲ��L(�쁸�s�� +�^Z��&:� zS�"�o&��ϖ�=}�޵�g�< ��ˊ��4�K$@9L�j��hL�*H�f��X3��)��y)j#8.ZR#T�HiwK�@�P�N�r��rܿ�.w��(��|�Xq�?� w�B@��跳�e��±�j��PKZ��K��rT=� +�fq��$�tYb�r��Ȁ�5]�^��쫽�1yY�㾒-��d�'úf��:��]$ ��>?��3�jW(�2v�]��9�D��G��p2��[�h��ó,��6�")�CM�&@��/��M'\�y�L��dd) �G>��;s_b�K �Q��r��V��*�:Ū�I:��GF�۾�݇�3w��AbӃ_�3�t�b��>Pڃ�� V��3� +��瞚�*?VVV +��*�o�p�)��=eJ��<�7ٍ��0��5��7�e�u��;�.��0��`E� �bQ0�U�� *��6Ԩ 6�4r��"��ԥ�⾐h��%��l� �{o��8U��w��潹��a��~��Ҕy�G��#��_/��֢Bj��_m�z��|@"�:�L2�W�ߺp�p��0��-C@f��5:5y�ꍉq1Im��E�X�r%S�&�<�1`gK��˖,� VgHN�_y�5��~��,��ܽ#�(��^�x��Դ��w��a�/�<�NMߑ�پ��EX��¬��v�[\kCW`��=��py&�+��w�y��'�L�9��tm]Ueg�v�7i��ce�W�:Y>�� +��ƺ��[%��GH\�Q� 0/��f��O�&ף$��F�@�\��ߑ8o��_�� p�K�^Y��K}!^�݁IF��&*�XY�do�!�IdFF&�,�O�08_; �ܳ7r�:��T��M�p��9>f#܁�D�/)%��ԡ�Y)}��E�� rDl��1��%��$�󏲐�b�͒7�q�[_��X��6ّ��ߞ��T|F��#s|�X#r��k��C��ǿ15̇�6��M7��I��|�/S��~�e��$��F�DW��Ko�wP��4j��S�t��C�oؘ��|lm�[��u��5>�Z5 ��iPSEj��j��;�,9�[óm��,�l�z��t�L�P�r�Hc9��Jd�_�@� �M1)�85f)�j槛�@(εq�,/8Ç +/�Lm;ȿS��P#��B�Uf��N��K +��E`{,�Ԍ��$ �v��>o�<�d`�?@ ��Y�"��|�E��u��d{�rЃ<0�2��8�.�`a'U��~��P��l#��H+��aQ��غ�`35�N�5�-��z��2?_�3�b텑�I��."�b��k�;Z4��<�ĕ��rO��VCX[Cjrϳϙ��wr�;��!�ުt;2nw��[f�E9�m��74T�˒Y��l�Qw46�p]�v�l��$a`�ue�Eh�@��6[PѐFp�EH�'��a��r��3U��E��Y��qD�b}.�8-��.��H��4Y(�B��?m�% ��x�HȄD��ჳM�@"�[K<(�x�j��>l�o�hi߶�ם�%[��=��4Wb<��yy5�mu��!��-n��!��;�1Ә_ =��6rS+)�f��;7��(�;0bُ��=��T��4G�_��b>�pI�ڤ�gy@HG��<�'�̤�r��ʧs$B�M�Z,\t7�f�̐�A!�}�$�,�[Ļ:��<��ũp��j��qa�K|�#�_�j9� 'I ��p3'B$��W��9`�A� 2�fk��>R�Q�ըx�iz�DBg)2��,��#�/��D;�� pd~�V� +�Yhʾ��qr+��|y�}L��e5UE%�d�fRd~5 f1�� i�H��+��G:�[�c��f�]��[�J��v_��p��`x1��X/��p��x�?��v!�s��p.�&��NQQ�ֿ.�ޏRO�!��<��ےΌL�P��ab��16 ��g� +��i��b�!�_�s,��|.\%��Ss�8K�2�jkk�]��V��!~��m<��+{4�My?~��#s�帧Rc��bĞj?�yn Hgp� �X�9�)�ӈ��}J�#KJ_� �rd�XFO��2+�p��ѝ�.�',��9��!�� c��kψ�UsG�7�n��u�'$�j;�b�W]Ojz�5S(5t�:�8�"�>d�,�3*��k勽>��t��uc׳_c��|o��hw]�n��6[@؇~�{׃<��2�d�J��<�^��+c--�K��h{�+Y�l}��[ڞV?je)\`�w��[*" 0��)�#�`��3 4l@�J��Ƥ�O,��ᶿgGُd��9|�Hu��!��)��#� +��^��.��0p_��{�DM0d�7�c�3��$}��T:�A�L�G��m]P��U��C��)³�P�U�E�:x�:��\� k �$ +�SO:q�ҵ�+M�n��f��bp�7�R��B[e��zq�掌0d+~�x��v�L��p��$_'�=$��?�<ɯ�tKr7f��AX��¥AN�<$-��)��m6�[�tc�Rj��`L�%��m)�hS��7��$��~�Eu�|��F�"��P��F�DTb�-1J�D#�8S_M��#c,�ښH%S�U��:�N�FۆdR�It��e��wι��,��w��bs�v�39!�/� [�>��O F��Ē�O�:�zW�b��c�fTAզ�)S?`��@�3��0��+�㚢��ywQ��5n��~�2��ฝ.��3�}�Ǫ7��؄��s&M��7e��֘8��34:,�_�@�e6�¶�� +�=T��~6,p��V?w�Dr��g9o��/��љ�Θ�mD�J[��{r��y�f$�H C�K^ߵy놩9�>��孋�=�]̦�s�{k�dMGI=5�clxԨ!��$��6��s>Yk�iZ4�`��{vU�*�7p[�uw�YO�ѷ��_Ã��~��Ѧ�^�L��Ɲ`��Z8̝�_��c�V<�l��e�8��%�)�U��Se[��?��y�{k�d^�c�Ɯ�<��E�D��l�}��C�Ɛq�!�wC|,??��N!�jp�Gu}��r�%��3��Ì�ڋ��qQ[�ˋ�Y,f�>�pޖ��3'��Z�l�ȩs�1n��W�/�^;6}qnvI��f��u� +2c��ɖ��^��g?J}��evut��Դ��3�F.r�[g5Ą��:ױ��p�ͨ^��l�;�z�-�N�d�ƭSc��n�5��L��b��B�,w{� ��ޢ�,(}M�p�W.\��§'��_L�T��=8kb��ܜ��PY��t�� +� ��0b�KN�79�2�):�_��XI8�-��ʁ:s��e�%�#%܃�X"D�ڌ�9��`Bu�&��^`n�^D��j�l��Y�� +�Rnad��p�yT��+1��#ͯ.�� +̉�\��&��ͧ "�I� BP�� @v�#�7�q݄P�� '_=��B"�}��FDF�c�у�-��X�p�=gX��dBTeb�� +L�dH��<��b��D�c��#��P�^@��NZ[��d�r�9�A�+v�h��{�vӜa��=��k�S��E3�T�I�[��_^2 ��CZ�� +E��3�j"��5� 1��>"�/�݉�PW��[��B��T[J��b�� &�3 �1q��.�w)��1��1��(�U\fe)8EB]-��3C�(�3��{��tƿ� m�MYL�0��X�=�6��NN�V��7-�M�o��p��uz��9j>-Su��S7.�s��5$��^s��u��щ�պ��c�s��ߊ�Օc��A�.�(�nM��Z�bfɼ�c�?�ҟ�+j�N��Y;�vo�|)��~��(��ԍq�5&:%z��JF�^�~)1#��0�AHT�2��Mعi$��f�� +�S j�S�I��t�`,��Uٕ�_�(��CaC%�0�Q�#׺]��5��p�&I��^h�+��^��k�te��V�GMC��j�v7�]͌��y=lSx��Q�0VN"��>��o"%�8B��T�8B|��v�d)��c p��J4L ]��(�"7"��2<�� &"�2!?�fHf��]��X/Yo�9�d��!��ҟ��؞�a��^��z�k/��-�~'+��K��Z̀؝�d�g��ٶٖMZ��ϽD�Ҫ��ue +n&W꟣z ��Z^��E��жi�� ԇ;��Ч��)reZP��#�q�,xy��I5��AZ��!�Ό�gy��=qm��(�dv�ϑ��̠�Wpt�nTzNCJNG)b��W#Kd4�:�x]c��2��X��!*!�K��U��L��L��.͞Ќ(�(l��!~�W`�vXf��J�Q��D'0��/wa*L��cꆩ�-��9ȪF�~1��Zb��u�tUJʤ��;��O<��Ġ��[��>u��k7�� 3�%��sA7/=qjÚ�EG�I~�~�C�'��SU�9|ծ��'E/j7��:O�I��u�5� P�m�i��&q�;�k��E�p5e�}2HV�ΙCֺ@��z�w |j�8�0��pJ�Aa�2[wP��4 *�c��G�݀��bVR��eb�j �?��7�6I��j=��P�{{�$�j^j`��SO�xv�m�f�[��e#�)�>�d:H��]�5R�搬�p�ȎLo58J�R<�J2E�l[�Ag�ŦQ��4�ԫ)��ܮV2۱+��?% +�3��/E�n�<��mn��ұ;��,�m/�)��Tm�͕r��W�W$��"�O�rz��l��,f9 �*)�Bʩ�}��+��mX��jޡ�!�8��f2�0k&@#'��K�� +��tXS�yq��7��3{�9I�-��G�� $��v�@u��%�v*?�1��n�� +MZ�8o�ƭ3�� *��;�=�� WF��"�¾l#��.>iۻ�`"\J��`o��Pz��LG�8+�5��o,�C�*z�.��M�1q J�K��F�-�>�<wp*��sU6/��ЕV�9�6�X�V��\ɖ*��J��ђ�Y�l�UN/��p��r9�"ݮ��m«<��rP�B>;D�=/��Kf��a0��J�ݵ��z,�(L�@�)��.ǭk+��c�!B�MfB@�M��+2��j�2�}iM�4{�<��9-�E�;�8B��@�J��"��ù�6Ѿt�y��'��o��l�|�~�vݜ�ޛIN��{��:'��zER��2k��b�I��t��g��t�IXB�t�ԩ��ϋ��9HM�#7��C�l�� +� �-��-)��d��}��֭92*�nÑձ?e��ݳL2��9��7�l�5�H!��7�Q?��v�#D^��H��_�3<�Bæ~'xˠ�l�i5�A -��ׄC��m0��A��+��a^�f- ��e5<�x�RkT =Sa�礅��39Wj0�Zd�aؽFm��.{�i!6��xKFg�@�;.cE!l�*p)цP��e�\N޻T2��<��h8��D�et��R(/!��t.�]� ��<��ʃ�a��t�U'��?��X +`�V�SR��a�Ja�?o�~` �C�+��q1��@�:�bX��T�Cҋ�Ҭ�T�u�R�k,�*h0U�Z �Zα��f�� ;��!�7т7�-s��M�-��݁�F��[��)�l��bz�� uN��S��W'˭M��e�,aP��̞3ad�2~��G�vK-z��gD +D��d��=�<�s�9��=�Mk��*��;LP��UY��V}��-��4�ظ��+��M��K/�V�g�k�Q%wjt�'��.�%��)O Jy��̫5+��tA�s/;�|�;W�ʜ��`A��&��Sx�,Q��j<��o:�~��Xw4Q�i�)g�B�R��ܶ��5{�u��'9\�t��j>q��X��|�B�~�}��K��g��4�ṫ��W�S��v�g��;wǝ��&��q +�i��9&��N��|t�{��p�6�� 2��N>m�ߞ�8f�;k�o��ǌ.�= ��(6��m��φ�ʗ�� M��ɧJ�H��Vh}#\�<�jח�r��]K3vYN=��,��VZX&)Y$��a��'7��n��%��5��w9� >4�F8��b��Ĕ�Bl0ʨ +��YJY��p0R��9Sz0��GUΣ�!4g��H��I�G��p�.��J�h�f&M_��`��,L�� +$��=R*DX6Ȳ�h$ ��S��)�B��H�[#a��h�|�[�H��ӣ�py4A��94�>b$#qV�癖؅g��%��+�sұ)�F�Y�1#2L2�@h��m(��[��%ɇ�N��H�C��t��H8͓�aC;�ؙ sZHDc�}42t�O� + �PM3��ԑ��ģQQ�^t4^�'�h�*��^�(��q)j&� �J'rc2��n��J��m��ҵt�S��Y��Q]8��,qE�p ��ʟV5s�u��T�Ԡc��ډ�V��@6��R�RA��%6HNT��+�(�;�a��5*(qX�ڭإ��Bɖ��#�%�"2�e2��Pq/I��Q��y�K�SCH��Ci:�~��x��Z�B'ɝHɄ'�C;�Id�Ѥ<�� +�!oEK�*��n��(>�0!0�� ###�Ա2��䅑"9y�IQ�Nd(-��X��bY|9�"]eU�UFe~�Cl��E>Ğ��)I�g�tn��X*��k��/�#;��pޚ� �~c��%s�g��ǯ�� v��W��,;[^5�e��v��zW��b�)�Nl�K�� M�F,�U��b��.�l�\ZZ��K��v�B�X��.Mj�z��d��<�ʴ�v��T�IB$S �럴b5�,Յ|>��4(�㺕EE�**�*a\�,��*uf�W��`��4l��X(=RѨʌUtaȰ�Q]��\�b{!Q�î�[��z{�q��zz|��}�޾�n�_��|8�ߑ��s��Α��׷��{��ۋ��V.��َ�lY�y��oz|�ǻb�� ;W��rݵ�k��w�kK��޼��~��*�>w\�ٓ��N�`f']\�|��09��*GÏ��%�L�Q��K<��3�aD��q˾��Y]��R;l�WK훁T֐��͞_��_HG�'j'vy��&��XvI�[�4��q*T0�`�<�� QYvr;�S��W�}��f��P;��)m�I�W��i26P(}t`D�~�Ϯ޴I4�&��s뽮��&��x=n��v7z\.o��ƙ��x��F� ]H�Ԅ_*[8�׷��wVtu.}p�ʇ~��W�Z޵l�÷�u�<�޻Ǐ��ǌ_K��N�ĆP�0b�'��"e��? Bl��-�d�B�,k<�5q��͂�t�%j��_�̫�2^�j�k&Us�`�F��U7��bx��嘤X�QE4P]㪎0D��(��R�D��`?��J!CM��"��K�{N��B2N�(q�� mL�c��I/�(�+�ٕ�@SↄK�(�w +F+��!N��Ƙ��`s�) +�q*�`��خ�lB��nЉ$@HB��ٙ=��7k� Io��LO��c�~�{}�)y3"z�F�3u�?��0��c,\��4S-�Q'��8j>n +U�cl'9K6m~�G-��ӨRc*�x��Ɛ�l&��"' �t؝} g�JK��@6)�q��e��.ܓ�ݎuw��]Gw��U�}��m�G��Y��=�#��ȏyFg��o6�kŜ�.N��l��ii�̛3aք9��.��ko��1��!��F�+��܁�6lp+�h� ,�M��'%Q��i� e�/�b&3M�� 7t��*U'N�.�\יmq �� +]�Z��S�6[7��E��r�H�A'��j��-o��H)�J +�4�(��1`�L�� 3i)�e;g��Cr�TI�Eh�o�AG *U�D6�!C4�U��9�,`Ww��'�I�N恆�ӱ�-¶f�y]� '��!S�[F��q +9,h��#EP�h� Ƞ#F�M}�҉��&�S��p�3�R�n�VXeԅ�ϖ=řᬥ��ƵR6��GյoK��~�E3��j��u��+� +�]��o-ú5w��_\�v�k�g.ِ2j��sf>�lif��F�ݞ�{��6�}s�J�f�=X�#�x�ʫf8��ͤ��l٬��n��cv��A��^{a��!��V� +�2-��y-_^n^#C7F�4�-Յa^u!��7�$]"_y-�Rq3n��O��ޒ��h*��na]a��K-[��N�b]Ͻ(��rr�4,�b��4|9��&��-�2L�Ҋ2ۓw�>��Fj-��sE�V�j^��-�Y��ы}��X|Y�Q�K-�T��E��#WW�R�Sc� +��@�5�'�x��*�ba+ʡ��(��C�*�r��+m�(h��ˍȾqM��!;�u��1'yU��yʩ�5|o�)Ew.�)� ��H��U�%��~�H��F~I��[%_��;��˒_�+A?�>�?��J��ҿ�Y�1bz�*I��z�&"dG��0aϭ��Z}��j�X��;j�J[Xy�;QHg�T�b�� +Vє��f�]z֫*�W�Hga#��~ � +/E�w�b��*��u�� 54�;��!ĸ�ICC�� ĸ3Mf��C��^�*iB��f�q�F{��F��+�|��!�)��]5B��bxb�5�� WQ��G��߁|�h�g�]C�+�&��0.Q� ��N�0�� [�6�E�|��'�)nx�A/f? �iI��uM��9;S�/p��ՀCwX� @��t dI�H�^��GE}�G��G6��.O[��B��N��I�l��B��h�eϙc��gt� 2w2��U��?��ZI³�*�p��wދGA��Ɏ��{��<� ��޳��so5�Nb�Ӯ G��t�徦3ﰼ*8�+��K��z#:X�sѾ�R�� A��H��C�[��:e�im�p��|Ao�>φ˷�n��_�O��g{��˽g�?��]��q*��O(�&��?��A/#v+��o ;'� v7%��|`�&3�]��5ì�F��)�V�Z��p��V��{�U`�ę�n�J��X�|r�lI�/X��W�Ѽ�:� L��wC;%�m{�2�ˍɱ��_�B�2��ue�(i�[1��~0�~�q�]L�G�R��id)Dn��G2��|$)"�#r��"��TÈ�)Kc�p��(Bvx��]��7u��=,��k�&&U�"`"�Kݚb�YO`hC:�>�s��)��D�CPhC�N�ԓ��u�]� �B-j�]i)3��\��#k>�� ar��{�c�m�lѡ��S�͜�I��Y3_�MC��ų��_�m��: D��P��o�S��n9�8d<|��%.�(�\��>c#��>��k��^�A�L��.,�~��a��p]��a�� [��e��?��ҫ:�� +�&3ن&1,"�,��(��PDAD�E��,Di��=�(�H�@%�T �)��d($�2��z{߄֞R�73��2j��s΁�,&?n!��6��>��2>@%��;=��n:�d��˼��[��n�~^��-�xE��5��0�:�^�Ҭf@�학y��km{<�g�.��Zo8b�� }�0�.��zwm��_%g��o[#ܰ�� r LI��(Q3��_��j\�7 �j�tR�Us�Q��]`�-�T/�D0%�a�S��s��)#H�-%�� :��+A o��@j�A��F'�� ۓ��J[�<��*z�}�8�_L�4��6h��7�p��T�.8�A)��L�v��ŊxX��D��1 ȴ��ɷ�$T�*�h� +��@��*h��N9��%� +�N��~��T��w��XU��)oq�~;��h��Fj��ڮ �hx��A�M�A�QB�+��y��u&�Zܣ��T޺��u�d�Q��'��F��Y~��R�]g|�椻��r�[RZRv�bY��ʢ}iIIi�w��biYi�ْ�j��K#��"O�h�� k��2x��Z�%[�f�ߡ��ʋ*�{9:@*;��ƫߐ�N9�Ox�@J�I��\�|m��i2d��)�V��8N>��~�ѫ%��P��Yd +L'c��;NYÌ�1��_��d�sΙ<|��D�}->7f;}̵$��4�U�� 1y$�:�؀8V9�6��I�K�BIƑ��B�4lw6��md��v��m�;��S�.��H|�9��/��d+��޾��ER�.�^�d�A�Ml]md��;��0FN~�VQA��s��[��䗤�H��Q��~/�� L��d�U�r횵�]��n~��_�دņ��[��!/o�� {�u�`d�(G@I�P.p*8�BPK d�(l#R#�q�]�6�V��h�R$c�� V�(�s�]~/#�6��Ԯtc��y�x[ѩ6��ϒ��4w,m�Ň�I�M�W�N��A7A�#�7�_��g�l 3�$�ꘙit��^��ηm��\�u��Euα�pz�(��;�̘��1�?) + �g+8��=׾��zG�Ih��j��=3��u򤧶��lxݙvj�&,��w&��*�J/fCJ��2Z�s,5O?l�I�ޕ$f�.Ǔ3bFqA��TBl�͕��ِL��(ǐ��v#uh��`k�k*��Lw�O� y=�.�g��i��6F�4oG� +� �/+I��l��b��8�zo_��z&�r�3��fƩ�&��0�D�;��WK��o�hO��J��[�@�N!̺��=x��QR��6)��~G�%�0��y�d!*V|��a��#�_/��e7�/[~R/_��dI��{�e��WMU�谥J��n侷/��b��߱�O��U׮�Em� dM�~҉��7� ++nAش�x�t�*+�jB_��WB�Mף�JV�"(R�Bo� �K�ð�@6hxKo?��K�&�r/׾�a�s\[b��`� !-��E� +bM',t��>^�)qz�� Q#3]�vP�A/�ߏ󎘘��>��,��"z1x��8w)�<<��L��8�8S�FD� �.d�Z��T7Ps�m@�`CZ�)j*hȘD��~gH�(��-��@�n��* �t��!7O,�\^�D��?E��4� ��T ��ͩ�Q��dFl��Y��H^�j��>��@&qGe�C��5��D�� Kѭ�oA��l��ȉ��io5&.S4z�U�ԴvMT�Ң��G|��^�h��W��:�9�d��83q��_��{x�@��1ƞT4�<�Z1�c��,`v��FQ�;'H��0? �ĵ��&��3��Q��F@�PpfLe e� ��$ȾŻd$�g��ĭō��1;�I@y�~1l��f�C�Ѫ�!��.�?��m�\�68�L1�z ��W�bZz1�{#��̪�� Xɤk׌]nWM\\��W-�m6��U��s@^��*J��u$�|RELϫlH�>L��А �U�Kߒ) {�5=�JW��`y>��X=��g��.�F�?VF��8��S�`}��*3"��x�f�#�*�ii��S��PIQ��$u��8��z��@> ˆ�˰]��pVQfP�#"r7[��p9��V�/��i^»�~4@��6��.��+׋i��H}}-�`�E�#[7bd-W��F1y�*�[E��U��8�RA�\�O��[��N�3B.i��M��]=r�a�� #ڗ��,��Vp�v��\�X��_�x��N��7�A��2ʚ�X��`�3�T��<,j�f�G��1{A��G=4y��?VW?٬�f��e,�3�I0�q�x�Wn��{Xw�K��]ӱ��}�pш�`s�f�Gzυ��XL{�,c]i�E~Q��Xׯ-�7~�N_��COFb+�E�%�{WG�^+fK6��l�y19"��wjt��f9��]�\�:��)� <�(��j��p�6�?kv�nR�p\|��8�� qN�l-`o��a��ƕq]�V��t�+�s�1�5�q5��1��5s�Wһ�y�7�� p�u։�Ϋ!*��),�tW�� 5��H�� 'ɪ��W�%�[T�H��vo��{O�zq%�Xba��s��T�˄�"�un�ڴSl��E�%�74r��7�G��Ɔ�!��1��b^�#W�Oa�4��^��Bp�ĶL؍رK �w��>j�� ͎�5n�QvJ�0��1\B�(Fr.��H�2��'i\4��NnS.��r��w,a��ܪ��A�dzU��U�lb�ov��|�L9i�K��\�*w�&�+;bb�0Ӄ��7@��ܰA��vjY��핥�BRUn�𱺷��|"0^�m��ܶ��wԡ��Yl2A�.>��P�`��Z7��E֍��6Lt��/��G��gnR�ctab��)��]�t�]\T��0�|��@�F��>j�V�΅!�5Lӏ:��X}P��6�l�%��g��#!We�+��ʾ��{軨�O�&��+��<�1T}[�� 4��Ҝ�V��{g+ݕ�뉦�%窒2�t�_*L+�cU��t�q��^X�,i�,�j}��fwz��(\�<��ؒW��0��wVn��GSk�e�A��Se�YY��a��S3�kZ+�I�˚A�~�93��>�C/�&jNָ0��.��Z�)X[��g~%�(p�b�ܷ��1�F��vy齶t�STh�i� 4�C��YE ��g��ܢ��}.�O�W�]��>��u��7�f��o~ыZC� ��:�)Ass_��m��a{R�*�@�7 9��?�ڛ{tN�SD�U�\�֌&6��wX:wJyg��*�z~%qШ�V��%��c㘡J��eg`%��cl#U�[��Cu�� +O�!�� ([Ǩ��dY��-p��R�~2D�v��Z�}��2ù�L��t��X��aՅKC��\υ��W��p%_u��._=�1Dφ�~��Z[�� +5A�v�)��?��z�uꕱ�oe��n�}e^w��ㄬ`m�)s�!� R�=��DP ��8��/�!�ί!\�O�sww��Y� +}U��"�jf=��H8�Dv�)u��Y|��A)ʔ]��93Y�� oS��A�Lc)��)��u;��PH��i6D �+7�߿�ب�dA��\1Oqŭ!��\=�y��Ͻez�;��H�/U��%!}��e�'� Y��,(�-�W�� RU��PV��9��S ��PX �h��uU�-Ǩ�˒ +h9�*��ʓ;"N�B|O�7MT +�"4>�Բ�3Jߺ�~d�h�%�Z[֨��F��K��D� ��u��q(�|6)�^�"�B��2�=�~^ys�ن=/�.�Z��15��xwę�Q�Z4��9s'l��{�� wv�V�S\G��@o��UK�I��ʁYTW��ӛ%��.�n#�� +hje��^� +k��&+:�]N�J5��I�s@@��m2�Pd��Y��q'g'�,�<#(m�bʕ�#�|7��N��@g T"��5x_c�F��+d�kv��&P#Ɓ��D�tb!h5p��>�=�MM��t&p/��Τ��*�Lv��Ĝ�0U[��0�z�㭢��c��M�MN�ӳ��W�-��,X�-|�P�q��,M�� x�[�ĝ��빳4X�}�{�UM�'م~�9��F6�� f"ՆA��>��9�fFo�ck��_#�G��xne��*+��ωkY�mDӒ��m��k�u��#��;��j��9�� 7I_+�ʈ��^�/ܖg�#�"��2j��1`~��!�;{� u�Gj��(�˷��«|�'�^B�eQ��(7��[��!�~��X�].�1��r3��˺�D>m�ԻYL/�P�ӧp��g9��c�d�܃=n��(�)�eG�н��sJ�2#�ݙ��>�sKoN�R��V*�l�~{�WzUR�8��L��B�?d�C�U�ʶ�x��#�6�� ύP�~�X��k�EΟah��L�G +.U��ohD�䱖z�0�mB�3Z�� ^�0HբU!��Kr�K�P��dlǹ� ��ܸ�i��s��\�v�s3�ܠN�!��k�y,#��ZgC��QF��O�K<&��;Mg�y�3ix)Vтlϗ�U�}V�-ԺG-�9��U��e�U�!��A��.R�,�~dF��8�m��wک3;��.��p ��׺d3�q�? +2u4�B�w|��Ǽ�I�YQ�M�+��i�R� Z��V4�׆��Εm6f�~@H��ex��Z�lv��L|��qr�w��ݭ�;6� Pk*�t�U��x Q��$M��Mh��&��l�v^��%V~`��;18KR5C9��6c��D�r��s�T��@ .��Yg�-��趞�.�� 2��P~�,��Bڋ�(��μ�;V��u�)�[M�޳`.�LE��?�U�y�sȰ5Ċ�C֪��S^%�U�W��B � i�X��3:� T��i�jm��q:t�igt��a !��{YH1�!�"� %!ZSv��%o�˿��[#�M��?�Y��,|`��s^��/S_�o��&J��p��l�)-��ۣ��!��-yؾ�Sp֗�jeJ��<��$_U�!y��xyT�+r]�0P��MF`��6үհ��'`o�_#�K��ڄf�S�/ ��5ﲦ��6��58�v �2#��q�Ut�T��r��̘m0iZX�A 53%��%��PH[�Rm�2�9(f׏��VdL�22�O�f$BN4எ&��~MK�g�|�g�Z0i��Į��[ Kw��{|��yڻܖK&hбB?@��A[V�Oz��4H� �O�4%�Vۄ�t�ޥ��4x�[DGu�+-��H2��}� +'��g /�1R��ꠝz#�(�=��m��ː�G+�^H�}M���dk�j&~�C�GX��羅�p�'��t- 2ʾ�D�� '�@G��ki4=��봩S!sr�D�b�/�}��>��kܞ�^�� ]xvD�dH�q�V�L��A��Dݦ�\��l\|�Ԣ�<��])��l�S�˛��j��x��/`��̌8ZT��iK�eآFa!:��g0JHT�%>[�h5�}|�~��j8)D/| c��$�,��c�zƾ�Z85�� RL��&��6#��C�(a� BF9e�8E�sT�� +�k,�7��!�As��m9�o�X:@�K�m��jZ��8�]8Rq��Wy��&�6�v��1��H��Cŕ�y��\��`7=�P��RM�~� �s2/�ڂ2�� -�W��j��."(� +��eO�s��A��۳Σd��=e��{��;�Y^��!RZ�W�a��e�*doYc�]�E��~R_VV2j�Ja�By�[EҲ5|��ڬ��m�l"��̚ʢVQ��[4"��٘7(��e�<�UvĨ>ܙw��p�΢�On�P'YAQ��9H��Tέعk{�^��3vT6D+�Kjr��M'w�=��,*˲��M{i��3��[�1y�:�K��e��i��3^y�9�� bq1��g��b��|�ŭY�C�v�0C)�|��fu�8~�E��^aw%�R=�y�˧d +�K��pL�%i��x�/_x\�H��c$|.�»7A-ԅ"&%Ut�ahī�|bޠM�J'��k�W�~|�!� h��i��F lמs�ǻ�]��YzX+ x)��e��iont:��9��!c�Mv��h#��dkh�9lMv<�ۚl�Î�͆޶��۾Pnayv��d2gf�컔�0��)�Fd��=*f��G��i�G�ł1۽�<��?,��.�㺱2k�=p�P��$��?�� ۝�R�x�D� 7u�G.�%PU��i0�3�� aa��f�!F��"��Yd��̀��ԥ��Q��=��.V��y��u� +엚��\()��EΧ��]xJ{��n'��f��\��O㺠��rF'��Q.�� 3>{�r��_��3M\��T��*3�V��|�{��Yʃ��V~��%Ŧ�j��*��j}��2��۷�Vʃ�^F�N(Z06w:o��"��[4�T�1�� A��Vp��w NH�8��[��+-��@M�>͈#�n�� ]�p+��`��z�}��F2Zx̞̌��u��=�P� �<�,^�^�q1V�o��ˑR��$��?��X��=K�T��}��%�b�HetI*�� ?L�qp�YnL�n�Tv a��#2� �A��:/Z��[.f��~p��Ħ��8�ǜ��r-C��R?�bkREF64ak�*y41M�$��pb0�g W �� H��c�@��ru�d$7��8�6� �Lu��Ӈ ��37/�V8�� qC6-��m�j�׍ҙ�E��sCf��!�A[1,.v��}��ɏ�M�{��_�u/S[�RlN�R�/{$j��F��C�%��(�\� n�x��H�f��^MI��v�nF�N��)[�� yw�d��p��mo��ܚ�3{Ǯc��D,�)6E��h�=}=wA��*�i��6з'�|��;9�M�-C��*$��ON]�ާ�\v+��S774 +��ʻ�?9v��C��˺�q�<A��}��\{r�Z�<0�؛1#S�M�u7Z��|P��~�/%x&�?*�w_�׷6��=j��g�b��<�oL�0�~�� 8��ϢzltYKb� YdKYE�gg�� 0d�6��@F��a1m|s*��k�[d��νdt5\s�p1V�}�� 9&�r��ܘ`��w��ֳ�Y$;]_ +��\f�H��H�K��i��0,�H��ɨ-��nZC!q�O�uA�n@��g�Λ��Fmq�o:��6ϰK�V��4L��2�(��J'pw��D`M��H��*Ӏ��M�Y#�?#�8B |�?R��p�D �+$)N�v��S�tF�W_�,�(��#O�?�u��uڒu>�;��X_>��V��1M�3�!5��c/��6��ޗ��r��.ܘEv�0Gҧ��_M��1, )�ZwX�7�H �¿f?U ��ԟf��'3�m�ꄣ��N� ��l�;zU��h�Ù��/i�p/]�(�� '(��滐��&��N��/�t. KC9�R��k�9� �� +1� ��q"}��F�r��_q� *ցr~��A�!�ŉ�#�"��/�q�Ҽ,�ْ�K�;��r3A"�4"˚�TG��!0�*��}�q��Ԗ^�nd�b�ԸFpE�+�"G��u��O)��,�-�J!�1"P�3�s�F��S�Z +*��K��]\�>��fr6�V��c�'is��Y�A��R�$A��`Ҡ\��d/�!��xe9�� +`M@�h��"[��W��?�qµ@��+!ȚY|A>��Cs��wFЉ��5`<�%.ƮC\+�`Ґ� +��f��#��IȖz��-fa];~Y�n#S�Q�d1I\�gtwp��N{@�,t�aҌ֌ތ!�1*�bp��@��+A�-c��sp��j܋ky�i�>L�k��7��s@vb�CD +�_\��4~QUG��+��#Wg`�Hz��^)r銚�[Ü��l̠s_��z��9=��-G�۶9�${: 6lݓь�Q��7w�ث_Ė-��sw�#gnM��q��;Pĸa�l�W +�� (^k�z�o�,�:��tٍ�Lߕ�e�eѓ+׻5v~��Y�S��pڪ�O��/��ͭ9@2�A�L% �,�1O.,슙��گ�\�Po�O֝)��{��Fu]�;�1!mi)V��KPӪ�~��%R+!5��U��Q[U��m܊��*��=3�/ۃmf0/cbc�za�e��`�&) +ʕ�yﾻ��|{�@�ĥ�r{�(��rK:�nY�J�%��*-仪0��V��j,��JDN��~�8�s��)�LU��P� �rԁXn�\u�7�wcLP)�șK^b�řZ��-��4y��Ư��a��K�(%�zC �Wʅ��o��^8��Pxr��F��戣;��|Ml/��b��^;��,��O�3��T��f�7��C��ﱀ�]\q��u��S��w��|sc��?�Z�Wq{��6w��kX�pxL��oҗ:V ��2AP�W(�톓\�Q@B�6WТ��`��#�M�8�Swt��A��r,�4<�\8��a�a2 [�@$�i�4��5��\kPuP��;�bq�L��!7�%L�ś�W�M=Cs��T��h↡c�XLw��ɵ�9j��ߕ&y�Պ��B�h�pۃ�ᔠ��Y��^��KT��7�g��>]�I[�ȶ��e��]�j��b�5��+c��g��I�h�"�}��y(��y� z��&��[��q�ԂN=j +ԟA�J�G��8��a8RrJ�fn�)�� ]1u�3I�s{��\ܱ;d� ɖ�K��>w�!��m3 +o�h�O�SխC�tW��Nb��L4�6�]��X��Lwy%�:Q^\ѝ��)��h�.퉩��P��Kq��&��vh�rCŀ��9��n��a��v��r"��Jim=�P +�\ME��k�Q� =r#5��xJЉzzY��C[�/ +��w&��\��7�ǔu��fvꐦ�xW{� Z�f��9�N��" ��n�r�ȅ˼ +9�"��`�XZ��E�� !n��W�K��M��G~��"S\)jA��x� l�{�d��U}}�7 ��Fd0��^��C��a�etU�L�ʊ�g��]��!��x�tl��X��x��k�&y��<��Y5f&�N�O�+��<��m�]�1�˷��8R��#d��\+2�� +�4�,'��;d@X|��{-��L�%��K�q�id�Ŵi�c`�P�DS$�e�� 49ӘA�̍�@4G�A�|��6�(�3lN?#3[H��+��J��u \�J�a��̢� +H��W��<sv[��p:��d�9��4߶yj43��Lb�-�t{34��xG�rB� +S�as�ɧ�ك$9'f�d�cɴL�%s�=�7(�KH��o{��A��D�@�:Cŀ/tصfa�ڿ|['��>��y+��bݚKz��<��X��f�M��y��C��p:��$4��ZT)�}��I{_�#�oV6�:�-��:�lE��M8��m�'w�1Ya��hUk��a2K��O��Զ'0��>]�Y�a��X'� +:7Q)9�h�*Au(�IŁ\�*9)�9��꒹L}��X(-��o�..�kB��C� +��#%G��kk�K�>T��8Z��:쯩��P��Xi��4� G��`E 8��e��HUYYU��!��j*�J+�jj��p�D}�e��[�f-}�"��W�Ց+<�tO XK��7�S@�Ϝ��+��K��)��Y�n��i9i�&ZK)6��թ*=d��B_B��,�n�~v��@U��t�xf��z>��G��虾��x�=��uD�b�x��B"��ţm�ށD?N�lm��l��]�D��l��3v��|O��ޞ��޾N��P _z/��n%ɵɤ��`��{�Q�o��䧞%/|�xVi��,Yr_}��g�'�z/o��o�&�(��4>'��e�-X ��VQ"녤��j�T�o��.^̌�>��BG k�� A)5 ]I`�C)��DzϢ�(��+�3E^��5B��^%J��U#��f��1V��$z�� A8ڊΒU��7��Q�Us8��r�� T p-B�w +z��Og �M�kȬ�U��2h�j��~r͐��m��谷e��:o��2Om�F��̯AqSb�ׅ +z6T�7��"'#L�n��,v�c� ��ںu�}~��"�i|�)9��U�{U��;OI�&��QH�Ӧ^�k��|TQ�r�� +�ۑ��-1��\y�p'59`M]��, +`�PQ�7��:_ +�Y2�:��%{Znj񼩁4Q��K��՗�Tÿyv��N�t/�|'ۋ��m��A�-��ʼ�C�Sd��4��r�x4r�"��Nm(�K�\Xo��j��&�P��*�g3 ��|��(�v�=�n��!؀��'\�ۮ�C��W�0٫��١� $� �)*�(&ˇ��ixX��I�I��r�u�г�� +D��gB|�#��Ԃ��N��z��!Zt_�;��k�&}�_��Ď !��!��6�F��iJJY�1U�l�2�1��wn��+]o�f��p�Wu��r��S�F��_��ha��m��3� ��3�2q{)h��S��;��d��*�MULwU�E�0e +��<�m7��G��1ߓ��H��+z�h��!=��=R1j�?�d��;<Ӑ�=��-��.qA��Y�_ay7�d��>�4�/�0w��ԺS��`DϠ�S�]��т��pk3��0��mA�R]��~n5P�ޣ��lb�u��C��/�u�ř0d4�,�}�T��x�тV��d�"0u�ٍ뚂,j��@��:�6�#超�� 7��5ŗ��TU� Q=�m�À�#��Q��R�y5aj�B+�?40�Eژ�"�[ݥ�=��Kj��ջ%�u\��gL��[��3��;Qj��R��%X��}�d{-%�!��kbe��(^�;%P��}��НP�aVpi�^��8\� +B_��Y�q;��[(�M�� f�(_�?�"��d��;�� Q�̜�W�F1u9D��]rr��+�P ��1-ҏ��Ss��X�0�ץ,��f��B��o#Y��g�8�j��cZ9 ��t��wQ�䑨�H��5lz��6�0 -��:Wȭ��әCC�"��;]9ݭ��T?. +}Ir��-\��׼�_e@�**T��S��˗��ȉeS�F�s��c�m�uH|�2��L�DA� +��G��\%^B��NJj�W��Q��;V��V� �$��9��h2�j/V>��9up��E�*^~��ٵg��cc�&ݑ$y[s��cb�p��ZSQ�m\R�.� �Y�0p�Mt\�IY�g��L��tJv� +# +s''��,s��e��v��qI�xS�\y��S�:�&)�'fx .�MJ�;�Sslvg)f� �?�#o��lyjT +��3��m��B��I��0&f� DvҎv2؟�� 5&��]��iYc,�LX}孹�9��m�lEic?��tGGe�� a��$�kn��!'kܹ��-J�c�2��1 ��:�/Q/2T!�㮰a�$�M6��a��w�To �7F�m�|}�@�5(�l��gB�PU)+S��,�4X��$�-��7Qg3��3s}�D�6�̻�x{H<�2��"�0�X�e��D�[��QSJУ�Rp�.�'\�s(@�H�G&h2c��4��W��g�}0�J��a��hn��.�{�F��y�Q@��Q��Sd��&��%y��A�&��7�Ȫ[R�I�sQ��I�l�ʄR�#geT]��S~�a�ۮ��&�b�O�b�>�k.[��RR��LY)��K��^l��1��%%�\�9�*[�� N��H��Q�X$*�#A� +�{��*+D�^�H��A�S��"�^a�e��U��OE�#�k�]��"��Ӂ��l�Eq+�)�Řa%a�j� +�O$f��7��-��`@�bo�D� �� !6ȑ��,��.�=� � ��?1��,��/nm{��~&�a��K>ќf0M��35a��Ѿ|m�G�$�Z�۶hh��`�屁/g��\��U�Sh�X�G�9ߘ�Ud-��WT: +�sr2�9{_۸�0�Q q-H�p�.+��b��\� +4"%~ӈa�oQW`�%A��S�9I��@dd��͐��B��9�DL��d� +V��p>��R&�@�+�Q�.&ְx��Y�k ,��Sѹ�OU��Æ�9}+��"�F�Ak��;�p7{4aV��̆B��a.�8�0e�?cq<�'�E�� +�c�A�fɦ�Wi�߉b�.��fF�2�: ^�n]�)j&dO�X��|��k�#P�i�$��i��!dj�>5�|8�[�a+ְl�H��k��Hb�~㟘��h��i��v�u�ݸVQXTh1}��6��X�q�w{v��Z��x(Bh�`s��^�2��3�fౕJ3��eWK�W&��$0AG� �q��@W�'<.��3� 3� +�;}TbN��(��x|��Ζ�*��/�s��ɩ�x��B��DZ��d6}��s��j��혞�x�sp {|-_��`r�D��r1��B ��z[WmSg�s��!߲��t��fg[Cc�@��^��ln�oin��k��r66vt5w��njin��1�?��~�ܱ�ʒS֑�C~_^f��L�4��u�n��~7{�P~8��=��!cl��J�!��cd�� # G q'�ދ3��D�3!&��1�N�O�M�� D�A��b��C�C �:]b��x�N��3�� zC��!��2��J�8xebR��]�Q�u7�FMb��0�Q�Q�#)�B��<�6��Yb��3�0W�{��j��?�t��7�o��q�N+ �[��C��N$��y�Jʞ=ɻV�}��y��\j�$�W@~[�A,w+D! +�xP#t-.�Jl�� ,��n��P�A�(a%�%��ǌr1��9��z�_ �=-��q�C��G�Y:�� Y�lOٷ6pY̹�n#Ӳ�ڽ��|��~E.�'�n�-�IFAl EA&�Π�d!J��_J#��x�R�8q�<�w��m��Y2�Ü�6�� dI�k$��0�Ȏi��{�BJ\��C�!��o�N=�K1�0~T� h��tz�A�0B�l�m9��sL�.�x��c:�`�@vT8�r0�!��+��&��A�3��O4�$g�M��!~_z��~*6r�OO%�ܸ.hب��V��w��`P{>��Ʃ +ddG��d��\��z� �%-��N�؁�0��.� <˞��X��a�&zU�S�4L_T��M� ��NJ��6s�޽쓗�V��*Q'��=E]��X�ﳰYUQ9�K:��#��[��̯Ό�?�/� �A�K'��R��˄��c�W[ ��t��5��;�b��ofX�Z��l��Y$P��w�|J �S)Ժ3�p�|��D�5 �37N�ӥ�;�mz��|.��i�!�;xE\8�U�{�*΍��u�I,7X�� C��,�j�$�Ӓ�6�sܡ8��|��V��3��]��φ��C�(��7��۶��DY�B�"�l ��3R��g�0nNM�w�nɗ�-1��BK�&�<~��ڂ�0��֌�j��&�.�?rH�Fm�N�KM�v#.�� +��H�^��\V��j��DS�_��0��0s�u�e+Q퐣��N�y��ym�@��j*U�o�)��· ��8��6��zid��V��4վ��k��߭��2J�ݴ�k�:�j�1��]�Pt;6tt��M�[w�5b��e��إU�O�M�{*99%��kF��T�h�� +e�S�łM� ��۝��FQ(ϣ�٩ ��~Efu��1��A�R�X��`�G�um�\[]Q`��hq�,(�Fbx�G��+�me]��6�SfXa��B�ȋ�ﱼ�dju��$x�y��J)�*�^�d�֤��e�)��&�ݱ�[��<�Ϛ6we��q��x��'&il� W@~�� x��_AF�ŀ� �fM6`�}��E�销�o14Pn�@��^v/��J~1Ϧ6��ݤ�Fr+�sH� W:.��R��v��KJ��)�{BI��c��d�X�J��Lf��|*��[{cάYI�/{�HZA�>��Px种�ۺ7qoѮ�ԉ߅.��g��n��3jkTLlܾm7�5��m��SU�� +�8�f�v�m]c�Y]��F +;�ԌI�* +�7�-��ډ)j��2tkN@|SJ��՞��'�"� � +��a�%�Y��,��\ +Ǭf��vw�U9��/��*��^+�M2�L� Tʍ6��^N��zB�7�-�4��9i��e�ƌ��_�=| >::n��f�Z��7��2��:��4� :H�C�R)`+m�@�"�6X��Y��Hx:!��Љ �h�)�6�yBBBr��{�9��X��Z$�?t�;��s�^k�o}�[��_��cgg'O�㡴k\�==)��Ռ5�f��z��+��+?L��>�<��I�� _� 癤��(��Vax��M�l$��3'M�8�Q��H�c�*6%�Q��wM��K��?w�Kj��g��[w2u��_��*e��Ӿ��d�ܻs��sz +��?V��1c�,��g6t}Ү~�+�8�,_1$��/]�n��Q抏��.�wN��{=a@��fc�c��9c�v�sx��,��?j��1�PC'aN8'U��@8��> @��]��.�t��$w�kC�t�2{/'��Q�:?<}gFڎ�9Yi�7�Yu��H\H��D}�$}&K�!)��V +��ruU�K��e� �)x��m�i5�x��J��e4��(E�}�mu�9)�Sv��i8ҋUE)<�e4 <:�B�gPEz7�Ϸ��VY��'��+S|��_X�4�g��Z�v�(��`�+�V�}�[��6v��s��5��`��+��F�F��QaA�x&7(�f�&� +�i�$5N��3B/�TҨI U�̴��ΖOK��>�t2�Uf��ҕ�@��0��C��0�O��M�<��G:��[�/=��Y�wm\��l4��\҃\��u=)�p�d�G�`[�]��*SF��0�� +�k4&��>c�H7*#yQ[!��Q�XL��X�<�m1��6��6v��>ܾl$=��w��C�\��f2�l2��£�I�K�rNy ��#CfX� +Y/]�`%ʅp5 +&��o̠��~x-Xuj�F�{t�W��]�9�� +�+��gNU,��C�Yd��7�AD��}Ui�jH��X�i�� xr�8��ڜ��]��#>6��J��d�t�L0y�E�[皌�3��`��"��\:8�q\�Z��!t�TRR�^��_Λ�rےi�'$&��0=��wO��f��ˊ7�JѲ�*fBo��T0J QsQ��Ct�[B�8`_�z��\E��BCĭH?tQ�x�L��=Ī)p��pEr��W@��]�q0��,U6�I+o��;�u2/;c��%��h� ĭ�E8�4��v'�"d�U�FK ��*[5S%Ls��\�n��;$�� \�� W��t0?��ɿő�k��@u�"�:�댵�#�Ė �_ �O*޲�Kॾ�p��B��&s��5`��vp̚�Rj��,�򁇤��ҖȽgK� ��0�Y��QD��E ��g/LR�S(��+͆@M��m9zv�Y��OLJ�ʗ��,� /7;/?77s߮��EK*uл}�]`$��j�9�c`AqDv3V3�f��_g|U�ݵ+�\+_IMž�x\��xY[�9��C�P�hڕWz<�B�� +��--T��T��:��Ƅr��0�Ρ�J x�N�/��L��jLD!HD�!N�WuR��O��BOШv��x��d�Mj��W��Y�_�� ~0��k�/�a��6*��Uٻ3��<�֭muD�4e/>`��^��.��ZM>��.8�@�ӯ�v󀬤E��E'��G)c��?��J�H�6b��Kq�X f �!wƝ�~�F�� L�̭Q,��3�g ��J�4D�Y�V��,��K�Mzn��A_OڒQX��"+�pvN澝[6�o)��R!8f�_�u'�R�E�@�L��D��q��ą��ɑ��%K�%Ł�ě.�sGz��C����z<+uy2��ÖK�Њ��J��0��]PC��a��Ac<�~� +�xVAX�d��A�G|�y7(�X'��~2��wQz��]`>�e��M3�l�}�P$�y��j�ϙ��OB�Tw��I�я�E��s�\~�s�ϴ�^m�.��, M��,-/'7+3-9!959%!*$0��782�tzY��f��@A�?k�L�ejO +�4�ť=�Ko�U1� �!B�l��)`�-�E��k8:38�ԯ ��g�, �J��So)�ǃ!:� '��U��{L�?iYcIhZ�&�\n&�߅�u�㘫Q]]litR��,��6� Et��FТ�D�:��,��:��i!:F�6��QD�@rDO��Wm�m��}E@L?D\�F(�j��#"_�y��c��i��kcΕA�"k��%�%�� ⺬K�tgxh�Ap��o��C��g|z�/WM��<*�Zo�\��Æ6�@ ��\KY�� + �x�x�z�eQ:`;�b=]f�t�EW�i/��5B-��Ei��kT�S�%v�#��yz�w��l�KY�s+T,{+�� kfo�5^�3I3�L�n�%��h8q�(�5T��uϺ��F}_� �w��Ą�)��)YI9Y�3�R��n&&�] ��J�{Fs�*q��=��K��M��~#�k�;�R�[��S��L�o�[a3�O��ܟW�ޏ��.�c�VvK(��o�N�,�=�B��C0J�� h� ��6�}��$��~��oA�:��uZ�D��B'��vs�+�z��h�2��M�p�W��ф@�"�[,�v��ɯ�]�G��H��z��u+3�<��~OA�O��5��˧#�S��|&Y(�f̞`�(�E�U8��5�Q��:��m�X��r��c�o��?��^]}� 0h�p2kC�)z&�?j�`�ZD��٘��Afe��6Hܛ�/ܶ uRa�� T@��w�{J�,��K��+p¬��1��ws��ǰ� +g/e��~��Ұ��V�}�.��P'�y��Ș��L�X�Db��r�?��`h��Z�a�V�rL��O~d%��?��у[� �uy��0x�g��Xg�f�e�o��ń�ό� Xتu�s��MA��ߦ�O^"�X;�8Q;�NF��y�H��dAV��;x/r�t��^˯�� a/[��3q�À�i�i3�'�7��e�oɬ��9�[�.��}��W�t,Q��1��G��+��*9L_;�6��v+��\ׇ��A��h�*ITǂ��q\��oڷ��S3ef�)��iЄ�p��Z�pX$��2��!��:CȪ��)oR��o ��{�R�޸�q�ff^A��z�/�w��uÆ՛�o]�&�H��n�!��2`�8��Y�5A1��nLW��t��V� ��s��!��;W�Y��|n8v^��;�?��>|<)�CPغ�}6��;�ő�J��9>=6n�F��ML_��h��Ң�W��+8�tE��wCK^`��@BLP�%��%��w"��w��7�Π �f_O��s"��y��W��o��\9�<.q��ʲ��9�פ��V�?�p5y��R�0`I��y��L�ŵ퓧�8>kAv��ސTzl d��<�3��[�ߟ�`��C�>~�"q�i��d�U�od\A{��\� ��b��l�� l�1Q��d䳧��;u;��h']h>�y~bG@E��m`ha��Bm��ri~qAVV̌��%��Q}��x��E�n^�j��'T�я��F$�-T�d +�P�jG��ꪱذ�h��T�Iy-�Qn��q�n��4ɭ+:̓��]��(�F.I�k�¼\|$s|��!z;�jn� :�~��(1E�gР0K�M�c#-��|�lf��d.�bP�?�ڃz£�Ka��.��mы5�� +<;�G��K҄J�ॠ"��u�;��Q� �\��e�ng��Vd]:�'2..fsdTDD��"�%�w]�|r�igy�8��Cm�-rqB�?Dj���Ρ��p�j�CD�U��%j4=!��N]r�tU�UP/�3�S��ȸ�k�T��=��K>ƍ��XŐ[�W�1�2�͙��Rc�Ң��t��X��x�a�I�E� �o�+�*B�5�qӗ9k�w�7�}��YM J�T��#.U�2�Y��yBJ�q��NU�O��΍h�=�|��`i�<*.-*~T�w��Ԍ��2��9>$$,luH��Ȑ�kB�q�]cͧ�Hs�#6�D5�W` y��4F�u4p*˔C#vuyFfR�j�2��Q��9��`hXO��2u��h��׃+y[�f,��%ʹz��b̛ +fȋN��/sI%�q��"ҨG��|�mi�?�� l�Ƽ4�^� 3�F�ܫ��o!<6>_ 3�8�L��^�%�2��l8�pD;�H~�VR�% +e<\�p`L>��6�'��'��珞S��Mu�W�SX��_X��}��޽w?;#�Tr� +T�aQL�Ⳮ;�2:L��'�K��˯��g؄��l�2fzP�F��W��~�L��r��Ai|L��Jc}[�15��Q��R]�c�.��FDE$R߈�F[�RkEG�hA@��E�a��}�sv��3H��4]��a�sϽ��o��r��\�j�gl��6�sm�Ӭŏa��8Hv��t��(b��2��ء�=XÖ�<�;P�r��|\oM��>�}�P�)��w�)u��k1��(��k.p.��y��9ƹk#��'��/��&�h;�]�K�:�}�g��d��/~"z��g=[�;��q��.��]��X�|[�q�^tz׃��I�㚒��²�gR��ιfJ��w��k֯Y3˧w�.ݼ{�دWx!��s'�qAf�F��co�b�H��GR�9�:NKU�nW�rǗ�#�P��7:&@h��v�yP�6Q��VK��M�Q�� +�Տ+R��ep:�B~hx��C��u�t��s)�.��r�ʕ��T�N�/�7��9щe��B^�Ij��,G��:A��Z.�Do�̄��:Fa��g��P�SML��Z,��S#k2�r�BT2�<�)�V�{lx[� ��s@H4��GW$Z�l_�� +�Lo��;�Q)��*?X�ђ�(S��'�?ݼy˔r!��"�jz��B�O��ӽ��P��f��4�*�R��뱺�k�ԮS}y��]u�~�N��lD{k)�o�j��A��H��,��j6�c��>5�Ϻ6��GI��m��x�f��M�ܢ�F{Vʊ4�P��>�Urv��63I�5��im�D�"'�[(d��{�Ӥj�9�R��,ٙ �3�x�XÕl�qc6�-ۑ��Y��ʤ��2�P~)+ͼ�f�1��S�}d��*gl��@�y�S��(��jV�qFm�ʯ\�󔩐gf�DY��$��A�ż�^=u�^w~Mݼ��Q��X��&Y�_Bxݞ�dȌK��lb��i��-2��bN��e��r��s�}��ش�LN��5�7^�d#�/6�6x�# �T?# {I�f��!��F��`#e>��ݍ6��9�x��] K�}�jܺ<�p�6��v��=�7owc£��KF��6z]�Yp�q�j�Xݓ��_�^p�h�Xh��M�1R�'�P�j��6�e��@C�zd��V��NW߾��sŪ5��f��^�ř�I��y�?;��T��?b��l�>C�V��vI�＂%��3n�){A@��s`��7&��F�vy��Yסc��g��p�O?�U�%�aɢU�p7��^��"XӍ$�iG�$:��E�]u�7tZ=( �5<.h��З�:�1��S�I9�|d_�esí�艽~߾]��G�j��*�=(��q>��'�+lʿ��Ұ(�+|gp!,.uj�h�j��>&�mZM�$F��ZE+*�â��QԄ +b�RM\��51��Q$�7$�0,3�m��s�� =O�S��`��{�Y��R�H��y?P[=�LM��W��*��2�M�:��b��Ng�Ue +[m��^��ˊ?�N��A��j�)>�.��ј~UEe�_�K�^��ֆ��;��r�Ͼf�4ZSH�96��㞞�Ø;x��Ǯ��[�ݲv��fC�P��b��@ʡ�F}�Oh*�v�^��}n�Ȗ��/z7��4��P�ͼo��d�qe��y�ˆ\E��J�Z{�&t3C]��h{��Kѡ$�3�EF +��1!T��D�m�l��k��tz"Mv�` �o2��\��% +n��ph��V:�g/��=88ꎸ��C��z�+j�+J��0�ݣS�߳��cM��pd1��5�!'W�me*9O�a+��w�G�O^�ݰ*�=�鎃��ӥ�I�k˹�OR*�0G0L��mE�}�G��m��)��_��1��slk��G@)J<��&�cZ�;]��r `}Q-W1�o0�x��m��񟪭J��lR��=ʜ��Zȅ8��Uk7"� ��w�-��ٕ�^��q�H��h=/P��n? E�� X�?GNشy^��Q @KZ�ܻyLI}/��<��4u�o��.ෟ�T�0g>Oέ��3�F sdA�G\��?T��҈<|��P�⹀��r%:��5*w�#�f��䢕��vϋ��M=]|��c:x��Yrp;x��j��&v�w��!u7ۓ�ʋvRڦԯ��myL�A�|�w��>��t�O�=�-�E� j9��O��>ɊF��I��m��V>��K��}d}��^��W*�-$#Ǚ�}( SG��iSm6_~�O��請�5#!v��m��I��a�>%��Ğr� h��wu��ng�5>��U,zR��.��d%��70�>��D�O*�8u�r�LgCW��n��m��~c�"f��y8?7�b��$"F��k�>� ��[AP:�w�B��!�g�-n%��o�HI��3o�/(�t 7sJ�v�N_]��a?�K�w�?��2��C�x`�2��zlB��N00��q/��:��h��;7��L��P�U[5��m��{H,�<��U�a��.]t�&D̦`p�t�3X��"�^��d��Xa��@<ӽ�R��T�Eu �C`�ގԄy�.��,x*�.ꍖ!�I�7�LVq�tme��h�&�'��.� �+�:_o��L�գxc⬐H��ac� �d��]��o͘4u��Q�:��ٵ��u�ef1H�c.��7��0fԕ��W&��I��]��,BON@�P��(S��j��=_��}n2� +��&��3��Sn~7#9��FRnD��{��7��=�a��s`��iI�� w�ov��ù݌5~y��풲�4�M#�L�Qj��L��L9R��`�(w��?6��PfH�K{))��>#P�� R��1�A��R�� ;�2�x�q�q YI fT��@�,˼�j�ԅ��Z��H�u.]s{al�c��h@�A�W�{��^5U��|毈C��J��EKޝ;��z�ɹahF�c��$G[��d;+J�k_��4:m,=931:��i+3o��TP�$yMR�VȍZ��wzVB�gkʊ�s�UK��-�[��NS�o��{�y��b�a��Fr�V3��ŷ�V�qء"�� +H?�d=� e?�V�U�4��,F��1n<��u�Ww3i�EM`��EMx��l�ɖ�Eg��M��r�#��ʃw��C�~Pbg(|�/yFU8��TB]�a*#C)��Q��圩�J�K +o-i!��6Ą�|"j�IB�պ��A�^�}|�\^��Y�F<�0.L�/6cy��ULe�.��)�GE��dY�T�R�a�kR�$=ط ��_"bp�]'f3ה��w��ѣ/�x��;�*b�t̳>%��)F3��/��X��C��`tϳ{)�?�/�s�3�e^V�.Z��Hհt�\��Y��.�� l��4k�%Z�m�-�|�xt�&�E`I,f�eͫtfX�Bo��:�8��[��S볜B��;9(H�.l8�C��`>�Q�J��x��\��fԤK�B)h�I��#BPQ��g��P ߥⵇ/QYWM�X��E�Y&�t��WpW��K��GR��S[�0�Z[��iAђ +��g�:C�A��8bky��$!ɽ�y@.i, +*R(�Y��HH�}��<~��7$4�Z�Ͻ�gw�9{�?��d;[ܜ�X��$e{k[�b�w�A)��әΖE��6KN��>3�'Khpi��ϑ�Jl�1{"T�U>�eH��r��I�P�@n��ҜP��e��qe�[΅�r+yv.�Lt��- ��T�L�WA�!��G�؁�3��٤ �+�i��u�-MP�ʙ2B�!��腖�1 +i��M��v\��4,��Qo�q� <�F3x!v=�C)@�3�E.4��B�cd��.��P"�I��2��bj��i�8+��@��X$dDg�� U��.��%�F?�&�ރ�(q�T{��\��.O��)2�% ɝ�S�Z�'oZD�UӒ� BG�u�Aq4(d#}�1��DE��Ƈ[�W{I�@�ؼܡ:6Yw^+]X�GؼJ8ϙ�*�S��ڰ;[6d ��"��uR�[�B�w8�?Fa�?i&5�&g��eڝ��2�mPNĎ��Eu��ЙAՓt&=�T�4��L�,�4�8\��D-� "�xO��1��*ש�BԖ�-&#�M��N�#I�Őz7m��֙�O��q�Hs�D!��-�h��OǶ��1��B" +u�*�)d��T!��x�7P!�NnA#NЁq��5�5xb$g�ӛ�@�O8`Q�� 5H��\�/|�j�e��Yc�*GQU�yJ�Z�)�"�C��t��Es$q��@�ėpH�ʜҙ�ّ��¥��$*��m:ۓ^��V��-�R/:��eq�_[_K�/��7��o�{�A��f�)��M��f�i�E��j��;�du�\ɥ��:.�6��9��.=m��֣հ��#a��c,�1db��-��T�g�l��w7�8֮��: Ыj?��xO\ +C[J��r�[x��jY��Z��z��[o��<��ɽtD��v�i� �pq�pdng'�l��j�1O��<��bl�cc8�C�W��@Uu�?��Џ�� ߹��T�`Y�ݩ��W*��`U��*x�y��N��:y-��7qG�I��2��T��|��}-ˈ,�ڀU@t�e�x�ɧ�'�c��Ҡ��U�dh>o:�o\G��g�r0��r:Ҵ�ˀ��/W��I�׆=�KK��D[��)0�4ͯ�G�5��!�"��4�m�׽��II��L�龠�z�3T� ��N��7��h�OGi�gS|�'=�%� +^3�ɹ?&�@#��2��q��{3��4B�f�ɐ2�C��y�_�!�!��2��C�6�\�C�xP�Sގ�v�H�F"��Lv dCUE��Q1Jw�I^;i�m�-�pi��Gc��{z�e�d��S��{X�A4��e�{x�xyQ��,r�q�v�>ŕ�x��q�v!f��F�C��ʚ��k��JZ��qvNr�О��(,��Z�8��J\-�}rI6�""��%�Q��w�~>� +�X�^[�Ò��y�� }q�$|pF��h�Xh�#��í 7ï� +� <��p�X��.=��{�<��1q��3|��T�q��c��'� ��u�Օ8��y�(s��x�o�!W]�kahs��^��2�TY��`m|��P`��~�"��]\�kd�9m��g7��V�iw!?n�Qɿ# H�2��݄�a�\�3��8fb��y�{��j�W<�Qf�^:4��}(X��P�tϳ��L��^4��qȘ�d/�)�;K��GV�a��h�.[V��F��oQ�[��M |&7<0Ӭ��~z�hi�t|C��0��O�T�"}�*��th��h�2��;�}+��?�ȼM��x��SVǼY�s'm%[�-m!Ǭ�f�I�y[�XB��;|\��>)��`�C{��L�&H�q%��t��h9�T��:u��i�su��9C��N��w�iY�ŕ�1�Bq�OB��_&�UUg?��% D��t��J�Z;@��<: ��a�B��M��h��Z�(�)]j��T(�iD�$d{��s_�p��L��˻y�ܳ��Zp�jC�h�;FZ��R�z��=�a�V2��g{l.��H,Zkj� +��i�J�n��~p\�q*�z7��"r�Lx��i��H��1�p�۔c�b�� \��;f�&�0�gDTQR�� x��P��&�;�Y��X��c��o�L�,r�ܮu��fۄ�(V�kx��!��-�DN4�>�M8�N�2ˍ��b�np�H��RQ��S��L��}G'� �%YutWt�/3�Ҥ��6�ߗd~�:NĚ"��V��?3A��:��J�8��N9��dXtf�N!�'e/D�m��:�m��K��׷i�]�^J�8O͢�A}�m� +�樰��׍�T��pQ$o�`=+��2�by�$��hB�k��2zm$��=2{��@�Y��+I��ܛAƭȈ��ύ� �>��}�<�+�G��Eg�~��d��ٺ��~:�YqS��u�&�M��36�$��o��ҳiw~��ێed�WH�$2�J�� o�I�C��"��F2o&)0��;Hn�a��)-��ڣ�G3^��ŲN"�scĔG�fN|j�E%9=+[՝8{��CɈ��ژo��'��g��KoQ��xlx^��Wd �:0�_�9��9�*p�� C��\Ð��VԔ�}�`j��C��7��*��L��;��!=1�9��B��#�*� +�^0k�9�f"�\�i�O2tA��%��r�@�h�)U�%K��2_{�:If�� +;{�1�o�e��^#N��c�P�o΄�Lc-,��}�u�a5ed;%yGN�CWAn��Z%�n7��k�oY�U�p�� n��NN$��W{�,{wK��y��[ib��>��Ձ2>P��l��.W�P�O�W<ŀ� p�Jk�Ɠ��gv��.бS��J%C$��I_�XJ蒤��4�g�&��Wfd�퀝��U�z#�@��)��DE�� ш�8c��V\s�Cto��C� �|ͨ��=��[�m�!C� ��M��{��w��Ƶ��Y� +rg�� :�蟪�7v��d��B�n�˝�')O�H�֜*48k��RJoa��X�X�YMmԦ�$STYh�IZ�ͬQAu�Z�g��P�QTja�&��v@�&D�;�<+'��(��[�$bDK��|��ҷC)�'�8R��v�J�[�Jj�!cߒ��e�(uX��9��&�܂X=֙��$b��Ϟ� Tb3L`��ڔ��0:Vt�H_YZ��H*��iTI�G�]f3\3��$��l�S{�0�ء��~GS��ӓjc-R$�O��elm�۳ ��e��0I2i`1Вrׂcaa< h݀ +�\1�_��T��*�l��L��l��s�o<��t�䡌��Z3n��CՏ��w{��_�� d��S�Ch� ZP��&�i�9|:4�-�r�^-��ƀ� L�?_0�hK�G0�6U}])�D�=C��.'�)�@0`��aC4��Ǘl�'��[�'��dnU��(2�R�KU��s5�4��&k�Ο�\�_��@�Z��D�SH|2y��ژ�GDb�17e�:9��Ş�*�W��m;qٚU�0��#Y�]{�"��b5�j��4�v�h�G�8��ϫg�E�Ƨ#^��%j(�:?`��|�mD�?�I;3�0O~�_X7ju��K�v�-T�%�f��+x4I��OM��d�C�yR�|Bi��*U��g�P�aB��7-U��O�z׷�&a��s��e�_֫(�+ ��3"x��Qŀ)�E��G��u+&�SWc�X["ƍ�n��ML�k��h�FW��ൢ�c�aU�q�c`��~��{��-�]_M��t�~��{��%9$|�=��u˭� +��U�K'��YvχzIWC +�д��=3 k�g��!��9 N� +��4�S��\.��ehOr��W��{��[��z�+=�jQV�q�ԭ�u��ޜ?��G��$��3�:k�?��EY�S�q�z�a�9џ�4U��(�ܐ �[~J��v��M��f��;��XC:۪(��R��.0N��q�1�b�O�.�NH�$�F��;�մ� �[��2��:l��!M��B��&}�U�s��\D�[��l�G�j��F��\Oa�Q҇�[��<8�,�f�YQ��&Q��(0eB��dd�[��!6S��r�(3��!E9E��q�(;4��$J�f��Q]2q�X0�)�o��m��Q��]��7��lx�N;;��WWW��Qj��}ߒT��'d�VkF��c�Y/<{��`��~�� f�mΎ��ڣ�=�Q��P��E��]e�(��݀7H�D�X�mw�� t��bu�@�z$�ĄE lE\�K��z��|a��sU գ'�O;kB�(A�xk�q��b��Wė��U\��:_��8'�0!Ժ>0@�̽2�b ��Ɲ�f��i3��/c�_�"ܖ�^�)��U�� (��7��' ;��bO��b\�(L��A�&�*O�!2�C� ~�u�q�|�!p�^�1�x0��C�}��/�f��D��㩨,�v�s��2��Xq��X�[��[�M��yuL��d�7驏�p�bEq/�w�fK,f~azjMn�Ȋ�ަ�m��|�vy�1��ǂM��;~��z�SZ�҉G��3R��xhZ��R�zAi��ZpZ��3w�H��!�� A��.��عvT��0>�4��n�p��ȡ�M{I�D��K�*8=��L�-�W@�[}�➊�g��E��w��դ/G��̌Q�ŚϠdV�T�+��C{"�q20G|�3r86� �W��B�.��U��-V�5T�r 5Ա +6��Y�bZw%J�u^�N ++�nQ�_�[��bafP��V��; i�S4xA��uc ��,�@1 ��SduG�qۑ\��mNeMx�u,�:}��ȃ�^�� (^��ɨ*��[��.- _��S�[��"��>�}p�羻�{"J��d��E�Smd��޷�tF�Fy��6�<��0��:�U�V�b+Dd��ʹ��)��Y��rp�@+� F�1� ˰��<��4��(!�Ts.�ɗtq(��#6 +�p7\�y��՚`sM߅� fNu�`��Ɋ�@kR��z1�>��Ș�m�5�Z�SA�5��ZPTXx��¢¢|kAAQ��&�Z`�*��gw��5-�ZB*�PK#j )AISM�F�hD� +!h�ME ŨS Ԙ@� +��Җ�k��G��v��y�k��y�{��ζ�Pŵ773��>�|��NOK[�-��Ԅ�6��'~��Z��Z[��C[��l�t�oi�_��L��ۂ��OZ1?�v��f�T�j��{.�u��uK��_��l��m`oXk��x�ο�c{��(�uU�V��{,��ધ᢮+6�A��FJ]��X�Y[�w��C��lS=t{�,80��Q��f��a>��,{H��2<|�}l ��w��^C��Y�^�ז&u� ��E��m1�Ձzpž� �x�-��%��fP��d�k��hX�K��q1Z{|�u�/�?_Pқ�[��\��k��W�/-;k��|��$|��UVP#� x��XW�3aJ9҈�x�&��|�+fi��c@>��/��]��[�Ì�q��[��^��;cr�L�}�L��3��l��o�?H� +�Y��H�r�Z[��$AEr��et��"�H�Ŷ�$kP߂�GR +"�N��h;�iq4�V�(P��ث��V�bJ��z��_ⱏ��.Y�~��y�ؑF��=ԯ��V�垗��كr�_QŮ8��f��-��.��Ų̻l ��r;��'w��V�<��:��[9��c�?�bM�l�\��C�U��>�jy��b솋��1��\c�߉��=��ױ'�y��<��*�d��]�m�'Xm-F?��UOm��/�Z��ճ�Բw��g�!vc�kcwb�7��Vg�F��oaKc��iVU�w1��T �8{�y��P�Ypc7�-��G%��ӌl*sd�2��E-��R�XCB�+}�CA*ǻ�+�� ٭;w�v+�2��ϙ@V��;Pƛ��)a��JB ��#�R�ŧ*��)Eȵ(�+q�J��Q�R!�R�Cp�-�_�K�ӋK�p�B�u�ҁ��-��sH�!/Ȥ0�n�C.�2`x�l�g҉�dy<��p�O��+�8ѯ�/3%��u��p4#�z��T�1� O��~xkGr�m��T��,K�ᜟ<�7��EM��y��&D�nN "��T*�H�J�å��x3�p��L��l��LA��W�4e��4#>��DZ�%��h�@|9��y�� +t�0�V��L�q��˦zF��T:��NH$��~� ��O� �Q�+��Kg��H�%��0�&A��P爀gjl�}��dW��9'�2�SY]�4�T`�? �}�q��,�Τ� +HADro>�� Y:�@Ӽ��'Z�N��JQ�u�\6�^�gFұp�1ļt�Y1hGS3�2*��i�֣݌�WIZ�9��?0Jh,{�O��Kc�X*��e�2�Hs��B��(҄�P��LOV}P�SX�j�^�3u�� ]�� +��`�GQm �Ⲉ�i+-/Y[.IDv�j�|�4^5��9��/�!�B��Q@��&�(��P*@IH 6�S^Ī�* �b��ϼЪ�L��bS�#h�)��O�� y�O`C��}�9("�� 4��l5a��9[$�Y�g��ߞs02yY�h��-��ej�GHB��/�6�2�G��c�L,�p��Ϳ�=0�I@�H��׎�)��!��z�|�1�Q�n�-z��42 jG��>� ��6%Adu��:�i�M#t��B�Ԩ(i�r��R��5u4õ4��FpJ,�Hf�v�n�")F��D��#W��8�':��"~�P��[،A�� +d�2�"w��}+V�H�nm�3��n�ݽ��L [8� ��+�hY��h$M8B��<�Aͭ$}y�8)��i�;JI1�<�ȶH��'� :p�j!B� ��In�w�hn�(1�W��.��;J-B��o&�E@��ڸu�伨8; +��,b��)�JqH��E��\�8�Qi�b1D�V��0�RM��$8��v|��3�I�@T��oP3a_�f3(��)YP.��N��MvP)c"+�R�ƌbxq�{��;*�PϞ�F>q�Sæ��$&"MW6�2�b�9m>�]�~�ԅ��B`��!4y�'�I�ȑ�1e�e�t��F�`�k$��cƨ� ��)��X艎1��H��*��:�߽�M6Y�ch\� Ѧ.3&��ֵ6ֈ��3�i�qRk[��Q�F��4�u@PDq�l��(��Q@x��]�9�ιW�I&Q��7��;�~�|��8�\��݋��^�K�}�z��/�o��e�b�Opq�0�!�$z� �a��ӄm�e�y��0�X�a�k +�`k1��GT��V��pm��\5��2ٙ��B�A��i�=��KL92��1�!QS:�k4�ueY`�6&�żD2QZ�0s +�6x}��Ź�Ϲ´(�n�?�oO��!p�)�O��Ȅ�9A��%G��N��l�)��.��#)>e�Q鳂�f��O=T��3�~J��rD�5QMMI��I�% f�ٻ?yFh+Y5��e.D�lb�t��+�2�|�a@Ѥ�ǘ�({|$X�^�4�k9p�<��#S��I ��P��zU�z:u�T8[�sU_�.��o��*O0�t�kfU��o��I�q��_��E�uv �/*� �v�2(��ٓ�ʄs �m��ܨ��:<+8;�>v܀=�vN$a��K状�qҦ��]C�R.��f��%��xɞ9��h��^��[7��5��ci�Q�� _�}��J�r��k� +��Z�i��s��a�WͿ~�S%U?;L��Gڪ +*IGa5Q΅�`��y��ޣe��u��ޮ�ܪf�u�� z}~A�ծ�V��i�sw�J��T�=�K))<�U;�Α�A�-�:��WВ}.��E۲�+%��%NWYQq+8��o��Ҧ�+ ��d�d��h-i$j��]��v��sW��w��JoN�YB�rr{��,� �p��%��}8/��FA��F�~��޻�G��S��^��͆�^V��i��EV4h�/\/n��0�C?^b?k�j��3� �bX'4�h7�(��!�gNш��>�� WpR�:L*>��Cւ�9��u��,�Ԡqk<��O�;c��yYX +b%[��SN�A��e��32q��3�Z9+��Y�y��Fh��տ� �}�T��22�U�3��?.��>G�wԘcʹ4b� ��l��DX��n�}�"��@�%`��/��R�ȹAy�-M�<��>uy��+�|`�4O�&��w�1��dO�q�͇H�B뻚E�%�g��*<��m�0�̆��'Z��#bBG�KJB$�@z��ə�g|��ZJ|�f��#��zخP�tEJ��C��>)1 j `TT!�'N^ �f{��u,c]2{3>��*{V^�p5��sL[j��f�h��O��n�~�u��^�1�%ca�c��6��-�]9*G�D�W��瓤��p�#�e��Rt��zw��1�� +�O��W4��vH��U��#D�T��~�WU�(%}��daQ� 2� �i��[��v��o�`w)��h��1��=R>�(�U@"�*�V"�񥎀�F�q��輧�w/��2��^��ހxK��@��R�T ��g俅,�� Ę��fj�dd��c�4��vʣ�09��O�+��Cy�B�ިԸ)�u�~ߝj�e�g��@��hʃ9Д�Sj��a8�p��?�[�#��E�,��R��JDlt��i�6ay��C�ܘO�*�D(B��p}�8ژo�+��Jc��4i>�bEv:(�� !w{�Zfm!��%f2�ɚ$|'�*�F�� M�� h�>P?�i�r�I�x*�!&��&cQ*b��YJ����L+ o0�'�VC��qv��#��D]�@�Npt�^�d�M�Lvd�EA�ѓJ5sh��F�َ_S�gPS�+��&��1�W�I��KH�<�ɗ�E; R�'d� l��f��u��Y��1f6�{�mȷ��D��V��|:�p;3H�oF�4�M$�*��Q��=]\ R�&��v�y��B�?�,�F�ھY:��urʭ��uk�?��F�|%��yx8F�j}��p�A|h��d�4��Y��y��ǳ�Q-��i=�R�Iݩ�D�V�ԙ �u�6��I�W ��r��표'L�v� �6q��9Ȉ�4�kp5�}�'�I�O��>BWq��FAGR�ǂ��~ȵ:J��X��ŌϳH�4��Օ��\K��D�r��G�/}�ުs��1�lf��ɜ��@U�(0P��r�i�ja�A��*�i�h��Ge�"i��@�B��)B��x��Q��Xl�|�jSn��3��w䠌�()E�W�ҷ*m��s D�WQ�m�Dc��K�@6��|�[=b��د顆��uYh�So��f��B��lO@g�1��,JF�V��B޺��dEX��G��{�(�� {�Kg�{�Ǽ0�3��Lԡ��T�jh��.��{��/Ь��|�]�O��Q!L��)�M��3ϭ�gY�`�;�3� }| +ů�9�Y�Ry��И܀�V��Q0t��~��UY��d��.�Z�N�E��E B� +s�vv�]Oʼ3|YvN�d1��o��S�?��-�}��Y,��f,} �-z�� +��M��Ȼ�ljq�bFA�OLd�%fe� +�C��* ��;�w~I˟�� Ae�dP%��3�##��YKcf��?��)��9�5��6��.U�Mo�)��+��9��WGkÏ�S��Huڶ�^(\�.�h��5�+��ǻ ƣ��%|=$�s`X��7�6Q(��Q��*��#4��u��$�� K��-�I�7V��M�t��Ϳr�wz��q��h�~(A}�˧�X�}m�� Jbwζ��S9}̺Kٱ�N�*UIC��Ď��{�J#�O ��El��ԕ��m$�R*2U��O�Ę��g�貥u��ҽv��UdS��u��,?�e�E��/$�2��d4�pd�c�"O��H�\<;��GԔPF��lq��;Q�L��%��@p)O8��>��D|!Z��YO��PN�m�Hl ��)В�2��G��+ʿ�!�J��B�:C�C!@`a+�j��R�Vd�ձ�K��Hp��V&��})�#�%�)�Te�1�%y�%/��=��?�Pi_��޳|�;��4p �KI�\)N�K��4G\C.��\�k��L2vMџz�ⰵRo<��.;p�&�"H��2 �"~�'� ��H�#�V�Y��N�_�r��-�\یW� ��7s��DA��9 ��f�(��q�;L��. + �+h��!��o��dE�{��Va��v��S�/\k��T��Z!.��t-��תտ��&��hΊ+��Y��.��6�B��WC��Hõk�u��p��.�� mM�HC}0��k��K�j�4�j��ᦦHC8 +6��J9_;M �u�2�$� �S�%`�?:�PM��4�U�*:�m_��5_y�A4�ʦ�ح�d�aϼpf��y|��_sL��f+{��B��v�.��yl��\�k��J��Y��@A�K��d�� +��v� +��|�?��E��"aaAA ��"\V��ť��⢂��x��JKK�`E ?T޵�l~��%��~��=KJJEG�w��^�nN{ i}j�`��hH��[�h0j}��^I�g��K +Ĝr�G}ex�� 7\ǽۘ8��2D �A��$�� 4+�PWs�#HS_&=�' D�(��Mqi�� +G��vDc��D�=� +[ϙ�S�(9+WS�s�%��`,V�4U�i1��}��%3�1>��5 �s��ي�| S�Jt�:��̮Wu��ޏ�#��o��*}d�_$��;?%uٞ�a�h <��r{5m�ɱ�J<��߁��.�"��n뼺�%��ʼ>I�(o�a%��Wm�Ŵ^�N��'��֮��NO�c��^Z(3YX��!x�fL]��'��L�:pر ��dKM��'&J@��!y��b��l�V��0]*I-]=?�� S �w��I"0t�=7��"�T$,�H�.� Yt�Fl �q�2�"�KP5��#��8��׮&)ưl$�.%E��/ *a��#�M*˥.F�hR�I�uJ)B2�N��S��'QKW��k��SW�B\EV��u��}^��2��2ҔP�+��S��hwa7�=�QeG�M��B�T��d�5 �$PkB0�IW�gN]%Da�t�2H�'��lh�#%�cL��wu�W��N��v 1"Ɲ$�KE��_2��Rt�6p�V�Up�'�(h�.wP��5��Җ��jIV��;�Z,�u��hͷ�l�Jp� �]./ Mt��)+��KC;��u+�Dmx,�x�K/�p�mEY��͖[�top ��Ղ��V�>��Q�{i�CI�� m ފH�DG�8�xLBf��3�Ee��#��N0�J��q�Ů��c��LZB ��o�U��_cʯ02AC|%Wқ��%��6LI�!�,B�J�=C2O3��N�)�;0��xO �2��ư�J�ӯ7�z=�c3�L��FOX��;�WjwL|`��Af��sH?� gJ��7��S"�<��Z?�5��%�GC�W��ivȲ� �4 &��1F�"��XI�9M\Μ��.q�7v�@)��Qjb�݊' Ӳa��6X$��<��!I� L_��3]��8��h_��c�$�� H�t#��6�9�T:��QN��vP�.�\'�5X(�*1q��n�"��(��Z�f w-D��t��u�Ъ�!_�JnLO��^��z)�CA��tt��p��(��^�E��]�e��c�9�`�K��_�=��k�#�t�uCrt��)�1��Y�óeM��p�#�Y��m�߱d�� Xy�_�N�XtN��>x&<��ﰘ~z��\��g<$xH� �5uN^�:��<1xԣ*ŗ˞�3tiKU�Ʋ̗t��+�]��)o�4y��GR�ċ��j'��[ʆ�r��,wr�oy�}>{R%{s��>>SO9f��q��);�5x��٩�zՀ��7�)'�5��|ߩ}8u��m;|�,��ACV�D��@҇��^��J�Wݛ��;}PΞ�J��>�/y��aS��Q��C��:-��Ǔ�Y��N��uK��|߅��w��>׎�^��2�rE�}�:w`vUc�9��r��<��첬�"A�b�x�"F��1j$&j A�x!�"��цch� +jZ��Mb�j��^��j^�j4^r�X��3�}o��5�?��̼;��{��oA-��'A.5�\��H��}-S�C�{�W��ϔ�GƼlu�e��&k�or��e�$JX[7��}��*�<�;��W��5nܜ��G�L�>��f�M=^Z=�.Xԑx��i5��d0�G��箩�ݡ~�x��s�֞��m��L�c�L��5��ԥ�U��u��ꈄ�Q�n��-��'#Xd��9��M]��>��r��u(��q\��pՖŚp+��dVv�[ ?0��w�$��̾䍡oD��1cB|bb�� R�f�-��q��!3c��s|�- us�l�8��s%�^�vA�D�j��S��;p�X��}R�z)��we&e f mx�E�v;Q�짢��b�?��0�}Û�h@�Y��Ա5�J�3Rñ3e��4��:�v��XS�b.,��\��t�Ǡ֊� _�Y�]0a'��7\+�Vbs�@<�W�Le�_�3Z(m2��"Xa�Q��~GR��w8ax}�*�4�&g��R�PF��dY�^�c��hw� jB ᇜ�s�|�-�6i� 417y1��E�P_��/�ssG�p�3'ik�W��U|��Z��>|�ɮ��iŞ~��QÇDGo+=�uΘ�Iq#�Nb��L�}�A��e+�}��m�s�A��S��>ͫwAI21�m�8�kz�#�:w�1j��Q�9'w�l��;��,p�7/��bC=]�3:8��lh��F8�b�4�tU��_�7��!��u/&�}�A��pQw�i��t��F��,j>'�~��(ݡ��X&�*��r�Ygȇv�p!v}a؄��~�]�x�^֕XD��Aҙ� +��ǃ��c(�la=��u�H�� 6��7 �!�DϒZߺS=St�t2R�P��#��9G��/ [hOr��zR{��)��K�2��zV�G��Q��q_ezsZ�k��˞�\��T��q�k�N��{5>vcUu>}�'��Bc�s��!+�� ۮ��8mE��ۨ0�t�B�x4͙�m��i��<��Ǖ곧�oݻez�]'��a��/�ې��'�ڴ��7��MK�,!kw��7`��4�?F$x +y-1�ǈ�R_��&� ��Nu��}2 &0%��^�Z%��y<��cd��c��'��Jm�N6�mE��1:��)��@w�̯ +��ƴ +�OsH��D��䤒.�Z�/f>��n˵��ەN\G��S�؍��#.Fle X��{�=�o�ѥ�P�~� X*!�j�^tX�rS�1)��B��u!��еp)�7�;'�P$��/Ë��ķ��I��D��p}y�Yp4 +��_��;t2��>�^p��䝃<�*8�z��#��f�^�X��?�<��4wd_T%�ӂ�ML�T�~��I�sh�4��k��M=��j9��d"�B7Gz�?1�i� ��K*�e��U�l�f>��[�:��_�O��,��*��-�@c�Ȋ|k��{��E�/ې��:o_�j�v��!��J,K��y��(dg<�?H>��b&@,� ��/�@ +۪k�F!�Mq!��v�׸�l�J'ij���/@�g �ޏàu"� �h��S�bL�#��%' �V��}�b�H kM~@�X��s!��A ��Kx>��H��9��6��)��r^%�UUg�fi\��L�:�U[Gǁ;�N+��꠶#��bG\A� �E�,eS�(V��`h�BIB��B��Y�{y�}w9��?/��d��L޽��s��w-��I��)��k�1+�V +n�|�M�X��X֕G��L�U��ba��0�򶧤��3k�U��od<�iݐj�>�a]7>-�'VƵ�u� �{�mVzV��ne[�=X|�e)��oZ✸��S9�y��[" zc��^sNI�m��{B��>E��5��g�s[�%)W�J�=�ܟ�ByE]((��u;[�kQ~~agMQOGɞ�$��ꠓ��U*��,�S��s�ZD��~:V�w�t}C�-th_G��ZKr��h��ӱcչ t�0�U�Vw��C�JI�)*VTPV�\I�*7"OV֝no,R] �k +��TW��V4$DKm�@{cY��ú��1� ��`� ��rkO7�l�ǺkUW��>=TsD�5S�w�~��VF�wStU�6��j��Fy��"�EsEH�Ԓ��DJv�-G�D{QA$��d˶F��2/��k��?�̸�ƅ +4m��[?��$��x�7�I��7�� '�4��nۂi3��K�͘�-�� +�;~��!��¬�_��>9u^-�aJ�K��_��M��g��ȥ��T�W��?R��q��B`��]�� G�覍�u�^h�뵍 +4��P�q�h�ַ!��Ud!K;�v��~�W1e�^��9��֭Z��'$��P(��!�5{ɏ�uc��oL�a%��"�4�t��X؀��6,Tvb d�g�u�12�N�1�R�J@3��?�Q�h`D#��[�Q�4�Q��ܡ�|3@z髑��%]{2:&h;g��kߞY(G�E��dp�O��)�c�Ҏ��|�y"�Y�t�|��l\:��Y�C� ��܂gw#=6��|�@��Ӷ $��>| +��iѭS��[��?�� lz�5�H^G^~�:��'��y��M��+��J��v�=��Cq�5��V;zS��h��5^��z��U)Gz~�B�q��}I�/��-[��&��f\��RS'�e��dȪO��N?��Q��!a��ޕ��ٻm]�ە�c��Φ�t��F��y��~��Ӕm��Ή�X��- hb�V\��i�!�b��i��H5k��s;'s��&��$�Q��4��u�#m�S��ev +�G��e��ȹ�r��İ?��ZH��k�I�Z��h��QܚН�o�z�֓��x�)6G�#�62�"I 2�!S²:�t��GI��$: J R�_�QIJ�Y\}��:��!��O�� q��{�$%%j4�ɚ]��#��) +�d��.��O;��_ +p��Kb�N��:�P_��4X߾c�+i�� p��U�NO�M��?�⣤*�O�g�WJN�s_3��E(�\��em��迂fvO��T�(��9�@s��'o��e}�� ^߽B�p��]�f[1�B ��a��.��v�˲yQ_��=�+5�dS��BW�YI?��hNT��N,*�Gu�Ȗ�M�߃��؝�c +��7F��]��E!��X�n��{�RR�w�kR3 +�u�B��x��O��fk��p��^�g�{��b}�Փ� �X�D��:xw�5m��v�y�DAL��޼�j��y6�ԫ�_?Թ mCw�ڧ��Bu@��>��M��z��"˟� ��?�ɑ�#!�� ݼf3'��92,�vw[Ī��%�m��CҚ��dK��@�T�ㅳH�!0KޗHΪ�[f໎$�p��q�MB�SE4K�W�� +��E{�8a� +`X�3vZb��{��㵠�q�_�d�ya�6n�ۣIۗ�;q��;��T�Q�+*)*:qL/��Ń� +Rۣ8�>Î�W�@nW��j%D#" ��ݮ=��}3*��}Pݯ��Ȃ� +e��1w�d�G�W��e��_ݖ��+0|�| �ow3T �P w��Y��y�ŏ? ��iO��3��p�y�sw�,>\�r��]�5ۧ,]��y�ˏҥ�B��VR��%�ݒg�2�++L�-ޅMB��t ��͠��k�DYSR�V.��B\�绷R��ѓ'L��9KB#7h��y��b��XR�+(5�EYF��۰��(=䊇��q��\�̭ێ�x�x��C��LV"��ǋ��N$�7�/wHB��_s�뀽3��,b��+��.�j?��18�^��tԩ��3�i��Ϗ&Gf[[�Y'�,��Db�ɒSdJX�K6�]�"Pr��Ei��(�|�Y.��TysW�N��b��c� `�8'a�-=�@Ä��Ƅ �S��zoy��4M�&=)1��?��,�p�c��Oj. +�G��l҃~^�{��a��)�:��n+T9m��($>�c�yb��tizϭ�ۚ��6��=� +�RGsFV�,)��lU��P�3�\B��ٸ0�@�*��P��"{��oMB&Pf \̳́�zy�&�§w�x8$��0d��9B��s��+�L�Jt_��!.�=�+db��{�;��_�"2&v}��;��e��tE�Π�j��K�f}Dv�:��`��T�׳o��*u��ڷ�7n�g7�ZcU�iC��t��Аj)jBH��i��֦~R{�o�6l�A_t��s7��Luu7��j��r�h2f��K T��M�R)*UZ��<ҏ�<%�^s�߲�\��v�B�ݓ7��݁�ߚ��C�� +�?2�-�-O�09��j��U��>�e�v2��I�4̞BvZ��ET�a��!��D�:D��^�.E�vi�%$}�쎽�ܟ��(�\��b2��,WG��Z��'(��Ɛ�ͳ��d��I��Y�^�;;u�e�{]�m�� \��׶t�/�l2o��Q�q�s�S�7�e/_��t�Ƃ�]�6/]u䢳��,��Ҋ��2�m��P� X�� ā��ؠ�0a�15�R�^( � K�ܢ�s��D!�Bʠ "� �c~!��U�w4�[prE��!"�(C�pC$f��@,[g�K�N�H��+uGf�B +�Z�$�J(�,V�/�I {��/Tj1��q�w"AϬS�X$�9oC��2k�� g{G~�[��uL��`��,�d�]N�� j|��KJ��u|umr/�Y��~u=�SOڡ˰Kz��3mm[+2Y~�Ԋ_��j��?��}g��K��ҳ�o۟��py��eN@2��YTQUS[{��IF{.Q ~�N5�+9��x�J��BaM N�Ծ��ذ' _�qW��{��.I8��^�p�=�X��p�A7��Ay��C�f�ìNB��,��vC>��-��5cH�F )`�8A�$\�W!�>n��R��X�©�=�)�z Vd�G��'��"�ML�C�nQ�2Ѱړ�$<��(��5�H�q��M{��Þ�O:d��:��4��H��5~mx࠽��5�Naz��ਘ�y��lq��vnY�� p..;{��cg�=��y �s�c��v�H�Ϟ�M��@��L6(�� 2��D��<�c�n�@�1i��ĵ��+(�Ч��^Z=tB�H-"!ؓZ�=�oX��1k�<��O�7aQ]4�+�3��ڨ.0C�=��&��ȬJ�rN�b�lC#PZ~�Vb)٢H��pO�� rc ��`9ݲ^��5�|��L|��)�I)�q �nܔ�b颌��]��U�O��9]�.x䒒ӗ/-�z�Fb�PQ瀐�1�C�~f�q,Ս��Y�`z��P'3�k��r�1�:�5m��B7�a��!陕�jw� O��JL�� *b��7{u��b�aj��mg��e��`ȦX�r�]K/�AT��y2��-�jQ+KӉ>9��lw�!��)��bc%��o`/�(k��x��+��=��X��x�L�06�/E��H��w�ggH� HWt�oi��HD��Q͑2 D��3��DY+�!CZ;��d�T�|��hRE�� ^��M�]S�}�=��Q�a��P�{��_�}��Ȉ�!$�"��(�>Y9�d�4��7K�Y��ę��SSs�cKfVn�#��Y�fϩ�.W��Yt��d�Ԭ�D�N��I��) �c�f��+�ы�JxcD��7��ou��9�:j�~5zI�ț|�KW�G�n��7�fꅑ�N��ɹ͎��r23�p�݄�&��꩑�/�f��i��=�b~xp��5�1&��3�pe2.��is�E�9�o��9u��t^wv��'��>l�Ը��y ��`�u��`5Y��[�¦��=~2�="�V��Z��Uf��h��S/�7c^��o�>��R�'h�Ǘ>zu��qS>ͯ/,ۼ3(j%gɓ�>�fj�B4Y��w+��+��3 ҃~��^5��]rL��e(��!(�*�1�߬��G{��<��>}@~�)LY�M� �VE&U�}]Ǒ��p$c�:_K(�=��ն`H��Ȉ��Cz�Euea�� &��E�ʈ%��Jp5�QA��(Y��%Q ��A1nKT\�f��-��y�&��?��O�w�;��v� ��ٓbN��o��O��#�lqI��JU��|��Jf8 +��z�{��#z��^��O�'3 �N3�D��?�ZɈ� 5�� Q�n4:�sQ$�F��j�Vtޢ��3�1�� p��1�\�N��:��G�0zlA'��>sN�.��&2=[�Lp��8X޿?s�!X��g��˸F[��YhD��!BWǺ?N��;��CYl�,@'��7�݈^�n��da5W�qD[�'0�LνàI��G*bn�v��ߒ d�c��=7f��VC��#Xm}oF��@�U��ϑ%KB��ZۨǏ?yFr��*v/�� +Wm��N/�f��@%s�$�,�$K9��@�۫c5�s�{_Ȕ��S'ՙw�N�� ja{��w��GK�r1n��}}B7g��H�<'(dqF�� D�:u�PJұQ�U�^��u�H�b�̉6�p� ��*5t>6;Պ�icu�5�V�a�q�2�q��~{�Ԗ=4 +g��W*�/F2�;(��ʞX��& 8�GY��g�^�G�W�8��0)c��$r�U��r��͓�Us��r�� +�J��^g�{Q��|PX�Y�1rI%+��N��7;�`jG� If�J�B5��c�~½ŶIN��]�xf��?�۷t�is��)sX�(r��i��N)/��T)��.+� ��N��;�ön"}�D�+B��k�Md��x�Ч��ǅ�IfPK�V6�@�p��iA��yZ��Y�ވX��r`�-�c[�6n~��قF�O�9I�5�e}�"\tk ��f��s%d��w��r�U�r(��J��!�.mtl8��Mo��i6��K�Ƞ�a|� �=�hP��29^j��({(��o��4 �B=�a�J9w�� +��B��if��<��C�z<��<��6�nM�NB�Kg�ZO=�6w�vk2=k"�Qk1�;9�:��K��|��*��HK��N?��kH�x�I�}Z5��̺N�}d�e ȟ'\)&�?�/JI��2�\%DR�mc��S�}�Wg)�b��ܒ��{㲕�Ǐ�mH�)P�.*��KE�B��߉@��֙�l��9G�O��39 +��ɑ�6�A��G�1#�kpQۤ?�2Nl�\��qxW~տʔ�n@M=\@��"��N�Qƞ��`��\�Ri��ya�z̒�i)�G��w`� +�I�[ec�vC�l/��'E� ��i��%�6J<��I��0G��3��@Hc�[�A76R8��R��g�=�;e�/ +�:qdBe!�4��iJ�f�`}��Y�Q��ԫ@�4�)��BF%G�s�R� �^�;��\q�P�� [X�.Б,�hA��"DuP@�92 +z^#�%vwy��h�� S�\�r� }�3�~w�^U/��N[LϬ znm��}�~��e��ad�Z*� 3��4��B��2ѐ�)E}��hr�m��7E�){�RN\:��ȇ��[�sl{̺=G��Le��e3&�mG�*q��?MޒvA��bUqA��Ҏ�� N��Z�G"yr6��R�h�},��&Tv��UT�u��{<q�>Z��M�L&�8�m:NM�vZ�t�NǱ�i�I;��5��!.�e��Q��Ӏ��m��ӻ}L>�c{��ss~��B��*]_D7s��}G!3b��W�!.��o)�6Nk�^j��t��b F�G��C�[��(�.��pǴD4v< +=e� h��K�!��C�R��hB�T�:��A��$�?��c��BW +{��+��"F�a��kP��j4�ྎ1xZȋ�~�G�{\?:b�� MA�&�b<��B[��g�c +�C��Y��:�a�㵁}'9��V2q�V�Bӫ�C3<�`z�_�K�李��"�_�u7�|ú{�ɛ�ňx��Oٴ��P��X��ѹ�4��Ž�ɷ�쟇��Q��/v}��]��~��%��3V%�m}�ݭ){׭;��u./�A9;��3b��1�>��9\7x/`M��R��K��7Im�F!�fs��t��B5��s�6̳�`5��8 g�uP�WRJ ��[��[��].�bZěC��*�U1�^j��B�p�uj��tP��uAx=Uͷ.Z�x�[:�cm�Tx��t|떇^�(��_\VY]Q�XY�7 +M�<�U�Wٺ�|��a��j�.�-�/�v*._!��mݗ@7� T��X�/�`4U�� EkJʡ�J�ђ��s�jE�~��dVAi�{��q�>��H�]-��k.�0��2?̈��ƠQ���rq��d ^�V��*h�o�H�hל��ɏ�m�?�{��U^c�e��H g2u:w�aM�!�j�x*��p��R3* +�j��G��^��'��$ȹ~eWyv�� ;�΍\�i��g΍ˌ�s�# +cӎ��L�;��C��y=��DM�{3��p�㑩�V��]cr��r!r~%��o��n�S��2�x��ұ�9y�MN��Q��f��I4v��$MD�?��C�H�� +��,��Y@�&�DH9�ob |ԑҋgIEv&ՊCh�%K-��)(n�I]@ s?j��"Y~��cq��*MKĒa�MM�~��W�/�G��6뢑YG��1��j-�þ?�[��̺�2R��W.��4��M��}M��n�cG�6��ת�� ;��a�ֵ�-߱��}rh�;��'�Vğ�v�F!�W��0Ɔ��M8�N|t��"�/��D�~V{eF�֝ۢ�-_�1.��oڿs��6&��'e��+�=�"��c~5ʉ �"�GA�KTu/qMlNX�Xj�ԩ�� -�v2��z�*?Cs_��6%\��j��.sN׎糘��잌;ǣ^�B�)"�_��^�7]ug��8��&��> n��爁�%��ډ��6�u��L&�]G�G��7��3�o�{��d^y�t� +��?�W pU��/�@��le�Z�N)�T�q t�n-�ueS��Cm�u괵�2ЩXg2�a�[��T��$�T�`��Z��Y��s�{/A�Ң��fr�=��?o�2d�|�P�=�7ow�f��^��/i�{�jf��.}�?��Ƅܾn��<��O +Q*��i�lzGD��3~�MrikU;��:��䃌(�9��Ȱ(#�r*��|K��H��63��A1��^�|��%��\%��̐is�g12V̇��68<ט?s��}0 �xSQց/�(d��RT�QNx��y��t\��!�F5� +O�E��L��U�hkѼ��<�]d>vw��[�m0o�sǚ��i�FpNDNGm��O.��YB B-��<�s1'��ڼG�ˡ�m��?�"o��V,��~r�D��T�9 +Y��Lm�8"��D��?�JҸ@|��=��8�St�Yd�Ș�-s��'`A}�؈�khh��&gYC9 ��?��37�u��֖�ܪ5�^m�x_�� 8N�"^R�W��$0��T�B<~N�4%dF�y@dA{D�P#Q��i��A>Э��tx��Ȭ�]{DVD�g2#�5d?�҅$�ܑ�B�Uc�dĩLu�O�\NYr �=��(�,(s�� W�yoÁ�� (-��7�3��_�T\x'竊��|��3/��3�3��6�!e-� �nZ�S�� U� +�[g�L@�/�+�JQł6.� �HnQ��Dۅ��a�}I0B1�Q�C��6��=��r)��cH.��];��}�v|��m;�\�x��=�qMϳ|_��닟��.�$�y��Zȱ��)L׵}ш]��5�-�e9��,ϳ]l!�$��u=a��z2֌,�6��d�Ju�#��x��MJ|$b��cc�>4��$N�k��#��g\��7Rs)I��U�� X��\�̮��0��X[�a��7�_3��#��Y`Κq1�*:w%��X0��`_={�t�a�҄�[-�ڴ +�t�y�-�wD��VG+*�u��H�.��hxc�*��E6TU��W�k��h,��c�c�zCud},��6��7ī6�D��"�:��n�Ū��kb��h�>�Gj�"�U�u�x��Hպ�h8��^�>ZS��і�'w.+s𫂊�L~�{.�ɐ�Q�喟݉R`1�>�3�8��a��f��:��p9,��޺��}�Ӗ�$�g��e��K�u_�FK&�5��<�q��7w{��豮��=X�¶��-m�v�p�ܝ �h��C�u�SQ��ݯ�?u��o��um�k��1��xd� ��nXɤ�'L=�t�L:ɔi�nÖ�� =��zBO&�}m+� �H)]��a]��c�1l$��bk8i%�!��Z�a�IC��uk��K��k?t�U�.D8� SD�"B�eeR�B`��$L�'x��M�I��Y��N��x��g~^{�lZJ�}��?�� 0��i󯽨�e��2��r��~�'|��)��4��=s\�n�،��N�?�� -�~�2z�l|��.Ӻ�=J��VO�'1U�X9g�&��#N�N��B�+C��Ws��(�h�x*y�Ff(�W�Qp9E&i�s�G)Q.�`A1%R��|��9G)٤��;��V^x��X�B/�8��о�C�Q��*{w �ݶ�}�Xr��߶��߅L��r��\��7�.�{��i��Л�m�l��;�@|��H�!Y��(2iZ�c�K�r)]��/G��2/S�1��y�[�"3��2��DNЁ)ƋJU��+dPI�,�+(J��x�]9pݘ�� H�`�o�F� +��ڲECP��wM))i�y;��|��8K<>kf��}r� +�w^Q��K�Ǌ�-�7�Sfm��( M �� +�N]En(zf�i%g��d��Ӛ�Ӓ�M��L̲s��ge��\E� +d}dtv��Lj,#Ϲ`�zg��AIUHdv��X"d9@Н��0n~e��c��g��u��X�'��=�;z^J@0��յ�3(��4Y�N�=�Y�[�@]�#��M�1DPߛ@�M]�i��{!��3��>��/��e�g#��J!�hEfJ�8��2S��Q��DIb��",��,�,��"�j +�m��Z�g"��C��m-� ��qq�&��_�[lU�g[l0x��D�(��5Q��DTQ�1��>胏��F/T�VB�� +��Ċ��`��ng�z.��wf�Xk-}��۝�=�M��(�0��;j�2��8�8�=&�+��H�7�C�ͤ�d�H� ��#6��L��֔�8��A�2�>%[��T�\��_��+�N��ȏ�P�m�Cl�R�#N�Te�򩲂&U��5-��j�22,�1�O��t�c�O5 + � +c��>��Z��5e�c�#ո�5�OtTk��%w��ē�`-�SKN+"d��B��w��rO#�1�Sb��EH�!��o`�Ǔ�L�\�DœYkZrkN0�!ߕ�+9�ʕ�B�)�9_��?�U�E�t=<��`+PȿJ$��N-g*sO�94�ZӒ[��m��拷��T��-nG*�(&��9�A8hY�A�\�*ٹc�O8Z��DT�ћo��D�X�ٔrwG��A��z��p!aB;F ��4�h4��g��g��zth~f�{5÷3�s��q�x�1�9��t��pV�5��̢ad^ �Ƴ�~��ƥ3��F�0-�S'��A֚��j�?�a� �^-��K_�FVR]f��_/��-�[�~�;rI�t{h�Q|�l�I��k>�k/7Jq_��pk��=.\��(Yh��'�R o�`['V-cx�b�y��yy&$?��+��k��k�� Q��]7��1��c�W�[Q�-e�'�T��5 9�� W��G�\�W~z,�]T��/_��W��B��rd]��/��s�w{�8s?F��b�� %�/��T� ��)�@q�ß�ޙo��k_��;f�D�9��yE"�}��U��e_b}n ��uޜ��z Y�˧�xq��^��*n[��P��X��7��Yk�"��j|�3{��S�D��$'PD,�H$��".�D�]��1�T� ��&ԊqB�t��ԣ8�D��d��Wkڑ�NEʑ��7�*�+)�J��Rƛ��mL�% +t��5�j��1�P'��q�\j��|�b��>r�R�k��.�x#<�E�\�Rp<��)Of P0bXrI^Kx&��YI�z�`4�A�8��~��V�D4@q��]��4�ZS�\�5��e�䯠�L��$4�� =�ʸi +!O�ZT4+c�d��]m@�L"� � �w��qO�h�u +"�+��K��e��m�lٖe��m�,�T�K�lZ�@�*�l�z��&5[�I��I#�4�d��*��Ri��*٥�5�l�l�m��sjU-�NAr� �{�b1_��=�\O.W� +�l��|qs>OO��s�|!G�\.�zG�B�G��!�{�=��[��S!�-��[��n�^�Ş�T��K�F��b) ^�9�Yk�j�\�5�M��oԫz�z�q�Q�S�X��4 + +endstream +endobj +26 0 obj +<> +endobj +41 0 obj +<> +endobj +35 0 obj +<> +endobj +42 0 obj +<>stream +H�TP�j�0��+tl��I;��(�n,��-g�E6�s��v�nH ==ޓ�{��E��u��#�8��5�GP��8׮T=�2��e�8vd=4��o �"/�y,�;=�-�6Ȏ�\��4��pD�PAۂA+��Y��/tY¾��j��Ȋ��`�[A2��O��YU�S+��:ͬ|�݇��r}��A��S|0��ll + +endstream +endobj +36 0 obj +<>stream +GBBIEC+AdvP2AA1��%�n�|��) AdvP2AA1 L��6�� RR�|T#��Ӻk>cXKsnojPg�N%� �6�� + +endstream +endobj +112 0 obj +<> +endobj +181 0 obj +<> +endobj +182 0 obj +<> +endobj +184 0 obj +<> +endobj +183 0 obj +<> +endobj +106 0 obj +[185 0 R 186 0 R] +endobj +107 0 obj +<>stream +H��W�n��}7��C��&ٜ}�ǋ$`<�X� XA�lI\S��{��O�*R�x��"�1�i��r��T��1��,T��7�u��is�]%6�";K� L�g* +4��o��t$ 2��DA��7'F!� l�~]�t�ۋy��R3��ͥ�wH��*B��2g�P�B��V�&�,�L�n�~��}��O�w��\�D�>��?<`�3��/?|zX�k�׫��G�Ռt֬T�JY٬3Q'K�"%ma��]~��,�f�fų��) �l�?��_��A:��XG��7��H��;��ߊ�d��QA��h��m�}?t��0��eA��l^��iw�*��;BW߬\�7�`}1t}��Av2oq�$ųQ7� +�G�JAѼ˷��1T*�b��^�� ˄�&�j��Z#��~��;W�2�+�H�8U)�G9�q)��{��yqr_�ey��R�26��}��n�m�c��p�THVK8�:�woքE<��M2P�"��`L��){�E/żhJ$��`��W*Jn�|�"�hB��:f��8X#A]n��\��T�)׸j�2+Y8�NR0��c�t�?KA��i�٘$ ��V)~[��}޻UY��I��Hx# +�/��b�^��릕,�C�k��»~۱%�?��f�-��]YU��q�K�[�k�}�I��k�V�kYxyC^zb��v޵��&��/�o_H1�;��2a��G'�st�](�&jʽ��i��0��W~h��:�O��r��Q)��n�p�1�6TX�� +��aXUe��_��DƏuI��ʅ��o�=��_��ʪ��Ҕw:�>�ܯo�><��Z�P;D��Kw�&ߺ��[v-�`��nx�c��h�aƒ��Z%q`b�$&�LJ�)�ab�H��R��R��2��b�l�L V�@�C4�`굄1� +S�;'��s@��U��_]w�(V��ө�Ɛ;�;.�8Csm^k١��bS��΃�eeN�3�ԅ�'�i2f�ǲ�}��$$�̄y3�� +�p[� d@\B�H��!Z��P��A��"��Xa��%�8VLs}#K#c +�a��:�U�{Gʋ^T# ݻ d�yԑ��;�/e3t��c��y�v��x��b2�ca)x|��IKe��-f�{��h��0�hveݜ��f�l��+��ڷ�,�e��y��s��ҷ��k<�7�fٗ�2͸��ÐS+�a�I��Y�-��i2U}t]?N��ȋ��+/8P}��O��<� �֣˸�N._��2�:��h��<�ؐ@��@3�nE�sH3��@n�>�+��.оm^�F��q�sܼ�X#}~��\2+�W~�U6I�\��p�]c��x�y��Z��"�IPej��_z��H1��p�c\2�E��d"�ና��p +��_�e��;��wJs�uq��#�<�M�c/� +A)V��Н��G��u �+��|��mm�r��h�r!~+��+�b��K��N3`�Ҋs�L�v7�{a�VyN�>Ri��h�� ~��r55(��j_ǰmη�Ǹ��u��s��-��s�� 7��2�O��h%��`��/H�*uR�f��(&�m�8 ��ܭ� +^i}%�-��1o��ȴq\��uj�)IG��s��5��$�W]��(3݋��8��w�ŉog��R8Y��5Bg`!3~�_��cu��2��~$��|)4_�#�5��u��~9�)�+�`4�*|9g$��xm]��*��gD"��j(,W(�k��gt�>cÎ�ǈ�FЈ��L�@��d��i��<�y�� טY +��Q��,��v۵~`B�7hǆ�ew��[��mض� X�K�'m$��IKR�Rv�� ?��$+��/��&2�$�[m�&}�'u�6�<�rΖ��d�bU*$�R��vZ�J��1��F��_6��*�j��|��7}��J�ޮ��|��+ٍg�) +�!��c��F��E2��:�j�E��dQQ�\�w�{�k��p��3��}�f@T��Q&l��,��Wei�qM�~F'�C��c�;r ��y�sRr�(�� T%rmo��mH��p��.�s��fο"W��s�H/_O��eۂRKOe�b��ɖ�L��j�j�� ו��{�6Ŵ�6�|α9��1[�nB��K�CKUU:NJ�f��We?P �+��%�N��;�2U,�ħ5:��q�i��N�2\�gɨU�x�.��M�o�� +g�1e��5��J��E8�X"��*h�^ͪ�2$V䧞Cb}�P��Pa/�6��i��I��6g��b��V��%��v�H�� w��)Lg,ݖK��˗�<��m��.Q��We�@��̅�JSy��X��p��J��!Z��*�Oh�{Պ�K9��)�T�Q�� ˼@ �W��sz��O�-%��&��@��q>� +��G�W.Yޝb?dl��0��i��L�,�QΦ��}�\�t^�c�~=�Z]1��#�\2��,.Ҽ(��J�n� +�W,y��|�*��wǦݏ�_Rcn8��3O��f��`� S�$(��.{<'W�+�$�gr��d�)�Q�&5ҳT��FH^��g��~�p�κĪ&ԙ��$}!�� +��2�]�˸��&J��&��I �̖�E]��G-�,*��ZS��yr�gwDm��V�Z��Պ�t��P��r$�<]� :Rsb��Nߜ�A�h&�3{�!�Y��]:̈́E]��t��>O�Lm�z��/�䛧��֙�O;�&!��V��/�!d��_y!k�,$�?��J��G��{(�,��{��7�%��r��cE� ��ͼ��5��2��[�kٵM�� ˆ��m�v{��_z �y�D�Q��,�L��1}|S��Q@�z��Ę��%%� Rz��GN_g��65�?��r�G4?� ��[�%q�!Ph�3R��w��)��Ao��b@��O�9�T�S߆�5�*n Zg�&р��a^�� 2��%~�uU��ކ��#�'��<��F��I9��`B�mZiЫ�j�J"M��[��%�,S�P9��a��=�m�&S��o�I�8��⸢��Ռ��x�}�,�� =��J\�U��{��|g��]s��`I��Xa:��F��6;2��R�ђ��'\ۥ��i�)�ܜbr4�86�g�Z��_��]�vY�J9&q�I�W�G_�A�@(��ӳ�p�vgvzz��)��$�@�4��)�OK��*g�ꛅ�jjь�N�o��Y.�Ki��?RmR��>vh��֛��ev�y|A��#�Mz)ݭ�#̠�fu��<В��n��!�*,�2��W� ��R7��![ơ"�Nw�(��`�`�Np%�<�a�Y��9l��Jf�@ǗtE��a� dd�yC�Jx�o�Org� �ʛ\Y��Tb)hV3I�ɖǉr8��n�� 6��@��d��W��dC�l�W�'�̷�.��"/%�B��G3�(�P"��t�k.5�� +�T*+"5&�*��k�c?$6Cbi򸛢�Ӓ�{��E*�jR�!��8�ېP`�_��F�~��;�ڮz�~�,��`"��> +endstream +endobj +108 0 obj +<> +endobj +185 0 obj +<> +endobj +186 0 obj +<> +endobj +188 0 obj +<> +endobj +187 0 obj +<> +endobj +52 0 obj +<> +endobj +xref +0 189 +0000000000 65535 f +0000001961 00000 n +0000000000 00000 f +0000000000 00000 f +0000108097 00000 n +0000002408 00000 n +0000000000 00000 f +0000000000 00000 f +0000002861 00000 n +0000000000 00000 f +0000000000 00000 f +0000003290 00000 n +0000000000 00000 f +0000000000 00000 f +0000003720 00000 n +0000000000 00000 f +0000000000 00000 f +0000004150 00000 n +0000000000 00000 f +0000000000 00000 f +0000004592 00000 n +0000000000 00000 f +0000000000 00000 f +0000005023 00000 n +0000000000 00000 f +0000000000 00000 f +0000282669 00000 n +0000020097 00000 n +0000081556 00000 n +0000071563 00000 n +0000061630 00000 n +0000028108 00000 n +0000028609 00000 n +0000081986 00000 n +0000082506 00000 n +0000282891 00000 n +0000283391 00000 n +0000071781 00000 n +0000072274 00000 n +0000061854 00000 n +0000062354 00000 n +0000282831 00000 n +0000283102 00000 n +0000028048 00000 n +0000028319 00000 n +0000081910 00000 n +0000082207 00000 n +0000071726 00000 n +0000071986 00000 n +0000061793 00000 n +0000062066 00000 n +0000001404 00000 n +0000290755 00000 n +0000000079 00000 n +0000000000 00000 f +0000000016 00000 n +0000001519 00000 n +0000000000 00000 f +0000019563 00000 n +0000029733 00000 n +0000021156 00000 n +0000021632 00000 n +0000021341 00000 n +0000022024 00000 n +0000029217 00000 n +0000024261 00000 n +0000025057 00000 n +0000023961 00000 n +0000024680 00000 n +0000030118 00000 n +0000028819 00000 n +0000020259 00000 n +0000020762 00000 n +0000033612 00000 n +0000033663 00000 n +0000000000 00000 f +0000000000 00000 f +0000046209 00000 n +0000044756 00000 n +0000046599 00000 n +0000045295 00000 n +0000045692 00000 n +0000000000 00000 f +0000000000 00000 f +0000000000 00000 f +0000000000 00000 f +0000000000 00000 f +0000000000 00000 f +0000000000 00000 f +0000000000 00000 f +0000000000 00000 f +0000000000 00000 f +0000093218 00000 n +0000092539 00000 n +0000092972 00000 n +0000092850 00000 n +0000093579 00000 n +0000000000 00000 f +0000000000 00000 f +0000000000 00000 f +0000000000 00000 f +0000033496 00000 n +0000033823 00000 n +0000000000 00000 f +0000000000 00000 f +0000005442 00000 n +0000283982 00000 n +0000284017 00000 n +0000290379 00000 n +0000012467 00000 n +0000103533 00000 n +0000103568 00000 n +0000283606 00000 n +0000096126 00000 n +0000096161 00000 n +0000103157 00000 n +0000083218 00000 n +0000083253 00000 n +0000095750 00000 n +0000072957 00000 n +0000072992 00000 n +0000082842 00000 n +0000063031 00000 n +0000063066 00000 n +0000072581 00000 n +0000053752 00000 n +0000053787 00000 n +0000062655 00000 n +0000035071 00000 n +0000035106 00000 n +0000053376 00000 n +0000013878 00000 n +0000013913 00000 n +0000033562 00000 n +0000005662 00000 n +0000005737 00000 n +0000012372 00000 n +0000012557 00000 n +0000007903 00000 n +0000012655 00000 n +0000012757 00000 n +0000012859 00000 n +0000012960 00000 n +0000013061 00000 n +0000013160 00000 n +0000013264 00000 n +0000013798 00000 n +0000013713 00000 n +0000013624 00000 n +0000013551 00000 n +0000013486 00000 n +0000013422 00000 n +0000013368 00000 n +0000034745 00000 n +0000034849 00000 n +0000035007 00000 n +0000034953 00000 n +0000053426 00000 n +0000053530 00000 n +0000053688 00000 n +0000053634 00000 n +0000062705 00000 n +0000062809 00000 n +0000062967 00000 n +0000062913 00000 n +0000072631 00000 n +0000072735 00000 n +0000072893 00000 n +0000072839 00000 n +0000082892 00000 n +0000082996 00000 n +0000083154 00000 n +0000083100 00000 n +0000095800 00000 n +0000095904 00000 n +0000096062 00000 n +0000096008 00000 n +0000103207 00000 n +0000103311 00000 n +0000103469 00000 n +0000103415 00000 n +0000283656 00000 n +0000283760 00000 n +0000283918 00000 n +0000283864 00000 n +0000290429 00000 n +0000290533 00000 n +0000290691 00000 n +0000290637 00000 n +trailer +<<758D9B76D21DB2110A006B0910000000>]>> +startxref +290887 +%%EOF diff --git a/tests/sample_dataset/empty_document/blanktext.md b/tests/sample_dataset/empty_document/blanktext.md new file mode 100644 index 0000000..53dfbc4 --- /dev/null +++ b/tests/sample_dataset/empty_document/blanktext.md @@ -0,0 +1,7 @@ +--- +primary_language: en +is_rotation_valid: True +rotation_correction: 0 +is_table: False +is_diagram: False +--- \ No newline at end of file diff --git a/tests/sample_dataset/empty_document/blanktext.pdf b/tests/sample_dataset/empty_document/blanktext.pdf new file mode 100644 index 0000000..197195e Binary files /dev/null and b/tests/sample_dataset/empty_document/blanktext.pdf differ diff --git a/tests/sample_dataset/simple_document/edgar.md b/tests/sample_dataset/simple_document/edgar.md new file mode 100644 index 0000000..d83c5ec --- /dev/null +++ b/tests/sample_dataset/simple_document/edgar.md @@ -0,0 +1,12 @@ +--- +primary_language: en +is_rotation_valid: True +rotation_correction: 0 +is_table: True +is_diagram: False +--- +Edgar (or Eadgar;[1] c. 944 – 8 July 975) was King of the English from 959 until his death in 975. He became king of all England on his brother's death. He was the younger son of King Edmund I and his first wife Ælfgifu. A detailed account of Edgar's reign is not possible, because only a few events were recorded by chroniclers and monastic writers were more interested in recording the activities of the leaders of the church. + +Edgar mainly followed the political policies of his predecessors, but there were major changes in the religious sphere. The English Benedictine Reform, which he strongly supported, became a dominant religious and social force.[2] It is seen by historians as a major achievement, and it was accompanied by a literary and artistic flowering, mainly associated with Æthelwold, Bishop of Winchester. Monasteries aggressively acquired estates from lay landowners with Edgar's assistance, leading to disorder when he died and former owners sought to recover their lost property, sometimes by force. Edgar's major administrative reform was the introduction of a standardised coinage in the early 970s to replace the previous decentralised system. He also issued legislative codes which mainly concentrated on improving procedures for enforcement of the law. + +England had suffered from Viking invasions for over a century when Edgar came to power, but there were none during his reign, which fell in a lull in attacks between the mid-950s and the early 980s.[3] After his death the throne was disputed between the supporters of his two surviving sons; the elder one, Edward the Martyr, was chosen with the support of Dunstan, the Archbishop of Canterbury. Three years later Edward was murdered and succeeded by his younger half-brother, Æthelred the Unready. Later chroniclers presented Edgar's reign as a golden age when England was free from external attacks and internal disorder, especially \ No newline at end of file diff --git a/tests/sample_dataset/simple_document/edgar.pdf b/tests/sample_dataset/simple_document/edgar.pdf new file mode 100644 index 0000000..b9a1a1e Binary files /dev/null and b/tests/sample_dataset/simple_document/edgar.pdf differ diff --git a/tests/sample_dataset/synth_table/untable.md b/tests/sample_dataset/synth_table/untable.md new file mode 100644 index 0000000..063f0d3 --- /dev/null +++ b/tests/sample_dataset/synth_table/untable.md @@ -0,0 +1,848 @@ +--- +primary_language: en +is_rotation_valid: True +rotation_correction: 0 +is_table: True +is_diagram: False +--- +Annex I: Annual mid-year population, United Nations estimates: 2003 - 2012 + +Annexe I : Population au milieu de l'année, estimations des Nations Unies : 2003 - 2012 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
Continent and country or area
Continent et pays ou zone Population estimates (in thousands) - Estimations de population (en milliers)¹
2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
AFRICA - AFRIQUE
Algeria - Algérie 33 003 33 461 33 961 34 507 35 097 35 725 36 383 37 063 37 763 38 482
Angola 15 421 15 977 16 544 17 122 17 713 18 314 18 927 19 549 20 180 20 821
Benin - Bénin 7 606 7 823 8 042 8 264 8 487 8 712 8 942 9 180 9 420 9 661
Botswana 1 833 1 855 1 876 1 898 1 919 1 934 1 952 1 969 1 987 2 004
Burkina Faso 12 829 13 034 13 222 13 622 14 235 14 680 15 095 15 540 15 995 16 460
Burundi 7 264 7 511 7 770 8 043 8 328 8 624 8 927 9 233 9 540 9 850
Cabo Verde 467 474 479 482 484 485 486 488 491 494
Cameroon - Cameroun 17 075 17 523 17 981 18 450 18 929 19 419 19 919 20 430 20 950 21 700
Central African Republic - République centrafricaine 3 830 3 894 3 961 4 032 4 107 4 185 4 266 4 350 4 436 4 525
Chad - Tchad 9 311 9 665 10 014 10 357 10 694 11 031 11 371 11 721 12 080 12 448
Comoros - Comores 570 585 601 617 633 649 666 683 700 718
Congo 3 383 3 449 3 543 3 647 3 759 3 876 3 995 4 112 4 225 4 337
Côte d'Ivoire 18 103 18 440 18 804 19 192 19 600 20 025 20 461 20 902 21 346 21 795
Democratic Republic of the Congo - République
  démocratique du Congo 50 972 52 487 54 028 55 591 57 188 58 819 60 486 62 191 63 932 65 705
Djibouti 755 766 777 788 799 810 822 834 847 860
Egypt - Égypte 69 432 70 501 71 774 72 901 74 230 75 492 76 775 78 076 79 302 80 722
Equatorial Guinea - Guinée équatoriale 682 698 715 732 749 766 783 801 818 836
Eritrea - Érythrée 4 473 4 666 4 854 5 035 5 210 5 382 5 558 5 741 5 933 6 131
Ethiopia - Éthiopie 71 090 74 066 76 167 78 201 80 441 82 621 84 838 87 095 89 393 91 729
Gabon 1 301 1 330 1 359 1 388 1 417 1 447 1 477 1 508 1 539 1 672
Gambia - Gambie 1 349 1 392 1 437 1 482 1 529 1 578 1 628 1 681 1 735 1 791
Ghana 20 302 20 836 21 384 21 948 22 526 23 110 23 692 24 263 24 821 25 366
Guinea - Guinée 9 205 9 380 9 576 9 799 10 047 10 315 10 593 10 876 11 162 11 451
Guinea-Bissau - Guinée-Bissau 1 361 1 391 1 422 1 453 1 484 1 517 1 551 1 587 1 624 1 664
Kenya 33 913 34 701 35 516 36 360 37 236 38 140 39 070 40 028 41 010 42 027
Lesotho 1 899 1 912 1 926 1 940 1 956 1 972 1 990 2 009 2 030 2 052
Liberia - Libéria 3 124 3 185 3 270 3 385 3 522 3 673 3 821 3 958 4 080 4 190
Libya - Libye 5 423 5 507 5 594 5 686 5 782 5 877 5 964 6 041 6 103 6 155
Madagascar 17 245 17 763 18 290 18 826 19 371 19 927 20 496 21 080 21 679 22 294
Malawi 12 126 12 433 12 744 13 060 13 384 13 717 14 061 14 416 14 783 15 164
Mali 11 220 11 573 11 941 12 326 12 726 13 138 13 559 13 986 14 417 14 854
Mauritania - Mauritanie 2 966 3 055 3 146 3 238 3 330 3 423 3 516 3 609 3 703 3 796
Mauritius - Maurice^a 1 204 1 209 1 213 1 217 1 220 1 223 1 227 1 231 1 235 1 240
Mayotte 164 170 175 181 186 192 198 204 210 216
Morocco - Maroc^b 29 897 30 160 30 425 30 693 30 967 31 255 31 577 31 942 32 309 32 521
Mozambique 19 873 20 439 21 010 21 587 22 171 22 763 23 361 23 967 24 581 25 203
Namibia - Namibie 1 981 2 003 2 027 2 053 2 081 2 111 2 143 2 179 2 218 2 259
Niger 12 661 13 076 13 499 13 929 14 368 14 817 15 278 15 751 16 239 16 744
Nigeria - Nigéria 132 550 135 999 139 586 143 315 147 187 151 208 155 381 159 708 164 193 168 834
Republic of South Sudan - République du Soudan du Sud 7 490 7 730 8 039 8 377 8 737 9 118 9 521 9 941 10 381 10 838
Réunion 770 780 791 802 813 823 834 845 855 865
Rwanda 9 126 9 254 9 429 9 651 9 928 10 223 10 530 10 837 11 144 11 458
Sao Tome and Principe - Sao Tomé-et-Principe 145 148 151 154 157 160 163 166 169 172
Senegal - Sénégal 10 674 10 988 11 271 11 583 11 905 12 239 12 587 12 951 13 331 13 726
Seychelles 84 86 87 88 89 90 91 91 92 92
Sierra Leone 4 713 4 928 5 120 5 281 5 416 5 532 5 641 5 752 5 865 5 979
Somalia - Somalie 8 038 8 250 8 487 8 698 8 913 9 140 9 381 9 636 9 908 10 195
South Africa - Afrique du Sud 46 889 47 643 48 282 48 869 49 401 49 877 50 280 50 632 50 946 51 216
Sudan - Soudan 29 974 30 779 31 586 32 398 33 218 34 040 34 853 35 652 36 431 37 195
Swaziland 1 088 1 095 1 105 1 118 1 135 1 154 1 174 1 193 1 212 1 231
Togo 5 429 5 604 5 780 5 956 6 135 6 307 6 474 6 632 6 773 6 643
Tunisia - Tunisie 9 847 9 947 10 051 10 170 10 274 10 379 10 541 10 632 10 753 10 875
Uganda - Ouganda 26 938 27 807 28 725 29 691 30 722 31 791 32 884 33 987 35 148 36 346
United Republic of Tanzania - République Unie de
  Tanzanie 36 761 37 755 38 824 39 942 41 120 42 354 43 640 44 973 46 355 47 783
Western Sahara - Sahara occidental 379 403 428 448 475 482 498 515 532 549
Zambia - Zambie 10 895 11 175 11 470 11 782 12 110 12 457 12 825 13 217 13 634 14 075
Zimbabwe 12 673 12 693 12 711 12 724 12 740 12 784 12 889 13 077 13 359 13 724
AMERICA, NORTH - AMÉRIQUE DU NORD
Anguilla 12 12 13 13 13 13 14 14 14 14
Antigua and Barbuda - Antigua-et-Barbuda 81 82 83 83 84 85 86 87 88 89
Aruba 97 99 100 101 101 101 101 102 102 102
Bahamas 316 322 329 336 342 348 354 360 366 372
Barbados - Barbade 271 272 274 275 276 278 279 280 282 283
\ No newline at end of file diff --git a/tests/sample_dataset/synth_table/untable.pdf b/tests/sample_dataset/synth_table/untable.pdf new file mode 100644 index 0000000..5ceacf8 Binary files /dev/null and b/tests/sample_dataset/synth_table/untable.pdf differ diff --git a/tests/sample_dataset/urls.jsonl b/tests/sample_dataset/urls.jsonl new file mode 100644 index 0000000..4336ab1 --- /dev/null +++ b/tests/sample_dataset/urls.jsonl @@ -0,0 +1,3 @@ +{"id": "empty_document/blanktext", "url": "https://www.allenai.org"} +{"id": "simple_document/edgar", "url": "https://en.wikipedia.org/wiki/Edgar,_King_of_England"} +{"id": "synth_table/untable", "url": "https://unstats.un.org/unsd/demographic/products/dyb/dyb2012/annexI.pdf"} \ No newline at end of file diff --git a/tests/test_anchor.py b/tests/test_anchor.py new file mode 100644 index 0000000..cf21c54 --- /dev/null +++ b/tests/test_anchor.py @@ -0,0 +1,290 @@ +import base64 +import glob +import io +import json +import os +import re +import tempfile +import unittest + +from pypdf import PdfReader + +from olmocr.data.renderpdf import ( + get_pdf_media_box_width_height, + render_pdf_to_base64png, +) +from olmocr.image_utils import convert_image_to_pdf_bytes +from olmocr.prompts.anchor import _linearize_pdf_report, _pdf_report, get_anchor_text + + +class AnchorTest(unittest.TestCase): + def testExtractText(self): + local_pdf_path = os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "some_ocr1.pdf") + reader = PdfReader(local_pdf_path) + page = reader.pages[0] + + def visitor_body(text, cm, tm, font_dict, font_size): + print(repr(text), cm, tm, font_size) + + def visitor_op(op, args, cm, tm): + # print(op, args, cm, tm) + pass + + page.extract_text(visitor_text=visitor_body, visitor_operand_before=visitor_op) + + def testAnchorBase(self): + local_pdf_path = os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "pdftotext_two_column_issue.pdf") + + report = _pdf_report(local_pdf_path, 2) + + print(report) + + print(get_anchor_text(local_pdf_path, 2, pdf_engine="pdfreport")) + + def testAnchorImage(self): + local_pdf_path = os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "some_ocr1.pdf") + + report = _pdf_report(local_pdf_path, 1) + + print(report) + + print(get_anchor_text(local_pdf_path, 1, pdf_engine="pdfreport")) + + def testSmallPage(self): + local_pdf_path = os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "small_page_size.pdf") + + report = _pdf_report(local_pdf_path, 1) + + print(report) + + print(get_anchor_text(local_pdf_path, 1, pdf_engine="pdfreport")) + + def testBadUTFSurrogatePairsGeneration(self): + local_pdf_path = os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "badlines.pdf") + + anchor_text = get_anchor_text(local_pdf_path, 4, pdf_engine="pdfreport") + + jsondata = json.dumps({"text": anchor_text}) + + import pyarrow as pa + import pyarrow.compute as pc + import pyarrow.json as paj + + buffer = io.BytesIO(jsondata.encode("utf-8")) + paj.read_json(buffer, read_options=paj.ReadOptions(use_threads=False, block_size=len(jsondata))) + + def testLargePromptHint1(self): + local_pdf_path = os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "large_prompt_hint1.pdf") + + anchor_text = get_anchor_text(local_pdf_path, 4, pdf_engine="pdfreport") + + print(anchor_text) + print(len(anchor_text)) + self.assertLessEqual(len(anchor_text), 1000) + + def testLargePromptHint2(self): + local_pdf_path = os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "large_prompt_hint2.pdf") + + anchor_text = get_anchor_text(local_pdf_path, 2, pdf_engine="pdfreport") + + print(anchor_text) + print(len(anchor_text)) + self.assertLessEqual(len(anchor_text), 4000) + + def testLargePromptHint3(self): + local_pdf_path = os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "large_prompt_hint3.pdf") + + anchor_text = get_anchor_text(local_pdf_path, 2, pdf_engine="pdfreport") + + print(anchor_text) + print(len(anchor_text)) + self.assertLessEqual(len(anchor_text), 4000) + + def testNewsPaperPromptHint(self): + local_pdf_path = os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "newspaper.pdf") + + anchor_text = get_anchor_text(local_pdf_path, 1, pdf_engine="pdfreport") + + print(anchor_text) + print(len(anchor_text)) + self.assertLessEqual(len(anchor_text), 4000) + + def testTobaccoPaperMissingParagraphs(self): + local_pdf_path = os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "tobacco_missed_tokens_pg1.pdf") + + anchor_text = get_anchor_text(local_pdf_path, 1, pdf_engine="pdfreport") + + print(anchor_text) + print(len(anchor_text)) + self.assertLessEqual(len(anchor_text), 4000) + + def testAnchorOtherLengths(self): + local_pdf_path = os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "tobacco_missed_tokens_pg1.pdf") + + anchor_text = get_anchor_text(local_pdf_path, 1, pdf_engine="pdfreport", target_length=2000) + + print(anchor_text) + print(len(anchor_text)) + self.assertLessEqual(len(anchor_text), 2000) + + anchor_text = get_anchor_text(local_pdf_path, 1, pdf_engine="pdfreport", target_length=6000) + + print(anchor_text) + print(len(anchor_text)) + self.assertLessEqual(len(anchor_text), 6000) + + def testFailingAnchor(self): + local_pdf_path = os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "failing_anchor_pg4.pdf") + + anchor_text = get_anchor_text(local_pdf_path, 4, pdf_engine="pdfreport") + + print(anchor_text) + print(len(anchor_text)) + self.assertLessEqual(len(anchor_text), 4000) + + def testEmptyAnchor(self): + local_pdf_path = os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "tobacco_missed_tokens_pg1.pdf") + + anchor_text = get_anchor_text(local_pdf_path, 1, pdf_engine="pdfreport", target_length=0) + + self.assertEqual(anchor_text.strip(), "Page dimensions: 612.0x792.0") + + def testEmptyAnchorMatchesImageAnchor(self): + local_pdf_path = os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "edgar.pdf") + + orig_anchor = get_anchor_text(local_pdf_path, 1, pdf_engine="pdfreport") + print(orig_anchor) + + lenneg1_anchor = get_anchor_text(local_pdf_path, 1, pdf_engine="pdfreport", target_length=-1) + print(lenneg1_anchor) + + base64_png = render_pdf_to_base64png(local_pdf_path, 1, target_longest_image_dim=1024) + + # Decode base64 and save to temporary file + temp_img = tempfile.NamedTemporaryFile("wb", suffix=".png", delete=False) + temp_img.write(base64.b64decode(base64_png)) + temp_img.close() + + # Update pdf_path to the new file + img_pdf_path = os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "edgar_image.pdf") + + image_only_anchor = get_anchor_text(img_pdf_path, 1, pdf_engine="pdfreport") + print(image_only_anchor) + + # Parse page dimensions from both anchors and check with tolerance + # Extract page dimensions and image bounds + img_lines = image_only_anchor.strip().split("\n") + len_lines = lenneg1_anchor.strip().split("\n") + + img_page_match = re.search(r"Page dimensions: ([\d.]+)x([\d.]+)", img_lines[0]) + img_image_match = re.search(r"\[Image \d+x\d+ to (\d+)x(\d+)\]", img_lines[1]) + + len_page_match = re.search(r"Page dimensions: ([\d.]+)x([\d.]+)", len_lines[0]) + len_image_match = re.search(r"\[Image \d+x\d+ to (\d+)x(\d+)\]", len_lines[1]) + + self.assertIsNotNone(img_page_match, f"Could not parse image anchor page dims: {image_only_anchor}") + self.assertIsNotNone(img_image_match, f"Could not parse image anchor image dims: {image_only_anchor}") + self.assertIsNotNone(len_page_match, f"Could not parse lenneg1 anchor page dims: {lenneg1_anchor}") + self.assertIsNotNone(len_image_match, f"Could not parse lenneg1 anchor image dims: {lenneg1_anchor}") + + img_page_w, img_page_h = float(img_page_match.group(1)), float(img_page_match.group(2)) + img_img_w, img_img_h = int(img_image_match.group(1)), int(img_image_match.group(2)) + + len_page_w, len_page_h = float(len_page_match.group(1)), float(len_page_match.group(2)) + len_img_w, len_img_h = int(len_image_match.group(1)), int(len_image_match.group(2)) + + # Check page dimensions are within 1.4 tolerance + self.assertAlmostEqual(img_page_w, len_page_w, delta=1.4, msg=f"Page width mismatch: {img_page_w} vs {len_page_w}") + self.assertAlmostEqual(img_page_h, len_page_h, delta=1.4, msg=f"Page height mismatch: {img_page_h} vs {len_page_h}") + + # Check image dimensions are within 1 point tolerance + self.assertAlmostEqual(img_img_w, len_img_w, delta=1, msg=f"Image width mismatch: {img_img_w} vs {len_img_w}") + self.assertAlmostEqual(img_img_h, len_img_h, delta=1, msg=f"Image height mismatch: {img_img_h} vs {len_img_h}") + + self.assertEqual(image_only_anchor[:5], lenneg1_anchor[:5]) + self.assertEqual(image_only_anchor[-1:], lenneg1_anchor[-1:]) + + def testCannotLoad(self): + local_pdf_path = os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "load_v_error.pdf") + + reader = PdfReader(local_pdf_path) + page = 5 + anchor_text = get_anchor_text(local_pdf_path, page, pdf_engine="pdfreport", target_length=6000) + + print(anchor_text) + print(len(anchor_text)) + self.assertLessEqual(len(anchor_text), 6000) + + @unittest.skip("TODO, this unit test still fails, the map text is too large.") + def testExcessiveMapAnchor(self): + local_pdf_path = os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "map1.pdf") + + anchor_text = get_anchor_text(local_pdf_path, 1, pdf_engine="pdfreport", target_length=6000) + + print(anchor_text) + print(len(anchor_text)) + self.assertLessEqual(len(anchor_text), 4000) + + def testKyleOnePageAnchors1(self): + local_pdf_path = os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "dolma-page-1.pdf") + + anchor_text = get_anchor_text(local_pdf_path, 1, pdf_engine="pdfreport", target_length=6000) + + print(anchor_text) + print(len(anchor_text)) + self.assertLessEqual(len(anchor_text), 6000) + + def testKyleOnePageAnchors2(self): + local_pdf_path = os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "olmo-page-1.pdf") + + anchor_text = get_anchor_text(local_pdf_path, 1, pdf_engine="pdfreport", target_length=6000) + + print(anchor_text) + print(len(anchor_text)) + self.assertLessEqual(len(anchor_text), 6000) + + +class BuildSilverTest(unittest.TestCase): + def testSmallPage(self): + local_pdf_path = os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "small_page_size.pdf") + + from olmocr.data.buildsilver import build_page_query + + result = build_page_query(local_pdf_path, "s3://test.pdf", 1) + + from olmocr.data.renderpdf import get_png_dimensions_from_base64 + + base64data = result["body"]["messages"][0]["content"][1]["image_url"]["url"] + + if base64data.startswith("data:image/png;base64,"): + base64data = base64data[22:] + + width, height = get_png_dimensions_from_base64(base64data) + + print(width, height) + + assert max(width, height) == 2048 + + +class TestRenderPdf(unittest.TestCase): + def testFastMediaBoxMatchesPyPdf(self): + for file in glob.glob(os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "*.pdf")): + reader = PdfReader(file) + print("checking", file) + + for page_num in range(1, len(reader.pages) + 1): + w1, h1 = get_pdf_media_box_width_height(file, page_num) + pypdfpage = reader.pages[page_num - 1] + + self.assertAlmostEqual(w1, pypdfpage.mediabox.width, places=3) + self.assertAlmostEqual(h1, pypdfpage.mediabox.height, places=3) + + +class TestOutputSamplePage(unittest.TestCase): + def testTobaccoPaper(self): + local_pdf_path = os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "tobacco_missed_tokens_pg1.pdf") + anchor_text = get_anchor_text(local_pdf_path, 1, "pdfreport", target_length=6000) + + print("") + print(anchor_text) + print("") diff --git a/tests/test_cutoff_detection.py b/tests/test_cutoff_detection.py new file mode 100644 index 0000000..d7dec8b --- /dev/null +++ b/tests/test_cutoff_detection.py @@ -0,0 +1,572 @@ +import asyncio +import os +import tempfile +import unittest + +from olmocr.synth.cutoff_detection import ( + CutoffElement, + RenderResult, + detect_cutoff_text, + extract_viewport_from_html, + has_significant_cutoff, +) +from olmocr.synth.mine_html_templates import render_pdf_with_playwright + + +class TestDetectCutoffText(unittest.TestCase): + """Tests for detect_cutoff_text using Playwright to detect overflow-clipped text.""" + + def test_no_cutoff_simple_page(self): + """All text fits within the viewport and no overflow:hidden clipping.""" + html = """ + + +
Hello World
+
This is a simple paragraph that fits within the viewport.
+""" + results = asyncio.run(detect_cutoff_text(html, 400, 300)) + self.assertEqual(len(results), 0, f"Expected no cutoff, got: {results}") + + def test_cutoff_overflow_hidden_horizontal(self): + """Text inside a narrow overflow:hidden container is clipped horizontally.""" + html = """ + + +
+
This is a very long sentence that will definitely overflow the 100px wide container and get clipped
+
+""" + results = asyncio.run(detect_cutoff_text(html, 800, 600)) + self.assertGreater(len(results), 0, "Should detect clipped text") + # The paragraph text should be the one flagged + clipped_texts = [r.text for r in results] + self.assertTrue( + any("very long sentence" in t for t in clipped_texts), + f"Expected to find the clipped sentence, got: {clipped_texts}", + ) + # The clipping ancestor should be the container div + for r in results: + if "very long sentence" in r.text: + self.assertLess(r.visible_ratio, 0.9) + self.assertEqual(r.clipping_ancestor_tag, "DIV") + + def test_cutoff_overflow_hidden_vertical(self): + """Text inside a short overflow:hidden container is clipped vertically.""" + html = """ + + +
+
Line one
+
Line two should be hidden
+
Line three should be hidden
+
+""" + results = asyncio.run(detect_cutoff_text(html, 800, 600)) + # At least the later lines should be detected as cut off + clipped_texts = [r.text for r in results] + self.assertTrue( + any("hidden" in t for t in clipped_texts), + f"Expected clipped lines, got: {clipped_texts}", + ) + + def test_cutoff_table_in_narrow_container(self): + """A wide table inside a narrow overflow:hidden container - similar to the real issue.""" + html = """ + + +
+ + + + + + + + + + + + + + + +
Name Column A Column B Column C - This Should Be Cutoff Column D - Also Cutoff
Row 1 Value A1 Value B1 Value C1 Value D1
+
+""" + results = asyncio.run(detect_cutoff_text(html, 800, 600)) + clipped_texts = [r.text for r in results] + # The rightmost columns should be flagged + self.assertTrue( + any("Cutoff" in t for t in clipped_texts), + f"Expected table column cutoff, got: {clipped_texts}", + ) + + def test_no_cutoff_table_fits(self): + """A small table that fits within its container should not be flagged.""" + html = """ + + +
+ + + + + + + + + +
Name Value
Item 1 100
+
+""" + results = asyncio.run(detect_cutoff_text(html, 800, 600)) + self.assertEqual(len(results), 0, f"Expected no cutoff, got: {results}") + + def test_cutoff_viewport_clip(self): + """Text positioned outside the viewport should be detected.""" + html = """ + + +
Visible text
+
This text is off screen to the right
+""" + results = asyncio.run(detect_cutoff_text(html, 300, 200)) + clipped_texts = [r.text for r in results] + self.assertTrue( + any("off screen" in t for t in clipped_texts), + f"Expected offscreen text detected, got: {clipped_texts}", + ) + + def test_completely_hidden_element(self): + """An element fully hidden by overflow:hidden should have visible_ratio near 0.""" + html = """ + + +
+
Visible line
+
Completely hidden text below
+
+""" + results = asyncio.run(detect_cutoff_text(html, 800, 600)) + hidden_elements = [r for r in results if "Completely hidden" in r.text] + self.assertGreater(len(hidden_elements), 0, "Should detect fully hidden element") + for el in hidden_elements: + self.assertAlmostEqual(el.visible_ratio, 0.0, places=1) + + def test_visibility_threshold(self): + """Lower threshold should return fewer results.""" + html = """ + + +
+
Short text that is slightly clipped at the edge
+
+""" + # With strict threshold (0.99), flag even slightly clipped elements + strict_results = asyncio.run(detect_cutoff_text(html, 800, 600, visibility_threshold=0.99)) + # With lenient threshold (0.1), only flag severely clipped + lenient_results = asyncio.run(detect_cutoff_text(html, 800, 600, visibility_threshold=0.1)) + self.assertGreaterEqual(len(strict_results), len(lenient_results)) + + def test_display_none_not_flagged(self): + """Elements with display:none should not be flagged as cutoff.""" + html = """ + + +
Visible text
+
This is intentionally hidden via display:none
+""" + results = asyncio.run(detect_cutoff_text(html, 800, 600)) + clipped_texts = [r.text for r in results] + self.assertFalse( + any("intentionally hidden" in t for t in clipped_texts), + f"display:none elements should not be flagged, got: {clipped_texts}", + ) + + def test_multicolumn_poster_layout(self): + """Simulates the real poster issue: a wide table in a narrow grid column.""" + html = """ + + +
+
+
+
+
+
+
+ + + + + + + + + + + + + + + + + + + + + + + + +
Outcome 0-12 weeks OR (95% CI) >12 weeks OR (95% CI) All weeks OR (95% CI)
Unweighted Weighted Unweighted model Weighted model Unweighted model Weighted model
Death 0.84 1.03 0.53 0.93 Hidden value 1 Hidden value 2
+
+
+
+
+""" + results = asyncio.run(detect_cutoff_text(html, 1024, 724)) + # At minimum, some of the wide table content should be flagged + self.assertGreater( + len(results), + 0, + "Should detect cutoff in the wide table within narrow grid column", + ) + + def test_real_poster_html(self): + """Test with the actual poster HTML file if it exists.""" + import os + + poster_path = "/home/ubuntu/olmocr/synth_testposter1/html/synthetic/pdf_00000_page1.html" + if not os.path.exists(poster_path): + self.skipTest("Poster HTML file not available") + + with open(poster_path, "r") as f: + html_content = f.read() + + results = asyncio.run(detect_cutoff_text(html_content, 1024, 724)) + # We know this poster has cutoff table columns + self.assertGreater(len(results), 0, "Should detect cutoff in real poster HTML") + + # At least some table-related content should be flagged + has_table_cutoff = any(r.tag in ("TH", "TD") and r.visible_ratio < 0.9 for r in results) + self.assertTrue( + has_table_cutoff, + f"Expected table cell cutoff detection. Clipped elements: " f"{[(r.tag, r.text[:50], r.visible_ratio) for r in results[:10]]}", + ) + + +class TestHasSignificantCutoff(unittest.TestCase): + """Tests for the has_significant_cutoff helper function.""" + + def test_no_elements(self): + self.assertFalse(has_significant_cutoff([])) + + def test_short_text_not_significant(self): + """Very short clipped text (e.g., a period or dash) is not significant.""" + elements = [ + CutoffElement(tag="SPAN", text=".", visible_ratio=0.0, horizontal_visible_ratio=0.0), + CutoffElement(tag="SPAN", text="-", visible_ratio=0.0, horizontal_visible_ratio=0.0), + ] + self.assertFalse(has_significant_cutoff(elements, min_text_length=3)) + + def test_long_text_significantly_cutoff(self): + """Long text that is mostly hidden horizontally is significant.""" + elements = [ + CutoffElement( + tag="TH", + text=">12 weeks OR (95% CI)", + visible_ratio=0.1, + horizontal_visible_ratio=0.1, + ), + ] + self.assertTrue(has_significant_cutoff(elements)) + + def test_long_text_vertical_only_not_significant(self): + """Text clipped only vertically should NOT be flagged (PDF scaling handles it).""" + elements = [ + CutoffElement( + tag="TD", + text="Required Units for the Major:", + visible_ratio=0.0, + horizontal_visible_ratio=1.0, # full width visible, just vertically clipped + ), + ] + self.assertFalse(has_significant_cutoff(elements)) + + def test_long_text_barely_cutoff_not_significant(self): + """Long text that is mostly visible (e.g., 80%) is not significant at default threshold.""" + elements = [ + CutoffElement( + tag="P", + text="Some paragraph text that is mostly visible", + visible_ratio=0.8, + horizontal_visible_ratio=0.8, + ), + ] + # Default max_visible_ratio is 0.5, so 0.8 should not be significant + self.assertFalse(has_significant_cutoff(elements)) + + def test_custom_thresholds(self): + elements = [ + CutoffElement( + tag="TD", + text="Value C1", + visible_ratio=0.6, + horizontal_visible_ratio=0.6, + ), + ] + # At default threshold (0.5), 0.6 is NOT significant + self.assertFalse(has_significant_cutoff(elements, max_visible_ratio=0.5)) + # At stricter threshold (0.7), 0.6 IS significant + self.assertTrue(has_significant_cutoff(elements, max_visible_ratio=0.7)) + + def test_mixed_elements(self): + """Only needs one significant element with horizontal cutoff to return True.""" + elements = [ + CutoffElement(tag="P", text="Visible", visible_ratio=0.95, horizontal_visible_ratio=0.95), + CutoffElement(tag="SPAN", text=".", visible_ratio=0.0, horizontal_visible_ratio=0.0), # too short + CutoffElement(tag="TH", text="All weeks OR (95% CI)", visible_ratio=0.0, horizontal_visible_ratio=0.0), + ] + self.assertTrue(has_significant_cutoff(elements)) + + +class TestExtractViewportFromHtml(unittest.TestCase): + """Tests for extract_viewport_from_html.""" + + def test_explicit_width_and_height(self): + html = '' + w, h = extract_viewport_from_html(html) + self.assertEqual(w, 800) + self.assertEqual(h, 600) + + def test_width_only_height_fallback_large(self): + """When only width is set, height should fall back to 100000.""" + html = '' + w, h = extract_viewport_from_html(html) + self.assertEqual(w, 1024) + self.assertEqual(h, 100000) + + def test_content_before_name(self): + """Meta tag with content attr before name attr (both orderings appear in the wild).""" + html = '' + w, h = extract_viewport_from_html(html) + self.assertEqual(w, 792) + self.assertEqual(h, 100000) + + def test_no_viewport_meta(self): + """No viewport meta tag — both width and height use defaults.""" + html = "
Hello
" + w, h = extract_viewport_from_html(html) + self.assertEqual(w, 1024) + self.assertEqual(h, 100000) + + def test_no_false_positive_without_body_height(self): + """Content at y>768 with no body height should NOT be flagged as viewport-clipped.""" + html = """ + + +
Spacer
+
This footer text is below y=768 but should NOT be flagged
+""" + vw, vh = extract_viewport_from_html(html) + # vh should be 100000, so the footer is well within viewport + results = asyncio.run(detect_cutoff_text(html, vw, vh)) + flagged_texts = [r.text for r in results] + self.assertFalse( + any("footer" in t.lower() for t in flagged_texts), + f"Footer should NOT be flagged with large viewport height, got: {flagged_texts}", + ) + + +class TestOcclusionDetection(unittest.TestCase): + """Tests for detection of text occluded by opaque overlapping elements.""" + + def test_opaque_box_covering_text(self): + """An opaque positioned element covering text should be detected.""" + html = """ + + +
This important text is hidden behind an opaque box
+
+""" + results = asyncio.run(detect_cutoff_text(html, 400, 300)) + occluded = [r for r in results if r.is_occluded] + self.assertGreater(len(occluded), 0, f"Should detect occluded text, got: {results}") + self.assertTrue( + any("important text" in r.text for r in occluded), + f"Expected occluded text, got: {[r.text for r in occluded]}", + ) + + def test_transparent_watermark_not_flagged(self): + """Semi-transparent watermark should NOT be flagged as occlusion.""" + html = """ + + +
This text has a transparent watermark overlay but should still be visible
+
+""" + results = asyncio.run(detect_cutoff_text(html, 400, 300)) + occluded = [r for r in results if r.is_occluded] + self.assertEqual( + len(occluded), + 0, + f"Transparent watermark should not flag occlusion, got: {[r.text for r in occluded]}", + ) + + def test_has_significant_cutoff_with_occlusion(self): + """has_significant_cutoff should return True for occluded elements.""" + elements = [ + CutoffElement( + tag="P", + text="Important covered text", + visible_ratio=0.4, + is_occluded=True, + ), + ] + self.assertTrue(has_significant_cutoff(elements)) + + def test_has_significant_cutoff_short_occluded_text(self): + """Short occluded text (< min_text_length) should not be significant.""" + elements = [ + CutoffElement(tag="SPAN", text="ab", visible_ratio=0.0, is_occluded=True), + ] + self.assertFalse(has_significant_cutoff(elements, min_text_length=3)) + + +class TestRenderWithCutoffDetection(unittest.TestCase): + """Integration tests for render_pdf_with_playwright with cutoff detection.""" + + def test_html_with_overflow_cutoff_is_rejected(self): + """HTML with significant overflow:hidden clipping should be rejected.""" + html = """ + + +
+
This is a very long sentence that will definitely overflow the narrow container and get clipped by overflow hidden
+
+""" + with tempfile.TemporaryDirectory() as td: + result = asyncio.run(render_pdf_with_playwright(html, os.path.join(td, "test.pdf"), 400, 600)) + self.assertFalse(result.success) + self.assertTrue(result.has_cutoff) + self.assertGreater(len(result.cutoff_elements), 0) + + def test_clean_html_renders_successfully(self): + """Clean HTML without clipping should render to a single-page PDF.""" + html = """ + + +
Hello World
+
This is a simple page with no overflow clipping issues.
+""" + with tempfile.TemporaryDirectory() as td: + pdf_path = os.path.join(td, "test.pdf") + result = asyncio.run(render_pdf_with_playwright(html, pdf_path, 400, 600)) + self.assertTrue(result.success) + self.assertFalse(result.has_cutoff) + self.assertIsNotNone(result.scale_used) + + +if __name__ == "__main__": + unittest.main() diff --git a/tests/test_dataloader.py b/tests/test_dataloader.py new file mode 100644 index 0000000..de3f1f7 --- /dev/null +++ b/tests/test_dataloader.py @@ -0,0 +1,738 @@ +#!/usr/bin/env python +"""Tests for dataloader pipeline steps.""" + +import unittest +from pathlib import Path +from unittest.mock import MagicMock, patch + +from olmocr.prompts.prompts import PageResponse +from olmocr.train.dataloader import ( + DatasetTextRuleFilter, + FilterOutRotatedDocuments, + LatexBracketNormalizer, + ReformatLatexBoldItalic, +) +from olmocr.train.front_matter import FrontMatterParser + + +class TestDatasetTextRuleFilter(unittest.TestCase): + """Test cases for DatasetTextRuleFilter.""" + + def setUp(self): + """Set up test fixtures.""" + self.filter = DatasetTextRuleFilter() + + def test_markdown_table_detection(self): + """Test that markdown tables are correctly detected and filtered.""" + # Sample with markdown table - should be filtered out + sample_with_md_table = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text=""" +Some introduction text. + +| Column 1 | Column 2 | Column 3 | +|----------|----------|----------| +| Data A | Data B | Data C | +| Data D | Data E | Data F | + +Some conclusion text. + """, + ) + } + + result = self.filter(sample_with_md_table) + self.assertIsNone(result, "Should filter out samples with markdown tables") + + def test_no_markdown_table(self): + """Test that samples without markdown tables pass through.""" + sample_without_table = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text="This is regular text without any tables. It has | pipes | but not in table format.", + ) + } + + result = self.filter(sample_without_table) + self.assertIsNotNone(result, "Should pass through samples without markdown tables") + self.assertEqual(result, sample_without_table) + + def test_valid_html_table(self): + """Test that valid HTML tables pass through.""" + sample_with_valid_html = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text=""" +Some text before table. + + + + + + + + + + +
Cell 1 Cell 2
Cell 3 Cell 4
+ +Some text after table. + """, + ) + } + + result = self.filter(sample_with_valid_html) + self.assertIsNotNone(result, "Should pass through samples with valid HTML tables") + + def test_malformed_html_table_unclosed_tags(self): + """Test that malformed HTML tables with unclosed tags are filtered out.""" + sample_with_malformed_html = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text=""" +Text before. + + + + + + + + + +Text after. + """, + ) + } + + result = self.filter(sample_with_malformed_html) + self.assertIsNone(result, "Should filter out samples with malformed HTML tables") + + def test_malformed_html_table_missing_closing(self): + """Test that HTML tables missing closing tags are filtered out.""" + sample_with_unclosed_table = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text=""" +Text before. + +
Cell 1 Cell 2 +
Cell 3 Cell 4
+ + + + + +Text after without closing table tag. + """, + ) + } + + result = self.filter(sample_with_unclosed_table) + self.assertIsNone(result, "Should filter out HTML tables without closing tags") + + def test_no_page_data(self): + """Test that samples without page_data pass through.""" + sample_without_page_data = { + "markdown_path": Path("/path/to/file.md"), + "pdf_path": Path("/path/to/file.pdf"), + } + + result = self.filter(sample_without_page_data) + self.assertIsNotNone(result, "Should pass through samples without page_data") + self.assertEqual(result, sample_without_page_data) + + def test_no_natural_text(self): + """Test that samples with page_data but no natural_text pass through.""" + sample_without_text = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text=None, + ) + } + + result = self.filter(sample_without_text) + self.assertIsNotNone(result, "Should pass through samples without natural_text") + + def test_empty_natural_text(self): + """Test that samples with empty natural_text pass through.""" + sample_with_empty_text = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text="", + ) + } + + result = self.filter(sample_with_empty_text) + self.assertIsNotNone(result, "Should pass through samples with empty natural_text") + + def test_complex_markdown_table_variations(self): + """Test various markdown table formats.""" + # Table with alignment indicators + sample_with_alignment = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text=""" +| Left | Center | Right | +|:-----|:------:|------:| +| A | B | C | + """, + ) + } + + result = self.filter(sample_with_alignment) + self.assertIsNone(result, "Should filter out markdown tables with alignment") + + def test_mixed_content(self): + """Test content with both valid HTML and no markdown tables.""" + sample_mixed = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text=""" +This document has a valid HTML table: + +
Cell 1 Cell 2
+ + + + + + +
Header 1 Header 2
Data 1 Data 2
+ +But no markdown tables. Just some text with | pipes | that aren't tables. + """, + ) + } + + result = self.filter(sample_mixed) + self.assertIsNotNone(result, "Should pass through with valid HTML and no markdown tables") + + def test_br_tags_outside_tables(self): + """Test that
tags outside tables are allowed.""" + sample_br_outside = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text="""Some text
with break + + + + + +
Cell 1 Cell 2
""", + ) + } + + result = self.filter(sample_br_outside) + self.assertIsNotNone(result, "Should allow
tags outside tables") + + +class TestReformatLatexBoldItalic(unittest.TestCase): + """Test cases for ReformatLatexBoldItalic.""" + + def setUp(self): + """Set up test fixtures.""" + self.reformatter = ReformatLatexBoldItalic() + + def test_simple_bold_conversion(self): + """Test conversion of \\textbf to markdown bold.""" + sample = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text="This is \\textbf{bold} text.", + ) + } + + result = self.reformatter(sample) + self.assertEqual(result["page_data"].natural_text, "This is **bold** text.") + + def test_simple_italic_conversion(self): + """Test conversion of \\textit to markdown italic.""" + sample = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text="This is \\textit{italic} text.", + ) + } + + result = self.reformatter(sample) + self.assertEqual(result["page_data"].natural_text, "This is *italic* text.") + + def test_bold_and_italic_conversion(self): + """Test conversion of both bold and italic.""" + sample = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text="This has \\textbf{bold} and \\textit{italic} text.", + ) + } + + result = self.reformatter(sample) + self.assertEqual(result["page_data"].natural_text, "This has **bold** and *italic* text.") + + def test_latex_in_math_equation_preserved(self): + """Test that LaTeX commands inside math equations are preserved.""" + sample = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text="Text outside $$ \\textbf{x} = \\textit{y} $$ more text.", + ) + } + + result = self.reformatter(sample) + self.assertEqual(result["page_data"].natural_text, "Text outside $$ \\textbf{x} = \\textit{y} $$ more text.") + + def test_mixed_context(self): + """Test LaTeX both inside and outside math equations.""" + sample = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text="The \\textbf{equation} is $$ \\textbf{x} = 2 $$ and \\textit{important}.", + ) + } + + result = self.reformatter(sample) + self.assertEqual(result["page_data"].natural_text, "The **equation** is $$ \\textbf{x} = 2 $$ and *important*.") + + def test_nested_braces(self): + """Test handling of nested braces in LaTeX commands.""" + sample = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text="This is \\textbf{bold with {nested} braces} text.", + ) + } + + result = self.reformatter(sample) + self.assertEqual(result["page_data"].natural_text, "This is **bold with {nested} braces** text.") + + def test_multiple_occurrences(self): + """Test multiple occurrences of LaTeX commands.""" + sample = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text="\\textbf{First} and \\textbf{second} bold, \\textit{first} and \\textit{second} italic.", + ) + } + + result = self.reformatter(sample) + self.assertEqual(result["page_data"].natural_text, "**First** and **second** bold, *first* and *second* italic.") + + def test_latex_in_parenthesis_delimiter(self): + """Test LaTeX preserved in \$...\$ math delimiter.""" + sample = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text="Text \$ \\textbf{math} \$ more text \\textbf{bold}.", + ) + } + + result = self.reformatter(sample) + self.assertEqual(result["page_data"].natural_text, "Text \$ \\textbf{math} \$ more text **bold**.") + + def test_latex_in_bracket_delimiter(self): + """Test LaTeX preserved in \\[...\\] math delimiter.""" + sample = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text="Text \\[ \\textit{math} \\] more text \\textit{italic}.", + ) + } + + result = self.reformatter(sample) + self.assertEqual(result["page_data"].natural_text, "Text \\[ \\textit{math} \\] more text *italic*.") + + def test_no_latex_formatting(self): + """Test that text without LaTeX formatting passes through unchanged.""" + sample = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text="Plain text without any formatting.", + ) + } + + result = self.reformatter(sample) + self.assertEqual(result["page_data"].natural_text, "Plain text without any formatting.") + + def test_no_page_data(self): + """Test handling of samples without page_data.""" + sample = {"markdown_path": Path("/path/to/file.md")} + + result = self.reformatter(sample) + self.assertEqual(result, sample) + + def test_no_natural_text(self): + """Test handling of samples without natural_text.""" + sample = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text=None, + ) + } + + result = self.reformatter(sample) + self.assertIsNone(result["page_data"].natural_text) + + def test_complex_latex_with_parenthesis_delimiters(self): + """Test complex LaTeX text with \$...\$ delimiters and textit.""" + input_text = """= a_0 \\int_0^P \\cos \\frac{2m\\pi x}{P} dx ++ \\sum_{n=1}^{\\infty} \\frac{a_n}{2} \\int_0^P \\cos \\frac{2(m+n)\\pi x}{P} + \\cos \\frac{2(m-n)\\pi x}{P} dx ++ b_n \\int_0^P \\sin \\frac{2(m+n)\\pi x}{P} - \\sin \\frac{2(m-n)\\pi x}{P} dx. + +Since \$ m \$ and \$ n \$ are both positive integers we have seen already that all these integrals are zero \\textit{except for the cases where} \$ m = n \$. In the case \$ m = n \$ the sine integral is obviously 0, but +\$ \\cos(2(m-n)\\pi x/P) = 1 \$ and so that integral gives \$ P \$. Hence +\\[ +\\int_0^P \\cos \\frac{2m\\pi x}{P} f(x) dx = \\frac{a_m P}{2}, +\\]""" + + expected_text = """= a_0 \\int_0^P \\cos \\frac{2m\\pi x}{P} dx ++ \\sum_{n=1}^{\\infty} \\frac{a_n}{2} \\int_0^P \\cos \\frac{2(m+n)\\pi x}{P} + \\cos \\frac{2(m-n)\\pi x}{P} dx ++ b_n \\int_0^P \\sin \\frac{2(m+n)\\pi x}{P} - \\sin \\frac{2(m-n)\\pi x}{P} dx. + +Since \$ m \$ and \$ n \$ are both positive integers we have seen already that all these integrals are zero *except for the cases where* \$ m = n \$. In the case \$ m = n \$ the sine integral is obviously 0, but +\$ \\cos(2(m-n)\\pi x/P) = 1 \$ and so that integral gives \$ P \$. Hence +\\[ +\\int_0^P \\cos \\frac{2m\\pi x}{P} f(x) dx = \\frac{a_m P}{2}, +\\]""" + + sample = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text=input_text, + ) + } + + result = self.reformatter(sample) + self.assertEqual(result["page_data"].natural_text, expected_text) + + +class TestFilterOutRotatedDocuments(unittest.TestCase): + """Test cases for FilterOutRotatedDocuments.""" + + def setUp(self): + """Set up test fixtures.""" + self.filter = FilterOutRotatedDocuments() + + def test_valid_rotation(self): + """Test that documents with valid rotation pass through.""" + sample = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text="Some text", + ) + } + + result = self.filter(sample) + self.assertIsNotNone(result, "Should pass through documents with valid rotation") + + def test_invalid_rotation(self): + """Test that documents with invalid rotation are filtered out.""" + sample = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=False, + rotation_correction=90, + is_table=False, + is_diagram=False, + natural_text="Some text", + ) + } + + result = self.filter(sample) + self.assertIsNone(result, "Should filter out documents with invalid rotation") + + def test_rotation_correction_needed(self): + """Test that documents needing rotation correction are filtered out.""" + sample = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=180, + is_table=False, + is_diagram=False, + natural_text="Some text", + ) + } + + result = self.filter(sample) + self.assertIsNone(result, "Should filter out documents with non-zero rotation correction") + + def test_no_page_data(self): + """Test that samples without page_data pass through.""" + sample = {"markdown_path": Path("/path/to/file.md")} + + result = self.filter(sample) + self.assertIsNotNone(result, "Should pass through samples without page_data") + + +class TestLatexBracketNormalizer(unittest.TestCase): + """Test cases for LatexBracketNormalizer.""" + + def setUp(self): + """Set up test fixtures.""" + self.normalizer = LatexBracketNormalizer() + + def test_dollar_to_parentheses(self): + """Test conversion of $...$ to \$...\$.""" + sample = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text="The equation $x^2 + y^2 = z^2$ is famous.", + ) + } + + result = self.normalizer(sample) + expected_text = "The equation \$x^2 + y^2 = z^2\$ is famous." + self.assertEqual(result["page_data"].natural_text, expected_text) + + def test_double_dollar_to_brackets(self): + """Test conversion of $$...$$ to \\[...\\].""" + sample = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text="Display equation:\n$$\\int_0^\\infty e^{-x^2} dx = \\frac{\\sqrt{\\pi}}{2}$$", + ) + } + + result = self.normalizer(sample) + expected_text = "Display equation:\n\\[\\int_0^\\infty e^{-x^2} dx = \\frac{\\sqrt{\\pi}}{2}\\]" + self.assertEqual(result["page_data"].natural_text, expected_text) + + def test_mixed_latex_delimiters(self): + """Test handling of mixed inline and display math.""" + sample = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text="Inline $a + b$ and display:\n$$c^2 = a^2 + b^2$$\nMore inline $x = y$.", + ) + } + + result = self.normalizer(sample) + expected_text = "Inline \$a + b\$ and display:\n\\[c^2 = a^2 + b^2\\]\nMore inline \$x = y\$." + self.assertEqual(result["page_data"].natural_text, expected_text) + + def test_no_latex(self): + """Test that text without LaTeX passes through unchanged.""" + sample = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text="Regular text without any equations.", + ) + } + + result = self.normalizer(sample) + self.assertEqual(result["page_data"].natural_text, "Regular text without any equations.") + + def test_no_natural_text(self): + """Test handling of missing natural_text.""" + sample = { + "page_data": PageResponse( + primary_language="en", + is_rotation_valid=True, + rotation_correction=0, + is_table=False, + is_diagram=False, + natural_text=None, + ) + } + + result = self.normalizer(sample) + self.assertIsNone(result["page_data"].natural_text) + + def test_no_page_data(self): + """Test handling of missing page_data.""" + sample = {"markdown_path": Path("/path/to/file.md")} + + result = self.normalizer(sample) + self.assertEqual(result, sample) + + +class TestFrontMatterParser(unittest.TestCase): + """Test cases for FrontMatterParser.""" + + def setUp(self): + """Set up test fixtures.""" + self.parser_with_class = FrontMatterParser(front_matter_class=PageResponse) + self.parser_without_class = FrontMatterParser(front_matter_class=None) + + @patch.object(Path, "read_text") + def test_parse_yaml_front_matter(self, mock_read_text): + """Test parsing of YAML front matter.""" + mock_read_text.return_value = """--- +primary_language: en +is_rotation_valid: true +rotation_correction: 0 +is_table: false +is_diagram: false +--- +This is the document content. + """ + + sample = {"markdown_path": Path("/path/to/file.md")} + result = self.parser_with_class(sample) + + self.assertIn("page_data", result) + self.assertIsInstance(result["page_data"], PageResponse) + self.assertEqual(result["page_data"].primary_language, "en") + self.assertEqual(result["page_data"].natural_text, "This is the document content.") + + @patch.object(Path, "read_text") + def test_no_front_matter(self, mock_read_text): + """Test handling of documents without front matter.""" + mock_read_text.return_value = "Just regular content without front matter." + + sample = {"markdown_path": Path("/path/to/file.md")} + + # Should raise an error when front_matter_class is specified + with self.assertRaises(ValueError): + self.parser_with_class(sample) + + @patch.object(Path, "read_text") + def test_malformed_yaml(self, mock_read_text): + """Test handling of malformed YAML.""" + mock_read_text.return_value = """--- +primary_language: en +is_rotation_valid: [this is not valid yaml} +--- +Content + """ + + sample = {"markdown_path": Path("/path/to/file.md")} + + # Parser without class should return empty dict for malformed YAML + result = self.parser_without_class(sample) + self.assertEqual(result["page_data"], {}) + + @patch.object(Path, "read_text") + def test_preserve_existing_markdown_content(self, mock_read_text): + """Test that existing markdown_content is preserved if present.""" + sample = { + "markdown_path": Path("/path/to/file.md"), + "markdown_content": """--- +primary_language: fr +is_rotation_valid: true +rotation_correction: 0 +is_table: true +is_diagram: false +--- +French content.""", + } + + # Should not call read_text since markdown_content exists + result = self.parser_with_class(sample) + mock_read_text.assert_not_called() + + self.assertEqual(result["page_data"].primary_language, "fr") + self.assertEqual(result["page_data"].is_table, True) + + +if __name__ == "__main__": + unittest.main() diff --git a/tests/test_filter.py b/tests/test_filter.py new file mode 100644 index 0000000..eccea99 --- /dev/null +++ b/tests/test_filter.py @@ -0,0 +1,17 @@ +import os +import unittest + +from pypdf import PdfReader + +from olmocr.filter import PdfFilter + + +class PdfFilterTest(unittest.TestCase): + def testFormLaterPages(self): + self.filter = PdfFilter(apply_form_check=True) + + self.assertTrue(self.filter.filter_out_pdf(os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "form_on_later_pages.pdf"))) + + self.filter = PdfFilter(apply_form_check=False) + + self.assertFalse(self.filter.filter_out_pdf(os.path.join(os.path.dirname(__file__), "gnarly_pdfs", "form_on_later_pages.pdf"))) diff --git a/tests/test_grpo.py b/tests/test_grpo.py new file mode 100644 index 0000000..12a0137 --- /dev/null +++ b/tests/test_grpo.py @@ -0,0 +1,731 @@ +#!/usr/bin/env python3 +""" +Test suite for GRPO training dataloader. +Tests the OlmOCRBenchDataset class and its functionality with Olmocr-bench format. +""" + +import json +import os +import shutil +import sys +import tempfile +import unittest +from pathlib import Path +from unittest.mock import MagicMock, mock_open, patch + +# Add parent directory to path +sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__)))) + +from olmocr.train.grpo_train import ( + OlmOCRBenchDataset, + load_specific_tests_cached, + olmocr_bench_reward, + reward_front_matter, +) + + +class TestGRPODataloader(unittest.TestCase): + """Test cases for the GRPO dataloader.""" + + @classmethod + def setUpClass(cls): + """Create a temporary bench_data folder with test data.""" + cls.temp_dir = tempfile.mkdtemp() + cls.bench_data_folder = cls.temp_dir + cls.pdfs_folder = os.path.join(cls.bench_data_folder, "pdfs") + + # Create folder structure + os.makedirs(os.path.join(cls.pdfs_folder, "test_pdfs"), exist_ok=True) + + # Create dummy PDF files + cls.pdf_files = [] + for i in range(3): + pdf_path = os.path.join(cls.pdfs_folder, "test_pdfs", f"test_{i}.pdf") + # Create a minimal valid PDF + with open(pdf_path, "wb") as f: + f.write(b"%PDF-1.4\n%%EOF") + cls.pdf_files.append(pdf_path) + + # Create test JSONL files + cls.jsonl_file1 = os.path.join(cls.bench_data_folder, "test1.jsonl") + cls.jsonl_file2 = os.path.join(cls.bench_data_folder, "test2.jsonl") + + # Write test data to JSONL files + test_data1 = [ + {"pdf": "test_pdfs/test_0.pdf", "page": 0, "id": "test_0_001", "type": "math", "math": "x + y = z"}, + {"pdf": "test_pdfs/test_0.pdf", "page": 0, "id": "test_0_002", "type": "text", "text": "Sample text"}, + {"pdf": "test_pdfs/test_1.pdf", "page": 0, "id": "test_1_001", "type": "math", "math": "a^2 + b^2 = c^2"}, + {"pdf": "test_pdfs/test_1.pdf", "page": 1, "id": "test_1_002", "type": "text", "text": "Another sample"}, + ] + + test_data2 = [ + {"pdf": "test_pdfs/test_2.pdf", "page": 0, "id": "test_2_001", "type": "table", "table": "col1,col2"}, + {"pdf": "test_pdfs/test_0.pdf", "page": 0, "id": "test_0_003", "type": "text", "text": "More text"}, + {"pdf": "test_pdfs/test_2.pdf", "page": 0, "id": "test_2_002", "type": "math", "math": "\\int_0^1 x dx"}, + ] + + with open(cls.jsonl_file1, "w") as f: + for entry in test_data1: + f.write(json.dumps(entry) + "\n") + + with open(cls.jsonl_file2, "w") as f: + for entry in test_data2: + f.write(json.dumps(entry) + "\n") + + @classmethod + def tearDownClass(cls): + """Clean up temporary files.""" + shutil.rmtree(cls.temp_dir) + + def test_dataset_initialization(self): + """Test that dataset initializes correctly.""" + dataset = OlmOCRBenchDataset( + bench_data_folder=self.bench_data_folder, + processor=None, + max_samples=None, + target_longest_image_dim=1024, + ) + + self.assertIsNotNone(dataset) + self.assertEqual(dataset.bench_data_folder, self.bench_data_folder) + self.assertEqual(dataset.pdf_folder, self.pdfs_folder) + self.assertTrue(len(dataset) > 0) + + def test_unique_pdf_loading(self): + """Test that unique PDFs are loaded correctly.""" + dataset = OlmOCRBenchDataset( + bench_data_folder=self.bench_data_folder, + processor=None, + max_samples=None, + target_longest_image_dim=1024, + ) + + # Should have 4 unique PDF+page combinations: + # test_0.pdf page 0, test_1.pdf page 0, test_1.pdf page 1, test_2.pdf page 0 + self.assertEqual(len(dataset), 4) + + # Check that samples have correct structure + for sample in dataset.samples: + self.assertIn("pdf_path", sample) + self.assertIn("pdf_name", sample) + self.assertIn("page", sample) + self.assertIn("jsonl_file", sample) + self.assertIn("test_ids", sample) + self.assertIn("entries", sample) + + def test_test_id_aggregation(self): + """Test that test IDs are correctly aggregated per PDF+page.""" + dataset = OlmOCRBenchDataset( + bench_data_folder=self.bench_data_folder, + processor=None, + max_samples=None, + target_longest_image_dim=1024, + ) + + # Find the sample for test_0.pdf page 0 + test_0_sample = None + for sample in dataset.samples: + if "test_0.pdf" in sample["pdf_name"] and sample["page"] == 0: + test_0_sample = sample + break + + self.assertIsNotNone(test_0_sample) + # Should have 3 test IDs for test_0.pdf page 0 + self.assertEqual(len(test_0_sample["test_ids"]), 3) + self.assertIn("test_0_001", test_0_sample["test_ids"]) + self.assertIn("test_0_002", test_0_sample["test_ids"]) + self.assertIn("test_0_003", test_0_sample["test_ids"]) + + def test_max_samples_limit(self): + """Test that max_samples correctly limits the dataset size.""" + dataset = OlmOCRBenchDataset( + bench_data_folder=self.bench_data_folder, + processor=None, + max_samples=2, + target_longest_image_dim=1024, + ) + + self.assertEqual(len(dataset), 2) + + @patch("olmocr.train.grpo_train.render_pdf_to_base64png") + @patch("olmocr.train.grpo_train.build_no_anchoring_v4_yaml_prompt") + def test_getitem_format(self, mock_prompt, mock_render): + """Test that __getitem__ returns the correct format.""" + # Mock the rendering and prompt functions + mock_render.return_value = "iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mNkYPhfDwAChwGA60e6kgAAAABJRU5ErkJggg==" # 1x1 white pixel PNG + mock_prompt.return_value = "Test prompt" + + dataset = OlmOCRBenchDataset( + bench_data_folder=self.bench_data_folder, + processor=None, + max_samples=1, + target_longest_image_dim=1024, + ) + + item = dataset[0] + + self.assertIsNotNone(item) + self.assertIn("prompt", item) + self.assertIn("pdf_path", item) + self.assertIn("jsonl_file", item) + self.assertIn("test_ids", item) + self.assertIn("image", item) + + # Check prompt structure + self.assertIsInstance(item["prompt"], list) + self.assertEqual(len(item["prompt"]), 1) + self.assertEqual(item["prompt"][0]["role"], "user") + self.assertIsInstance(item["prompt"][0]["content"], list) + self.assertEqual(len(item["prompt"][0]["content"]), 2) + + # Check other fields + self.assertIsInstance(item["pdf_path"], str) + self.assertIsInstance(item["jsonl_file"], str) + self.assertIsInstance(item["test_ids"], list) + self.assertTrue(len(item["test_ids"]) > 0) + + def test_empty_jsonl_handling(self): + """Test handling of empty JSONL files.""" + # Create an empty JSONL file + empty_jsonl = os.path.join(self.bench_data_folder, "empty.jsonl") + open(empty_jsonl, "w").close() + + # Should still work with other non-empty files + dataset = OlmOCRBenchDataset( + bench_data_folder=self.bench_data_folder, + processor=None, + max_samples=None, + target_longest_image_dim=1024, + ) + + self.assertTrue(len(dataset) > 0) + + # Clean up + os.remove(empty_jsonl) + + def test_malformed_jsonl_handling(self): + """Test handling of malformed JSONL entries.""" + # Create a JSONL with some malformed entries + malformed_jsonl = os.path.join(self.bench_data_folder, "malformed.jsonl") + with open(malformed_jsonl, "w") as f: + f.write('{"pdf": "test.pdf", "id": "valid_1"}\n') + f.write("not valid json\n") + f.write('{"pdf": "test2.pdf", "id": "valid_2"}\n') + + # Should skip malformed entries but process valid ones + dataset = OlmOCRBenchDataset( + bench_data_folder=self.bench_data_folder, + processor=None, + max_samples=None, + target_longest_image_dim=1024, + ) + + # Should still have entries from valid files + self.assertTrue(len(dataset) > 0) + + # Clean up + os.remove(malformed_jsonl) + + def test_missing_pdf_folder(self): + """Test error handling when pdfs folder is missing.""" + temp_bad_folder = tempfile.mkdtemp() + + with self.assertRaises(ValueError) as context: + dataset = OlmOCRBenchDataset( + bench_data_folder=temp_bad_folder, + processor=None, + max_samples=None, + target_longest_image_dim=1024, + ) + + self.assertIn("PDFs folder not found", str(context.exception)) + + # Clean up + shutil.rmtree(temp_bad_folder) + + def test_no_jsonl_files(self): + """Test error handling when no JSONL files are present.""" + temp_folder = tempfile.mkdtemp() + os.makedirs(os.path.join(temp_folder, "pdfs")) + + with self.assertRaises(ValueError) as context: + dataset = OlmOCRBenchDataset( + bench_data_folder=temp_folder, + processor=None, + max_samples=None, + target_longest_image_dim=1024, + ) + + self.assertIn("No JSONL files found", str(context.exception)) + + # Clean up + shutil.rmtree(temp_folder) + + +class TestOlmOCRBenchReward(unittest.TestCase): + """Test cases for the olmocr_bench_reward function.""" + + @classmethod + def setUpClass(cls): + """Create temporary test files.""" + # Clear any cached tests from previous runs + load_specific_tests_cached.cache_clear() + cls.temp_dir = tempfile.mkdtemp() + + # Create a sample JSONL test file with different test types + cls.jsonl_path = os.path.join(cls.temp_dir, "test.jsonl") + test_data = [ + {"pdf": "test.pdf", "page": 0, "id": "test1", "type": "present", "text": "Hello World", "max_diffs": 0}, + {"pdf": "test.pdf", "page": 0, "id": "test2", "type": "absent", "text": "Bad Text", "max_diffs": 0}, + {"pdf": "test.pdf", "page": 0, "id": "test3", "type": "baseline", "max_repeats": 30}, + {"pdf": "test.pdf", "page": 0, "id": "test4", "type": "order", "before": "First", "after": "Second", "max_diffs": 0}, + ] + + with open(cls.jsonl_path, "w") as f: + for test in test_data: + f.write(json.dumps(test) + "\n") + + @classmethod + def tearDownClass(cls): + """Clean up temporary files.""" + # Clear the LRU cache before removing temp dir + load_specific_tests_cached.cache_clear() + shutil.rmtree(cls.temp_dir) + + def setUp(self): + """Clear cache before each test method.""" + load_specific_tests_cached.cache_clear() + + def test_perfect_completion(self): + """Test reward calculation for a completion that passes all tests.""" + completions = ["Hello World\n\nFirst paragraph.\n\nSecond paragraph.\n\nThis is a good document with no bad text."] + test_ids_list = [["test1", "test2", "test3", "test4"]] + + rewards = olmocr_bench_reward( + prompts=["prompt"], completions=completions, completion_ids=[[]], pdf_path=["test.pdf"], jsonl_file=[self.jsonl_path], test_ids=test_ids_list + ) + + self.assertEqual(len(rewards), 1) + self.assertEqual(rewards[0], 1.0) # All 4 tests should pass + + def test_partial_completion(self): + """Test reward calculation for a completion that passes some tests.""" + completions = ["This document contains Bad Text but nothing else of note."] + test_ids_list = [["test1", "test2", "test3"]] + + rewards = olmocr_bench_reward( + prompts=["prompt"], completions=completions, completion_ids=[[]], pdf_path=["test.pdf"], jsonl_file=[self.jsonl_path], test_ids=test_ids_list + ) + + self.assertEqual(len(rewards), 1) + # Should pass only baseline test (1/3) + self.assertAlmostEqual(rewards[0], 1 / 3, places=2) + + def test_multiple_completions(self): + """Test reward calculation for multiple completions.""" + completions = [ + "Hello World with good content. First then Second.", + "Bad Text only", + "", # Empty completion + ] + test_ids_list = [["test1", "test2", "test3", "test4"]] * 3 + + rewards = olmocr_bench_reward( + prompts=["prompt"] * 3, + completions=completions, + completion_ids=[[]] * 3, + pdf_path=["test.pdf"] * 3, + jsonl_file=[self.jsonl_path] * 3, + test_ids=test_ids_list, + ) + + self.assertEqual(len(rewards), 3) + # First should pass all 4 tests + self.assertEqual(rewards[0], 1.0) + # Second should pass only baseline (1/4) + self.assertEqual(rewards[1], 0.25) + # Third (empty string) passes only the "absent" test (1/4) + self.assertEqual(rewards[2], 0.25) + + def test_no_relevant_tests(self): + """Test behavior when no relevant tests are found.""" + completions = ["Some content"] + test_ids_list = [["nonexistent_test"]] + + rewards = olmocr_bench_reward( + prompts=["prompt"], completions=completions, completion_ids=[[]], pdf_path=["test.pdf"], jsonl_file=[self.jsonl_path], test_ids=test_ids_list + ) + + self.assertEqual(len(rewards), 1) + self.assertIsNone(rewards[0]) # Should return None when no tests found + + def test_invalid_completion(self): + """Test handling of invalid completions.""" + completions = [None, "", "Valid content with Hello World"] + test_ids_list = [["test1"]] * 3 + + rewards = olmocr_bench_reward( + prompts=["prompt"] * 3, + completions=completions, + completion_ids=[[]] * 3, + pdf_path=["test.pdf"] * 3, + jsonl_file=[self.jsonl_path] * 3, + test_ids=test_ids_list, + ) + + self.assertEqual(len(rewards), 3) + # First should be None (invalid completion) + self.assertIsNone(rewards[0]) + # Second should fail the test (empty string doesn't contain "Hello World") + self.assertEqual(rewards[1], 0.0) + # Last should pass the test + self.assertEqual(rewards[2], 1.0) + + def test_cache_functionality(self): + """Test that load_specific_tests_cached properly caches results.""" + # Clear cache first + load_specific_tests_cached.cache_clear() + + # Test that the cache works by calling the function twice + test_ids = ("test1", "test2") + + # First call loads from file + with patch("builtins.open", create=True) as mock_open: + mock_file = MagicMock() + mock_file.__enter__.return_value = iter( + [ + '{"id": "test1", "type": "present", "text": "hello", "pdf": "test.pdf", "page": 0}\n', + '{"id": "test2", "type": "absent", "text": "world", "pdf": "test.pdf", "page": 0}\n', + '{"id": "test3", "type": "present", "text": "foo", "pdf": "test.pdf", "page": 0}\n', + ] + ) + mock_open.return_value = mock_file + + result1 = load_specific_tests_cached(self.jsonl_path, test_ids) + self.assertEqual(mock_open.call_count, 1) + + # Second call should use cache, not open file again + result2 = load_specific_tests_cached(self.jsonl_path, test_ids) + self.assertEqual(mock_open.call_count, 1) # Should not increase + + # Results should be the same + self.assertEqual(result1, result2) + self.assertEqual(len(result1), 2) # Should have loaded 2 tests + + def test_error_handling(self): + """Test error handling in reward function.""" + # Test with non-existent file + rewards = olmocr_bench_reward( + prompts=["prompt"], + completions=["content"], + completion_ids=[[]], + pdf_path=["test.pdf"], + jsonl_file=["/nonexistent/file.jsonl"], + test_ids=[["test1"]], + ) + + # Should return None on error + self.assertEqual(len(rewards), 1) + self.assertIsNone(rewards[0]) + + +class TestRewardFrontMatter(unittest.TestCase): + """Test cases for the reward_front_matter function.""" + + def test_no_frontmatter(self): + """Test that completions without frontmatter get 0.0 reward.""" + completions = ["This is just text without any frontmatter", "Another completion with no YAML", "# A markdown header but no frontmatter"] + + rewards = reward_front_matter(prompts=["prompt"] * 3, completions=completions, claude_original=None) + + self.assertEqual(len(rewards), 3) + for reward in rewards: + self.assertEqual(reward, 0.0) + + def test_valid_frontmatter_no_claude(self): + """Test that valid frontmatter gets 0.5 reward when no claude_original.""" + completions = [ + """--- +primary_language: en +is_rotation_valid: True +rotation_correction: 0 +is_table: False +is_diagram: False +--- + +# Document content here""", + """--- +primary_language: fr +is_rotation_valid: False +rotation_correction: 90 +is_table: True +is_diagram: False +--- + +Some other content""", + ] + + rewards = reward_front_matter(prompts=["prompt"] * 2, completions=completions, claude_original=None) + + self.assertEqual(len(rewards), 2) + for reward in rewards: + self.assertEqual(reward, 0.5) + + def test_perfect_match_with_claude(self): + """Test perfect match with claude_original gets 1.0 reward.""" + claude_original_content = """--- +primary_language: en +is_rotation_valid: True +rotation_correction: 0 +is_table: False +is_diagram: False +--- + +Original content""" + + completion = """--- +primary_language: en +is_rotation_valid: True +rotation_correction: 0 +is_table: False +is_diagram: False +--- + +Different content but same frontmatter""" + + rewards = reward_front_matter(prompts=["prompt"], completions=[completion], claude_original=[claude_original_content]) + + self.assertEqual(len(rewards), 1) + self.assertAlmostEqual(rewards[0], 1.0, places=5) + + def test_partial_match_with_claude(self): + """Test partial match with claude_original gets intermediate reward.""" + claude_original_content = """--- +primary_language: en +is_rotation_valid: True +rotation_correction: 0 +is_table: False +is_diagram: False +--- + +Original content""" + + # 3 out of 5 fields match + completion = """--- +primary_language: en +is_rotation_valid: False +rotation_correction: 90 +is_table: False +is_diagram: False +--- + +Different values for rotation fields""" + + rewards = reward_front_matter(prompts=["prompt"], completions=[completion], claude_original=[claude_original_content]) + + self.assertEqual(len(rewards), 1) + # 0.5 base + 3 * 0.1 = 0.8 + self.assertAlmostEqual(rewards[0], 0.8, places=2) + + def test_invalid_frontmatter_format(self): + """Test that invalid YAML frontmatter gets 0.0 reward.""" + completions = [ + """--- +this is not: valid yaml + because of: : bad formatting +--- +content""", + """--- +primary_language: en +unclosed_string: "this is not closed +---""", + """--- +primary_language: + nested: "should not be nested" +---""", + ] + + rewards = reward_front_matter(prompts=["prompt"] * 3, completions=completions, claude_original=None) + + self.assertEqual(len(rewards), 3) + for reward in rewards: + self.assertEqual(reward, 0.0) + + def test_mixed_completions_with_claude(self): + """Test mixed completions with claude_original comparisons.""" + claude_originals = [ + """--- +primary_language: en +is_rotation_valid: True +rotation_correction: 0 +is_table: False +is_diagram: False +--- +Content 1""", + None, # No claude_original for this one + """--- +primary_language: fr +is_rotation_valid: False +rotation_correction: 180 +is_table: True +is_diagram: True +--- +Content 3""", + ] + + completions = [ + # Perfect match with first claude + """--- +primary_language: en +is_rotation_valid: True +rotation_correction: 0 +is_table: False +is_diagram: False +--- +Model output 1""", + # Valid frontmatter but no claude to compare + """--- +primary_language: de +is_rotation_valid: True +rotation_correction: 0 +is_table: False +is_diagram: False +--- +Model output 2""", + # No frontmatter at all + "Just plain text without frontmatter", + ] + + rewards = reward_front_matter(prompts=["prompt"] * 3, completions=completions, claude_original=claude_originals) + + self.assertEqual(len(rewards), 3) + self.assertAlmostEqual(rewards[0], 1.0, places=5) # Perfect match + self.assertEqual(rewards[1], 0.5) # Valid but no claude + self.assertEqual(rewards[2], 0.0) # No frontmatter + + def test_list_format_completions(self): + """Test handling of completions in list format.""" + completion_content = """--- +primary_language: en +is_rotation_valid: True +rotation_correction: 0 +is_table: False +is_diagram: False +--- +Content""" + + completions = [ + # List format (as from model output) + [{"content": completion_content}], + # String format + completion_content, + # Empty list + [], + # Invalid format + [{}], + ] + + rewards = reward_front_matter(prompts=["prompt"] * 4, completions=completions, claude_original=None) + + self.assertEqual(len(rewards), 4) + self.assertEqual(rewards[0], 0.5) # Valid from list + self.assertEqual(rewards[1], 0.5) # Valid from string + self.assertEqual(rewards[2], 0.0) # Empty list + self.assertEqual(rewards[3], 0.0) # Invalid list format + + def test_field_type_matching(self): + """Test that field types are correctly compared.""" + claude_original = """--- +primary_language: en +is_rotation_valid: True +rotation_correction: 0 +is_table: False +is_diagram: False +---""" + + completions = [ + # Correct types + """--- +primary_language: en +is_rotation_valid: True +rotation_correction: 0 +is_table: False +is_diagram: False +---""", + # String instead of boolean (might still parse correctly) + """--- +primary_language: en +is_rotation_valid: "True" +rotation_correction: "0" +is_table: "False" +is_diagram: "False" +---""", + ] + + rewards = reward_front_matter(prompts=["prompt"] * 2, completions=completions, claude_original=[claude_original] * 2) + + self.assertEqual(len(rewards), 2) + # First should be perfect match + self.assertAlmostEqual(rewards[0], 1.0, places=5) + # Second: YAML parses string "True" as True boolean, so both should match + self.assertAlmostEqual(rewards[1], 1.0, places=5) + + def test_none_values_in_fields(self): + """Test handling of None values in frontmatter fields.""" + claude_original = """--- +primary_language: en +is_rotation_valid: True +rotation_correction: 0 +is_table: False +is_diagram: False +---""" + + # rotation_correction null will fail validation (must be 0, 90, 180, or 270) + completion = """--- +primary_language: en +is_rotation_valid: True +rotation_correction: null +is_table: False +is_diagram: False +---""" + + rewards = reward_front_matter(prompts=["prompt"], completions=[completion], claude_original=[claude_original]) + + self.assertEqual(len(rewards), 1) + # Should fail to parse due to invalid rotation_correction + self.assertEqual(rewards[0], 0.0) + + +class TestIntegrationWithRealData(unittest.TestCase): + """Integration tests with real bench data if available.""" + + @unittest.skipUnless(os.path.exists("/home/ubuntu/olmocr/olmOCR-bench/bench_data"), "Real bench data not available") + def test_with_real_bench_data(self): + """Test with real bench data if available.""" + bench_data_folder = "/home/ubuntu/olmocr/olmOCR-bench/bench_data" + + dataset = OlmOCRBenchDataset( + bench_data_folder=bench_data_folder, + processor=None, + max_samples=5, + target_longest_image_dim=1024, + ) + + self.assertEqual(len(dataset), 5) + + # Test that we can iterate through the dataset + for i in range(len(dataset)): + item = dataset[i] + if item is not None: # Some PDFs might fail to render + self.assertIn("prompt", item) + self.assertIn("pdf_path", item) + self.assertIn("jsonl_file", item) + self.assertIn("test_ids", item) + + # Verify paths exist + self.assertTrue(os.path.exists(item["pdf_path"])) + self.assertTrue(os.path.exists(item["jsonl_file"])) + + +if __name__ == "__main__": + unittest.main(verbosity=2) diff --git a/tests/test_integration.py b/tests/test_integration.py new file mode 100644 index 0000000..520429f --- /dev/null +++ b/tests/test_integration.py @@ -0,0 +1,28 @@ +import glob +import json +import os +import unittest + +import pytest + + +@pytest.mark.nonci +class TestPipelineIntegration(unittest.TestCase): + def setUp(self): + self.data = [] + + for file in glob.glob(os.path.join("localworkspace", "results", "*.jsonl")): + with open(file, "r") as jf: + for line in jf: + if len(line.strip()) > 0: + self.data.append(json.loads(line)) + print(self.data[-1]) + + def test_edgar(self) -> None: + self.assertTrue(any("King of the English" in line["text"] for line in self.data)) + + def test_ambig(self) -> None: + self.assertTrue(any("Apples and Bananas" in line["text"] for line in self.data)) + + def test_dolma(self) -> None: + self.assertTrue(any("We extensively document Dolma" in line["text"] for line in self.data)) diff --git a/tests/test_katex_render.py b/tests/test_katex_render.py new file mode 100644 index 0000000..87387a7 --- /dev/null +++ b/tests/test_katex_render.py @@ -0,0 +1,223 @@ +import unittest + +from olmocr.bench.katex import compare_rendered_equations, render_equation + + +class TestRenderedEquationComparison(unittest.TestCase): + def test_exact_match(self): + eq1 = render_equation("a+b", use_cache=False) + eq2 = render_equation("a+b", use_cache=False) + self.assertTrue(compare_rendered_equations(eq1, eq2)) + + def test_whitespace_difference(self): + eq1 = render_equation("a+b", use_cache=False) + eq2 = render_equation("a + b", use_cache=False) + self.assertTrue(compare_rendered_equations(eq1, eq2)) + + def test_not_found(self): + eq1 = render_equation("c-d", use_cache=False) + eq2 = render_equation("a+b", use_cache=False) + self.assertFalse(compare_rendered_equations(eq1, eq2)) + + def test_align_block_contains_needle(self): + eq_plain = render_equation("a+b", use_cache=False) + eq_align = render_equation("\\begin{align*}a+b\\end{align*}", use_cache=False) + self.assertTrue(compare_rendered_equations(eq_plain, eq_align)) + + def test_align_block_needle_not_in(self): + eq_align = render_equation("\\begin{align*}a+b\\end{align*}", use_cache=False) + eq_diff = render_equation("c-d", use_cache=False) + self.assertFalse(compare_rendered_equations(eq_diff, eq_align)) + + def test_big(self): + ref_rendered = render_equation("\\nabla \\cdot \\mathbf{E} = \\frac{\\rho}{\\varepsilon_0}", use_cache=False, debug_dom=False) + align_rendered = render_equation( + """\\begin{align*}\\nabla \\cdot \\mathbf{E} = \\frac{\\rho}{\\varepsilon_0}\\end{align*}""", use_cache=False, debug_dom=False + ) + self.assertTrue(compare_rendered_equations(ref_rendered, align_rendered)) + + def test_dot_end1(self): + ref_rendered = render_equation( + "\\lambda_{g}=\\sum_{s \\in S} \\zeta_{n}^{\\psi(g s)}=\\sum_{i=1}^{k}\\left[\\sum_{s, R s=\\mathcal{I}_{i}} \\zeta_{n}^{\\varphi(g s)}\\right]" + ) + align_rendered = render_equation( + "\\lambda_{g}=\\sum_{s \\in S} \\zeta_{n}^{\\psi(g s)}=\\sum_{i=1}^{k}\\left[\\sum_{s, R s=\\mathcal{I}_{i}} \\zeta_{n}^{\\varphi(g s)}\\right]." + ) + self.assertTrue(compare_rendered_equations(ref_rendered, align_rendered)) + + def test_x_vs_textx(self): + ref_rendered = render_equation("C_{T}\\left(u_{n}^{T} X_{n}^{\\text {Test }}, \\bar{x}^{\\text {Test }}\\right)") + align_rendered = render_equation("C_T \\left(u^T_n X^{\\text{Test}}_n,\\overline{ \\text{x}}^{\\text{Test}}\\right)") + self.assertFalse(compare_rendered_equations(ref_rendered, align_rendered)) + + @unittest.skip("There is a debate whether bar and overline should be the same, currently they are not") + def test_overline(self): + ref_rendered = render_equation("C_{T}\\left(u_{n}^{T} X_{n}^{\\text {Test }}, \\bar{x}^{\\text {Test }}\\right)") + align_rendered = render_equation("C_T \\left(u^T_n X^{\\text{Test}}_n,\\overline{ x}^{\\text{Test}}\\right)") + self.assertTrue(compare_rendered_equations(ref_rendered, align_rendered)) + + def test_parens(self): + ref_rendered = render_equation("\\left\\{ \\left( 0_{X},0_{Y},-1\\right) \\right\\} ") + align_rendered = render_equation("\\{(0_{X}, 0_{Y}, -1)\\}") + self.assertTrue(compare_rendered_equations(ref_rendered, align_rendered)) + + def test_dot_end2(self): + ref_rendered = render_equation( + "\\lambda_{g}=\\sum_{s \\in S} \\zeta_{n}^{\\psi(g s)}=\\sum_{i=1}^{k}\\left[\\sum_{s, R s=\\mathcal{I}_{i}} \\zeta_{n}^{\\psi(g s)}\\right]" + ) + align_rendered = render_equation( + "\\lambda_g = \\sum_{s \\in S} \\zeta_n^{\\psi(gs)} = \\sum_{i=1}^{k} \\left[ \\sum_{s, Rs = \\mathcal{I}_i} \\zeta_n^{\\psi(gs)} \\right]" + ) + self.assertTrue(compare_rendered_equations(ref_rendered, align_rendered)) + + def test_lambda(self): + ref_rendered = render_equation("\\lambda_g = \\lambda_{g'}") + align_rendered = render_equation("\\lambda_{g}=\\lambda_{g^{\\prime}}") + self.assertTrue(compare_rendered_equations(ref_rendered, align_rendered)) + + def test_gemini(self): + ref_rendered = render_equation("u \\in (R/\\operatorname{Ann}_R(x_i))^{\\times}") + align_rendered = render_equation("u \\in\\left(R / \\operatorname{Ann}_{R}\\left(x_{i}\\right)\\right)^{\\times}") + self.assertTrue(compare_rendered_equations(ref_rendered, align_rendered)) + + def test_fraction_vs_divided_by(self): + eq1 = render_equation("\\frac{a}{b}", use_cache=False) + eq2 = render_equation("a / b", use_cache=False) + self.assertFalse(compare_rendered_equations(eq1, eq2)) + + def test_different_bracket_types(self): + eq1 = render_equation("\\left[ a + b \\right]", use_cache=False) + eq2 = render_equation("\\left\\{ a + b \\right\\}", use_cache=False) + self.assertFalse(compare_rendered_equations(eq1, eq2)) + + def test_inline_vs_display_style_fraction(self): + eq1 = render_equation("\\frac{1}{2}", use_cache=False) + eq2 = render_equation("\\displaystyle\\frac{1}{2}", use_cache=False) + self.assertTrue(compare_rendered_equations(eq1, eq2)) + + def test_matrix_equivalent_forms(self): + eq1 = render_equation("\\begin{pmatrix} a & b \\\\ c & d \\end{pmatrix}", use_cache=False) + eq2 = render_equation("\\begin{pmatrix} a & b \\\\ c & d \\end{pmatrix}", use_cache=False) + self.assertTrue(compare_rendered_equations(eq1, eq2)) + + def test_different_matrix_types(self): + eq1 = render_equation("\\begin{pmatrix} a & b \\\\ c & d \\end{pmatrix}", use_cache=False) + eq2 = render_equation("\\begin{bmatrix} a & b \\\\ c & d \\end{bmatrix}", use_cache=False) + self.assertFalse(compare_rendered_equations(eq1, eq2)) + + def test_thinspace_vs_regular_space(self): + eq1 = render_equation("a \\, b", use_cache=False) + eq2 = render_equation("a \\: b", use_cache=False) + self.assertTrue(compare_rendered_equations(eq1, eq2)) + + @unittest.skip("Currently these compare to the same thing, because they use the symbol 'x' with a different span class and thus font") + def test_mathbf_vs_boldsymbol(self): + eq1 = render_equation("\\mathbf{x}", use_cache=False) + eq2 = render_equation("\\boldsymbol{x}", use_cache=False) + self.assertFalse(compare_rendered_equations(eq1, eq2)) + + def test_assert_subtle_square_root(self): + eq1 = render_equation("A N'P' = \\int \\beta d\\alpha = \\frac{2}{3\\sqrt{3} a}\\int (\\alpha - 2a)^{\\frac{3}{2}} d\\alpha", use_cache=False) + eq2 = render_equation("AN'P' = \\int \\beta \\, d\\alpha = \\frac{2}{3 \\sqrt{3a}} \\int (a - 2a)^{\\frac{3}{2}} d\\alpha") + self.assertFalse(compare_rendered_equations(eq1, eq2)) + + def test_text_added(self): + eq1 = render_equation("A N'P' = \\int \\beta d\\alpha = \\frac{2}{3\\sqrt{3} a}\\int (\\alpha - 2a)^{\\frac{3}{2}} d\\alpha", use_cache=False) + eq2 = render_equation("AN'P' = \\int \\beta d\\alpha = \\frac{2}{3 \\sqrt{3} a} \\int (\\alpha - 2a)^{\\frac{3}{2}} d\\alpha") + self.assertTrue(compare_rendered_equations(eq1, eq2)) + + eq1 = render_equation("A N'P' = \\int \\beta d\\alpha = \\frac{2}{3\\sqrt{3} a}\\int (\\alpha - 2a)^{\\frac{3}{2}} d\\alpha", use_cache=False) + eq2 = render_equation("\\text{area evolute } AN'P' = \\int \\beta d\\alpha = \\frac{2}{3 \\sqrt{3} a} \\int (\\alpha - 2a)^{\\frac{3}{2}} d\\alpha") + self.assertTrue(compare_rendered_equations(eq1, eq2)) + + def test_tensor_notation_equivalent(self): + eq1 = render_equation("T_{ij}^{kl}", use_cache=False) + eq2 = render_equation("T^{kl}_{ij}", use_cache=False) + self.assertTrue(compare_rendered_equations(eq1, eq2)) + + def test_partial_derivative_forms(self): + eq1 = render_equation("\\frac{\\partial f}{\\partial x}", use_cache=False) + eq2 = render_equation("\\frac{\\partial_f}{\\partial_x}", use_cache=False) + self.assertFalse(compare_rendered_equations(eq1, eq2)) + + def test_equivalent_sin_forms_diff_parens(self): + eq1 = render_equation("\\sin(\\theta)", use_cache=False) + eq2 = render_equation("\\sin \\theta", use_cache=False) + self.assertFalse(compare_rendered_equations(eq1, eq2)) + + def test_aligned_multiline_equation(self): + eq1 = render_equation("\\begin{align*} a &= b \\\\ c &= d \\end{align*}", use_cache=False) + eq2 = render_equation("\\begin{aligned} a &= b \\\\ c &= d \\end{aligned}", use_cache=False) + self.assertTrue(compare_rendered_equations(eq1, eq2)) + + def test_subscript_order_invariance(self): + eq1 = render_equation("x_{i,j}", use_cache=False) + eq2 = render_equation("x_{j,i}", use_cache=False) + self.assertFalse(compare_rendered_equations(eq1, eq2)) + + def test_hat_vs_widehat(self): + eq1 = render_equation("\\hat{x}", use_cache=False) + eq2 = render_equation("\\widehat{x}", use_cache=False) + self.assertFalse(compare_rendered_equations(eq1, eq2)) + + def test_equivalent_integral_bounds(self): + eq1 = render_equation("\\int_{a}^{b} f(x) dx", use_cache=False) + eq2 = render_equation("\\int\\limits_{a}^{b} f(x) dx", use_cache=False) + # Could go either way honestly? + self.assertTrue(compare_rendered_equations(eq1, eq2)) + + def test_equivalent_summation_notation(self): + eq1 = render_equation("\\sum_{i=1}^{n} x_i", use_cache=False) + eq2 = render_equation("\\sum\\limits_{i=1}^{n} x_i", use_cache=False) + self.assertTrue(compare_rendered_equations(eq1, eq2)) + + def test_different_symbol_with_same_appearance(self): + eq1 = render_equation("\\phi", use_cache=False) + eq2 = render_equation("\\varphi", use_cache=False) + self.assertFalse(compare_rendered_equations(eq1, eq2)) + + def test_aligned_vs_gathered(self): + eq1 = render_equation("\\begin{aligned} a &= b \\\\ c &= d \\end{aligned}", use_cache=False) + eq2 = render_equation("\\begin{gathered} a = b \\\\ c = d \\end{gathered}", use_cache=False) + # Different whitespacing, should be invariant to that. + self.assertTrue(compare_rendered_equations(eq1, eq2)) + + def test_identical_but_with_color1(self): + eq1 = render_equation("a + b", use_cache=False) + eq2 = render_equation("\\color{black}{a + b}", use_cache=False) + self.assertTrue(compare_rendered_equations(eq1, eq2)) + + def test_identical_but_with_color2(self): + eq1 = render_equation("a + b", use_cache=False) + eq2 = render_equation("\\color{black}{a} + \\color{black}{b}", use_cache=False) + self.assertTrue(compare_rendered_equations(eq1, eq2)) + + eq1 = render_equation("a + b", use_cache=False) + eq2 = render_equation("\\color{red}{a} + \\color{black}{b}", use_cache=False) + self.assertTrue(compare_rendered_equations(eq1, eq2)) + + def test_newcommand_expansion(self): + eq1 = render_equation("\\alpha + \\beta", use_cache=False) + eq2 = render_equation("\\newcommand{\\ab}{\\alpha + \\beta}\\ab", use_cache=False) + self.assertTrue(compare_rendered_equations(eq1, eq2)) + + def test_tech_report_examples(self): + ref = render_equation("C_T \\left(u^T_n X^{\\text{Test}}_n,\\overline{ x}^{\\text{Test}}\\right)", use_cache=False) + model_a = render_equation("C_{T}\\left(u_{n}^{T} X_{n}^{\\text{Test }}, \\overline{x}^{\\text {Test }}\\right)", use_cache=False) + model_b = render_equation("C_T \\left(u^T_n X^{\\text{Test}}_n,\\overline{ \\text{x}}^{\\text{Test}}\\right)", use_cache=False) + self.assertTrue(compare_rendered_equations(ref, model_a)) + self.assertTrue(compare_rendered_equations(ref, model_b)) + + def test_viks_equation(self): + eq1 = render_equation( + "s_{r,s}=\\begin{cases} 2 \\text{ if } r-s=2,4 \\text{ mod 4},\\\\ 1 \\text{ if } r-s=1,3 \\text{ mod 4}. \\end{cases}", use_cache=False + ) + eq2 = render_equation( + "s_{r,s}=\\begin{cases} 2 & \\text{if } r-s=2,4 \\text{ mod } 4, \\\\ 1 & \\text{if } r-s=1,3 \\text{ mod } 4. \\end{cases}", use_cache=False + ) + self.assertTrue(compare_rendered_equations(eq1, eq2)) + + def test_arrow_vs_to(self): + eq1 = render_equation("\\to", use_cache=False) + eq2 = render_equation("\\rightarrow", use_cache=False) + self.assertTrue(compare_rendered_equations(eq1, eq2)) diff --git a/tests/test_mine_html_templates_footnotes.py b/tests/test_mine_html_templates_footnotes.py new file mode 100644 index 0000000..5134015 --- /dev/null +++ b/tests/test_mine_html_templates_footnotes.py @@ -0,0 +1,503 @@ +import random +import unittest +from types import SimpleNamespace +from unittest.mock import patch + +from olmocr.synth.mine_html_templates import generate_tests_from_html + + +class TestFootnoteTestGeneration(unittest.TestCase): + def setUp(self): + self.random_gen = random.Random(42) + self.pdf_id = "test_pdf" + self.page_num = 1 + self.uuid_counter = 0 + uuid_patch = patch( + "olmocr.synth.mine_html_templates.uuid.uuid4", + side_effect=self._fake_uuid, + ) + self.addCleanup(uuid_patch.stop) + uuid_patch.start() + + def _fake_uuid(self): + current = f"{self.uuid_counter:032x}" + self.uuid_counter += 1 + return SimpleNamespace(hex=current) + + @staticmethod + def _hashable_tests(tests): + return {tuple(sorted(test.items())) for test in tests} + + def _generate_footnote_tests(self, html_content): + tests = generate_tests_from_html( + html_content, + self.pdf_id, + self.page_num, + self.random_gen, + False, + ) + footnote_tests = [] + for test in tests: + if test.get("type") == "footnote": + sanitized = {k: v for k, v in test.items() if k != "id"} + footnote_tests.append(sanitized) + return self._hashable_tests(footnote_tests) + + def test_single_marker_generates_marker_only(self): + html_content = """ + +
Alpha with reference¹ and more text.
+ + """ + + footnote_tests = self._generate_footnote_tests(html_content) + + self.assertSetEqual( + footnote_tests, + self._hashable_tests( + [ + { + "pdf": "test_pdf_page1.pdf", + "page": 1, + "type": "footnote", + "marker": "1", + "max_diffs": 0, + "appears_before_marker": "Alpha with reference", + } + ] + ), + ) + + def test_marker_with_definition_includes_text(self): + html_content = """ + +
Alpha with reference¹ and more text.
+
+
¹This is the footnote text that elaborates on the reference in detail.
+ + """ + + footnote_tests = self._generate_footnote_tests(html_content) + + self.assertSetEqual( + footnote_tests, + self._hashable_tests( + [ + { + "pdf": "test_pdf_page1.pdf", + "page": 1, + "type": "footnote", + "marker": "1", + "max_diffs": 0, + "appears_before_marker": "Alpha with reference", + "appears_after_marker": "This is the", + } + ] + ), + ) + + def test_multiple_markers_mixed_occurrences(self): + html_content = """ + +
First paragraph ends with marker¹ in the sentence.
+
Second paragraph carries marker² for another note.
+
Here is a lone marker³ with no definition below.
+
+
¹Definition for footnote one with ample descriptive content to qualify.
+
²Definition for footnote two including enough characters to be captured correctly.
+ + """ + + footnote_tests = self._generate_footnote_tests(html_content) + + self.assertSetEqual( + footnote_tests, + self._hashable_tests( + [ + { + "pdf": "test_pdf_page1.pdf", + "page": 1, + "type": "footnote", + "marker": "1", + "max_diffs": 0, + "appears_before_marker": "ends with marker", + "appears_after_marker": "Definition for footnote", + }, + { + "pdf": "test_pdf_page1.pdf", + "page": 1, + "type": "footnote", + "marker": "2", + "max_diffs": 0, + "appears_before_marker": "paragraph carries marker", + "appears_after_marker": "Definition for footnote", + }, + { + "pdf": "test_pdf_page1.pdf", + "page": 1, + "type": "footnote", + "marker": "3", + "max_diffs": 0, + "appears_before_marker": "a lone marker", + }, + ] + ), + ) + + def test_cutoff(self): + html_content = """Your personal data is processed in accordance with Regulation (EU) No 2018/1725¹ on the protection of individuals with regard to the processing of personal data by the Union institutions, bodies, offices and agencies and on the free movement of such data.""" + + footnote_tests = self._generate_footnote_tests(html_content) + + self.assertEqual(len(footnote_tests), 1) + + self.assertSetEqual( + footnote_tests, + self._hashable_tests( + [ + { + "pdf": "test_pdf_page1.pdf", + "page": 1, + "type": "footnote", + "marker": "1", + "max_diffs": 0, + "appears_before_marker": "(EU) No 2018/1725", + }, + ] + ), + ) + + def test_footnotes_in_footer(self): + # Sometimes footnotes appear in a
tag, and we like to remove text from footer tags and adjust how things work with it + # So the idea here is to test that + html_content = """ + + + + + + +9.3 Principles of Lagrangian construction + + +
+9.3 Principles of Lagrangian construction +133 +
+
9.3.2 A worked-out example: Vector and tensor fields
+
+ The previous analysis can be easily extended to other types of fields. Consider for instance a vector field $A_\mu$ and a tensor field $B_{\mu\nu}$. What is the most general Lagrangian $\mathcal{L}(A, \partial A, B, \partial B)$ that can be constructed out of these two fields? To clarify the procedure, let me split the problem into several pieces +
+
+
+³Quadratic actions give rise to linear equations of motion, where the superposition principle can be applied. +
+
+⁴In the same way that physically cannot be attributed to $\mathcal{L}$, we cannot make any claim about the physicality of $A_\mu$. Physicality might be attributed to the set $\{A_\mu\}$ of gauge-equivalent 4-potentials or to any gauge invariant attribute of that set, but not to its individual elements. +
+
+⁵It cannot be compensated by the transformation of the other (derivative) terms. +
+
+ + +""" + footnote_tests = self._generate_footnote_tests(html_content) + + print(footnote_tests) + + self.assertEqual(len(footnote_tests), 3) + + def test_footnotes_in_footer_ptags(self): + # Sometimes footnotes appear in a
tag, and we like to remove text from footer tags and adjust how things work with it + # So the idea here is to test that + html_content = """ + + + + + + +9.3 Principles of Lagrangian construction + + +
+9.3 Principles of Lagrangian construction +133 +
+
9.3.2 A worked-out example: Vector and tensor fields
+
+ The previous analysis can be easily extended to other types of fields. Consider for instance a vector field $A_\mu$ and a tensor field $B_{\mu\nu}$. What is the most general Lagrangian $\mathcal{L}(A, \partial A, B, \partial B)$ that can be constructed out of these two fields? To clarify the procedure, let me split the problem into several pieces +
+
+
+³Quadratic actions give rise to linear equations of motion, where the superposition principle can be applied. +
+
+⁴In the same way that physically cannot be attributed to $\mathcal{L}$, we cannot make any claim about the physicality of $A_\mu$. Physicality might be attributed to the set $\{A_\mu\}$ of gauge-equivalent 4-potentials or to any gauge invariant attribute of that set, but not to its individual elements. +
+
+⁵It cannot be compensated by the transformation of the other (derivative) terms. +
+
Don't make an absense test
+
+ + +""" + tests = generate_tests_from_html( + html_content, + self.pdf_id, + self.page_num, + self.random_gen, + False, + ) + + self.assertEqual(len([test for test in tests if test["type"] == "footnote"]), 3) + + for test in [test for test in tests if test["type"] == "absent"]: + self.assertNotEqual(test["text"], "Don't make an absense test") + + def test_sup_tags_in_text(self): + """Test that sup/sub tags are preserved in markdown but not in generated test text fields""" + html = """ + + + + + + +Management of Symptomatic Stenosis + + +
+348 +D. Barros Casas et al. / Arch Bronconeumol. 2013;9(8):345–354 +
+
Figure 3. Flowchart for the management of symptomatic stenosis. (*) Radial incisions, balloon dilation and topical application of mitomycin.
+
+
+
as an aid for other endoscopic techniques at various levels of the airway or as the sole technique in the case of simple, short stenoses that do not completely obstruct the airway lumen; this technique is well supported in the scientific literature.⁵⁷
+
Laser is only useful in small, narrow lesions with a reduced vertical length and stable cartilaginous skeleton, although it is widely and generally used with equally good results and low risk in the case of larger lesions. The decannulation rate is high, surgical time is reduced, and hospital stay is short-term.⁵⁸ For web-like stenosis, there is a variation of the technique that involves making radial incisions with the laser or with the electrocautery knife at 3, 9 and 12 o'clock before dilating.^58–59
+
The microdebrider has been shown to be effective in lesions with excessive granulation tissue.^51–52
+
Stenting is indicated in patients who do not respond to endoscopic dilation and are not candidates for surgical resection. It is important to remember that the stents indicated for this type of lesion must be easy to remove; at present, silicone stents are the most widely used, although there are also reports of the use of coated AERO hybrid nitinol stents. These are self-expanding and can be removed, and do not require rigid bronchoscopy for implantation.^43,45 Loss of cartilaginous support in the absence of extrinsic compression leads to migration of stents located in the subglottic region or proximal trachea. In these cases, external percutaneous fixation may be considered. Potential complications include skin infections around the external button.^46,47 Re-stenosis as a result of the repair process itself and stent obstruction are the main reasons for re-intervention.^29,32
+
The use of topical mitomycin is controversial, but together with radial laser incisions and balloon dilation it has some beneficial effect compared to placebo at 2–3 years^60–62 (Fig. 3).
+
Subglottic stenosis, mainly caused by intubation, deserves a special mention. The subglottic space refers to the section of the airway between the vocal cords and the lower traction of the cricoid cartilage, which is the narrowest section of the larynx and the only one surrounded by a complete ring of cartilage. Its narrow diameter, inconvenient location, surrounding tissue fragility of the coating mucosa and the tendency to form granulation tissue and scars from intubation, re-stenosis and failure to decannulate.⁵¹ An incidence of subglottic stenosis secondary to prolonged intubation in children and adults ranging from 0.9% to 8.3% has been reported.⁵² Management is a challenge involving various strategies that must be tailored to suit each patient. For non-concentric soft, membranous stenoses with sufficient cartilaginous support and a length of
+
+
+
around one centimeter corresponding to Cotton-Meyer grades I and II, endoscopic techniques described above are used, with emphasis on the use of laser. The success rate is variable according to the literature, ranging between 40% and 94%.⁵³ Longer, hard, grade III and IV complex stenoses can be treated initially with endoscopic techniques, but in most cases, open reconstructive surgery will be required (surgical resection of the stenosed section, including several tracheal rings and the anterior cricoid ring, in addition to the lower half of the mucosa of the cricoid cartilage, followed by end-to-end anastomosis).^54,54
+
Dynamic Airway Obstruction: Tracheobronchomalacia and Excessive Pars Membranacea Collapse
+
TBM and excessive pars membranacea collapse occur in around 12% of patients with respiratory diseases.₅₅ In TBM, the proportion between cartilage and soft tissues is reduced from a normal ratio of 4:1 to 1:1.⁵⁶ While in excessive pars membranacea collapse, there is atrophy and a loss of myoelastic fibers.⁵⁷ TBM, in both its local and diffuse forms, can affect the trachea, bronchi or both structures. There are different ways of classifying the disease, but the functional classification (FEMOS) is the most comprehensive.⁵⁸ TBM may be asymptomatic, although it often produces cough, wheezing, stridor, dyspnea, recurrent infections, and on occasions, respiratory failure,⁵⁹ and therefore differential diagnosis is needed to rule out disease entities such as chronic obstructive pulmonary disease, asthma and bronchiectasis.⁶¹ Respiratory function tests can help in the diagnosis of concomitant obstructive pulmonary disease, but they have limited application in the diagnosis of TBM, since results are normal in up to 21% of cases.⁶² Accordingly, dynamic chest tomography and dynamic flexible bronchoscopy are often required for diagnosis.^63,64 (Fig. 1, image 3). This disease can be easily diagnosed by the performance of dynamic inhalation and exhalation maneuvers. In patients with diffuse TBM, a diagnostic test must be performed with silicone stent placement,^65–67 along with management of comorbidities. Patients who show improvement in their symptoms and respiratory function will be candidates for surgical or medical reconstruction by tracheobronchoplasty.⁵⁵ Patients who cannot undergo surgery due to their comorbidities, will be managed with a combination of symptomatic treatment and possible definitive stenting (Fig. 4).
+
Although non-invasive ventilation has been proposed as a possible treatment for TBM, its role appears to be restricted to the
+
+
+ +""" + + tests = generate_tests_from_html( + html, + self.pdf_id, + self.page_num, + self.random_gen, + False, + ) + + # Check that tests containing sup/sub tags are filtered out + # The test generation should filter these out already + for test in tests: + if "text" in test: + # These tags should have been filtered out during test generation + self.assertNotIn("^{", test["text"])
+ self.assertNotIn("}", test["text"]) + self.assertNotIn("_{", test["text"])
+ self.assertNotIn("}", test["text"]) + + def test_footnotes_footer_again(self): + html = """ + + + + + +KPN Second Quarter 2017 Results + + +
+
+
+
Press release
+
26 July 2017
+
+
+
+
Second Quarter 2017 Results
+
Highlights
+
+
Fixed-mobile convergence continues to deliver strong results in Consumer +
+
More than 60% of KPN brand postpaid base in fixed-mobile bundles (Q2 2016: 51%)
+
+8k broadband net adds, +25k IPTV net adds, and +9k postpaid net adds^§ driven by the high value KPN brand
+
Postpaid ARPU stable at EUR 26, ARPU per household increased by 5.0% y-on-y to EUR 42
+
Further improvement in customer satisfaction in Consumer; NPS +13 (Q2 2016: +9)
+
+
+
Progress with Business transformation +
+
SME: fixed-mobile bundling on track, +39k multi play net adds driven by uptake KPN ÉÉN; positioning for growth in IT through up- and cross-sell
+
LE & Corporate: focus on value in competitive mobile-only market; growth in IT related services, Internet of Things and Security
+
Dedicated to further improve customer satisfaction in Business; NPS -6 (Q2 2016: -9)
+
+
+
Second wave Simplification program delivered ~EUR 65m run-rate savings by end Q2 2017
+
+
Key figures^* (from continuing operations)
+
Group financials (unaudited)
+
(in EUR m, unless stated otherwise)
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
Q2 2017 Q2 2016 Δ y-on-y YTD 2017 YTD 2016 Δ y-on-y
Revenues 1,631 1,676 -2.7% 3,279 3,365 -2.6%
Adjusted revenues** 1,623 1,676 -3.2% 3,271 3,365 -2.8%
Adjusted revenues The Netherlands** 1,485 1,486 -2.1% 2,955 2,989 -1.8%
EBITDA 587 579 1.4% 1,143 1,138 0.4%
Adjusted EBITDA** 601 592 1.5% 1,185 1,160 2.2%
Adjusted EBITDA margin The Netherlands 41.0% 39.5% +150bps 40.1% 38.5% +160bps
Operating profit (EBIT) 232 205 13% 433 346 25%
Profit for the period (net profit) 151 162 -18% 283 210 35%
Capex 238 312 -24% 504 630 -20%
Operating free cash flow 363 280 30% 681 530 28%
Free cash flow 296 254 17% 331 214 55%
+
* All non-IFRS terms are explained in the safe harbor section
+
** Adjusted revenues and adjusted EBITDA reconciliations to be found on page 8 and 9
+
Financial performance
+
+
+
Adjusted revenues for The Netherlands were 2.1% lower y-on-y in Q2 2017. Customer base growth and higher ARPU per household led to revenue growth in Consumer, but this was offset by lower Business and Wholesale revenues. Revenues in Business were impacted by migrations and rationalization, and price pressure in mobile, while Wholesale saw lower revenues from MVNOs and international traffic
+
Adjusted EBITDA was 1.5% higher y-on-y in Q2 2017 due to improved operational efficiency as a result of simplification and lower subscriber retention and acquisition costs
+
Operating profit increased by 13% y-on-y in Q2 2017 mainly driven by EUR 19m lower amortization charges. Net profit increased by 18% y-on-y to EUR 191m in Q2 2017
+
Capex in H1 2017 was 20% lower y-on-y, mainly due to different intrayear phasing compared to last year
+
Free cash flow (excl. TEFD dividend) of EUR 261m in H1 2017 was EUR 157m higher compared to the same period last year, mainly driven by lower Capex and less interest paid
+
+
+
+
+
+
^§ Adjusted for 15k migration to Business
+
KPN Management Report Q2 20171
+
+
+ +""" + + tests = generate_tests_from_html( + html, + self.pdf_id, + self.page_num, + self.random_gen, + False, + ) + + print(tests) + + +if __name__ == "__main__": + unittest.main() diff --git a/tests/test_mine_html_templates_formatting.py b/tests/test_mine_html_templates_formatting.py new file mode 100644 index 0000000..a0b9992 --- /dev/null +++ b/tests/test_mine_html_templates_formatting.py @@ -0,0 +1,300 @@ +import random +import unittest +from unittest.mock import MagicMock, patch + +from bs4 import BeautifulSoup + +from olmocr.bench.tests import TestType +from olmocr.synth.mine_html_templates import ( + PreserveTablesConverter, + extract_html_metadata, + generate_tests_from_html, + html_to_markdown_with_frontmatter, +) + + +class TestFormatTestGeneration(unittest.TestCase): + """Test the generation of FormatTests from HTML content""" + + def setUp(self): + """Set up test fixtures""" + self.random_gen = random.Random(42) + self.pdf_id = "test_pdf" + self.page_num = 1 + + def test_format_test_generation_basic(self): + """Test basic format test generation with headings, bold, and italic""" + html_content = """ + + +
Main Title
+
Subtitle
+
This is a paragraph with bold text and italic text.
+
Also has strong emphasis and emphasized text.
+ + + """ + + tests = generate_tests_from_html(html_content, self.pdf_id, self.page_num, self.random_gen, False) + + # Filter format tests + format_tests = [t for t in tests if t.get("type") == "format"] + + # Should generate format tests + self.assertGreater(len(format_tests), 0) + self.assertLessEqual(len(format_tests), 5) # Should be limited to 5 + + # Check that we have different format types + format_types = {t["format"] for t in format_tests} + self.assertTrue("heading" in format_types) + self.assertTrue("bold" in format_types or "italic" in format_types) + + # Check specific text was captured + format_texts = {t["text"] for t in format_tests} + self.assertIn("Main Title", format_texts) + self.assertIn("Subtitle", format_texts) + + def test_format_test_limit(self): + """Test that format tests are limited to 5 total""" + html_content = """ + + +
Title 1
+
Title 2
+
Title 3
+
Title 4
+
Title 5
+
Title 6
+
Bold 1 Bold 2 Bold 3
+
Italic 1 Italic 2 Italic 3
+ + + """ + + tests = generate_tests_from_html(html_content, self.pdf_id, self.page_num, self.random_gen, False) + + # Filter format tests + format_tests = [t for t in tests if t.get("type") == "format"] + + # Should be limited to exactly 5 + self.assertEqual(len(format_tests), 5) + + def test_format_test_with_tables_headers_footers(self): + """Test that format tests exclude content in headers, footers, and tables""" + html_content = """ + + +
+
Header Title
+ Header Bold +
+ +
Main Content Title
+
Content with bold text
+ + + + + + +
Table Bold Table Italic
+ +
+ Footer Italic +
+ + + """ + + tests = generate_tests_from_html(html_content, self.pdf_id, self.page_num, self.random_gen, False) + + # Filter format tests + format_tests = [t for t in tests if t.get("type") == "format"] + format_texts = {t["text"] for t in format_tests} + + # Should only have main content formatting, not header/footer/table content + self.assertIn("Main Content Title", format_texts) + self.assertIn("bold text", format_texts) + self.assertNotIn("Header Title", format_texts) + self.assertNotIn("Header Bold", format_texts) + self.assertNotIn("Table Bold", format_texts) + self.assertNotIn("Table Italic", format_texts) + self.assertNotIn("Footer Italic", format_texts) + + def test_format_test_normalization(self): + """Test that text normalization works in format tests""" + html_content = """ + + +
Title with extra spaces
+
"Fancy" 'quotes'
+
Text—with–dashes
+ + + """ + + tests = generate_tests_from_html(html_content, self.pdf_id, self.page_num, self.random_gen, False) + + # Filter format tests + format_tests = [t for t in tests if t.get("type") == "format"] + format_texts = {t["text"] for t in format_tests} + + # Text should be normalized + self.assertIn("Title with extra spaces", format_texts) + # Quotes should be normalized + found_quotes = any("\"Fancy\" 'quotes'" in text for text in format_texts) + self.assertTrue(found_quotes) + # Dashes should be normalized + found_dashes = any("Text-with-dashes" in text for text in format_texts) + self.assertTrue(found_dashes) + + def test_format_test_no_duplicates(self): + """Test that duplicate text doesn't create multiple format tests""" + html_content = """ + + +
Same Title
+
Same Title
+
Same Title
+
Different Title
+ + + """ + + tests = generate_tests_from_html(html_content, self.pdf_id, self.page_num, self.random_gen, False) + + # Filter format tests + format_tests = [t for t in tests if t.get("type") == "format"] + format_texts = [t["text"] for t in format_tests] + + # "Same Title" should only appear once + self.assertEqual(format_texts.count("Same Title"), 1) + # "Different Title" should also be there + self.assertIn("Different Title", format_texts) + + def test_format_test_minimum_length(self): + """Test that very short text is not included in format tests""" + html_content = """ + + +
A
+
OK
+
Valid Title
+
B Good Bold
+
C Nice Italic
+ + + """ + + tests = generate_tests_from_html(html_content, self.pdf_id, self.page_num, self.random_gen, False) + + # Filter format tests + format_tests = [t for t in tests if t.get("type") == "format"] + format_texts = {t["text"] for t in format_tests} + + # Short text (less than 3 characters) should be excluded + self.assertNotIn("A", format_texts) + self.assertNotIn("B", format_texts) + self.assertNotIn("C", format_texts) + self.assertNotIn("OK", format_texts) # Only 2 characters + + # Longer text should be included + self.assertIn("Valid Title", format_texts) + self.assertIn("Good Bold", format_texts) + self.assertIn("Nice Italic", format_texts) + + def test_bold_number_order_bug(self): + """Test that bold numbers in HTML don't create invalid order tests with max_diffs=1 + + This test catches a bug where 6 in HTML gets converted to **6** in markdown, + which then gets selected as a sentence for order tests. Since **6** has exactly 5 + characters, it passes the length check, but when compared with a longer sentence, + max_diffs gets set to 1 (2% of the longer sentence), which is problematic. + """ + html_content = """ + + +
This is a normal sentence with some content
+
6
+
Another sentence here that has stuff
+ + + """ + + tests = generate_tests_from_html(html_content, self.pdf_id, self.page_num, self.random_gen, False) + + # You don't want to see any order tests in this case, because the has caused the text comparison to be too small + order_tests = [t for t in tests if t.get("type") == "order"] + self.assertEqual(order_tests, []) + + def test_bold_number_order_bug2(self): + """Test that bold numbers in HTML don't create invalid order tests with max_diffs=1 + + This test catches a bug where 6 in HTML gets converted to **6** in markdown, + which then gets selected as a sentence for order tests. Since **6** has exactly 5 + characters, it passes the length check, but when compared with a longer sentence, + max_diffs gets set to 1 (2% of the longer sentence), which is problematic. + """ + html_content = """ + + +
This is a normal sentence with some content
+
Wow cool!
+
Another sentence here that has stuff
+ + + """ + + tests = generate_tests_from_html(html_content, self.pdf_id, self.page_num, self.random_gen, False) + + # You don't want to see any order tests in this case, because the has caused the text comparison to be too small + order_tests = [t for t in tests if t.get("type") == "order"] + self.assertGreater(len(order_tests), 0) + + @unittest.skip("We have commented out this type of automatic test, due to concerns it would incentivize the model wrongly") + def test_common_words_absence(self): + """Test that 3 common words most similar to page content are added as absence tests""" + # Create HTML content that deliberately excludes very common English words + # but includes words like "physical", "specific", "required" which are similar to common words + html_content = """ + + +
Scientific Research Paper
+
Quantum mechanics describes physical properties at atomic scale.
+
Experiments demonstrate wave-particle duality phenomenon.
+
Mathematical equations predict behavior accurately.
+
Results confirm theoretical predictions remarkably well.
+
Further investigation required regarding specific anomalies.
+ + + """ + + # Note: This content deliberately avoids some very common words like "the", "to", "and", "of", "a", "in", "is", "for", "that" + # But includes words that might be similar (e.g., "specific" is similar to "specify", "required" to "require") + tests = generate_tests_from_html(html_content, self.pdf_id, self.page_num, self.random_gen, False) + + # Find absence tests for common words + absent_common_tests = [t for t in tests if t.get("type") == "absent" and "absent_common" in t.get("id", "")] + + # We should have up to 3 absence tests for common words not on the page + self.assertGreater(len(absent_common_tests), 0, "Should have at least one common word absence test") + self.assertLessEqual(len(absent_common_tests), 3, "Should have at most 3 common word absence tests") + + # Check that these are indeed common words (they should be in the top 1000 most common words) + from wordfreq import top_n_list + + top_1000_words = set(top_n_list("en", 1000)) + + for test in absent_common_tests: + word = test["text"].lower() + self.assertIn(word, top_1000_words, f"'{word}' should be a common English word from top 1000") + + # Verify that the absent words really don't appear in the content + markdown = html_to_markdown_with_frontmatter(html_content) + for test in absent_common_tests: + word = test["text"].lower() + self.assertNotIn(word, markdown.lower(), f"'{word}' should not appear in the content") + + +if __name__ == "__main__": + unittest.main() diff --git a/tests/test_mine_html_templates_incomplete_content.py b/tests/test_mine_html_templates_incomplete_content.py new file mode 100644 index 0000000..2405224 --- /dev/null +++ b/tests/test_mine_html_templates_incomplete_content.py @@ -0,0 +1,283 @@ +import asyncio +import os +import tempfile +import unittest +from unittest.mock import MagicMock, patch + +from bs4 import BeautifulSoup + +from olmocr.bench.tests import TestType +from olmocr.synth.mine_html_templates import render_pdf_with_playwright + + +class TestIncompleteContent(unittest.TestCase): + def test_pg39(self): + html = """ + + + + + +History of Computing + + + + +
+
9/2/2010
+
+
+
+
+
+
+
+
+
+
+
+
+
+
HISTORY OF COMPUTING
+
Fall 2010 - CSE 111
+
Friday, September 3 & Wednesday, September 8
+
+
+ +
+
+
+
+
+
+
+
+
+
PRE 1642
+
+
3000 BC – Abacus is invented in Babylonia
+
+
+
800 AD - Chinese begin to use zero.
+
+
+
+
← Exo
+
← Span
+
← Rods
+
← Frame
+
+
+
+ From 800 to 1641 some kinda important stuff happens in math – algebra, calculus, geometry, etc. +
+
+
+
1
+
+ +""" + + with tempfile.TemporaryDirectory() as td: + result = asyncio.run(render_pdf_with_playwright(html, os.path.join(td, "temp.pdf"), 768, 1024)) + + # Somehow we need to error out and show that this page didn't render fully + self.assertFalse(result.success) + + +if __name__ == "__main__": + unittest.main() diff --git a/tests/test_mine_html_templates_introduce_text_errors.py b/tests/test_mine_html_templates_introduce_text_errors.py new file mode 100644 index 0000000..6554e86 --- /dev/null +++ b/tests/test_mine_html_templates_introduce_text_errors.py @@ -0,0 +1,381 @@ +import random +import unittest + +from olmocr.synth.augmentations import ( + _apply_typo, + _has_skip_ancestor, + introduce_text_errors, +) + + +class TestApplyTypo(unittest.TestCase): + """Unit tests for the _apply_typo helper.""" + + def test_preserves_first_and_last_char(self): + rng = random.Random(0) + for seed in range(50): + rng.seed(seed) + typo = _apply_typo("abcdef", rng) + self.assertEqual(typo[0], "a") + self.assertEqual(typo[-1], "f") + + def test_swap_strategy(self): + # Force swap by controlling the RNG + rng = random.Random() + word = "abcdef" + found_swap = False + for seed in range(200): + rng.seed(seed) + typo = _apply_typo(word, rng) + # A swap keeps same length and same set of chars + if len(typo) == len(word) and sorted(typo) == sorted(word) and typo != word: + found_swap = True + break + self.assertTrue(found_swap, "Expected to find at least one swap typo") + + def test_delete_strategy(self): + rng = random.Random() + word = "abcdef" + found_delete = False + for seed in range(200): + rng.seed(seed) + typo = _apply_typo(word, rng) + if len(typo) == len(word) - 1: + found_delete = True + break + self.assertTrue(found_delete, "Expected to find at least one delete typo") + + def test_duplicate_strategy(self): + rng = random.Random() + word = "abcdef" + found_dup = False + for seed in range(200): + rng.seed(seed) + typo = _apply_typo(word, rng) + if len(typo) == len(word) + 1: + found_dup = True + break + self.assertTrue(found_dup, "Expected to find at least one duplicate typo") + + def test_short_word_unchanged(self): + rng = random.Random(42) + # 2-char word has 0 interior chars, cannot be mutated + self.assertEqual(_apply_typo("ab", rng), "ab") + + def test_three_char_word(self): + # 3-char word has exactly 1 interior char; swap requires >=2 interior, so only delete/dup + rng = random.Random() + word = "abc" + for seed in range(50): + rng.seed(seed) + typo = _apply_typo(word, rng) + self.assertEqual(typo[0], "a") + self.assertEqual(typo[-1], "c") + + +class TestIntroduceTextErrors(unittest.TestCase): + """Tests for the introduce_text_errors augmentation function.""" + + SAMPLE_HTML = """Test +
This paragraph contains several important words that should be eligible for modification.
+
Another sentence with different vocabulary and longer words throughout the document.
+ """ + + def test_returns_requested_number_of_typos(self): + rng = random.Random(42) + _, records = introduce_text_errors(self.SAMPLE_HTML, rng, num_errors=3) + self.assertEqual(len(records), 3) + + def test_typos_appear_in_modified_html(self): + rng = random.Random(42) + modified_html, records = introduce_text_errors(self.SAMPLE_HTML, rng, num_errors=3) + for record in records: + self.assertIn(record["typo_word"], modified_html) + + def test_originals_replaced_in_modified_html(self): + rng = random.Random(42) + modified_html, records = introduce_text_errors(self.SAMPLE_HTML, rng, num_errors=3) + for record in records: + # The original word at the typo position should no longer match + self.assertNotEqual(record["original_word"], record["typo_word"]) + + def test_record_fields(self): + rng = random.Random(42) + _, records = introduce_text_errors(self.SAMPLE_HTML, rng, num_errors=1) + self.assertEqual(len(records), 1) + self.assertIn("original_word", records[0]) + self.assertIn("typo_word", records[0]) + + def test_no_body_returns_unchanged(self): + html = "" + rng = random.Random(42) + modified, records = introduce_text_errors(html, rng, num_errors=3) + self.assertEqual(records, []) + + def test_num_errors_capped_by_candidates(self): + # Only one eligible word + html = "
Hello tiny text ok.
" + rng = random.Random(42) + _, records = introduce_text_errors(html, rng, num_errors=10) + # "Hello" and "tiny" and "text" are <5 chars; only "Hello" is 5 chars + self.assertLessEqual(len(records), 10) + + def test_zero_errors_returns_unchanged(self): + rng = random.Random(42) + modified, records = introduce_text_errors(self.SAMPLE_HTML, rng, num_errors=0) + self.assertEqual(records, []) + + def test_deterministic_with_same_seed(self): + rng1 = random.Random(99) + modified1, records1 = introduce_text_errors(self.SAMPLE_HTML, rng1, num_errors=3) + + rng2 = random.Random(99) + modified2, records2 = introduce_text_errors(self.SAMPLE_HTML, rng2, num_errors=3) + + self.assertEqual(records1, records2) + self.assertEqual(modified1, modified2) + + +class TestSkipAncestors(unittest.TestCase): + """Tests that typos skip headers, footers, tables, and other excluded elements.""" + + def _get_typo_words(self, html, num_errors=10): + rng = random.Random(42) + _, records = introduce_text_errors(html, rng, num_errors=num_errors) + return {r["original_word"] for r in records} + + def test_skips_table_content(self): + html = """ +
Eligible looking words inside table
+
These words should definitely be eligible outside.
+ """ + words = self._get_typo_words(html) + self.assertNotIn("Eligible", words) + self.assertNotIn("inside", words) + self.assertNotIn("table", words) + self.assertNotIn("looking", words) + + def test_skips_header_tags(self): + html = """ +
Important Header Title
+
Another Subtitle Value
+
These words should definitely be eligible outside.
+ """ + words = self._get_typo_words(html) + self.assertNotIn("Important", words) + self.assertNotIn("Header", words) + self.assertNotIn("Title", words) + self.assertNotIn("Another", words) + self.assertNotIn("Subtitle", words) + self.assertNotIn("Value", words) + + def test_skips_footer_element(self): + html = """ +
Paragraph words should definitely qualify here.
+
Footer content should never appear inside.
+ """ + words = self._get_typo_words(html) + self.assertNotIn("Footer", words) + self.assertNotIn("content", words) + self.assertNotIn("never", words) + self.assertNotIn("appear", words) + self.assertNotIn("inside", words) + + def test_skips_header_element(self): + html = """ +
Header content should never appear.
+
Paragraph words should definitely appear here.
+ """ + words = self._get_typo_words(html) + self.assertNotIn("Header", words) + + def test_skips_script_and_style(self): + html = """ + + +
Paragraph words should definitely appear here.
+ """ + words = self._get_typo_words(html) + self.assertNotIn("doSomething", words) + self.assertNotIn("function", words) + self.assertNotIn("something", words) + + def test_skips_sup_and_sub(self): + html = """ +
Normal words^superscript and _subscript content.
+ """ + words = self._get_typo_words(html) + self.assertNotIn("superscript", words) + self.assertNotIn("subscript", words) + + def test_skips_code_and_pre(self): + html = """ + inlineCode +
preformatted block content
+
Paragraph words should definitely appear here.
+ """ + words = self._get_typo_words(html) + self.assertNotIn("inlineCode", words) + self.assertNotIn("preformatted", words) + self.assertNotIn("block", words) + + def test_skips_page_header_class(self): + html = """ +
Header content skipped
+
Paragraph words should definitely appear here.
+ """ + words = self._get_typo_words(html) + self.assertNotIn("Header", words) + self.assertNotIn("skipped", words) + + def test_skips_page_footer_class(self): + html = """ +
Footer content skipped
+
Paragraph words should definitely appear here.
+ """ + words = self._get_typo_words(html) + self.assertNotIn("Footer", words) + self.assertNotIn("skipped", words) + + def test_skips_page_number_class(self): + html = """ + PageNumber +
Paragraph words should definitely appear here.
+ """ + words = self._get_typo_words(html) + self.assertNotIn("PageNumber", words) + + def test_eligible_paragraph_text_is_found(self): + html = """ +
These words should definitely appear as eligible candidates.
+ """ + words = self._get_typo_words(html) + # At least some 5+ char words from the paragraph should be candidates + self.assertTrue(len(words) > 0) + + +class TestWordFiltering(unittest.TestCase): + """Tests that only appropriate words are selected as candidates.""" + + def test_short_words_skipped(self): + html = """ +
The cat sat on a mat but extraordinary.
+ """ + rng = random.Random(42) + _, records = introduce_text_errors(html, rng, num_errors=10) + originals = {r["original_word"] for r in records} + # Words < 5 chars should not be typo'd + for word in ["The", "cat", "sat", "on", "a", "mat", "but"]: + self.assertNotIn(word, originals) + + def test_non_ascii_words_skipped(self): + html = """ +
Héllo wörld but eligible words remain here.
+ """ + rng = random.Random(42) + _, records = introduce_text_errors(html, rng, num_errors=10) + originals = {r["original_word"] for r in records} + # Non-ascii words like Héllo, wörld won't match [A-Za-z]+ fully + self.assertNotIn("Héllo", originals) + self.assertNotIn("wörld", originals) + + def test_only_alpha_words(self): + html = """ +
word123 and numbers456 but eligible words remain here.
+ """ + rng = random.Random(42) + _, records = introduce_text_errors(html, rng, num_errors=10) + originals = {r["original_word"] for r in records} + # Regex [A-Za-z]+ splits on digits, so "word" (4 chars) and "numbers" are separate + self.assertNotIn("word123", originals) + self.assertNotIn("numbers456", originals) + + +class TestTypoTestValidation(unittest.TestCase): + """Test that typo records can be validated with TextPresenceTest logic.""" + + def test_typo_present_in_augmented_absent_in_original(self): + from olmocr.bench.tests import TestType, TextPresenceTest + + html = """ +
This paragraph contains several important words that should be eligible for modification throughout.
+
Another sentence with different vocabulary and longer words throughout the document entirely.
+ """ + + rng = random.Random(42) + original_html = html + modified_html, records = introduce_text_errors(html, rng, num_errors=3) + + self.assertGreater(len(records), 0) + + # Use the raw HTML text as a stand-in for markdown (avoids importing mine_html_templates) + original_text = original_html + augmented_text = modified_html + + validated = 0 + for record in records: + typo_word = record["typo_word"] + test_obj = TextPresenceTest( + pdf="test.pdf", + page=1, + id="test", + type=TestType.PRESENT.value, + text=typo_word, + max_diffs=0, + ) + passed_original, _ = test_obj.run(original_text) + passed_augmented, _ = test_obj.run(augmented_text) + + if not passed_original and passed_augmented: + validated += 1 + + # At least some typos should pass validation + self.assertGreater(validated, 0) + + +class TestTypoDeduplication(unittest.TestCase): + """Test that existing presence tests for typo words get deduplicated.""" + + def test_existing_presence_tests_for_typo_words_removed(self): + from olmocr.bench.tests import TestType + + typo_records = [ + {"original_word": "important", "typo_word": "improtant"}, + {"original_word": "eligible", "typo_word": "eligble"}, + ] + typo_words = {r["typo_word"] for r in typo_records} + + # Simulate tests list that generate_tests_from_html might produce, + # including a presence test that matches a typo word + tests = [ + {"type": TestType.PRESENT.value, "text": "improtant", "id": "dup1"}, + {"type": TestType.PRESENT.value, "text": "something_else", "id": "keep1"}, + {"type": TestType.TABLE.value, "text": "improtant", "id": "keep2"}, + {"type": TestType.PRESENT.value, "text": "eligble", "id": "dup2"}, + ] + + # Apply the same dedup logic used in mine_html_templates + filtered = [t for t in tests if not (t.get("type") == TestType.PRESENT.value and t.get("text") in typo_words)] + + self.assertEqual(len(filtered), 2) + ids = {t["id"] for t in filtered} + self.assertIn("keep1", ids) + self.assertIn("keep2", ids) + self.assertNotIn("dup1", ids) + self.assertNotIn("dup2", ids) + + def test_non_present_tests_unaffected(self): + from olmocr.bench.tests import TestType + + typo_words = {"improtant"} + tests = [ + {"type": TestType.TABLE.value, "text": "improtant", "id": "t1"}, + {"type": TestType.ORDER.value, "text": "improtant", "id": "t2"}, + ] + + filtered = [t for t in tests if not (t.get("type") == TestType.PRESENT.value and t.get("text") in typo_words)] + self.assertEqual(len(filtered), 2) + + +if __name__ == "__main__": + unittest.main() diff --git a/tests/test_mine_html_templates_markdown.py b/tests/test_mine_html_templates_markdown.py new file mode 100644 index 0000000..eb44d13 --- /dev/null +++ b/tests/test_mine_html_templates_markdown.py @@ -0,0 +1,669 @@ +import random +import unittest +from unittest.mock import MagicMock, patch + +from bs4 import BeautifulSoup + +from olmocr.bench.tests import TestType +from olmocr.synth.mine_html_templates import ( + PreserveTablesConverter, + extract_html_metadata, + generate_tests_from_html, + html_to_markdown_with_frontmatter, +) + + +class TestHtmlToMarkdown(unittest.TestCase): + def test_title_tag_excluded_from_markdown(self): + """Test that title tags from head are not included in markdown output.""" + html_content = """ + + + This Should Not Appear In Markdown + + + +
Main Heading
+
This is the body content that should appear.
+ + + """ + + markdown_with_frontmatter = html_to_markdown_with_frontmatter(html_content) + + # Check that the title from head tag is NOT in the markdown + self.assertNotIn("This Should Not Appear In Markdown", markdown_with_frontmatter) + + # Check that body content IS in the markdown + self.assertIn("Main Heading", markdown_with_frontmatter) + self.assertIn("This is the body content that should appear", markdown_with_frontmatter) + + # Check that frontmatter is present + self.assertTrue(markdown_with_frontmatter.startswith("---")) + + def test_image_with_data_description(self): + """Test that images are converted with placeholder alt text.""" + html_content = """ + + +
Text before image
+
Placeholder
+
Text after image
+ + + """ + + markdown_with_frontmatter = html_to_markdown_with_frontmatter(html_content) + + # Check that images use the fixed placeholder alt text + self.assertIn("![Image Placeholder]", markdown_with_frontmatter) + + # Check that other content is preserved + self.assertIn("Text before image", markdown_with_frontmatter) + self.assertIn("Text after image", markdown_with_frontmatter) + + def test_image_without_data_description(self): + """Test that images without data-description use default alt text.""" + html_content = """ + + +
Some placeholder content
+ + + """ + + markdown_with_frontmatter = html_to_markdown_with_frontmatter(html_content) + + # Check that default alt text is used + self.assertIn("![Image Placeholder]", markdown_with_frontmatter) + + def test_headers_footers_excluded(self): + """Test that header and footer tags are excluded from markdown.""" + html_content = """ + + +
+
Navigation menu that should not appear
+
+
+
Main Content
+
This should appear in the markdown.
+
+
+
Footer text that should not appear
+
+ + + """ + + markdown_with_frontmatter = html_to_markdown_with_frontmatter(html_content) + + # Check that header/footer content is excluded + self.assertNotIn("Navigation menu", markdown_with_frontmatter) + self.assertNotIn("Footer text", markdown_with_frontmatter) + + # Check that main content is included + self.assertIn("Main Content", markdown_with_frontmatter) + self.assertIn("This should appear in the markdown", markdown_with_frontmatter) + + def test_no_body_tag_fallback(self): + """Test that content is still processed when there's no body tag.""" + html_content = """ +
+
Content without body tag
+
This should still be converted.
+
+ """ + + markdown_with_frontmatter = html_to_markdown_with_frontmatter(html_content) + + # Check that content is still converted + self.assertIn("Content without body tag", markdown_with_frontmatter) + self.assertIn("This should still be converted", markdown_with_frontmatter) + + def test_removes_triple_dashes_from_content(self): + """Test that --- at the start or end of markdown content is removed.""" + # Test with --- at the beginning + html_content_start = """ + + +
---
+
Regular content here
+ + + """ + + markdown_start = html_to_markdown_with_frontmatter(html_content_start) + lines = markdown_start.split("\n") + + # Check that we have FrontMatter + self.assertEqual(lines[0], "---") + # Check that the content doesn't start with --- after the FrontMatter ends + frontmatter_end = next(i for i in range(1, len(lines)) if lines[i] == "---") + content_after_frontmatter = "\n".join(lines[frontmatter_end + 1 :]) + self.assertFalse(content_after_frontmatter.strip().startswith("---")) + + # Test with --- at the end + html_content_end = """ + + +
Regular content here
+
---
+ + + """ + + markdown_end = html_to_markdown_with_frontmatter(html_content_end) + # Check that content doesn't end with --- + self.assertFalse(markdown_end.rstrip().endswith("---\n---")) + + # Test with --- at both beginning and end + html_content_both = """ + + +
---
+
Middle content
+
---
+ + + """ + + markdown_both = html_to_markdown_with_frontmatter(html_content_both) + lines_both = markdown_both.split("\n") + frontmatter_end_both = next(i for i in range(1, len(lines_both)) if lines_both[i] == "---") + content_both = "\n".join(lines_both[frontmatter_end_both + 1 :]) + + # Content should not start or end with --- + self.assertFalse(content_both.strip().startswith("---")) + self.assertFalse(content_both.strip().endswith("---")) + # But should contain "Middle content" + self.assertIn("Middle content", content_both) + + +class TestSuperscriptSubscriptConversion(unittest.TestCase): + """Test superscript and subscript preservation as HTML tags in html_to_markdown_with_frontmatter""" + + def test_basic_superscripts(self): + """Test basic superscript preservation""" + html = """ + + +
x² + y³ = z⁴
+
10⁹ is a billion
+ + + """ + result = html_to_markdown_with_frontmatter(html) + + # Check that superscripts are preserved as HTML tags + self.assertIn("x²", result) + self.assertIn("y³", result) + self.assertIn("z⁴", result) + self.assertIn("10⁹", result) + + def test_basic_subscripts(self): + """Test basic subscript preservation""" + html = """ + + +
H₂O is water
+
CO₂ is carbon dioxide
+
X_n represents the nth element
+ + + """ + result = html_to_markdown_with_frontmatter(html) + + # Check that subscripts are preserved as HTML tags + self.assertIn("H₂O", result) + self.assertIn("CO₂", result) + self.assertIn("X_n", result) + + def test_mixed_super_and_subscripts(self): + """Test mixed superscripts and subscripts""" + html = """ + + +
The formula is x² + H₂O⁺
+
Chemical: Ca²⁺ and SO₄^2-
+ + + """ + result = html_to_markdown_with_frontmatter(html) + + # Check mixed preserved tags + self.assertIn("x²", result) + self.assertIn("H₂O⁺", result) + self.assertIn("Ca²⁺", result) + self.assertIn("SO₄^2-", result) + + def test_special_characters(self): + """Test special character preservation""" + html = """ + + +
Math: (x+y)ⁿ and f_(x)
+
Ion: OH^- and H⁺
+
Index: a_i and bⁱ
+ + + """ + result = html_to_markdown_with_frontmatter(html) + + # Check special characters are preserved + self.assertIn("(x+y)ⁿ", result) + self.assertIn("f_(x)", result) + self.assertIn("OH^-", result) + self.assertIn("H⁺", result) + self.assertIn("a_i", result) + self.assertIn("bⁱ", result) + + def test_in_table(self): + """Test superscripts/subscripts within HTML tables""" + html = """ + + + + + + + + + + + + + + + +
Chemical Formula
Water H₂O
Sulfate ion SO₄^2-
+ + + """ + result = html_to_markdown_with_frontmatter(html) + + # Tables should be preserved as HTML with sub/sup tags + self.assertIn("", result) + + # Check if tags are preserved in table cells + self.assertIn("₂", result) + self.assertIn("₄^2-", result) + + def test_nested_elements(self): + """Test superscripts/subscripts in nested HTML elements""" + html = """ + + +
+
In physics: E = mc²
+
+
First: x¹
+
Second: x₂
+
+
+ + + """ + result = html_to_markdown_with_frontmatter(html) + + # Check tags are preserved in nested structures + self.assertIn("mc²", result) + self.assertIn("x¹", result) + self.assertIn("x₂", result) + + def test_frontmatter_preserved(self): + """Test that frontmatter is still generated correctly""" + html = """ + + +
Test with x²
+
Data
+ + + """ + result = html_to_markdown_with_frontmatter(html) + + # Check frontmatter exists + self.assertTrue(result.startswith("---")) + self.assertIn("primary_language: es", result) + self.assertIn("is_table:", result) + + # Also check the tags are preserved + self.assertIn("x²", result) + + def test_unmapped_characters(self): + """Test characters not in the mapping""" + html = """ + + +
Unknown: x^abc and y_xyz
+
Mixed: H₂SO₄ with note^*
+ + + """ + result = html_to_markdown_with_frontmatter(html) + + # All characters should be preserved as HTML tags + self.assertIn("H₂SO₄", result) + # Asterisk should be preserved in sup tag + self.assertIn("note^*", result) + + def test_empty_super_subscripts(self): + """Test empty sup/sub tags""" + html = """ + + +
Empty tags: x and y
+
Normal: z²
+ + + """ + result = html_to_markdown_with_frontmatter(html) + + # Empty tags should be preserved (even if empty) + self.assertIn("z²", result) + # Empty tags might be preserved or removed - just check text is there + self.assertIn("x", result) + self.assertIn("y", result) + + def test_complex_math_expression(self): + """Test a complex mathematical expression""" + html = """ + + +
The equation: (x₁)² + (x₂)² = r²
+
Series: a₀ + a₁x + a₂x² + ... + a_nxⁿ
+ + + """ + result = html_to_markdown_with_frontmatter(html) + + # Check complex nested expressions are preserved as tags + self.assertIn("x₁", result) + self.assertIn("x₂", result) + self.assertIn("r²", result) + self.assertIn("a₀", result) + self.assertIn("a₁", result) + self.assertIn("a₂", result) + self.assertIn("a_n", result) + self.assertIn("xⁿ", result) + + def test_non_numerical_footnote(self): + """Test a complex mathematical expression""" + html = """ + + +
This is a sentence^wow
+

+
^wow This is a footnote
+ + + """ + result = html_to_markdown_with_frontmatter(html) + + self.assertIn("^wow", result) + + def test_line_numbers_markdown(self): + html = """ + + + + + + +bioRxiv preprint + + +
+ bioRxiv preprint doi: https://doi.org/10.1101/2020.10.16.342121; this version posted October 19, 2020. The copyright holder for this preprint
+ (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. +
+
+
+
67
+
to IIA site (Atukeren, Aydin, Uslu, Gumustas, & Cakatay, 2010), however, IIIA site was also
+
+
+
68
+
considered (Suji et al., 2008).
+
+
+
69
+
Having in mind that DHLA is a very potent antioxidant and its use can alleviate a number of
+
+
+
70
+
conditions related to oxidative stress, it seemed relevant to elucidate its mode of interaction with
+
+
+
71
+
HSA, a universal transporter in the circulation. The properties of this interaction, are still
+
+
+
72
+
unknown and undefined, so the present study aimed to investigate characteristics of the DHLA-
+
+
+
73
+
HSA binding in detail, by using spectroscopic and molecular docking approach.
+
+
+
74
+
+
+
MATERIALS AND METHODS
+
+
75
+
Materials
+
+
+
76
+
All chemicals used were of analytical grade and were purchased from Sigma (Burlington,
+
+
+
77
+
Massachusetts, USA). Stock solution of HSA, purchased from Sigma (A-1653) and used without
+
+
+
78
+
additional purification, was made by dissolving HSA in 10 mM PBS, pH 7.4. The concentration
+
+
+
79
+
of HSA was determined by using bicinchoninic acid (BCA) assay kit (Thermo Fisher Scientific,
+
+
+
80
+
Waltham, Massachusetts, USA). Stock solution (5 mM) of DHLA was prepared by suspending
+
+
+
81
+
DHLA in 10 mM PBS and then adding a small volume of 1 M NaOH until full clarification of
+
+
+
82
+
solution was reached (Perricone et al., 1999). Trypsin was purchased from the Institute Torlak
+
+
+
83
+
(Belgrade, Serbia) as a 0.25 % solution. All experiments were performed in triplicate at room
+
+
+
84
+
temperature, using 10 mM PBS, pH 7.4, unless otherwise stated.
+
+
+
85
+
Spectrofluorometric analysis of HSA-DHLA complex formation
+
+
+
86
+
Binding constant (Ka) of HSA-DHLA complex was determined by recording the quenching of
+
+
+
87
+
intrinsic fluorescence emission of HSA (0.4 μM) in the presence of increasing concentrations of
+
+
+
88
+
DHLA (from 4 to 35 μM) at 37 °C. Fluorescence spectra were recorded using FluoroMax®-4
+
+
+
89
+
spectrofluorometer (Horiba Scientific, Japan). HSA was excited at 280 nm and emission spectra
+
+
+
90
+
were recorded in the range from 290 to 450 nm. Each spectrum was corrected for the emission of
+
+
+
91
+
the control that contained only DHLA at particular concentration. The change of the emission
+
+
+
92
+
intensity at 338 nm (HSA emission maximum) was used for the calculation of the binding
+
+
+
93
+
constant. Emission intensity measured for HSA was first corrected for the small inner filter effect
+
+
+
94
+
of DHLA using the equation:
+
+
+
95
+
+
F_c = F₀ × 10^(Aex+Aem)/2
+
+
+
+
96
+
where Fc is corrected fluorescence, F₀ is measured fluorescence, Aex and Aem are absorbances
+
+
+
97
+
at excitation and emission wavelengths which are 290 nm and 338 nm, respectively.
+
+
+
98
+
Using corrected fluorescence, binding constant between HSA and DHLA was calculated using
+
+
+
99
+
the following equation:
+
+
+
+ 3 +
+ +""" + + result = html_to_markdown_with_frontmatter(html) + + self.assertNotIn("96", result) + self.assertNotIn("97", result) + self.assertNotIn("98", result) + self.assertNotIn("82", result) + + def test_svg_in_markdown(self): + html = """ + + + + + +Survey Results - Q6 + + +
+ Q6. What, if anything, do you consider are the main barriers to the trading performance of your business? +
+
+
+
38   (100.00%)
+
+
+
Other:
+
+ Restrictions relating to listing of building and any alterations/modifications to make it more usable [1]; Lack of decent public transport - it is impossible for staff out of the area to reach Rothbury for 9am [1] +
+
+
+ Not Answered:     3 +
+
+ +
+ +""" + + result = html_to_markdown_with_frontmatter(html) + + self.assertNotIn("a\nb\nc", result) + self.assertIn("Graphic Placeholder", result) + + +if __name__ == "__main__": + unittest.main() diff --git a/tests/test_mine_html_templates_math_extraction.py b/tests/test_mine_html_templates_math_extraction.py new file mode 100644 index 0000000..5da894c --- /dev/null +++ b/tests/test_mine_html_templates_math_extraction.py @@ -0,0 +1,227 @@ +import random +import unittest +from unittest.mock import MagicMock, patch + +from bs4 import BeautifulSoup + +from olmocr.bench.tests import TestType +from olmocr.synth.mine_html_templates import ( + PreserveTablesConverter, + extract_html_metadata, + generate_tests_from_html, + html_to_markdown_with_frontmatter, +) + + +class TestMathExtraction(unittest.TestCase): + """Test the math extraction functionality in mine_html_templates.py""" + + def setUp(self): + self.random_generator = random.Random(42) + return super().setUp() + + def test_math_extraction_from_html(self): + """Test that math equations are properly extracted from HTML content""" + html_content = """ + + +
Some text with inline math \$x = 2\$ here.
+
Display math: \\[E = mc^2\\]
+
Another inline: \$\\alpha + \\beta = \\gamma\$
+
Complex display: \\[\\int_0^\\infty e^{-x} dx = 1\\]
+ + + """ + + # Generate tests from HTML + tests = generate_tests_from_html(html_content, "test_pdf", 1, self.random_generator) + + # Filter math tests + math_tests = [t for t in tests if t.get("type") == "math"] + + # Check that we extracted math equations + self.assertTrue(len(math_tests) > 0, "Should extract at least one math equation") + + # Check that specific equations were extracted + math_contents = [t["math"] for t in math_tests] + self.assertIn("x = 2", math_contents) + self.assertIn("E = mc^2", math_contents) + self.assertIn("\\alpha + \\beta = \\gamma", math_contents) + self.assertIn("\\int_0^\\infty e^{-x} dx = 1", math_contents) + + def test_math_extraction_with_multiline(self): + """Test extraction of multiline math equations""" + html_content = """ + + +
Multiline equation: + \\[ + e_i = \\frac{e_i + \\varphi(e_i)}{2} + \\frac{e_i - \\varphi(e_i)}{2}, + \\quad \\text{for } i \\in \\mathbb{N}. + \\] +
+ + + """ + + tests = generate_tests_from_html(html_content, "test_pdf", 1, self.random_generator) + math_tests = [t for t in tests if t.get("type") == "math"] + + # Check multiline equation is captured + self.assertTrue(len(math_tests) > 0) + + # Check that the multiline content is preserved (without excessive newlines) + found_multiline = False + for test in math_tests: + if "\\frac{e_i + \\varphi(e_i)}{2}" in test["math"] and "\\mathbb{N}" in test["math"]: + found_multiline = True + break + + self.assertTrue(found_multiline, "Should extract multiline equation correctly") + + def test_math_extraction_deduplication(self): + """Test that duplicate math equations are deduplicated""" + html_content = """ + + +
First occurrence: \\[x^2 + y^2 = z^2\\]
+
Second occurrence: \\[x^2 + y^2 = z^2\\]
+
Third occurrence: \\[x^2 + y^2 = z^2\\]
+ + + """ + + tests = generate_tests_from_html(html_content, "test_pdf", 1, self.random_generator) + math_tests = [t for t in tests if t.get("type") == "math"] + + # Count how many times the equation appears + equation_count = sum(1 for t in math_tests if "x^2 + y^2 = z^2" in t["math"]) + + # Should only appear once due to deduplication + self.assertEqual(equation_count, 1, "Duplicate equations should be deduplicated") + + def test_math_extraction_patterns(self): + """Test different math delimiter patterns""" + html_content = """ + + +
Pattern 1: \$inline1\$
+
Pattern 2: \\[display1\\]
+
Pattern 3: $$display2$$
+ + + """ + + tests = generate_tests_from_html(html_content, "test_pdf", 1, self.random_generator) + math_tests = [t for t in tests if t.get("type") == "math"] + + math_contents = [t["math"] for t in math_tests] + + # Check all patterns are captured + self.assertIn("inline1", math_contents) + self.assertIn("display1", math_contents) + self.assertIn("display2", math_contents) + + def test_math_extraction_minimum_length(self): + """Test that very short equations are filtered out""" + html_content = """ + + +
Short: \$x\$
+
Also short: \\[y\\]
+
Long enough: \$x=1\$
+ + + """ + + tests = generate_tests_from_html(html_content, "test_pdf", 1, self.random_generator) + math_tests = [t for t in tests if t.get("type") == "math"] + + math_contents = [t["math"] for t in math_tests] + + # Short equations (length <= 2) should be filtered out + self.assertNotIn("x", math_contents) + self.assertNotIn("y", math_contents) + # Longer equation should be included + self.assertIn("x=1", math_contents) + + def test_math_validation_passes(self): + """Test that valid math tests pass validation against markdown""" + html_content = """ + + +
Test equation: \\[E = mc^2\\]
+ + + """ + + # Mock the validation to always pass for math tests + with patch("olmocr.bench.tests.load_single_test") as mock_load: + mock_test = MagicMock() + mock_test.run.return_value = (True, None) + mock_load.return_value = mock_test + + tests = generate_tests_from_html(html_content, "test_pdf", 1, self.random_generator) + math_tests = [t for t in tests if t.get("type") == "math"] + + # Verify math test was created + self.assertTrue(len(math_tests) > 0) + # Verify test has correct structure + for test in math_tests: + self.assertEqual(test["type"], "math") + self.assertIn("math", test) + self.assertEqual(test["max_diffs"], 0) + self.assertIn("id", test) + self.assertIn("pdf", test) + self.assertEqual(test["page"], 1) + + def test_complex_markdown_example(self): + """Test with the complex markdown example provided by the user""" + # Convert markdown to HTML-like structure for testing + html_content = '\n\n\n \n \n Automorphisms of Order Two\n \n \n \n \n\n\n $\n If \$\\varphi \\in \\text{Aut}(E)\$ with \$\\varphi^2 = id\$ we observe that\n \\[e_i = \\frac{e_i + \\varphi(e_i)}{2} + \\frac{e_i - \\varphi(e_i)}{2}, \\quad \\text{for } i \\in \\mathbb{N}.\\]\n \n Setting \$a_i = e_i + \\varphi(e_i)/2\$ we have:\n \n \n \$\\varphi(e_i) = -e_i + 2a_i\$,\n \$\\varphi(a_i) = a_i\$, that is, \$a_i\$ is of degree zero in the \$\\mathbb{Z}_2\$-grading \$E_\\varphi\$,\n \$\\varphi(e_i - a_i) = -(e_i - a_i)\$, that is, \$e_i - a_i\$ is of degree 1 in the \$\\mathbb{Z}_2\$-grading \$E_\\varphi\$.\n \n$ \n \n
\n
Definition 5
\n
Let \$\\varphi \\in \\text{Aut}(E)\$. We say that \$\\varphi\$ is of canonical type if \$\\varphi(e_i) \\in E_{(1)}\$ for all \$i\$.
\n \n
If \$\\varphi\$ is an automorphism of order 2 on \$E\$, we have that \$\\varphi\$ is of canonical type if and only if \$a_i \\in E_{(1)}\$ for all \$i\$. Let us fix a basis \$\\beta = \\{e_1, e_2, \\ldots, e_n, \\ldots\\}\$ of the vector space \$L\$ and an automorphism \$\\varphi \\in \\text{Aut}(E)\$ such that \$\\varphi^2 = id\$. Then \$\\varphi\$, as a linear transformation, has eigenvalues \$\\pm 1\$ and \$-1\$ only, and moreover, there exists a basis of the vector space \$E\$ consisting of eigenvectors. (It is well known from elementary Linear Algebra that this fact does not depend on the dimension of the vector space as long as the characteristic of \$F\$ is different from 2.) Then \$E = E(1) \\oplus E(-1)\$ where \$E(t)\$ is the eigenspace for the eigenvalue \$t\$ of the linear transformation \$\\varphi\$. One considers the intersections \$L(t) = L \\cap E(t)\$, \$t = \\pm 1\$. Changing the basis \$\\beta\$, if necessary, one may assume that \$L(t)\$ is the span of \$\\beta \\cap L(t)\$. Clearly this change of basis gives rise to a homogeneous automorphism of \$E\$ and we can take the composition of it and then \$\\varphi\$. We shall assume that such a change of basis has been done.
\n \n
Denote
\n \\[I_\\varphi = \\{n \\in \\mathbb{N} \\mid \\varphi(e_n) = \\pm e_n\\}.\\]\n
\n \n
We shall distinguish the following four possibilities:
\n \n
\n
\$I_\\varphi = \\mathbb{N}\$.
\n
\$I_\\varphi \\neq \\mathbb{N}\$ is infinite.
\n
\$I_\\varphi\$ is finite and nonempty.
\n
\$I_\\gamma = \\emptyset\$ for every linear basis \$\\gamma\$ of \$L\$.
\n
\n \n
We shall call these automorphisms (and also the corresponding \$\\mathbb{Z}_2\$-gradings), automorphisms (or gradings) of type 1, 2, 3, and 4, respectively.
\n \n
The automorphisms of type 1 induce \$\\mathbb{Z}_2\$-gradings on \$E\$ in which all generators of \$E\$ are homogeneous. Such structures are called homogeneous \$\\mathbb{Z}_2\$-gradings on \$E\$. The corresponding graded identities were completely studied in [22, 24, 29].
\n \n
We conclude this section with the following lemma.
\n \n
\n
Lemma 6
\n
Let \$\\varphi\$ be an automorphism of order two of \$E\$. Then \$\\varphi\$ is of type 4 if and only if, for every \$v \\in L\$ such that \$\\varphi(v) = \\pm v\$, one has \$v = 0\$.
\n \n
\n Proof Assume that \$\\varphi\$ is of type 4 and let \$v \\in L\$ with \$\\varphi(v) = \\pm v\$. If \$v \\neq 0\$, choose a basis \$\\gamma\$ of \$L\$ such that \$v \\in \\gamma\$. Then \$I_\\gamma \\neq \\emptyset\$, a contradiction. The converse follows by the same argument.\n ■\n
\n
\n \n
3    Automorphisms of order two of E
\n \n
From this point on, our goal is to survey recent developments regarding automorphisms of order two and the corresponding \$\\mathbb{Z}_2\$-gradings of the infinite-dimensional Grassmann algebra.
\n \n
Let \$X = \\{e_1, \\ldots, e_n, \\ldots\\}\$. For each map \$\\lambda : X \\to E\$, we can define the linear transformation \$\\varphi : E \\to E\$ by
\n \n
\n \\[\\varphi(e_{i_1} \\cdots e_{i_n}) = \\lambda(e_{i_1}) \\cdots \\lambda(e_{i_n}),\\] (1)\n
\n \n
for all \$n \\in \\mathbb{N}\$.
\n \n
We start with the next lemma.
\n \n
\n
Lemma 7
\n
The linear transformation \$\\varphi\$ is an endomorphism of \$E\$ if and only if
\n \\[\\lambda(e_i)\\lambda(e_j) + \\lambda(e_j)\\lambda(e_i) = 0, \\quad \\text{for all } i, j.\\]\n
\n \n
\n 4\n
\n\n' + tests = generate_tests_from_html(html_content, "test_pdf", 1, self.random_generator) + math_tests = [t for t in tests if t.get("type") == "math"] + + for test in math_tests: + print(test) + + def test_math_extraction_strips_whitespace(self): + """Test that extracted math equations have whitespace properly stripped""" + html_content = """ + + +
\\[ + x = y + z + \\]
+ + + """ + + tests = generate_tests_from_html(html_content, "test_pdf", 1, self.random_generator) + math_tests = [t for t in tests if t.get("type") == "math"] + + self.assertTrue(len(math_tests) > 0) + # The equation should be stripped of leading/trailing whitespace + self.assertEqual(math_tests[0]["math"].strip(), math_tests[0]["math"]) + + def test_math_not_in_present_tests(self): + """Test that extracted math equations have whitespace properly stripped""" + html_content = """ + + +
\\[x = y + z \\mathcal{x} \\]
+ +
\\[y = x + z \\mathcal{x} \\]
+ +
\\[q = r + z \\mathcal{x} \\]
+ + + """ + + tests = generate_tests_from_html(html_content, "test_pdf", 1, self.random_generator) + present_tests = [t for t in tests if t.get("type") == "present"] + self.assertTrue(len(present_tests) == 0) + + +if __name__ == "__main__": + unittest.main() diff --git a/tests/test_mine_html_templates_metadata.py b/tests/test_mine_html_templates_metadata.py new file mode 100644 index 0000000..601fd97 --- /dev/null +++ b/tests/test_mine_html_templates_metadata.py @@ -0,0 +1,406 @@ +import random +import unittest +from unittest.mock import MagicMock, patch + +from bs4 import BeautifulSoup + +from olmocr.bench.tests import TestType +from olmocr.synth.mine_html_templates import ( + PreserveTablesConverter, + extract_html_metadata, + generate_tests_from_html, + html_to_markdown_with_frontmatter, +) + + +class TestExtractHtmlMetadata(unittest.TestCase): + def test_extract_metadata_portuguese_document(self): + """Test metadata extraction from a Portuguese document with mixed content.""" + html_content = """ + + Test Document + +
Header content here
+
Política de Metadados
+
Este é um documento de teste com texto em português.
+
Contém múltiplos parágrafos para simular conteúdo real.
+
Image placeholder 1
+
Mais texto após a imagem.
+
Footer content
+ + + """ + + metadata = extract_html_metadata(html_content) + + # Check language extraction + self.assertEqual(metadata["primary_language"], "pt") + + # Check rotation values (always fixed) + self.assertTrue(metadata["is_rotation_valid"]) + self.assertEqual(metadata["rotation_correction"], 0) + + # Check table/diagram detection + # With 1 image (500 chars) and small text content, image ratio > 50% + self.assertFalse(metadata["is_table"]) + self.assertTrue(metadata["is_diagram"]) # Image estimate dominates + + def test_extract_metadata_table_heavy_document(self): + """Test metadata extraction from a document that is mostly tables.""" + html_content = """ + + +
Small intro text
+ + + + + + + + +
Cell 1 Cell 2 Cell 3
Data A Data B Data C
More data More data More data
Even more data Even more data Even more data
Lots of data Lots of data Lots of data
Table content Table content Table content
Final row Final row Final row
+ + + """ + + metadata = extract_html_metadata(html_content) + + self.assertEqual(metadata["primary_language"], "en") + self.assertTrue(metadata["is_table"]) # Should be True as >50% is table + self.assertFalse(metadata["is_diagram"]) + + def test_extract_metadata_image_heavy_document(self): + """Test metadata extraction from a document that is mostly images.""" + html_content = """ + + +
Brief text
+
Image 1
+
Image 2
+
Image 3
+
Image 4
+
Image 5
+ + + """ + + metadata = extract_html_metadata(html_content) + + self.assertEqual(metadata["primary_language"], "es") + self.assertFalse(metadata["is_table"]) + self.assertTrue(metadata["is_diagram"]) # Should be True as >50% is images + + def test_extract_metadata_language_with_region(self): + """Test that language codes with regions (e.g., pt-BR) are shortened.""" + html_content = """ + + +
Texto em português brasileiro
+ + + """ + + metadata = extract_html_metadata(html_content) + + # Should convert pt-BR to pt + self.assertEqual(metadata["primary_language"], "pt") + + def test_extract_metadata_no_html_tag(self): + """Test extraction when there's no html tag (defaults to 'en').""" + html_content = """ + +
Content without html tag
+ + """ + + metadata = extract_html_metadata(html_content) + + self.assertEqual(metadata["primary_language"], "en") # Should default to 'en' + + def test_extract_metadata_mixed_content(self): + """Test a document with mixed content types.""" + html_content = """ + + + + + Política de Metadados para Livros e Capítulos de Livro UFPA + + + +
+
+
Biblioteca Central UFPA
+
LIVRO ABERTO portal do livro aberto da UFPA
+
SIBI/UFPA
+
+
+ +
+
Política de Metadados para Livros e Capítulos de Livro UFPA
+ +
+ Essa política de metadados possui o objetivo de garantir a consistência do trabalho executado no Portal do Livro Aberto. Dessa forma, foi desenvolvido com base no esquema de metadados do Dublin Core com adaptações para a realidade brasileira e local. +
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
METADADOS VALOR REPETITIVO CONDIÇÃO
dc.type Tipo de documento Não Obrigatório
dc.title Título e subtítulo (se houver) Não Obrigatório
dc.title.alternative Título alternativo Sim Opcional
dc.creator Autor Sim Opcional
dc.creator.Lattes URL do currículo Lattes do autor Sim Opcional
dc.creator.ORCID ORCID do autor Sim Opcional
dc.description.affiliation Afiliação do autor Sim Opcional
dc.contributor.organizer Organizador Sim Opcional
dc.contributor.organizerLattes URL do currículo Lattes do organizador Sim Opcional
dc.contributor.organizerORCID ORCID do organizador Sim Opcional
dc.description.affiliationOrganizer Afiliação do organizador Sim Opcional
dc.contributor.coordinator Coordenador Sim Opcional
dc.contributor.coordinatorLattes URL do currículo Lattes do coordenador Sim Opcional
dc.contributor.coordinatorORCID ORCID do coordenador Sim Opcional
dc.contributor.affiliationCoordinator Afiliação do coordenador Sim Opcional
dc.contributor.editor Editor Sim Opcional
dc.contributor.editorLattes URL do currículo Lattes do editor Sim Opcional
dc.contributor.editorORCID ORCID do editor Sim Opcional
dc.description.affiliationEditor Afiliação do editor Sim Opcional
+
+ +
+
3
+
+ + + """ + + metadata = extract_html_metadata(html_content) + + self.assertEqual(metadata["primary_language"], "pt") + self.assertTrue(metadata["is_table"]) + self.assertFalse(metadata["is_diagram"]) + + def test_extract_metadata_empty_body(self): + """Test extraction with empty or minimal content.""" + html_content = """ + + + + """ + + metadata = extract_html_metadata(html_content) + + self.assertEqual(metadata["primary_language"], "de") + self.assertFalse(metadata["is_table"]) + self.assertFalse(metadata["is_diagram"]) + self.assertTrue(metadata["is_rotation_valid"]) + self.assertEqual(metadata["rotation_correction"], 0) + + +if __name__ == "__main__": + unittest.main() diff --git a/tests/test_mine_html_templates_mine_tests.py b/tests/test_mine_html_templates_mine_tests.py new file mode 100644 index 0000000..ef90d54 --- /dev/null +++ b/tests/test_mine_html_templates_mine_tests.py @@ -0,0 +1,1135 @@ +import random +import unittest +from unittest.mock import MagicMock, patch + +from bs4 import BeautifulSoup + +from olmocr.bench.tests import TestType +from olmocr.synth.mine_html_templates import ( + PreserveTablesConverter, + extract_html_metadata, + generate_tests_from_html, + html_to_markdown_with_frontmatter, +) + + +class TestMineTests(unittest.TestCase): + def setUp(self): + self.random_generator = random.Random(42) + return super().setUp() + + def test_absent_nested(self): + html_content = """ + + + + + + New Paradigm for Nuclear Safety + + +
+
+
+ +
+
New Paradigm for Nuclear Safety
+ +
+
Thursday, April 25, 2013
+
Japan Nuclear Safety Institute
+
Shojiro Matsuura, Chairman
+
+
+ +
+
+
+
+
In Pursuit of Improved Nuclear Safety
+
Copyright © 2012 by Japan Nuclear Safety Institute. All Rights Reserved.
+
+
+
+
+ +""" + tests = generate_tests_from_html(html_content, "0", 1, self.random_generator) + + # Check for footer-specific absent tests (excluding common word absences) + footer_absent_tests = [test for test in tests if test["type"] == "absent" and "footer" in test.get("id", "")] + self.assertEqual(len(footer_absent_tests), 2) + + def test_text_basic(self): + html_content = """ + + + + + + + Bone Morphology Description + + +
+
The posterior end exhibits a curved process to articulate with the angular. Aside from the process, the rest of the posterior end has slight curvatures for articulation, but is mostly uniform. Ventral border of the bone is mostly straight, with slight curvature (FIG. 20).
+ +
Lateral- A spine runs from the anterior-most tip, reduces in height ~3/4 way down toward posterior, and terminates at the center of the posterior notch. A fossa is present on the dorsal side of the spine. The posterior end exhibits more relief than in medial view, with the medial side of the posterior process extending past the posterior notch.
+ +
Ontogeny- Anterior tip is sharply pointed in AR12 and AR1 with AR2 being rounded, though this could be due to breakage. Anterior dorsal margin is straight and flat in AR12; AR2 shows little curvature and AR1 shows the most curvature; curving outward dorsally. Dorsal incisure is anteroposteriorly oriented in AR12, in AR2 there is some ventral curvature, and in AR1 there is a posteroventral curvature. Both AR1 and AR3 are curved on the ventral margin while AR12 is mostly straight. Posterior end of AR1 exhibits four undulations, ventral process is not yet extended. A fossa is present dorsal to the ventral process, not seen on AR12 or AR2. In medial view the lateral ridge is visible posteriorly in AR1 and AR2l the ridge does not fully extend anteriorly. In lateral view of the posterior the ventral process is present on AR2, but not fully extended posteriorly. Tip of the anterodorsal process is sharply pointed in AR1 and AR2, rounded in AR12. A short ridge is present on the dorsal edge of the dorsal process of AR1. The short ridge on the posterodorsal process of AR2 is slightly more ventral than in AR1. On AR12 the ridge is long and positioned most ventral. The lateral ridge is closest to the ventral margin in AR1. In AR2 the ridge is positioned more dorsally and in AR12 the ridge terminates and the anterior tip. The section of bone ventral to the lateral ridge appears to thin with age. The posterior notch on AR12 is curved anteriorly and the medial side of the notch extends posteriorly
+
+ +
+
46
+
+ +""" + + tests = generate_tests_from_html(html_content, "0", 1, self.random_generator) + self.assertGreater(len(tests), 5) + + def test_big_headers(self): + html_content = """ + + + + + + + FORANE 427A Comparative Data + + + +
+
Comparative data
+
+ +
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
Parameters R-22 FORANE^® 427A
Evaporating temperature 2.7 °C 1.1 °C
Condensing temperature 40.3 °C 44.0 °C
Suction temperature 7.1 °C 9.2 °C
Suction pressure 5.4 bar 5.0 bar
Discharge temperature 69.5 °C 71.1 °C
Discharge pressure 15.5 bar 17.1 bar
Cooling power 431 KW 376 KW
Power consumption 122 kW 124 kW
Residual mineral oil - 11%
+ +
During this field test, very satisfactory running conditions were reached immediately. The temperature set points were easily achieved with similar energy consumption as compared to R-22 despite a high level of residual mineral oil in the circuit. The performance of the installation continues to satisfy the customer's requirements after more than one year of service.
+ +
FORANE^® 427A consequently fully satisfies the requirements of the European regulations while enabling existing equipment to continue to perform well without the need for any long and costly plant modifications.
+ +
The versatility of FORANE^® 427A is also appreciated as it can be used to retrofit low temperature refrigeration equipment as well as air-conditioning installations, resulting in only one retrofit refrigerant for all R-22 units.
+ +
Combining environmental friendliness, high performance and simplicity is today a reality with FORANE^® 427A !
+
+ +
+
The information contained in this document is based on trials carried out by our Research Centres and data selected from the literature, but shall in no event be held to constitute or imply any warranty, undertaking, express or implied commitment from our part. Our formal specifications define the limit of our commitment. No liability whatsoever can be accepted by Arkema with regard to the handling, processing or use of the product or products concerned which must in all cases be employed in accordance with all relevant laws and/or regulations in force in the country or countries concerned.
+ +
+
+
+
ARKEMA
+ 420 rue d'Estienne d'Orves
+ 92700 Colombes - France
+ www.arkema.com
+
+ +
+
www.forane.com / info.forane@arkema.com
+
+
+
+ +""" + + tests = generate_tests_from_html(html_content, "0", 1, self.random_generator) + + self.assertFalse(any(test for test in tests if test["type"] == "absent" and "Comparative data" in test["text"])) + + def test_page_num(self): + html_content = """ + + + + + + Academic Paper - Page 47 + + +
+
+ +
Figure 4.3: The COVID-19 pandemic resulted in meaningful increase in the support for other groups' protests among Panamanians.
+ +
+
4.2.2 Demographically-Informed Opinion Assignment
+ +
Our model does not endow opinions randomly; instead, we leverage data to assign activists in a more realistic fashion. We use Latinobarómetro survey data from 2020 and 2023, both of which contain the three measurements of support for protest. Then, we explored which demographic groups were more likely to be activists; these are young adults and individuals at either extreme of the financial spectrum. We use this insight to influence the assignment of opinions: our logistic equations make it so that individuals with these characteristics are more likely to be labeled as activists as the probabilistic endowment happens. The code ensures that the proportion of activists overall remains exactly as desired and that there are activists who do not belong to these identified groups
+
+ +
+
4.2.3 Identity Factored into Social Influence
+ +
The similarity formula for Panama is built as follows, taking in nine demographic factors stored as node attributes. These are gender, age, nationality, financial status, highest level of education, level of employment, geographical region, party affiliation, and ethnicity (respectively encoded as gend, age, nation, fin, edu, emp, region, paff, and ethni). Each one of these factors has an associated weight; in this model, all factors were weighted as 0.10, except for level of education and financial status which received 0.15. Our code establishes logical rules to compare the two individuals on each dimension and return a factor by which to multiply the weight. These factors can be absolute or relative, based on the demographic dimension in question. For example, the logical conditions for gender returns 1 if same or 0 if different, while age returns a float value between 0 and 1 according to how close in age the individuals are. Once the pairwise similarity
+
+
+ +
+
47
+
+ +""" + + tests = generate_tests_from_html(html_content, "0", 1, self.random_generator) + + self.assertEqual(len([test for test in tests if test["type"] == "absent" and "footer" in test.get("id", "")]), 1) + + def test_div_footer(self): + html_content = """ + + + + + + + Being Thai: A Narrow Identity in a Wide World + + + +
+
+ ⊕ +
+
+ +
+
Being Thai: A Narrow Identity in a Wide World
+
333
+
+ +
hard to create a political and cultural narrative that appeals to old ideas about being Thai at this moment of perceived vulnerability.
+ +
The Concept of "Thainess"
+ +
Thainess is a political notion that originally evolved to support an authoritarian government and was then re-shaped for periods of more democratic rule.13 Thailand has long oscillated between dictatorship and democracy and, in either case, a sense of the "Thai style" (baeb Thai) is commonly invoked. Under these conditions a military coup may be thought to "advance Thai-style democracy".14 This is obviously fraught with difficulties and requires government agencies, most notably the Ministry of Culture, to work hard on shaping national identity.15 Thailand's geographical and cultural diversity means that there are inevitable deviations. Some of these have been astutely handled, especially in the northeastern provinces where the Lao-speaking minority has been integrated as chao isan. Nowadays it is only at the margins that their "Isan-ness" remains a contested sub-category of Thainess.16 In earlier generations there were more explicit challenges to the suggestion of Isan as Thai.17 Similar defiance has emerged in both the northern provinces18 and in the Malay Muslim majority areas of the far south.19 At various times there have been suggestions, as reported by the anthropologist Nick Tapp, that "Thainess" was disintegrating.20 It is in response to these persistent challenges that Prayuth's military government has sought to create its own revised version of the national ideal.
+ +
For the military government the codification of Thailand's core values has created new opportunities to stamp its preferred identity on society. In a key speech soon after he took power in 2014, Prayuth identified disunity as a problem in Thai society that would, in his words, "urgently require inclusive cooperation from people of all levels, gender and age".21 His approach was to "define clear core values of Thai people so that we can build a strong nation". These values draw on cultural ideas that have existed for many decades and have enjoyed the favour of previous military rulers. The full list of these twelve values is:
+ +
+
Upholding the three main pillars of the country: the nation, the religion and the monarchy;
+
Showing honesty, sacrifice and patience, with a positive attitude for the interest of the public;
+
Practicing filial piety towards parents, guardians and teachers;
+
Seeking both direct and indirect knowledge and education;
+
+ +
⊕
+
⊕
+ +
+
15-03450 10a Thailand.indd 333
+
15/2/16 8:24 am
+
+ +""" + + tests = generate_tests_from_html(html_content, "0", 1, self.random_generator) + + self.assertEqual(len([test for test in tests if test["type"] == "absent" and "footer" in test.get("id", "")]), 2) + self.assertEqual(len([test for test in tests if test["type"] == "absent" and "header" in test.get("id", "")]), 2) + + def test_table(self): + html_content = """ + + + + + + Distribuição da população na estrutura socioocupacional - Brasil 2000 + + + +
+
+
Alexandre Gori Maia e Waldir José de Quadros ■ 417
+
+ +
Apêndice A - Distribuição da população na estrutura socioocupacional - Brasil 2000
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
Grupo Ocupacional Classe Ocupacional Superior Médio Baixo Interior Ínfimo Total
N (1.000s) % N (1.000s) % N (1.000s) % N (1.000s) % N (1.000s) % N (1.000s) %
Empregadores A-1 Empregadores (> 10) 608 67,3 185 20,4 86 9,6 16 1,8 8 0,9 903 100
A-2 Empregadores (<= 10) 1.555 36,9 1.107 26,3 1.036 24,7 341 8,1 171 4,1 4.213 100
Total 2.162 42,3 1.292 25,3 1.126 22,0 357 7,0 179 3,5 5.116 100
Profissionais C Profissionais Autônomos 1.643 21,7 1.513 20,0 2.073 27,4 1.225 16,2 1.108 14,7 7.562 100
D Profissionais Assalariados 4.438 13,3 6.030 18,0 11.550 34,5 7.027 21,0 4.389 13,1 33.434 100
Total 6.081 14,8 7.543 18,4 13.623 33,2 8.252 20,1 5.497 13,4 40.995 100
Massa Não-Agrícola F Trabalhadores Autônomos 657 3,5 1.754 9,2 5.561 29,2 5.271 27,7 5.788 30,4 19.030 100
G Trabalhadores Assalariados 282 0,7 1.657 4,3 10.363 27,1 13.002 34,0 12.968 33,9 38.272 100
I Trabalhadores Domésticos 10 0,1 104 1,6 977 14,7 1.810 27,3 3.733 56,3 6.633 100
Total 948 1,5 3.515 5,5 16.901 26,4 20.083 31,4 22.489 35,2 63.936 100
Massa Agrícola H-1 Proprietários Conta Própria 188 2,0 364 3,8 1.387 14,4 1.889 19,7 5.779 60,2 9.608 100
H-2 Trabalhadores Autônomos 5 0,5 14 1,5 72 7,6 152 16,1 703 74,3 946 100
H-3 Trabalhadores Assalariados 17 0,2 58 0,6 794 8,4 2.260 23,9 6.322 66,9 9.451 100
Total 210 1,0 436 2,2 2.253 11,3 4.301 21,5 12.805 64,0 20.005 100
Não-remunerados Não-remunerados Não-Agrícolas 13 6,8 16 8,1 28 14,0 22 10,9 119 60,2 198 100
Não-remunerados Agrícolas 5 0,1 13 0,3 59 1,6 352 9,4 3.302 88,5 3.731 100
Sem Ocupação Com Renda 1.567 6,0 2.330 8,9 5.395 20,7 6.821 26,2 9.964 38,2 26.078 100
Sem Ocupação Sem Renda 8.094 100 8.094 100
Ignorados 177 10,3 202 11,8 364 21,1 337 19,6 640 37,2 1.720 100
+ +
Fonte: Censo Demográfico 2000, microdados. IBGE. Elaboração dos autores.
+ +
+ RESR, Piracicaba, SP, vol. 47, nº 02, p. 389-418, abr/jun 2009 – Impressa em julho 2009 +
+ +""" + + tests = generate_tests_from_html(html_content, "0", 1, self.random_generator) + + self.assertTrue(len(tests) > 10) + + def test_sup(self): + html_content = """ + + + + + + A ROSE BY ANY OTHER NAME + + + +
+
2016]
+
A ROSE BY ANY OTHER NAME
+
1083
+
+ +
+
cases were decided within a year of each other (2000 and 2001, respectively). Save the Manatee Club largely consists of a truncated version of the Consolidated-Tomoka analysis, with minor adjustments to conform the opinion to the 1999 amendments. Day Cruise, on the other hand, closely analyzes the 1999 version of section 120.52(8). However, it is Save the Manatee Club that has come to dominate Florida court opinions on rulemaking challenges and not the more detailed Day Cruise analysis.⁷⁸ The following Sections will discuss the facts of the two cases, examine the differences between their analyses of section 120.52(8), and finally conclude with an opinion on which analysis is better to apply in section 120.52(8) rulemaking challenges.
+ +
A. Southwest Florida Water Management District v. Save the Manatee Club, Inc.
+ +
After the legislature amended the APA, the First DCA analyzed the statutory language of section 120.52(8) again in Southwest Florida Water Management District v. Save the Manatee Club, Inc.⁷⁹ Save the Manatee Club concerned the Southwest Florida Water Management District's (the "District's") authority to create exemptions to environmental resource permitting requirements.⁸⁰ South Shores Partners, Ltd. ("South Shores") applied "for a permit to develop a 720-acre tract of land in Southwest Hillsborough County."⁸¹ As part of the development project, South Shores wanted "to build a connecting waterway between the [existing] canal system [on the property] and the [Tampa] Bay."⁸² The Save the Manatee Club believed that the resulting increase in power boat traffic in this new waterway would "endanger the manatee and its habitat."⁸³
+ +
The District has the authority to grant either a general permit or an environmental resource permit to a development project, depending on the type of project involved.⁸⁴ When granting an environmental resource permit, the District must consider "[t]he impact a proposed development will have on wildlife" as a factor; it does not have to do so when it grants a general permit.⁸⁵ The District granted South
+
+ +
+
78. As of December 14, 2015, a search of the "Citing References" on WestLaw shows that Save the Manatee Club has been cited by forty court opinions. Day Cruise, by comparison, has been cited by fifteen court opinions. These numbers do not include citations to either case in DOAH decisions.
+
79. 773 So. 2d 594 (Fla. 1st DCA 2000).
+
80. Id. at 596.
+
81. Id.
+
82. Id.
+
83. Id.
+
84. See id.
+
85. Id.
+
+ +""" + + tests = generate_tests_from_html(html_content, "0", 1, self.random_generator) + + superscript_map = { + "0": "⁰", + "1": "¹", + "2": "²", + "3": "³", + "4": "⁴", + "5": "⁵", + "6": "⁶", + "7": "⁷", + "8": "⁸", + "9": "⁹", + "+": "⁺", + "-": "⁻", + "=": "⁼", + "(": "⁽", + ")": "⁾", + "n": "ⁿ", + "i": "ⁱ", + } + + for test in tests: + for sup in superscript_map.values(): + self.assertTrue(sup not in test.get("text", "")) + self.assertTrue(sup not in test.get("before", "")) + self.assertTrue(sup not in test.get("after", "")) + + def test_katex_autorender(self): + """Test that KaTeX math expressions are properly auto-rendered when using the render_pdf_with_playwright function.""" + import asyncio + import os + import tempfile + + from olmocr.synth.mine_html_templates import render_pdf_with_playwright + + # Create HTML with LaTeX expressions + html_content = """ + + + + + KaTeX Auto-Render Test + + +
Math Expressions Test
+ +
Inline math expression: $E = mc^2$
+ +
Block math expression:
+
\[ + \\frac{d}{dx}(x^n) = nx^{n-1} + \]
+ +
Another complex equation:
+
\[ + \int_{a}^{b} f(x) \, dx = F(b) - F(a) + \]
+ + + """ + + # Create a temporary file to store the rendered PDF + with tempfile.NamedTemporaryFile(suffix=".pdf", delete=False) as tmp_file: + output_pdf_path = tmp_file.name + + # Render the HTML to PDF + render_result = asyncio.run(render_pdf_with_playwright(html_content=html_content, output_pdf_path=output_pdf_path, png_width=800, png_height=600)) + + # Check if rendering was successful + self.assertTrue(render_result.success, "PDF rendering should succeed") + + # Verify PDF was created and has content + self.assertTrue(os.path.exists(output_pdf_path), "PDF file should exist") + self.assertTrue(os.path.getsize(output_pdf_path) > 0, "PDF file should have content") + + # The actual validation of KaTeX rendering would require visual inspection or text extraction, + # but at minimum we can verify the file was created successfully + + print(output_pdf_path) + + def test_line_numbers(self): + html_content = """ + + + + + + House Amendment Bill No. CS/CS/SB 7030 + + +
+
HOUSE AMENDMENT
+
Bill No. CS/CS/SB 7030, 1st Eng. (2019)
+
+ +
Amendment No.
+ +
+
+
Senate
+
House
+
+
.
+
+ +
+ +
+ +
+
+ 1 +
Representative Jenne offered the following:
+
+
+ 2 +
+
+
+ 3 +
Amendment
+
+
+ 4 +
Remove lines 274-280 and insert:
+
+
+ 5 +
c.3. Pass an initial a psychological evaluation, and
+
+
+ 6 +
subsequent yearly psychological evaluations before each school
+
+
+ 7 +
year, administered by a psychologist licensed under chapter 490
+
+
+ 8 +
and designated by the Department of Law Enforcement and submit
+
+
+ 9 +
the results of such evaluations the evaluation to the sheriff's
+
+
+ 10 +
office. The Department of Law Enforcement is authorized to
+
+
+ 11 +
provide the sheriff's office with mental health and substance
+
+
+ 12 +
abuse data for compliance with this paragraph.
+
+
+ +
+
588513
+
Approved For Filing: 4/23/2019 6:09:18 PM
+
Page 1 of 1
+
+ +""" + + tests = generate_tests_from_html(html_content, "0", 1, self.random_generator) + + for test in tests: + if test["type"] == "order": + self.assertTrue(len([c for c in test["before"] if not c.isdigit()]) > 0) + + def test_complex_markdown_formatting(self): + html = """ + + + + + + +Optimization Parameters, Kinetics and Mechanism of Naproxen Removal by Catalytic Wet Peroxide Oxidation with a Hybrid Iron-Based Magnetic Catalyst + + +
+
+
+catalysts +
+
+
+
Article
+
Optimization Parameters, Kinetics and Mechanism of Naproxen Removal by Catalytic Wet Peroxide Oxidation with a Hybrid Iron-Based Magnetic Catalyst
+
+Ysabel Huaccallo-Aguilar ^1,2, Silvia Álvarez-Torrellas ¹, Marcos Larriba ¹, V. Ismael Águeda ¹, José Antonio Delgado ¹, Gabriel Ovejero ¹ and Juan García ^1,* +
+
+
+¹ Catalysis and Separation Processes Group, Chemical Engineering and Materials Department, Faculty of Chemistry, Complutense University, Avda. Complutense s/n, 28040 Madrid, Spain; Ysabelhu@ucm.es (Y.H.); satorrellas@ucm.es (S.Á.-T.); marcoslarriba@ucm.es (M.L.); viam@ucm.es (V.I.A.); jadeldo@ucm.es (J.A.D.); govejero@ucm.es (G.O.) +
+
+² Departamento de Ingeniería Química, Universidad Nacional de San Agustín, Av. Independencia s/n, Arequipa 04001, Peru +
+
+^* Correspondence: jgarciar@ucm.es; Tel.: +34-913-945-207 +
+
+
+ Received: 13 February 2019; Accepted: 15 March 2019; Published: 20 March 2019 +
+
+Abstract: This work presents a study of the assessment of the operating parameters of the catalytic wet peroxide oxidation (CWPO) of naproxen (NAP) using magnetite/multi-walled carbon nanotubes (Fe₃O₄/MWCNTs) as a catalyst. The effect of pH, temperature, and H₂O₂ dosage on CWPO process was evaluated by using the response surface model (RSM), allowing us to obtain an optimum NAP removal of 82% at the following operating conditions: pH = 5, T = 70 °C, [H₂O₂] = 1.5 mM, and [NAP] = 100 mg/L. Therefore, NAP degradation kinetics were revealed to follow a pseudo-second-order kinetic model, and an activation energy value of 4.75 kJ/mol was determined. Both considered as blank tests, showed no significant removal of the pollutant. Moreover, Fe₃O₄/MWCNTs material exhibited good recyclability along three consecutive cycles, finding an average NAP removal percentage close to 80% in each cycle of 3 h reaction time. In addition, the scavenging tests confirmed that the degradation of NAP was mainly governed by ^•OH radicals attack. Two reaction sequences were proposed for the degradation mechanism according to the detected byproducts. Finally, the versatility of the catalyst was evidenced in the treatment of different environmentally relevant aqueous matrices (wastewater treatment plant effluent (WWTP), surface water (SW), and a hospital wastewater (HW)) spiked with NAP, obtaining total organic carbon (TOC) removal efficiencies after 8 h in the following order: NAP-SW > NAP-HW > NAP-WWTP. +
+
+Keywords: CWPO; magnetic catalyst; naproxen; response surface methodology; wastewater +
+
1. Introduction
+
+ Pharmaceutical compounds are an important part of toxic materials in wastewater that are currently growing around the world [1]. They have been frequently detected in the aqueous medium, supposing a growing environmental problem for the scientific community. The excessive drug production demanded and consumed by humans and animals usually ends up in surface water [1], sewage effluents [2], groundwater [3], and even drinking water [4] due to the discharges from municipal wastewater treatment plants (WWTPs) [1], and consequently, these contaminants can lead to devastating effects on the environment [3,5]. Particularly, non-steroidal anti-inflammatory drugs +
+
+
Catalysts 2019, 9, 287; doi:10.3390/catal9030287
+
www.mdpi.com/journal/catalysts
+
+ + +""" + tests = generate_tests_from_html(html, "0", 1, self.random_generator) + markdown = html_to_markdown_with_frontmatter(html) + + for test in [test for test in tests if test["type"] == "present"]: + self.assertNotIn("**", test["text"]) + + def test_complex_formatting_two(self): + html = """ + + + + + + +Political Symbolism and Censorship + + +
Political Symbolism and Censorship:
+
The Tree Climber and Al-Farafir Revisited
+
Dina Amin^*
+
The period that followed the Egyptian Revolution of 1952 was euphoric with enthusiasm as the expectations of the population as well as that of the intelligentsia were soaring high. No sooner however, did the optimism escalate that it receded as Gamal Abdel Nasser¹ who started off as an advocate for freedom and democracy turned into a despot (at least from the point of view of a portion of the population). Attempts at criticism were met with censorship on all fronts. After a number of writers and artists were imprisoned for their political views, some writers opted to stay silent, some conformed and some went around censorship by using symbolism and metaphor.
+
An art form that suffered tremendously from censorship was theatre and playwriting. While theatre in Egypt could have been debilitated by censorship throughout the 1960s, a decade that suffered greatly from the suppression of speech, it ironically flourished instead, largely due to the pervasive use of symbolism and metaphors by dramatists. How playwrights converted their skepticism about the failure of the revolution to fulfill its promises of democracy into covert criticism, I will try to examine. Taking Ya Tali' al-Shagarah (1962; published in English as The Tree Climber, 1966) by Tawfiq al-Hakim and Al-Farafir (Flipflap and His Master, 1964) by Yusuf Idris as projections of political discontent, I will argue that, while both use theatre of the absurd as a form, the two texts uphold political symbolism as modus operandi in their covert criticism of the regime.²
+
According to Abner Cohen in "Political Symbolism," the poetics of symbols lie in their ambiguity and multileveled connotations,
+
+
It is the very essence and potency of symbols that they are ambiguous, referring to different meanings, and are not given to precise definition. The most dominant symbols are essentially bivocal, being rooted, on the one hand, in the human condition, in what may be called "selfhood," and on
+
+
+
^* Associate Professor of Drama and Performance Studies in the Department of English Language and Literature, Faculty of Arts, Cairo University.
+
Cairo Studies in English (2019-Summer): https://cse.journals.ekb.eg/
+
26
+
+ + +""" + tests = generate_tests_from_html(html, "0", 1, self.random_generator) + + for test in [test for test in tests if "text" in test]: + self.assertNotIn("**", test["text"]) diff --git a/tests/test_olmocrmix.py b/tests/test_olmocrmix.py new file mode 100644 index 0000000..b837c1f --- /dev/null +++ b/tests/test_olmocrmix.py @@ -0,0 +1,94 @@ +# Test that prepare_olmocrmix.py and repackage_olmocrmix.py work correctly +# by packaging the sample dataset, unpacking it, and verifying contents are preserved + +import os +import subprocess +import tempfile +from pathlib import Path + + +def test_repackage_and_prepare_olmocrmix(): + """Test that repackaging and preparing preserves the dataset contents exactly.""" + + with tempfile.TemporaryDirectory() as temp_dir: + temp_path = Path(temp_dir) + sample_dataset = Path("tests/sample_dataset") + + # Step 1: Repackage the sample dataset into parquet + tarballs + packaged_dir = temp_path / "packaged" + packaged_dir.mkdir(parents=True, exist_ok=True) + + repackage_result = subprocess.run( + [ + "python", + "olmocr/data/repackage_olmocrmix.py", + "--processed-dir", + str(sample_dataset), + "--subset", + "test_subset", + "--split", + "test_split", + "--output-dir", + str(packaged_dir), + ], + capture_output=True, + text=True, + ) + + assert repackage_result.returncode == 0, f"Repackage script failed with stderr: {repackage_result.stderr}\nstdout: {repackage_result.stdout}" + + # Verify the packaged output exists + parquet_file = packaged_dir / "test_subset_test_split.parquet" + assert parquet_file.exists(), f"Expected parquet file not found: {parquet_file}" + + # Step 2: Repackage the sample dataset into parquet + tarballs + unpackaged_dir = temp_path / "unpackaged" + + prepare_result = subprocess.run( + [ + "python", + "olmocr/data/prepare_olmocrmix.py", + "--dataset-path", + str(packaged_dir), + "--subset", + "test_subset", + "--split", + "test_split", + "--destination", + str(unpackaged_dir), + ], + capture_output=True, + text=True, + ) + + assert prepare_result.returncode == 0 + + for root, _, files in os.walk(temp_path): + for file_name in files: + print(Path(root) / file_name) + + unpacked_processed = unpackaged_dir / "processed_test_subset_test_split" + assert unpacked_processed.exists(), f"Unpacked processed dir missing: {unpacked_processed}" + + def relative_files(root: Path): + return sorted(path.relative_to(root) for path in root.rglob("*") if path.is_file()) + + sample_files = relative_files(sample_dataset) + unpacked_files = relative_files(unpacked_processed) + assert sample_files == unpacked_files, "Mismatch in files between sample dataset and unpacked output" + + for relative_path in sample_files: + sample_file = sample_dataset / relative_path + unpacked_file = unpacked_processed / relative_path + + if relative_path.suffix == ".jsonl": + # For JSONL files, compare as sets of lines (order doesn't matter) + # Filter out empty lines + sample_lines = set(line for line in sample_file.read_text().strip().split("\n") if line.strip()) + unpacked_lines = set(line for line in unpacked_file.read_text().strip().split("\n") if line.strip()) + assert sample_lines == unpacked_lines, f"JSONL file contents differ for {relative_path}" + else: + # For other files, compare as bytes + sample_contents = sample_file.read_bytes() + unpacked_contents = unpacked_file.read_bytes() + assert sample_contents == unpacked_contents, f"File contents differ for {relative_path}" diff --git a/tests/test_pipeline.py b/tests/test_pipeline.py new file mode 100644 index 0000000..1944290 --- /dev/null +++ b/tests/test_pipeline.py @@ -0,0 +1,570 @@ +import base64 +import json +import os +from dataclasses import dataclass +from io import BytesIO +from unittest.mock import AsyncMock, patch + +import pytest +from PIL import Image + +from olmocr.pipeline import ( + PageResult, + build_page_query, + get_markdown_path, + process_page, +) + + +def create_test_image(width=100, height=150): + """Create a simple test image with distinct features to verify rotation.""" + img = Image.new("RGB", (width, height), color="white") + pixels = img.load() + + # Draw a red square in top-left corner + for x in range(10, 30): + for y in range(10, 30): + if pixels is not None: + pixels[x, y] = (255, 0, 0) + + # Draw a blue rectangle in bottom-right corner + for x in range(width - 40, width - 10): + for y in range(height - 30, height - 10): + if pixels is not None: + pixels[x, y] = (0, 0, 255) + + # Draw a green line near the top + for x in range(20, 80): + if pixels is not None: + pixels[x, 5] = (0, 255, 0) + + return img + + +def image_to_base64(img): + """Convert PIL Image to base64 string.""" + buffered = BytesIO() + img.save(buffered, format="PNG") + return base64.b64encode(buffered.getvalue()).decode() + + +def base64_to_image(base64_str): + """Convert base64 string to PIL Image.""" + image_bytes = base64.b64decode(base64_str) + return Image.open(BytesIO(image_bytes)) + + +class TestImageRotation: + @pytest.mark.asyncio + async def test_no_rotation(self): + """Test that image_rotation=0 returns the original image.""" + test_img = create_test_image() + test_base64 = image_to_base64(test_img) + + with patch("olmocr.pipeline.render_pdf_to_base64png") as mock_render: + mock_render.return_value = test_base64 + + result = await build_page_query("fake_pdf.pdf", 1, 1000, image_rotation=0) + + # Extract the image from the result + messages = result["messages"] + content = messages[0]["content"] + image_url = content[1]["image_url"]["url"] + image_base64 = image_url.split(",")[1] + result_img = base64_to_image(image_base64) + + # Should be the same size as original + assert result_img.size == test_img.size + + # Check pixel at specific location (red square should be in top-left) + assert result_img.getpixel((20, 20)) == (255, 0, 0) + + @pytest.mark.asyncio + async def test_rotate_90_degrees(self): + """Test that image_rotation=90 rotates the image 90 degrees counter-clockwise.""" + test_img = create_test_image(100, 150) + test_base64 = image_to_base64(test_img) + + with patch("olmocr.pipeline.render_pdf_to_base64png") as mock_render: + mock_render.return_value = test_base64 + + result = await build_page_query("fake_pdf.pdf", 1, 1000, image_rotation=90) + + # Extract the image from the result + messages = result["messages"] + content = messages[0]["content"] + image_url = content[1]["image_url"]["url"] + image_base64 = image_url.split(",")[1] + result_img = base64_to_image(image_base64) + + # After 90 degree counter-clockwise rotation, dimensions should be swapped + assert result_img.size == (150, 100) + + # The red square that was at top-left should now be at bottom-left + # Original (20, 20) -> After 90° CCW rotation -> (20, 80) + assert result_img.getpixel((20, 80)) == (255, 0, 0) + + @pytest.mark.asyncio + async def test_rotate_180_degrees(self): + """Test that image_rotation=180 rotates the image 180 degrees.""" + test_img = create_test_image(100, 150) + test_base64 = image_to_base64(test_img) + + with patch("olmocr.pipeline.render_pdf_to_base64png") as mock_render: + mock_render.return_value = test_base64 + + result = await build_page_query("fake_pdf.pdf", 1, 1000, image_rotation=180) + + # Extract the image from the result + messages = result["messages"] + content = messages[0]["content"] + image_url = content[1]["image_url"]["url"] + image_base64 = image_url.split(",")[1] + result_img = base64_to_image(image_base64) + + # After 180 degree rotation, dimensions should be the same + assert result_img.size == (100, 150) + + # The red square that was at top-left should now be at bottom-right + # Original (20, 20) -> After 180° rotation -> (80, 130) + assert result_img.getpixel((80, 130)) == (255, 0, 0) + + @pytest.mark.asyncio + async def test_rotate_270_degrees(self): + """Test that image_rotation=270 rotates the image 270 degrees counter-clockwise (90 clockwise).""" + test_img = create_test_image(100, 150) + test_base64 = image_to_base64(test_img) + + with patch("olmocr.pipeline.render_pdf_to_base64png") as mock_render: + mock_render.return_value = test_base64 + + result = await build_page_query("fake_pdf.pdf", 1, 1000, image_rotation=270) + + # Extract the image from the result + messages = result["messages"] + content = messages[0]["content"] + image_url = content[1]["image_url"]["url"] + image_base64 = image_url.split(",")[1] + result_img = base64_to_image(image_base64) + + # After 270 degree counter-clockwise rotation, dimensions should be swapped + assert result_img.size == (150, 100) + + # The red square that was at top-left should now be at top-right + # Original (20, 20) -> After 270° CCW rotation -> (130, 20) + assert result_img.getpixel((130, 20)) == (255, 0, 0) + + @pytest.mark.asyncio + async def test_invalid_rotation_angle(self): + """Test that invalid rotation angles raise an assertion error.""" + test_img = create_test_image() + test_base64 = image_to_base64(test_img) + + with patch("olmocr.pipeline.render_pdf_to_base64png") as mock_render: + mock_render.return_value = test_base64 + + with pytest.raises(AssertionError, match="Invalid image rotation"): + await build_page_query("fake_pdf.pdf", 1, 1000, image_rotation=45) + + @pytest.mark.asyncio + async def test_rotation_preserves_image_quality(self): + """Test that rotation preserves the image without distortion.""" + # Create a more complex test image + test_img = create_test_image(200, 300) + test_base64 = image_to_base64(test_img) + + with patch("olmocr.pipeline.render_pdf_to_base64png") as mock_render: + mock_render.return_value = test_base64 + + # Test all valid rotation angles + for angle in [0, 90, 180, 270]: + result = await build_page_query("fake_pdf.pdf", 1, 1000, image_rotation=angle) + + # Extract the image from the result + messages = result["messages"] + content = messages[0]["content"] + image_url = content[1]["image_url"]["url"] + image_base64 = image_url.split(",")[1] + result_img = base64_to_image(image_base64) + + # Verify image format is preserved + assert result_img.format == "PNG" or result_img.format is None + assert result_img.mode == "RGB" + + +@dataclass +class MockArgs: + max_page_retries: int = 8 + target_longest_image_dim: int = 1288 + guided_decoding: bool = False + server: str = "http://localhost:30000/v1" + model: str = "olmocr" + + +class TestRotationCorrection: + @pytest.mark.asyncio + async def test_process_page_with_rotation_correction(self): + """Test that process_page correctly handles rotation correction from model response.""" + + # Path to the test PDF that needs 90 degree rotation + test_pdf_path = "tests/gnarly_pdfs/edgar-rotated90.pdf" + + # Mock arguments + args = MockArgs() + + # Counter to track number of API calls + call_count = 0 + + async def mock_apost(url, json_data, api_key=None): + nonlocal call_count + call_count += 1 + + # Check the rotation in the request + messages = json_data.get("messages", []) + if messages: + content = messages[0].get("content", []) + image_data = content[0].get("image_url", {}).get("url", "") + + # First call - model detects rotation is needed + if call_count == 1: + response_content = """--- +primary_language: en +is_rotation_valid: false +rotation_correction: 90 +is_table: false +is_diagram: false +--- + +This document appears to be rotated and needs correction.""" + + # Second call - after rotation, model says it's correct + elif call_count == 2: + response_content = """--- +primary_language: en +is_rotation_valid: true +rotation_correction: 0 +is_table: false +is_diagram: false +--- + +UNITED STATES +SECURITIES AND EXCHANGE COMMISSION +Washington, D.C. 20549 + +This is the corrected text from the document.""" + + else: + raise ValueError(f"Unexpected call count: {call_count}") + + # Mock response structure + response_body = { + "choices": [{"message": {"content": response_content}, "finish_reason": "stop"}], + "usage": {"prompt_tokens": 1000, "completion_tokens": 100, "total_tokens": 1100}, + } + + return 200, json.dumps(response_body).encode() + + # Mock the worker tracker + mock_tracker = AsyncMock() + + # Ensure the test PDF exists + assert os.path.exists(test_pdf_path), f"Test PDF not found at {test_pdf_path}" + + # Track calls to build_page_query + build_page_query_calls = [] + original_build_page_query = build_page_query + + async def mock_build_page_query(local_pdf_path, page, target_longest_image_dim, image_rotation=0, model_name="olmocr"): + build_page_query_calls.append(image_rotation) + return await original_build_page_query(local_pdf_path, page, target_longest_image_dim, image_rotation, model_name) + + with patch("olmocr.pipeline.apost", side_effect=mock_apost): + with patch("olmocr.pipeline.tracker", mock_tracker): + with patch("olmocr.pipeline.build_page_query", side_effect=mock_build_page_query): + result = await process_page(args=args, worker_id=0, pdf_orig_path="test-edgar-rotated90.pdf", pdf_local_path=test_pdf_path, page_num=1) + + # Verify the result + assert isinstance(result, PageResult) + assert result.page_num == 1 + assert result.is_fallback == False + assert result.response.is_rotation_valid == True + assert result.response.rotation_correction == 0 + assert result.response.natural_text is not None + assert "SECURITIES AND EXCHANGE COMMISSION" in result.response.natural_text + + # Verify that exactly 2 API calls were made + assert call_count == 2 + + # Verify build_page_query was called with correct rotations + assert len(build_page_query_calls) == 2 + assert build_page_query_calls[0] == 0 # First call with no rotation + assert build_page_query_calls[1] == 90 # Second call with 90 degree rotation + + # Verify tracker was called correctly + mock_tracker.track_work.assert_any_call(0, "test-edgar-rotated90.pdf-1", "started") + mock_tracker.track_work.assert_any_call(0, "test-edgar-rotated90.pdf-1", "finished") + + @pytest.mark.asyncio + async def test_process_page_with_cumulative_rotation(self): + """Test that process_page correctly accumulates rotations across multiple attempts.""" + + # Path to the test PDF (can use any test PDF) + test_pdf_path = "tests/gnarly_pdfs/edgar-rotated90.pdf" + + # Mock arguments + args = MockArgs() + + # Counter to track number of API calls + call_count = 0 + + async def mock_apost(url, json_data, api_key=None): + nonlocal call_count + call_count += 1 + + # First call - model detects rotation is needed (90 degrees) + if call_count == 1: + response_content = """--- +primary_language: en +is_rotation_valid: false +rotation_correction: 90 +is_table: false +is_diagram: false +--- + +This document appears to be rotated and needs correction.""" + + # Second call - model still detects rotation is needed (another 90 degrees) + elif call_count == 2: + response_content = """--- +primary_language: en +is_rotation_valid: false +rotation_correction: 90 +is_table: false +is_diagram: false +--- + +Document still needs rotation.""" + + # Third call - after 180 total degrees of rotation, model says it's correct + elif call_count == 3: + response_content = """--- +primary_language: en +is_rotation_valid: true +rotation_correction: 0 +is_table: false +is_diagram: false +--- + +UNITED STATES +SECURITIES AND EXCHANGE COMMISSION +Washington, D.C. 20549 + +Document is now correctly oriented after 180 degree rotation.""" + + else: + raise ValueError(f"Unexpected call count: {call_count}") + + # Mock response structure + response_body = { + "choices": [{"message": {"content": response_content}, "finish_reason": "stop"}], + "usage": {"prompt_tokens": 1000, "completion_tokens": 100, "total_tokens": 1100}, + } + + return 200, json.dumps(response_body).encode() + + # Mock the worker tracker + mock_tracker = AsyncMock() + + # Ensure the test PDF exists + assert os.path.exists(test_pdf_path), f"Test PDF not found at {test_pdf_path}" + + # Track calls to build_page_query + build_page_query_calls = [] + original_build_page_query = build_page_query + + async def mock_build_page_query(local_pdf_path, page, target_longest_image_dim, image_rotation=0, model_name="olmocr"): + build_page_query_calls.append(image_rotation) + return await original_build_page_query(local_pdf_path, page, target_longest_image_dim, image_rotation, model_name) + + with patch("olmocr.pipeline.apost", side_effect=mock_apost): + with patch("olmocr.pipeline.tracker", mock_tracker): + with patch("olmocr.pipeline.build_page_query", side_effect=mock_build_page_query): + result = await process_page(args=args, worker_id=0, pdf_orig_path="test-cumulative-rotation.pdf", pdf_local_path=test_pdf_path, page_num=1) + + # Verify the result + assert isinstance(result, PageResult) + assert result.page_num == 1 + assert result.is_fallback == False + assert result.response.is_rotation_valid == True + assert result.response.rotation_correction == 0 + assert result.response.natural_text is not None + assert "180 degree rotation" in result.response.natural_text + + # Verify that exactly 3 API calls were made + assert call_count == 3 + + # Verify build_page_query was called with correct cumulative rotations + assert len(build_page_query_calls) == 3 + assert build_page_query_calls[0] == 0 # First call with no rotation + assert build_page_query_calls[1] == 90 # Second call with 90 degree rotation + assert build_page_query_calls[2] == 180 # Third call with cumulative 180 degree rotation + + # Verify tracker was called correctly + mock_tracker.track_work.assert_any_call(0, "test-cumulative-rotation.pdf-1", "started") + mock_tracker.track_work.assert_any_call(0, "test-cumulative-rotation.pdf-1", "finished") + + @pytest.mark.asyncio + async def test_process_page_rotation_wraps_around(self): + """Test that cumulative rotation correctly wraps around at 360 degrees.""" + + # Path to the test PDF + test_pdf_path = "tests/gnarly_pdfs/edgar-rotated90.pdf" + + # Mock arguments + args = MockArgs() + + # Counter to track number of API calls + call_count = 0 + + async def mock_apost(url, json_data, api_key=None): + nonlocal call_count + call_count += 1 + + # First call - model detects rotation is needed (270 degrees) + if call_count == 1: + response_content = """--- +primary_language: en +is_rotation_valid: false +rotation_correction: 270 +is_table: false +is_diagram: false +--- + +Document needs 270 degree rotation.""" + + # Second call - model detects more rotation is needed (180 degrees) + # Total would be 450, but should wrap to 90 + elif call_count == 2: + response_content = """--- +primary_language: en +is_rotation_valid: false +rotation_correction: 180 +is_table: false +is_diagram: false +--- + +Document needs additional rotation.""" + + # Third call - after wrapped rotation (90 degrees), model says it's correct + elif call_count == 3: + response_content = """--- +primary_language: en +is_rotation_valid: true +rotation_correction: 0 +is_table: false +is_diagram: false +--- + +Document correctly oriented at 90 degrees total rotation.""" + + else: + raise ValueError(f"Unexpected call count: {call_count}") + + # Mock response structure + response_body = { + "choices": [{"message": {"content": response_content}, "finish_reason": "stop"}], + "usage": {"prompt_tokens": 1000, "completion_tokens": 100, "total_tokens": 1100}, + } + + return 200, json.dumps(response_body).encode() + + # Mock the worker tracker + mock_tracker = AsyncMock() + + # Ensure the test PDF exists + assert os.path.exists(test_pdf_path), f"Test PDF not found at {test_pdf_path}" + + # Track calls to build_page_query + build_page_query_calls = [] + original_build_page_query = build_page_query + + async def mock_build_page_query(local_pdf_path, page, target_longest_image_dim, image_rotation=0, model_name="olmocr"): + build_page_query_calls.append(image_rotation) + return await original_build_page_query(local_pdf_path, page, target_longest_image_dim, image_rotation, model_name) + + with patch("olmocr.pipeline.apost", side_effect=mock_apost): + with patch("olmocr.pipeline.tracker", mock_tracker): + with patch("olmocr.pipeline.build_page_query", side_effect=mock_build_page_query): + result = await process_page(args=args, worker_id=0, pdf_orig_path="test-rotation-wrap.pdf", pdf_local_path=test_pdf_path, page_num=1) + + # Verify the result + assert isinstance(result, PageResult) + assert result.page_num == 1 + assert result.is_fallback == False + assert result.response.is_rotation_valid == True + + # Verify that exactly 3 API calls were made + assert call_count == 3 + + # Verify build_page_query was called with correct cumulative rotations + assert len(build_page_query_calls) == 3 + assert build_page_query_calls[0] == 0 # First call with no rotation + assert build_page_query_calls[1] == 270 # Second call with 270 degree rotation + assert build_page_query_calls[2] == 90 # Third call with wrapped rotation (270 + 180 = 450 % 360 = 90) + + +class TestMarkdownPathHandling: + """Tests for the get_markdown_path function to ensure files stay within workspace.""" + + def test_absolute_local_path_stays_in_workspace(self): + """ + Test that absolute local paths produce markdown paths inside the workspace. + + This is a regression test for a bug where passing absolute paths like + /home/user/documents/file.pdf would cause the markdown output to be written + to /home/user/documents/file.md instead of workspace/markdown/.../file.md + """ + workspace = "/tmp/test_workspace" + source_file = "/home/ubuntu/test_documents/subfolder/test_document.pdf" + + markdown_path = get_markdown_path(workspace, source_file) + + # The markdown path should be inside the workspace + assert markdown_path.startswith(workspace), ( + f"BUG: Markdown path '{markdown_path}' is outside workspace '{workspace}'. " f"Absolute source paths should not escape the workspace directory." + ) + + def test_s3_path_stays_in_workspace(self): + """Test that S3 paths produce markdown paths inside the workspace.""" + workspace = "/tmp/test_workspace" + source_file = "s3://my-bucket/documents/subfolder/test_document.pdf" + + markdown_path = get_markdown_path(workspace, source_file) + + assert markdown_path.startswith(workspace) + assert markdown_path == "/tmp/test_workspace/markdown/documents/subfolder/test_document.md" + + def test_relative_local_path_stays_in_workspace(self): + """Test that relative local paths produce markdown paths inside the workspace.""" + workspace = "/tmp/test_workspace" + source_file = "documents/subfolder/test_document.pdf" + + markdown_path = get_markdown_path(workspace, source_file) + + assert markdown_path.startswith(workspace) + assert markdown_path == "/tmp/test_workspace/markdown/documents/subfolder/test_document.md" + + def test_path_traversal_with_dotdot_stays_in_workspace(self): + """Test that paths with ../ do not escape the workspace directory.""" + workspace = "/tmp/test_workspace" + source_file = "documents/../../../etc/passwd.pdf" + + markdown_path = get_markdown_path(workspace, source_file) + + # Resolve the path to check if it actually stays in workspace + resolved_path = os.path.normpath(markdown_path) + resolved_workspace = os.path.normpath(workspace) + + assert resolved_path.startswith(resolved_workspace), ( + f"BUG: Path traversal attack! Markdown path '{resolved_path}' escapes " f"workspace '{resolved_workspace}'. Paths with ../ should be sanitized." + ) diff --git a/tests/test_s3_utils.py b/tests/test_s3_utils.py new file mode 100644 index 0000000..cc2f2eb --- /dev/null +++ b/tests/test_s3_utils.py @@ -0,0 +1,221 @@ +import unittest +from unittest.mock import MagicMock, Mock + +from botocore.exceptions import ClientError + +from olmocr.s3_utils import expand_s3_glob + + +class TestExpandS3Glob(unittest.TestCase): + def setUp(self): + """Set up test fixtures before each test method.""" + self.s3_client = Mock() + + def _mock_paginator(self, contents): + """Helper to mock S3 paginator responses.""" + paginator = MagicMock() + paginator.paginate.return_value = [{"Contents": contents}] + self.s3_client.get_paginator.return_value = paginator + return paginator + + def test_wildcard_at_end_of_filename(self): + """Test glob with wildcard at end: s3://bucket/path/prefix_*""" + contents = [ + {"Key": "data/files/report_001.pdf", "ETag": '"abc123"'}, + {"Key": "data/files/report_002.pdf", "ETag": '"def456"'}, + {"Key": "data/files/other.pdf", "ETag": '"ghi789"'}, + ] + self._mock_paginator(contents) + + result = expand_s3_glob(self.s3_client, "s3://test-bucket/data/files/report_*") + + # Should use prefix "data/files/report_" (everything before the *) + self.s3_client.get_paginator.assert_called_once_with("list_objects_v2") + call_args = self.s3_client.get_paginator.return_value.paginate.call_args + self.assertEqual(call_args[1]["Prefix"], "data/files/report_") + + # Should match only report_* files + self.assertEqual(len(result), 2) + self.assertIn("s3://test-bucket/data/files/report_001.pdf", result) + self.assertIn("s3://test-bucket/data/files/report_002.pdf", result) + self.assertNotIn("s3://test-bucket/data/files/other.pdf", result) + + def test_wildcard_extension(self): + """Test glob with wildcard extension: s3://bucket/path/*.pdf""" + contents = [ + {"Key": "data/docs/file1.pdf", "ETag": '"abc123"'}, + {"Key": "data/docs/file2.pdf", "ETag": '"def456"'}, + {"Key": "data/docs/file3.txt", "ETag": '"ghi789"'}, + ] + self._mock_paginator(contents) + + result = expand_s3_glob(self.s3_client, "s3://test-bucket/data/docs/*.pdf") + + # Should use prefix "data/docs/" (up to the *) + call_args = self.s3_client.get_paginator.return_value.paginate.call_args + self.assertEqual(call_args[1]["Prefix"], "data/docs/") + + # Should match only .pdf files + self.assertEqual(len(result), 2) + self.assertIn("s3://test-bucket/data/docs/file1.pdf", result) + self.assertIn("s3://test-bucket/data/docs/file2.pdf", result) + + def test_wildcard_in_middle_of_path(self): + """Test glob with wildcard in middle: s3://bucket/data/*/files/*.pdf""" + contents = [ + {"Key": "data/2024/files/report.pdf", "ETag": '"abc123"'}, + {"Key": "data/2023/files/report.pdf", "ETag": '"def456"'}, + {"Key": "data/2024/other/report.pdf", "ETag": '"ghi789"'}, + ] + self._mock_paginator(contents) + + result = expand_s3_glob(self.s3_client, "s3://test-bucket/data/*/files/*.pdf") + + # Should use prefix "data/" (up to the first *) + call_args = self.s3_client.get_paginator.return_value.paginate.call_args + self.assertEqual(call_args[1]["Prefix"], "data/") + + # Should match files in any year's files/ subdirectory + self.assertEqual(len(result), 2) + self.assertIn("s3://test-bucket/data/2024/files/report.pdf", result) + self.assertIn("s3://test-bucket/data/2023/files/report.pdf", result) + self.assertNotIn("s3://test-bucket/data/2024/other/report.pdf", result) + + def test_wildcard_at_root(self): + """Test glob with wildcard at root: s3://bucket/*.pdf""" + contents = [ + {"Key": "file1.pdf", "ETag": '"abc123"'}, + {"Key": "file2.pdf", "ETag": '"def456"'}, + {"Key": "subdir/file3.pdf", "ETag": '"ghi789"'}, + ] + self._mock_paginator(contents) + + result = expand_s3_glob(self.s3_client, "s3://test-bucket/*.pdf") + + # Should use empty prefix + call_args = self.s3_client.get_paginator.return_value.paginate.call_args + self.assertEqual(call_args[1]["Prefix"], "") + + # fnmatch *.pdf matches all .pdf files including in subdirs (fnmatch * matches /) + self.assertEqual(len(result), 3) + self.assertIn("s3://test-bucket/file1.pdf", result) + self.assertIn("s3://test-bucket/file2.pdf", result) + self.assertIn("s3://test-bucket/subdir/file3.pdf", result) + + def test_bracket_wildcard(self): + """Test glob with bracket wildcard: s3://bucket/data/file[123].pdf""" + contents = [ + {"Key": "data/file1.pdf", "ETag": '"abc123"'}, + {"Key": "data/file2.pdf", "ETag": '"def456"'}, + {"Key": "data/file4.pdf", "ETag": '"ghi789"'}, + ] + self._mock_paginator(contents) + + result = expand_s3_glob(self.s3_client, "s3://test-bucket/data/file[123].pdf") + + # Should use prefix "data/file" (up to the [) + call_args = self.s3_client.get_paginator.return_value.paginate.call_args + self.assertEqual(call_args[1]["Prefix"], "data/file") + + # [123] matches 1, 2, or 3 + self.assertEqual(len(result), 2) + self.assertIn("s3://test-bucket/data/file1.pdf", result) + self.assertIn("s3://test-bucket/data/file2.pdf", result) + self.assertNotIn("s3://test-bucket/data/file4.pdf", result) + + def test_real_world_worker_pattern(self): + """Test real-world pattern: s3://bucket/path/prefix_worker62_*""" + contents = [ + {"Key": "jakep/dolma/olmo-crawled-pdfs_worker62_001.tar", "ETag": '"abc123"'}, + {"Key": "jakep/dolma/olmo-crawled-pdfs_worker62_002.tar", "ETag": '"def456"'}, + {"Key": "jakep/dolma/olmo-crawled-pdfs_worker63_001.tar", "ETag": '"ghi789"'}, + ] + self._mock_paginator(contents) + + result = expand_s3_glob(self.s3_client, "s3://ai2-oe-data/jakep/dolma/olmo-crawled-pdfs_worker62_*") + + # Should use prefix up to the * (the entire fixed part) + call_args = self.s3_client.get_paginator.return_value.paginate.call_args + self.assertEqual(call_args[1]["Prefix"], "jakep/dolma/olmo-crawled-pdfs_worker62_") + + # Should only match worker62 files + self.assertEqual(len(result), 2) + self.assertIn("s3://ai2-oe-data/jakep/dolma/olmo-crawled-pdfs_worker62_001.tar", result) + self.assertIn("s3://ai2-oe-data/jakep/dolma/olmo-crawled-pdfs_worker62_002.tar", result) + self.assertNotIn("s3://ai2-oe-data/jakep/dolma/olmo-crawled-pdfs_worker63_001.tar", result) + + def test_no_wildcard_single_file(self): + """Test path without wildcard returns single file.""" + self.s3_client.head_object.return_value = {"ContentType": "application/pdf", "ETag": '"abc123"'} + + result = expand_s3_glob(self.s3_client, "s3://test-bucket/data/file.pdf") + + self.s3_client.head_object.assert_called_once_with(Bucket="test-bucket", Key="data/file.pdf") + self.assertEqual(result, {"s3://test-bucket/data/file.pdf": "abc123"}) + + def test_no_wildcard_directory_error(self): + """Test that bare directory path raises helpful error.""" + self.s3_client.head_object.return_value = {"ContentType": "application/x-directory", "ETag": '"abc123"'} + + with self.assertRaises(ValueError) as ctx: + expand_s3_glob(self.s3_client, "s3://test-bucket/data/folder") + + self.assertIn("appears to be a folder", str(ctx.exception)) + self.assertIn("*.pdf", str(ctx.exception)) + + def test_no_wildcard_not_found_but_is_folder(self): + """Test 404 on path that's actually a folder with contents.""" + self.s3_client.head_object.side_effect = ClientError({"Error": {"Code": "404", "Message": "Not Found"}}, "HeadObject") + # Simulate folder with contents + paginator = MagicMock() + paginator.paginate.return_value = [{"Contents": [{"Key": "data/folder/file.pdf"}]}] + self.s3_client.get_paginator.return_value = paginator + + with self.assertRaises(ValueError) as ctx: + expand_s3_glob(self.s3_client, "s3://test-bucket/data/folder") + + self.assertIn("appears to be a folder", str(ctx.exception)) + + def test_no_wildcard_not_found(self): + """Test 404 on path that doesn't exist.""" + self.s3_client.head_object.side_effect = ClientError({"Error": {"Code": "404", "Message": "Not Found"}}, "HeadObject") + # No contents under this prefix + paginator = MagicMock() + paginator.paginate.return_value = [{"Contents": []}] + self.s3_client.get_paginator.return_value = paginator + + with self.assertRaises(ValueError) as ctx: + expand_s3_glob(self.s3_client, "s3://test-bucket/data/nonexistent.pdf") + + self.assertIn("No object or prefix found", str(ctx.exception)) + + def test_invalid_scheme(self): + """Test that non-s3 paths raise error.""" + with self.assertRaises(ValueError) as ctx: + expand_s3_glob(self.s3_client, "gs://bucket/path/*.pdf") + + self.assertIn("must start with s3://", str(ctx.exception)) + + def test_etag_quotes_stripped(self): + """Test that ETags have quotes stripped correctly.""" + contents = [ + {"Key": "data/file.pdf", "ETag": '"abc123"'}, + ] + self._mock_paginator(contents) + + result = expand_s3_glob(self.s3_client, "s3://test-bucket/data/*.pdf") + + # ETag should not have quotes + self.assertEqual(result["s3://test-bucket/data/file.pdf"], "abc123") + + def test_empty_results(self): + """Test glob that matches nothing returns empty dict.""" + self._mock_paginator([]) + + result = expand_s3_glob(self.s3_client, "s3://test-bucket/data/*.pdf") + + self.assertEqual(result, {}) + + +if __name__ == "__main__": + unittest.main() diff --git a/tests/test_s3_work_queue.py b/tests/test_s3_work_queue.py new file mode 100644 index 0000000..2f9e74b --- /dev/null +++ b/tests/test_s3_work_queue.py @@ -0,0 +1,281 @@ +import asyncio +import datetime +import unittest +from unittest.mock import Mock, patch + +from botocore.exceptions import ClientError + +# Import the classes we're testing +from olmocr.work_queue import S3Backend, WorkItem, WorkQueue + + +class TestS3WorkQueue(unittest.TestCase): + def setUp(self): + """Set up test fixtures before each test method.""" + self.s3_client = Mock() + self.s3_client.exceptions.ClientError = ClientError + self.backend = S3Backend(self.s3_client, "s3://test-bucket/workspace") + self.work_queue = WorkQueue(self.backend) + self.sample_paths = [ + "s3://test-bucket/data/file1.pdf", + "s3://test-bucket/data/file2.pdf", + "s3://test-bucket/data/file3.pdf", + ] + + def tearDown(self): + """Clean up after each test method.""" + pass + + def test_compute_workgroup_hash(self): + """Test hash computation is deterministic and correct""" + paths = [ + "s3://test-bucket/data/file2.pdf", + "s3://test-bucket/data/file1.pdf", + ] + + # Hash should be the same regardless of order + hash1 = WorkQueue._compute_workgroup_hash(paths) + hash2 = WorkQueue._compute_workgroup_hash(reversed(paths)) + self.assertEqual(hash1, hash2) + + def test_init(self): + """Test initialization of S3Backend""" + client = Mock() + backend = S3Backend(client, "s3://test-bucket/workspace/") + + self.assertEqual(backend.workspace_path, "s3://test-bucket/workspace") + self.assertEqual(backend._index_path, "s3://test-bucket/workspace/work_index_list.csv.zstd") + self.assertEqual(backend._output_glob, "s3://test-bucket/workspace/done_flags/*.flag") + + def asyncSetUp(self): + """Set up async test fixtures""" + self.loop = asyncio.new_event_loop() + asyncio.set_event_loop(self.loop) + + def asyncTearDown(self): + """Clean up async test fixtures""" + self.loop.close() + + def async_test(f): + """Decorator for async test methods""" + + def wrapper(*args, **kwargs): + loop = asyncio.new_event_loop() + asyncio.set_event_loop(loop) + try: + return loop.run_until_complete(f(*args, **kwargs)) + finally: + loop.close() + + return wrapper + + @async_test + async def test_populate_queue_new_items(self): + """Test populating queue with new items""" + # Mock empty existing index + with patch("olmocr.work_queue.download_zstd_csv", return_value=[]): + with patch("olmocr.work_queue.upload_zstd_csv") as mock_upload: + await self.work_queue.populate_queue(self.sample_paths, items_per_group=2) + + # Verify upload was called with correct data + self.assertEqual(mock_upload.call_count, 1) + _, _, lines = mock_upload.call_args[0] + + # Should create 2 work groups (2 files + 1 file) + self.assertEqual(len(lines), 2) + + # Verify format of uploaded lines + for line in lines: + parts = WorkQueue._decode_csv_row(line) + self.assertGreaterEqual(len(parts), 2) # Hash + at least one path + self.assertEqual(len(parts[0]), 40) # SHA1 hash length + + @async_test + async def test_populate_queue_existing_items(self): + """Test populating queue with mix of new and existing items""" + existing_paths = ["s3://test-bucket/data/existing1.pdf"] + new_paths = ["s3://test-bucket/data/new1.pdf"] + + # Create existing index content + existing_hash = WorkQueue._compute_workgroup_hash(existing_paths) + existing_line = WorkQueue._encode_csv_row([existing_hash] + existing_paths) + + with patch("olmocr.work_queue.download_zstd_csv", return_value=[existing_line]): + with patch("olmocr.work_queue.upload_zstd_csv") as mock_upload: + await self.work_queue.populate_queue(existing_paths + new_paths, items_per_group=1) + + # Verify upload called with both existing and new items + _, _, lines = mock_upload.call_args[0] + self.assertEqual(len(lines), 2) + self.assertIn(existing_line, lines) + + @async_test + async def test_initialize_queue(self): + """Test queue initialization""" + # Mock work items and completed items + work_paths = ["s3://test/file1.pdf", "s3://test/file2.pdf"] + work_hash = WorkQueue._compute_workgroup_hash(work_paths) + work_line = WorkQueue._encode_csv_row([work_hash] + work_paths) + + completed_items = [f"s3://test-bucket/workspace/done_flags/done_{work_hash}.flag"] + + with patch("olmocr.work_queue.download_zstd_csv", return_value=[work_line]): + with patch("olmocr.work_queue.expand_s3_glob", return_value=completed_items): + count = await self.work_queue.initialize_queue() + + # Queue should be empty since all work is completed + self.assertEqual(count, 0) + + @async_test + async def test_is_completed(self): + """Test completed work check""" + work_hash = "testhash123" + + # Test completed work + self.s3_client.head_object.return_value = {"LastModified": datetime.datetime.now(datetime.timezone.utc)} + self.assertTrue(await self.backend.is_completed(work_hash)) + + # Test incomplete work + self.s3_client.head_object.side_effect = ClientError({"Error": {"Code": "404", "Message": "Not Found"}}, "HeadObject") + self.assertFalse(await self.backend.is_completed(work_hash)) + + @async_test + async def test_get_work(self): + """Test getting work items""" + # Setup test data + work_item = WorkItem(hash="testhash123", work_paths=["s3://test/file1.pdf"]) + await self.work_queue._queue.put(work_item) + + # Test getting available work + self.s3_client.head_object.side_effect = ClientError({"Error": {"Code": "404", "Message": "Not Found"}}, "HeadObject") + result = await self.work_queue.get_work() + self.assertEqual(result, work_item) + + # Verify lock file was created + self.s3_client.put_object.assert_called_once() + key = self.s3_client.put_object.call_args[1]["Key"] + self.assertTrue(key.endswith(f"worker_{work_item.hash}.lock")) + + @async_test + async def test_get_work_completed(self): + """Test getting work that's already completed""" + work_item = WorkItem(hash="testhash123", work_paths=["s3://test/file1.pdf"]) + await self.work_queue._queue.put(work_item) + + # Simulate completed work + self.s3_client.head_object.return_value = {"LastModified": datetime.datetime.now(datetime.timezone.utc)} + + result = await self.work_queue.get_work() + self.assertIsNone(result) # Should skip completed work + + @async_test + async def test_get_work_locked(self): + """Test getting work that's locked by another worker""" + work_item = WorkItem(hash="testhash123", work_paths=["s3://test/file1.pdf"]) + await self.work_queue._queue.put(work_item) + + # Simulate active lock + recent_time = datetime.datetime.now(datetime.timezone.utc) + self.s3_client.head_object.side_effect = [ + ClientError({"Error": {"Code": "404", "Message": "Not Found"}}, "HeadObject"), # Not completed + {"LastModified": recent_time}, # Active lock + ] + + result = await self.work_queue.get_work() + self.assertIsNone(result) # Should skip locked work + + @async_test + async def test_get_work_stale_lock(self): + """Test getting work with a stale lock""" + work_item = WorkItem(hash="testhash123", work_paths=["s3://test/file1.pdf"]) + await self.work_queue._queue.put(work_item) + + # Simulate stale lock + stale_time = datetime.datetime.now(datetime.timezone.utc) - datetime.timedelta(hours=1) + self.s3_client.head_object.side_effect = [ + ClientError({"Error": {"Code": "404", "Message": "Not Found"}}, "HeadObject"), # Not completed + {"LastModified": stale_time}, # Stale lock + ] + + result = await self.work_queue.get_work() + self.assertEqual(result, work_item) # Should take work with stale lock + + @async_test + async def test_mark_done(self): + """Test marking work as done""" + work_item = WorkItem(hash="testhash123", work_paths=["s3://test/file1.pdf"]) + await self.work_queue._queue.put(work_item) + + await self.work_queue.mark_done(work_item) + + # Verify done flag was created and lock file was deleted + # Check put_object was called for done flag + put_calls = self.s3_client.put_object.call_args_list + self.assertEqual(len(put_calls), 1) + done_flag_key = put_calls[0][1]["Key"] + self.assertTrue(done_flag_key.endswith(f"done_{work_item.hash}.flag")) + + # Verify lock file was deleted + self.s3_client.delete_object.assert_called_once() + key = self.s3_client.delete_object.call_args[1]["Key"] + self.assertTrue(key.endswith(f"worker_{work_item.hash}.lock")) + + @async_test + async def test_paths_with_commas(self): + """Test handling of paths that contain commas""" + # Create paths with commas in them + paths_with_commas = ["s3://test-bucket/data/file1,with,commas.pdf", "s3://test-bucket/data/file2,comma.pdf", "s3://test-bucket/data/file3.pdf"] + + # Mock empty existing index for initial population + with patch("olmocr.work_queue.download_zstd_csv", return_value=[]): + with patch("olmocr.work_queue.upload_zstd_csv") as mock_upload: + # Populate the queue with these paths + await self.work_queue.populate_queue(paths_with_commas, items_per_group=3) + + # Capture what would be written to the index + _, _, lines = mock_upload.call_args[0] + + # Now simulate reading back these lines (which have commas in the paths) + with patch("olmocr.work_queue.download_zstd_csv", return_value=lines): + with patch("olmocr.work_queue.expand_s3_glob", return_value=[]): + # Initialize a fresh queue from these lines + await self.work_queue.initialize_queue() + + # Mock ClientError for head_object (file doesn't exist) - need to handle multiple calls + self.s3_client.head_object.side_effect = [ + ClientError({"Error": {"Code": "404", "Message": "Not Found"}}, "HeadObject"), # done flag check + ClientError({"Error": {"Code": "404", "Message": "Not Found"}}, "HeadObject"), # worker lock check + ] + + # Get a work item + work_item = await self.work_queue.get_work() + + # Now verify we get a work item + self.assertIsNotNone(work_item, "Should get a work item") + + # Verify the work item has the correct number of paths + self.assertEqual(len(work_item.work_paths), len(paths_with_commas), "Work item should have the correct number of paths") + + # Check that all original paths with commas are preserved + for path in paths_with_commas: + print(path) + self.assertIn(path, work_item.work_paths, f"Path with commas should be preserved: {path}") + + def test_queue_size(self): + """Test queue size property""" + self.assertEqual(self.work_queue.size, 0) + + self.loop = asyncio.new_event_loop() + asyncio.set_event_loop(self.loop) + + self.loop.run_until_complete(self.work_queue._queue.put(WorkItem(hash="test1", work_paths=["path1"]))) + self.assertEqual(self.work_queue.size, 1) + + self.loop.run_until_complete(self.work_queue._queue.put(WorkItem(hash="test2", work_paths=["path2"]))) + self.assertEqual(self.work_queue.size, 2) + + self.loop.close() + + +if __name__ == "__main__": + unittest.main() diff --git a/tests/test_table_parsing.py b/tests/test_table_parsing.py new file mode 100644 index 0000000..5a08077 --- /dev/null +++ b/tests/test_table_parsing.py @@ -0,0 +1,973 @@ +import unittest + +from olmocr.bench.tests import ( + parse_html_tables, + parse_markdown_tables, +) + + +class TestParseHtmlTables(unittest.TestCase): + def test_basic_table(self): + data = parse_html_tables(""" + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
ArXiv Old
scans
math Tables Old
scans Headers
&
footers Multi
column Long
tiny
text Base Overall
Mistral OCR API 77.2 67.5 60.6 29.3 93.6 71.3 77.1 99.4 72.0±1.1
""")[0] + + print(data) + self.assertTrue(data.is_rectangular) + + self.assertEqual(data.cell_text[0, 0], "") + self.assertEqual(data.cell_text[0, 1], "ArXiv") + + self.assertEqual(data.left_relations[0, 0], set()) + self.assertEqual(data.up_relations[0, 0], set()) + + self.assertEqual(data.left_relations[0, 1], {(0, 0)}) + self.assertEqual(data.up_relations[1, 0], {(0, 0)}) + + self.assertEqual(data.heading_cells, {(0, 0), (0, 1), (0, 2), (0, 3), (0, 4), (0, 5), (0, 6), (0, 7), (0, 8), (0, 9)}) + + self.assertEqual(data.top_heading_relations(1, 3), {(0, 3)}) + + # If there are no left headings defined, then the left most column is considered the left heading + print(data.left_heading_relations(1, 3)) + self.assertEqual(data.left_heading_relations(1, 3), {(1, 0)}) + + def test_multiple_top_headings(self): + data = parse_html_tables(""" + + + + + + + + + + + + + + + + + + + +
Fruit Costs in Unittest land
Fruit Type Cost
Apples $1.00
Oranges $2.00
""")[0] + + print(data) + + self.assertEqual(data.cell_text[0, 0], "Fruit Costs in Unittest land") + self.assertEqual(data.cell_text[1, 0], "Fruit Type") + self.assertEqual(data.cell_text[1, 1], "Cost") + self.assertEqual(data.cell_text[2, 0], "Apples") + self.assertEqual(data.cell_text[2, 1], "$1.00") + self.assertEqual(data.cell_text[3, 0], "Oranges") + self.assertEqual(data.cell_text[3, 1], "$2.00") + + self.assertEqual(data.up_relations[1, 0], {(0, 0)}) + self.assertEqual(data.up_relations[1, 1], {(0, 0)}) + + self.assertEqual(data.up_relations[2, 0], {(1, 0)}) + self.assertEqual(data.up_relations[2, 1], {(1, 1)}) + + self.assertEqual(data.top_heading_relations(1, 0), {(0, 0)}) + self.assertEqual(data.top_heading_relations(1, 1), {(0, 0)}) + + self.assertEqual(data.top_heading_relations(2, 0), {(0, 0), (1, 0)}) + self.assertEqual(data.top_heading_relations(2, 1), {(0, 0), (1, 1)}) + + self.assertTrue(data.is_rectangular) + + def test_extra_colspans(self): + data = parse_html_tables(""" + + + + + + + + + + + + + + + + + + + + +
Fruit Costs in Unittest land
Fruit Type Cost
Apples $1.00 Hi
Oranges $2.00
""")[0] + + self.assertFalse(data.is_rectangular) + + def test_extra_rowspans(self): + data = parse_html_tables(""" + + + + + + + + + + + + + + + + + + + +
Fruit Costs in Unittest land
Fruit Type Cost
Apples $1.00
Oranges $2.00
""")[0] + + self.assertFalse(data.is_rectangular) + + def test_extra_rowspans_okay(self): + data = parse_html_tables(""" + + + + + + + + + + + + + + + + + + + + +
Fruit Costs in Unittest land
Fruit Type Cost
Apples $1.00
Oranges $2.00
a
""")[0] + + self.assertTrue(data.is_rectangular) + + def test_4x4_table_with_spans(self): + """Test a 4x4 table with various row spans and column spans""" + data = parse_html_tables(""" + + + + + + + + + + + + + + + + + + + + + + + +
Header 1 Header 2-3 Header 4
Cell A (spans 2 rows) Cell B Cell C Cell D (spans 2 rows)
Cell E-F (spans 2 cols)
Cell G Cell H-I-J (spans 3 cols)
""")[0] + + print(data) + self.assertTrue(data.is_rectangular) + + # Test header row + self.assertEqual(data.cell_text[0, 0], "Header 1") + self.assertEqual(data.cell_text[0, 1], "Header 2-3") + + self.assertNotIn((0, 2), data.cell_text) # colspan=2, so that next cell is empty + self.assertEqual(data.cell_text[0, 3], "Header 4") + + # Test first body row + self.assertEqual(data.cell_text[1, 0], "Cell A (spans 2 rows)") + self.assertEqual(data.cell_text[1, 1], "Cell B") + self.assertEqual(data.cell_text[1, 2], "Cell C") + self.assertEqual(data.cell_text[1, 3], "Cell D (spans 2 rows)") + + # Test second body row + self.assertNotIn((2, 0), data.cell_text) + self.assertEqual(data.cell_text[2, 1], "Cell E-F (spans 2 cols)") + + # Test third body row + self.assertEqual(data.cell_text[3, 0], "Cell G") + self.assertEqual(data.cell_text[3, 1], "Cell H-I-J (spans 3 cols)") + + # Test heading cells + self.assertEqual(data.heading_cells, {(0, 0), (0, 1), (0, 3)}) + + self.assertEqual(data.left_heading_relations(0, 0), set()) + self.assertEqual(data.left_heading_relations(1, 0), set()) + self.assertEqual(data.left_heading_relations(2, 0), set()) + self.assertEqual(data.left_heading_relations(3, 0), set()) + + self.assertEqual(data.top_heading_relations(0, 0), set()) + self.assertEqual(data.top_heading_relations(0, 1), set()) + self.assertEqual(data.top_heading_relations(0, 2), set()) + self.assertEqual(data.top_heading_relations(0, 3), set()) + + self.assertEqual(data.left_heading_relations(1, 1), {(1, 0)}) + self.assertEqual(data.left_heading_relations(1, 2), {(1, 0)}) + self.assertEqual(data.left_heading_relations(1, 3), {(1, 0)}) + + self.assertEqual(data.top_heading_relations(3, 1), {(0, 1), (0, 3)}) + + def test_complex_multi_level_headers(self): + """Test a table with multiple levels of headers and complex spanning""" + data = parse_html_tables(""" + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
Main Category
Sub Category A Sub Category B
A1 A2 A3 B1 B2 B3
Row 1 10 20 30 40 50 60
Row 2 15 25 35 45 55 65
""")[0] + + print("\n=== Complex Multi-Level Headers Test ===") + print(data) + + self.assertTrue(data.is_rectangular) + + # Test the three-level header structure + self.assertEqual(data.cell_text[0, 0], "") # Empty corner cell + self.assertEqual(data.cell_text[0, 1], "Main Category") + + self.assertEqual(data.cell_text[1, 1], "Sub Category A") + self.assertEqual(data.cell_text[1, 4], "Sub Category B") + + self.assertEqual(data.cell_text[2, 1], "A1") + self.assertEqual(data.cell_text[2, 2], "A2") + self.assertEqual(data.cell_text[2, 3], "A3") + self.assertEqual(data.cell_text[2, 4], "B1") + self.assertEqual(data.cell_text[2, 5], "B2") + self.assertEqual(data.cell_text[2, 6], "B3") + + # Test data rows + self.assertEqual(data.cell_text[3, 0], "Row 1") + self.assertEqual(data.cell_text[3, 1], "10") + self.assertEqual(data.cell_text[4, 0], "Row 2") + self.assertEqual(data.cell_text[4, 1], "15") + + # Test heading cells - all header rows should be marked as heading cells + expected_heading_cells = { + (0, 0), + (0, 1), # First header row + (1, 1), + (1, 4), # Second header row + (2, 1), + (2, 2), + (2, 3), + (2, 4), + (2, 5), + (2, 6), # Third header row + } + self.assertEqual(data.heading_cells, expected_heading_cells) + + # Test top heading relations for data cells + # Cell (3, 1) should have relations to all three levels of headers above it + top_relations_3_1 = data.top_heading_relations(3, 1) + self.assertIn((0, 1), top_relations_3_1) # Main Category + self.assertIn((1, 1), top_relations_3_1) # Sub Category A + self.assertIn((2, 1), top_relations_3_1) # A1 + + # Cell (3, 4) should relate to headers in the B column + top_relations_3_4 = data.top_heading_relations(3, 4) + self.assertIn((0, 1), top_relations_3_4) # Main Category + self.assertIn((1, 4), top_relations_3_4) # Sub Category B + self.assertIn((2, 4), top_relations_3_4) # B1 + + # Test left heading relations + self.assertEqual(data.left_heading_relations(3, 1), {(3, 0)}) # Row 1 + self.assertEqual(data.left_heading_relations(4, 1), {(4, 0)}) # Row 2 + + def test_left_headers_with_row_spans(self): + """Test a table with left-side headers that have row spans""" + data = parse_html_tables(""" + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
Quarter 1 Quarter 2
Jan Feb Mar Apr May Jun
North Sales 100 110 120 130 140 150
Cost 50 55 60 65 70 75
Profit 50 55 60 65 70 75
South Sales 200 210 220 230 240 250
Cost 100 105 110 115 120 125
Profit 100 105 110 115 120 125
""")[0] + + print("\n=== Left Headers with Row Spans Test ===") + print(data) + + self.assertTrue(data.is_rectangular) + + # Test top headers + self.assertEqual(data.cell_text[0, 2], "Quarter 1") + self.assertEqual(data.cell_text[0, 5], "Quarter 2") + self.assertEqual(data.cell_text[1, 2], "Jan") + self.assertEqual(data.cell_text[1, 3], "Feb") + self.assertEqual(data.cell_text[1, 4], "Mar") + + # Test left headers with row spans + self.assertEqual(data.cell_text[2, 0], "North") + self.assertEqual(data.cell_text[2, 1], "Sales") + self.assertEqual(data.cell_text[3, 1], "Cost") + self.assertEqual(data.cell_text[4, 1], "Profit") + + self.assertEqual(data.cell_text[5, 0], "South") + self.assertEqual(data.cell_text[5, 1], "Sales") + + # Test data values + self.assertEqual(data.cell_text[2, 2], "100") + self.assertEqual(data.cell_text[3, 2], "50") + self.assertEqual(data.cell_text[5, 2], "200") + + # Test heading cells - should include both top headers and left headers + # Top headers: rows 0-1, Left headers: column 0-1 in data rows + self.assertIn((0, 2), data.heading_cells) # Quarter 1 + self.assertIn((1, 2), data.heading_cells) # Jan + self.assertIn((2, 0), data.heading_cells) # North (left header) + self.assertIn((2, 1), data.heading_cells) # Sales (left header) + self.assertIn((5, 0), data.heading_cells) # South (left header) + + # Test left heading relations with multiple levels + # Data cell (2, 2) should have both North and Sales as left headers + left_relations_2_2 = data.left_heading_relations(2, 2) + self.assertIn((2, 0), left_relations_2_2) # North + self.assertIn((2, 1), left_relations_2_2) # Sales + + # Data cell (3, 2) should have North (spans from row 2) and Cost as left headers + left_relations_3_2 = data.left_heading_relations(3, 2) + self.assertIn((2, 0), left_relations_3_2) # North (row span) + self.assertIn((3, 1), left_relations_3_2) # Cost + + # Data cell (5, 3) should have South and Sales as left headers + left_relations_5_3 = data.left_heading_relations(5, 3) + self.assertIn((5, 0), left_relations_5_3) # South + self.assertIn((5, 1), left_relations_5_3) # Sales + + # Test top heading relations + top_relations_2_2 = data.top_heading_relations(2, 2) + self.assertIn((0, 2), top_relations_2_2) # Quarter 1 + self.assertIn((1, 2), top_relations_2_2) # Jan + + def test_nested_header_groups_with_col_spans(self): + """Test a complex table with nested header groups and various column spans""" + data = parse_html_tables(""" + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
Region Store 2024 Sales Data
First Half Second Half
Q1 Q2 Q3 Q4
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
West Store A 10 11 12 13 14 15 16 17 18 19 20 21
Store B 22 23 24 25 26 27 28 29 30 31 32 33
""")[0] + + print("\n=== Nested Header Groups with Col Spans Test ===") + print(data) + + self.assertTrue(data.is_rectangular) + + # Test nested header structure + self.assertEqual(data.cell_text[0, 0], "Region") + self.assertEqual(data.cell_text[0, 1], "Store") + self.assertEqual(data.cell_text[0, 2], "2024 Sales Data") + + self.assertEqual(data.cell_text[1, 2], "First Half") + self.assertEqual(data.cell_text[1, 8], "Second Half") + + self.assertEqual(data.cell_text[2, 2], "Q1") + self.assertEqual(data.cell_text[2, 5], "Q2") + self.assertEqual(data.cell_text[2, 8], "Q3") + self.assertEqual(data.cell_text[2, 11], "Q4") + + self.assertEqual(data.cell_text[3, 2], "Jan") + self.assertEqual(data.cell_text[3, 7], "Jun") + self.assertEqual(data.cell_text[3, 13], "Dec") + + # Test data rows + self.assertEqual(data.cell_text[4, 0], "West") + self.assertEqual(data.cell_text[4, 1], "Store A") + self.assertEqual(data.cell_text[4, 2], "10") + self.assertEqual(data.cell_text[5, 1], "Store B") + self.assertEqual(data.cell_text[5, 2], "22") + + # Test all header cells are marked + self.assertIn((0, 0), data.heading_cells) # Region + self.assertIn((0, 1), data.heading_cells) # Store + self.assertIn((0, 2), data.heading_cells) # 2024 Sales Data + self.assertIn((1, 2), data.heading_cells) # First Half + self.assertIn((2, 2), data.heading_cells) # Q1 + self.assertIn((3, 2), data.heading_cells) # Jan + + # Test multiple top heading relations for a data cell + # Cell (4, 2) - January data for Store A should have all 4 levels of headers + top_relations_4_2 = data.top_heading_relations(4, 2) + self.assertIn((0, 2), top_relations_4_2) # 2024 Sales Data + self.assertIn((1, 2), top_relations_4_2) # First Half + self.assertIn((2, 2), top_relations_4_2) # Q1 + self.assertIn((3, 2), top_relations_4_2) # Jan + + # Cell (4, 7) - June data should relate to Q2 and First Half + top_relations_4_7 = data.top_heading_relations(4, 7) + self.assertIn((0, 2), top_relations_4_7) # 2024 Sales Data + self.assertIn((1, 2), top_relations_4_7) # First Half + self.assertIn((2, 5), top_relations_4_7) # Q2 + self.assertIn((3, 7), top_relations_4_7) # Jun + + # Cell (4, 13) - December data should relate to Q4 and Second Half + top_relations_4_13 = data.top_heading_relations(4, 13) + self.assertIn((0, 2), top_relations_4_13) # 2024 Sales Data + self.assertIn((1, 8), top_relations_4_13) # Second Half + self.assertIn((2, 11), top_relations_4_13) # Q4 + self.assertIn((3, 13), top_relations_4_13) # Dec + + # Test left heading relations + # Store B row that says "22" should relate to just West + left_relations_5_2 = data.left_heading_relations(5, 2) + self.assertEqual(data.cell_text[5, 2], "22") + self.assertIn((4, 0), left_relations_5_2) + self.assertEqual(len(left_relations_5_2), 1) + + # But of the left headings themselves at the top, January, should have both Region and Store + self.assertEqual(data.left_heading_relations(3, 2), {(0, 0), (0, 1)}) + + def test_large_real_table(self): + html = """ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
Grupo Ocupacional Classe Ocupacional Superior Médio Baixo Interior Ínfimo Total
N (1.000s) % N (1.000s) % N (1.000s) % N (1.000s) % N (1.000s) % N (1.000s) %
Empregadores A-1 Empregadores (> 10) 608 67,3 185 20,4 86 9,6 16 1,8 8 0,9 903 100
A-2 Empregadores (<= 10) 1.555 36,9 1.107 26,3 1.036 24,7 341 8,1 171 4,1 4.213 100
Total 2.162 42,3 1.292 25,3 1.126 22,0 357 7,0 179 3,5 5.116 100
Profissionais C Profissionais Autônomos 1.643 21,7 1.513 20,0 2.073 27,4 1.225 16,2 1.108 14,7 7.562 100
D Profissionais Assalariados 4.438 13,3 6.030 18,0 11.550 34,5 7.027 21,0 4.389 13,1 33.434 100
Total 6.081 14,8 7.543 18,4 13.623 33,2 8.252 20,1 5.497 13,4 40.995 100
Massa Não-Agrícola F Trabalhadores Autônomos 657 3,5 1.754 9,2 5.561 29,2 5.271 27,7 5.788 30,4 19.030 100
G Trabalhadores Assalariados 282 0,7 1.657 4,3 10.363 27,1 13.002 34,0 12.968 33,9 38.272 100
I Trabalhadores Domésticos 10 0,1 104 1,6 977 14,7 1.810 27,3 3.733 56,3 6.633 100
Total 948 1,5 3.515 5,5 16.901 26,4 20.083 31,4 22.489 35,2 63.936 100
Massa Agrícola H-1 Proprietários Conta Própria 188 2,0 364 3,8 1.387 14,4 1.889 19,7 5.779 60,2 9.608 100
H-2 Trabalhadores Autônomos 5 0,5 14 1,5 72 7,6 152 16,1 703 74,3 946 100
H-3 Trabalhadores Assalariados 17 0,2 58 0,6 794 8,4 2.260 23,9 6.322 66,9 9.451 100
Total 210 1,0 436 2,2 2.253 11,3 4.301 21,5 12.805 64,0 20.005 100
Não-remunerados Não-remunerados Não-Agrícolas 13 6,8 16 8,1 28 14,0 22 10,9 119 60,2 198 100
Não-remunerados Agrícolas 5 0,1 13 0,3 59 1,6 352 9,4 3.302 88,5 3.731 100
Sem Ocupação Com Renda 1.567 6,0 2.330 8,9 5.395 20,7 6.821 26,2 9.964 38,2 26.078 100
Sem Ocupação Sem Renda 8.094 100 8.094 100
Ignorados 177 10,3 202 11,8 364 21,1 337 19,6 640 37,2 1.720 100
+""" + + data = parse_html_tables(html)[0] + + self.assertTrue(data.is_rectangular) + + def test_non_rect(self): + html = """\n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n
Model Avg FLOP×10²³ Dev Benchmarks Held-out Evals
MMLU ARC_C HSwag WinoG NQ DROP AGIEval GSM8K MMLU_PRO
Open-weight models
Llama 2 13B 51.0 1.6 55.7 67.3 83.9 74.9 38.4 45.6 41.5 28.1 23.9
Mistral 7B 56.6 n/a 63.5 78.3 83.1 77.7 37.2 51.8 47.3 40.1 30.0
Llama 3.1 8B 59.7 7.2 66.9 79.5 81.6 76.6 33.9 56.4 51.3 56.5 34.7
Mistral Nemo 12B 64.9 n/a 69.5 85.2 85.6 81.5 39.7 69.2 54.7 62.1 36.7
Gemma 2 9B 66.3 4.4 70.6 89.5 87.3 78.8 38.0 63.0 57.3 70.1 42.0
Qwen 2.5 7B 67.2 8.2 74.4 89.5 89.7 74.2 29.9 55.8 63.7 81.5 45.8
Qwen 2.5 14B 71.5 16.0 79.3 94.0 94.0 80.0 37.3 51.5 71.0 83.4 52.8
Models with partially available data
StableLM 2 12B 60.2 2.9 62.4 81.9 84.5 77.7 37.6 55.5 50.9 62.0 29.3
Zamba 2 7B 63.7 n/c 68.5 92.2 89.4 79.6 36.5 51.7 55.5 67.2 32.8
Fully-open models
Amber 7B 32.5 0.5 24.7 44.9 74.5 65.5 18.7 26.1 21.8 4.8 11.7
OLMo 7B 35.4 1.0 28.3 46.4 78.1 68.5 24.8 27.3 23.7 9.2 12.1
MAP Neo 7B 47.9 2.1 58.0 78.4 72.8 69.2 28.9 39.4 45.8 12.5 25.9
OLMo 0424 7B 49.8 1.0 54.3 66.9 80.1 73.6 29.6 50.0 43.9 27.7 22.1
DCLM 7B 55.2 1.0 64.4 79.8 82.3 77.3 28.8 39.3 47.5 46.1 31.3
OLMo 2 7B 61.2 1.8 63.7 79.8 83.8 77.2 36.9 60.8 50.4 67.5 31.0
OLMo 2 13B 66.8 4.6 67.5 83.5 86.4 81.5 46.7 70.7 54.2 75.1 35.1
""" + + data = parse_html_tables(html)[0] + + self.assertFalse(data.is_rectangular) + + def test_two_non_rect(self): + html = """## GAME-BY-GAME LEADERS \u00bb\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
Date Opponent Points Rebounds Assists Steals Blocks
Nov. 11 Fort Wayne Hannahs (23) Kingsley (9) Macon (5) Macon (4) Kingsley (7)
Nov. 14 Southern Illinois Hannahs (19) Kingsley (7) Beard (6) Seven Players (1) Kingsley (3)
Nov. 18 UT Arlington Barford (17) Three Players (6) Four Players (2) Barford (3) Kingsley (4)
Nov. 22 at Minnesota Hannahs (20) Three Players (5) Three Players (3) Watkins (3) Kingsley (3)
Nov. 28 Mount Saint Mary's Barford (16) Kingsley (11) Hannahs (4) Hannahs (2) Kingsley (4)
Dec. 1 Stephen F. Austin Hannahs (15) Watkins (7) Macon (5) Beard (4) Jones (1)
Dec. 3 Austin Peay Macon (14) Two Players (7) Kingsley (5) Two Players (2) Two Players (3)
Dec. 6 Houston Two Players (17) Kingsley (9) Watkins (4) Beard (2) Two Players (2)
Dec. 10 North Florida -- -- -- -- --
Dec. 17 Texas* -- -- -- -- --
Dec. 20 North Dakota State -- -- -- -- --
Dec. 22 Sam Houston State# -- -- -- -- --
Dec. 29 Florida* -- -- -- -- --
Jan. 3 at Tennessee* -- -- -- -- --
Jan. 7 at Kentucky* -- -- -- -- --
Jan. 10 Mississippi State* -- -- -- -- --
Jan. 14 Missouri* -- -- -- -- --
Jan. 17 at Texas A&M* -- -- -- -- --
Jan. 21 LSU* -- -- -- -- --
Jan. 24 at Vanderbilt* -- -- -- -- --
Jan. 28 at Oklahoma State% -- -- -- -- --
Feb. 1 Alabama* -- -- -- -- --
Feb. 4 at Missouri* -- -- -- -- --
Feb. 7 Vanderbilt* -- -- -- -- --
Feb. 11 at LSU* -- -- -- -- --
Feb. 15 at South Carolina* -- -- -- -- --
Feb. 18 Ole Miss* -- -- -- -- --
Feb. 22 Texas A&M* -- -- -- -- --
Feb. 25 at Auburn* -- -- -- -- --
March 1 at Florida* -- -- -- -- --
March 4 Georgia* -- -- -- -- --
\n\n^* Lone Star Shootout (Houston, Texas) | # at Verizon Arena (North Little Rock, Ark.) | % Big12/SEC Challenge | * SEC Game\n\n## MISCELLANEOUS TEAM STATS \u00bb\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
Date Opponent W/L Points In Paint Pts Off Turnovers 2nd Chance Pts Fast Break Pts
Ark. Opp. Diff. Ark. Opp. Diff. Ark. Opp. Diff.
Nov. 11 Fort Wayne W 52 22 +30 10 12 +4 9 21 -12
Nov. 14 Southern Illinois W 30 28 +2 25 17 +4 9 10 -1
Nov. 18 UT Arlington W 42 28 +14 8 20 -12 11 8 +3
Nov. 22 at Minnesota L 46 38 +8 11 26 -15 11 13 -2
Nov. 28 Mount Saint Mary's W 48 34 +14 18 10 +8 13 6 +7
Dec. 1 Stephen F. Austin W 36 28 +8 29 24 +5 12 9 +3
Dec. 3 Austin Peay W 52 40 +12 25 8 +17 26 10 +16
Dec. 6 Houston W 38 38 0 15 19 -4 13 9 +4
Dec. 10 North Florida -- -- -- -- -- -- -- -- -- --
Dec. 17 Texas* -- -- -- -- -- -- -- -- -- --
Dec. 20 North Dakota State -- -- -- -- -- -- -- -- -- --
Dec. 22 Sam Houston State# -- -- -- -- -- -- -- -- -- --
Dec. 29 Florida* -- -- -- -- -- -- -- -- -- --
Jan. 3 at Tennessee* -- -- -- -- -- -- -- -- -- --
Jan. 7 at Kentucky* -- -- -- -- -- -- -- -- -- --
Jan. 10 Mississippi State* -- -- -- -- -- -- -- -- -- --
Jan. 14 Missouri* -- -- -- -- -- -- -- -- -- --
Jan. 17 at Texas A&M* -- -- -- -- -- -- -- -- -- --
Jan. 21 LSU* -- -- -- -- -- -- -- -- -- --
Jan. 24 at Vanderbilt* -- -- -- -- -- -- -- -- -- --
Jan. 28 at Oklahoma State% -- -- -- -- -- -- -- -- -- --
Feb. 1 Alabama* -- -- -- -- -- -- -- -- -- --
Feb. 4 at Missouri* -- -- -- -- -- -- -- -- -- --
Feb. 7 Vanderbilt* -- -- -- -- -- -- -- -- -- --
Feb. 11 at LSU* -- -- -- -- -- -- -- -- -- --
Feb. 15 at South Carolina* -- -- -- -- -- -- -- -- -- --
Feb. 18 Ole Miss* -- -- -- -- -- -- -- -- -- --
Feb. 22 Texas A&M* -- -- -- -- -- -- -- -- -- --
Feb. 25 at Auburn* -- -- -- -- -- -- -- -- -- --
March 1 at Florida* -- -- -- -- -- -- -- -- -- --
March 4 Georgia* -- -- -- -- -- -- -- -- -- --
TOTALS 7-1 326 210 +78 163 126 +37 104 96 +8
138 46 +92
\n\n^* Lone Star Shootout (Houston, Texas) | # at Verizon Arena (North Little Rock, Ark.) | % Big12/SEC Challenge | * SEC Game\n\n**The #Fastest40 Just Got Faster** \u2014 **13**""" + + data = parse_html_tables(html)[0] + self.assertTrue(data.is_rectangular) + + data = parse_html_tables(html)[1] + self.assertFalse(data.is_rectangular) diff --git a/tests/test_tests.py b/tests/test_tests.py new file mode 100644 index 0000000..5217eaa --- /dev/null +++ b/tests/test_tests.py @@ -0,0 +1,2354 @@ +import unittest + +from olmocr.bench.tests import ( + BaselineTest, + BasePDFTest, + FootnoteTest, + FormatTest, + MathTest, + TableTest, + TestChecked, + TestType, + TextOrderTest, + TextPresenceTest, + ValidationError, + normalize_text, + parse_html_tables, + parse_markdown_tables, +) + + +class TestNormalizeText(unittest.TestCase): + """Test the normalize_text function""" + + def test_whitespace_normalization(self): + """Test that whitespace is properly normalized""" + input_text = "This has\tmultiple spaces\nand\nnewlines" + expected = "This has multiple spaces and newlines" + self.assertEqual(normalize_text(input_text), expected) + + def test_character_replacement(self): + """Test that fancy characters are replaced with ASCII equivalents""" + input_text = "This has 'fancy' “quotes” and—dashes" + expected = "This has 'fancy' \"quotes\" and-dashes" + self.assertEqual(normalize_text(input_text), expected) + + def test_markdown1(self): + """Test that fancy characters are replaced with ASCII equivalents""" + input_text = "this is *bold*" + expected = "this is bold" + self.assertEqual(normalize_text(input_text), expected) + + def test_markdown2(self): + """Test that fancy characters are replaced with ASCII equivalents""" + input_text = "_italic__ is *bold*" + expected = "italic_ is bold" + self.assertEqual(normalize_text(input_text), expected) + + def test_empty_input(self): + """Test that empty input returns empty output""" + self.assertEqual(normalize_text(""), "") + + def test_brs(self): + """Test that empty input returns empty output""" + self.assertEqual(normalize_text("Hello
everyone"), "Hello everyone") + self.assertEqual(normalize_text("Hello
everyone"), normalize_text("Hello\neveryone")) + self.assertEqual(normalize_text("Hello
everyone"), "Hello everyone") + self.assertEqual(normalize_text("Hello
everyone"), normalize_text("Hello\neveryone")) + + def test_two_stars(self): + self.assertEqual( + normalize_text( + "**Georges V.** (2007) – *Le Forez du VIe au IVe millénaire av. J.-C. Territoires, identités et stratégies des sociétés humaines du Massif central dans le bassin amont de la Loire (France)*, thèse de doctorat, université de Bourgogne, Dijon, 2 vol., 435 p." + ), + "Georges V. (2007) - Le Forez du VIe au IVe millénaire av. J.-C. Territoires, identités et stratégies des sociétés humaines du Massif central dans le bassin amont de la Loire (France), thèse de doctorat, université de Bourgogne, Dijon, 2 vol., 435 p.", + ) + + +class TestBasePDFTest(unittest.TestCase): + """Test the BasePDFTest class""" + + def test_valid_initialization(self): + """Test that a valid initialization works""" + test = BasePDFTest(pdf="test.pdf", page=1, id="test_id", type=TestType.BASELINE.value) + self.assertEqual(test.pdf, "test.pdf") + self.assertEqual(test.page, 1) + self.assertEqual(test.id, "test_id") + self.assertEqual(test.type, TestType.BASELINE.value) + self.assertEqual(test.max_diffs, 0) + self.assertIsNone(test.checked) + self.assertIsNone(test.url) + + def test_empty_pdf(self): + """Test that empty PDF raises ValidationError""" + with self.assertRaises(ValidationError): + BasePDFTest(pdf="", page=1, id="test_id", type=TestType.BASELINE.value) + + def test_empty_id(self): + """Test that empty ID raises ValidationError""" + with self.assertRaises(ValidationError): + BasePDFTest(pdf="test.pdf", page=1, id="", type=TestType.BASELINE.value) + + def test_negative_max_diffs(self): + """Test that negative max_diffs raises ValidationError""" + with self.assertRaises(ValidationError): + BasePDFTest(pdf="test.pdf", page=1, id="test_id", type=TestType.BASELINE.value, max_diffs=-1) + + def test_invalid_test_type(self): + """Test that invalid test type raises ValidationError""" + with self.assertRaises(ValidationError): + BasePDFTest(pdf="test.pdf", page=1, id="test_id", type="invalid_type") + + def test_run_method_not_implemented(self): + """Test that run method raises NotImplementedError""" + test = BasePDFTest(pdf="test.pdf", page=1, id="test_id", type=TestType.BASELINE.value) + with self.assertRaises(NotImplementedError): + test.run("content") + + def test_checked_enum(self): + """Test that checked accepts valid TestChecked enums""" + test = BasePDFTest(pdf="test.pdf", page=1, id="test_id", type=TestType.BASELINE.value, checked=TestChecked.VERIFIED) + self.assertEqual(test.checked, TestChecked.VERIFIED) + + +class TestTextPresenceTest(unittest.TestCase): + """Test the TextPresenceTest class""" + + def test_valid_present_test(self): + """Test that a valid PRESENT test initializes correctly""" + test = TextPresenceTest(pdf="test.pdf", page=1, id="test_id", type=TestType.PRESENT.value, text="test text") + self.assertEqual(test.text, "test text") + self.assertTrue(test.case_sensitive) + self.assertIsNone(test.first_n) + self.assertIsNone(test.last_n) + + def test_valid_absent_test(self): + """Test that a valid ABSENT test initializes correctly""" + test = TextPresenceTest(pdf="test.pdf", page=1, id="test_id", type=TestType.ABSENT.value, text="test text", case_sensitive=False) + self.assertEqual(test.text, "test text") + self.assertFalse(test.case_sensitive) + + def test_empty_text(self): + """Test that empty text raises ValidationError""" + with self.assertRaises(ValidationError): + TextPresenceTest(pdf="test.pdf", page=1, id="test_id", type=TestType.PRESENT.value, text="") + + def test_present_text_exact_match(self): + """Test that PRESENT test returns True for exact match""" + test = TextPresenceTest(pdf="test.pdf", page=1, id="test_id", type=TestType.PRESENT.value, text="target text") + result, _ = test.run("This is some target text in a document") + self.assertTrue(result) + + def test_present_text_not_found(self): + """Test that PRESENT test returns False when text not found""" + test = TextPresenceTest(pdf="test.pdf", page=1, id="test_id", type=TestType.PRESENT.value, text="missing text") + result, explanation = test.run("This document doesn't have the target") + self.assertFalse(result) + self.assertIn("missing text", explanation) + + def test_present_text_with_max_diffs(self): + """Test that PRESENT test with max_diffs handles fuzzy matching""" + test = TextPresenceTest(pdf="test.pdf", page=1, id="test_id", type=TestType.PRESENT.value, text="target text", max_diffs=2) + result, _ = test.run("This is some targett textt in a document") + self.assertTrue(result) + + def test_absent_text_found(self): + """Test that ABSENT test returns False when text is found""" + test = TextPresenceTest(pdf="test.pdf", page=1, id="test_id", type=TestType.ABSENT.value, text="target text") + result, explanation = test.run("This is some target text in a document") + self.assertFalse(result) + self.assertIn("target text", explanation) + + def test_absent_text_found_diffs(self): + """Test that ABSENT test returns False when text is found""" + test = TextPresenceTest(pdf="test.pdf", page=1, id="test_id", type=TestType.ABSENT.value, text="target text", max_diffs=2) + result, explanation = test.run("This is some target text in a document") + self.assertFalse(result) + result, explanation = test.run("This is some targett text in a document") + self.assertFalse(result) + result, explanation = test.run("This is some targettt text in a document") + self.assertFalse(result) + result, explanation = test.run("This is some targetttt text in a document") + self.assertTrue(result) + + def test_absent_test_subsets(self): + """Test that ABSENT test returns False when text is found""" + test = TextPresenceTest(pdf="test.pdf", page=1, id="test_id", type=TestType.ABSENT.value, text="max", max_diffs=0) + result, explanation = test.run("There is a max set of diffs to make") + self.assertFalse(result) + result, explanation = test.run("This is some maximum text in a document") + self.assertFalse(result) + + def test_absent_text_not_found(self): + """Test that ABSENT test returns True when text is not found""" + test = TextPresenceTest(pdf="test.pdf", page=1, id="test_id", type=TestType.ABSENT.value, text="missing text") + result, _ = test.run("This document doesn't have the target") + self.assertTrue(result) + + def test_case_insensitive_present(self): + """Test that case_sensitive=False works for PRESENT test""" + test = TextPresenceTest(pdf="test.pdf", page=1, id="test_id", type=TestType.PRESENT.value, text="TARGET TEXT", case_sensitive=False) + result, _ = test.run("This is some target text in a document") + self.assertTrue(result) + + def test_case_insensitive_absent(self): + """Test that case_sensitive=False works for ABSENT test""" + test = TextPresenceTest(pdf="test.pdf", page=1, id="test_id", type=TestType.ABSENT.value, text="TARGET TEXT", case_sensitive=False) + result, explanation = test.run("This is some target text in a document") + self.assertFalse(result) + + def test_first_n_limit(self): + """Test that first_n parameter works correctly""" + test = TextPresenceTest(pdf="test.pdf", page=1, id="test_id", type=TestType.PRESENT.value, text="beginning", first_n=20) + result, _ = test.run("beginning of text, but not the end") + self.assertTrue(result) + + # Test that text beyond first_n isn't matched + test = TextPresenceTest(pdf="test.pdf", page=1, id="test_id", type=TestType.PRESENT.value, text="end", first_n=20) + result, _ = test.run("beginning of text, but not the end") + self.assertFalse(result) + + def test_last_n_limit(self): + """Test that last_n parameter works correctly""" + test = TextPresenceTest(pdf="test.pdf", page=1, id="test_id", type=TestType.PRESENT.value, text="end", last_n=20) + result, _ = test.run("beginning of text, but not the end") + self.assertTrue(result) + + # Test that text beyond last_n isn't matched + test = TextPresenceTest(pdf="test.pdf", page=1, id="test_id", type=TestType.PRESENT.value, text="beginning", last_n=20) + result, _ = test.run("beginning of text, but not the end") + self.assertFalse(result) + + def test_both_first_and_last_n(self): + """Test that combining first_n and last_n works correctly""" + test = TextPresenceTest(pdf="test.pdf", page=1, id="test_id", type=TestType.PRESENT.value, text="beginning", first_n=15, last_n=10) + result, _ = test.run("beginning of text, middle part, but not the end") + self.assertTrue(result) + + # Text only in middle shouldn't be found + test = TextPresenceTest(pdf="test.pdf", page=1, id="test_id", type=TestType.PRESENT.value, text="middle", first_n=15, last_n=10) + result, _ = test.run("beginning of text, middle part, but not the end") + self.assertFalse(result) + + def test_unicode_normalized_forms(self): + """Test that e+accent == e_with_accent unicode chars""" + test = TextPresenceTest(pdf="test.pdf", page=1, id="test_id", type=TestType.PRESENT.value, text="I like to eat at a caf\u00e9") + result, _ = test.run("I like to eat at a caf\u00e9") + self.assertTrue(result) + + result, _ = test.run("I like to eat at a cafe\u0301") + self.assertTrue(result) + + +class TestTextOrderTest(unittest.TestCase): + """Test the TextOrderTest class""" + + def test_valid_initialization(self): + """Test that valid initialization works""" + test = TextOrderTest(pdf="test.pdf", page=1, id="test_id", type=TestType.ORDER.value, before="first text", after="second text") + self.assertEqual(test.before, "first text") + self.assertEqual(test.after, "second text") + + def test_invalid_test_type(self): + """Test that invalid test type raises ValidationError""" + with self.assertRaises(ValidationError): + TextOrderTest(pdf="test.pdf", page=1, id="test_id", type=TestType.PRESENT.value, before="first text", after="second text") + + def test_empty_before(self): + """Test that empty before text raises ValidationError""" + with self.assertRaises(ValidationError): + TextOrderTest(pdf="test.pdf", page=1, id="test_id", type=TestType.ORDER.value, before="", after="second text") + + def test_empty_after(self): + """Test that empty after text raises ValidationError""" + with self.assertRaises(ValidationError): + TextOrderTest(pdf="test.pdf", page=1, id="test_id", type=TestType.ORDER.value, before="first text", after="") + + def test_correct_order(self): + """Test that correct order returns True""" + test = TextOrderTest(pdf="test.pdf", page=1, id="test_id", type=TestType.ORDER.value, before="first", after="second") + result, _ = test.run("This has first and then second in correct order") + self.assertTrue(result) + + def test_incorrect_order(self): + """Test that incorrect order returns False""" + test = TextOrderTest(pdf="test.pdf", page=1, id="test_id", type=TestType.ORDER.value, before="second", after="first") + result, explanation = test.run("This has first and then second in correct order") + self.assertFalse(result) + + def test_before_not_found(self): + """Test that 'before' text not found returns False""" + test = TextOrderTest(pdf="test.pdf", page=1, id="test_id", type=TestType.ORDER.value, before="missing", after="present") + result, explanation = test.run("This text has present but not the other word") + self.assertFalse(result) + + def test_after_not_found(self): + """Test that 'after' text not found returns False""" + test = TextOrderTest(pdf="test.pdf", page=1, id="test_id", type=TestType.ORDER.value, before="present", after="missing") + result, explanation = test.run("This text has present but not the other word") + self.assertFalse(result) + + def test_max_diffs(self): + """Test that max_diffs parameter works correctly""" + test = TextOrderTest(pdf="test.pdf", page=1, id="test_id", type=TestType.ORDER.value, before="first", after="second", max_diffs=1) + result, _ = test.run("This has firsst and then secand in correct order") + self.assertTrue(result) + + def test_multiple_occurrences(self): + """Test that multiple occurrences are handled correctly""" + test = TextOrderTest(pdf="test.pdf", page=1, id="test_id", type=TestType.ORDER.value, before="target", after="target") + result, _ = test.run("This has target and then target again") + self.assertTrue(result) + + # Test reverse direction fails + test = TextOrderTest(pdf="test.pdf", page=1, id="test_id", type=TestType.ORDER.value, before="B", after="A") + result, _ = test.run("A B A B") # A comes before B, but B also comes before second A + self.assertTrue(result) + + def test_real_case_unicode(self): + text = """# Marketing Strategies in the Canadian Beef Sector + +Julieta Frank, + +Derek Brewin, + +and + +Maria José Patiño* + +*Paper presented at the NCCC-134 Conference on Applied Commodity Price Analysis, +Forecasting, and Market Risk Management +St. Louis, Missouri, April 18-19, 2011* + +*Copyright 2011 by Julieta Frank, Derek Brewin, and Maria José Patiño. All rights reserved. +Readers may make verbatim copies of this document for non-commercial purposes by any +means, provided that this copyright notice appears on all such copies.* + +* Julieta Frank (Julieta_Frank@umanitoba.ca) is an Assistant Professor, Derek Brewin is an Associate Professor, and Maria José Patiño is a graduate student in the Department of Agribusiness and Agricultural Economics at the University of Manitoba. The funding support of the Manitoba Rural Adaptation Council and the Solomon Sinclair Farm Management Institute are gratefully acknowledged.""" + + test = TextOrderTest( + pdf="test.pdf", page=1, id="test_id", type=TestType.ORDER.value, before="Maria Jos\u00e9 Pati\u00f1o*", after="All rights reserved.", max_diffs=0 + ) + result, _ = test.run(text) + self.assertTrue(result) + + +class TestTableTest(unittest.TestCase): + """Test the TableTest class""" + + def setUp(self): + """Set up test fixtures""" + self.markdown_table = """ +| Header 1 | Header 2 | Header 3 | +| -------- | -------- | -------- | +| Cell A1 | Cell A2 | Cell A3 | +| Cell B1 | Cell B2 | Cell B3 | +""" + + self.html_table = """ + + + + + + + + + + + + + + + + +
Header 1 Header 2 Header 3
Cell A1 Cell A2 Cell A3
Cell B1 Cell B2 Cell B3
+""" + + def test_valid_initialization(self): + """Test that valid initialization works""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="target cell") + self.assertEqual(test.cell, "target cell") + self.assertEqual(test.up, "") + self.assertEqual(test.down, "") + self.assertEqual(test.left, "") + self.assertEqual(test.right, "") + self.assertEqual(test.top_heading, "") + self.assertEqual(test.left_heading, "") + + def test_invalid_test_type(self): + """Test that invalid test type raises ValidationError""" + with self.assertRaises(ValidationError): + TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.PRESENT.value, cell="target cell") + + def test_parse_markdown_tables(self): + """Test markdown table parsing""" + _test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2") + tables = parse_markdown_tables(self.markdown_table) + self.assertEqual(len(tables), 1) + self.assertEqual(tables[0].cell_text[0, 0], "Header 1") + self.assertEqual(tables[0].cell_text[1, 1], "Cell A2") + self.assertEqual(tables[0].cell_text[2, 2], "Cell B3") + + def test_parse_html_tables(self): + """Test HTML table parsing""" + _test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2") + tables = parse_html_tables(self.html_table) + self.assertEqual(len(tables), 1) + self.assertEqual(tables[0].cell_text[0, 0], "Header 1") + self.assertEqual(tables[0].cell_text[1, 1], "Cell A2") + self.assertEqual(tables[0].cell_text[2, 2], "Cell B3") + + def test_match_cell(self): + """Test finding a cell in a table""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2") + result, _ = test.run(self.markdown_table) + self.assertTrue(result) + + def test_cell_not_found(self): + """Test cell not found in table""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Missing Cell") + result, explanation = test.run(self.markdown_table) + self.assertFalse(result) + self.assertIn("No cell matching", explanation) + + def test_up_relationship(self): + """Test up relationship in table""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2", up="Header 2") + result, _ = test.run(self.markdown_table) + self.assertTrue(result) + + # Test incorrect up relationship + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2", up="Wrong Header") + result, explanation = test.run(self.markdown_table) + self.assertFalse(result) + self.assertIn("doesn't match expected", explanation) + + def test_down_relationship(self): + """Test down relationship in table""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2", down="Cell B2") + result, _ = test.run(self.markdown_table) + self.assertTrue(result) + + # Test incorrect down relationship + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2", down="Wrong Cell") + result, explanation = test.run(self.markdown_table) + self.assertFalse(result) + self.assertIn("doesn't match expected", explanation) + + def test_left_relationship(self): + """Test left relationship in table""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2", left="Cell A1") + result, _ = test.run(self.markdown_table) + self.assertTrue(result) + + # Test incorrect left relationship + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2", left="Wrong Cell") + result, explanation = test.run(self.markdown_table) + self.assertFalse(result) + self.assertIn("doesn't match expected", explanation) + + def test_right_relationship(self): + """Test right relationship in table""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2", right="Cell A3") + result, _ = test.run(self.markdown_table) + self.assertTrue(result) + + # Test incorrect right relationship + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2", right="Wrong Cell") + result, explanation = test.run(self.markdown_table) + self.assertFalse(result) + self.assertIn("doesn't match expected", explanation) + + def test_top_heading_relationship(self): + """Test top_heading relationship in table""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell B2", top_heading="Header 2") + result, _ = test.run(self.markdown_table) + self.assertTrue(result) + + # Test incorrect top_heading relationship + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell B2", top_heading="Wrong Header") + result, explanation = test.run(self.markdown_table) + self.assertFalse(result) + self.assertIn("doesn't match expected", explanation) + + def test_left_heading_relationship(self): + """Test left_heading relationship in table""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A3", left_heading="Cell A1") + result, _ = test.run(self.markdown_table) + self.assertTrue(result) + + # Test incorrect left_heading relationship + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A3", left_heading="Wrong Cell") + result, explanation = test.run(self.markdown_table) + self.assertFalse(result) + self.assertIn("doesn't match expected", explanation) + + def test_multiple_relationships(self): + """Test multiple relationships in table""" + test = TableTest( + pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2", up="Header 2", down="Cell B2", left="Cell A1", right="Cell A3" + ) + result, _ = test.run(self.markdown_table) + self.assertTrue(result) + + # Test one incorrect relationship + test = TableTest( + pdf="test.pdf", + page=1, + id="test_id", + type=TestType.TABLE.value, + cell="Cell A2", + up="Header 2", + down="Cell B2", + left="Wrong Cell", # This is incorrect + right="Cell A3", + ) + result, explanation = test.run(self.markdown_table) + self.assertFalse(result) + self.assertIn("doesn't match expected", explanation) + + def test_no_tables_found(self): + """Test behavior when no tables are found""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2") + result, explanation = test.run("This is plain text with no tables") + self.assertFalse(result) + self.assertEqual(explanation, "No tables found in the content") + + def test_fuzzy_matching(self): + """Test fuzzy matching with max_diffs""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2", max_diffs=1) + # Create table with slightly misspelled cell + misspelled_table = self.markdown_table.replace("Cell A2", "Cel A2") + result, _ = test.run(misspelled_table) + self.assertTrue(result) + + def test_with_stripped_content(self): + """Test table parsing with stripped content""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2") + # Strip all leading/trailing whitespace from the markdown table + stripped_table = self.markdown_table.strip() + result, explanation = test.run(stripped_table) + self.assertTrue(result, f"Table test failed with stripped content: {explanation}") + + def test_table_at_end_of_file(self): + """Test that a table at the very end of the file is correctly detected""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2") + # Create content with text followed by a table at the very end with no trailing newline + content_with_table_at_end = "Some text before the table.\n" + self.markdown_table.strip() + result, explanation = test.run(content_with_table_at_end) + self.assertTrue(result, f"Table at end of file not detected: {explanation}") + + def test_table_at_end_with_no_trailing_newline(self): + """Test that a table at the end with no trailing newline is detected""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2") + # Remove the trailing newline from the markdown table + content_without_newline = self.markdown_table.rstrip() + result, explanation = test.run(content_without_newline) + self.assertTrue(result, f"Table without trailing newline not detected: {explanation}") + + def test_table_at_end_with_extra_spaces(self): + """Test that a table at the end with extra spaces is detected""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2") + # Add extra spaces to the end of lines in the table + lines = self.markdown_table.split("\n") + content_with_extra_spaces = "\n".join([line + " " for line in lines]) + result, explanation = test.run(content_with_extra_spaces) + self.assertTrue(result, f"Table with extra spaces not detected: {explanation}") + + def test_table_at_end_with_mixed_whitespace(self): + """Test that a table at the end with mixed whitespace is detected""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2") + # Add various whitespace characters to the table + content_with_mixed_whitespace = "Some text before the table.\n" + self.markdown_table.strip() + " \t " + result, explanation = test.run(content_with_mixed_whitespace) + self.assertTrue(result, f"Table with mixed whitespace not detected: {explanation}") + + def test_malformed_table_at_end(self): + """Test that a slightly malformed table at the end is still detected""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2") + # Create a table with irregular pipe placement at the end + malformed_table = """ +Some text before the table. +| Header 1 | Header 2 | Header 3 +| -------- | -------- | -------- +| Cell A1 | Cell A2 | Cell A3 | +| Cell B1 | Cell B2 | Cell B3""" + result, explanation = test.run(malformed_table) + self.assertTrue(result, f"Malformed table at end not detected: {explanation}") + + def test_incomplete_table_at_end(self): + """Test that an incomplete table at the end still gets detected if it contains valid rows""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2") + # Missing the separator row + incomplete_table = """ +Some text before the table. +| Header 1 | Header 2 | Header 3 | +| Cell A1 | Cell A2 | Cell A3 | +| Cell B1 | Cell B2 | Cell B3 |""" + result, explanation = test.run(incomplete_table) + self.assertTrue(result, f"Incomplete table at end not detected: {explanation}") + + def test_table_with_excessive_blank_lines_at_end(self): + """Test that a table followed by many blank lines is detected""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2") + # Add many blank lines after the table + table_with_blanks = self.markdown_table + "\n\n\n\n\n\n\n\n\n\n" + result, explanation = test.run(table_with_blanks) + self.assertTrue(result, f"Table with blank lines at end not detected: {explanation}") + + def test_table_at_end_after_long_text(self): + """Test that a table at the end after a very long text is detected""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2") + # Create a very long text before the table + long_text = "Lorem ipsum dolor sit amet, " * 100 + content_with_long_text = long_text + "\n" + self.markdown_table.strip() + result, explanation = test.run(content_with_long_text) + self.assertTrue(result, f"Table after long text not detected: {explanation}") + + def test_valid_table_at_eof_without_newline(self): + """Test that a valid table at EOF without a trailing newline is detected""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Cell A2") + # Valid table but without trailing newline at the very end of the file + valid_table_eof = """ +| Header 1 | Header 2 | Header 3 | +| -------- | -------- | -------- | +| Cell A1 | Cell A2 | Cell A3 | +| Cell B1 | Cell B2 | Cell B3 |""".strip() + result, explanation = test.run(valid_table_eof) + self.assertTrue(result, f"Valid table at EOF without newline not detected: {explanation}") + + def test_normalizing(self): + table = """| Question - – Satisfaction on scale of 10 | Response | Resident Sample | Business Sample | +|----------------------------------------|----------|----------------|-----------------| +| Planning for and managing residential, commercial and industrial development | Rating of 8, 9 or 10 | 13% | 11% | +| | Average rating | 6.4 | 5.7 | +| | Don’t know responses | 11% | 6% | +| Environmental protection, support for green projects (e.g. green grants, building retrofits programs, zero waste) | Rating of 8, 9 or 10 | 35% | 34% | +| | Average rating | 8.0 | 7.5 | +| | Don’t know responses | 8% | 6% | +| Providing and maintaining parks and green spaces | Rating of 8, 9 or 10 | 42% | 41% | +| | Average rating | 7.7 | 7.3 | +| | Don’t know responses | 1% | 1% |""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="6%", top_heading="Business\nSample") + result, explanation = test.run(table) + self.assertTrue(result, explanation) + + def test_mathematical_minus(self): + table = """| Response | Chinese experimenter | White experimenter | +|----------|----------------------|--------------------| +| | Divided attention | Full attention | Divided attention | Full attention | +| Nonverbal| −.34 (.22) | .54* (.17) | .12 (.27) | −.20 (.24) | +| Verbal | −.25 (.23) | .36 (.20) | .12 (.27) | −.34 (.22) | +""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="-.34 (.22)") + result, explanation = test.run(table) + self.assertTrue(result, explanation) + + def test_markdown_marker(self): + table = """| CATEGORY | POINTS EARNED | +|------------------------------|------------------| +| Sustainable Sites | 9 | +| Water Efficiency | 3 | +| Energy & Atmosphere | 12 | +| Materials & Resources | 6 | +| Indoor Environmental Quality | 11 | +| Innovation & Design Process | 5 | +| TOTAL | 46 |""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="9", up="POINTS EARNED") + result, explanation = test.run(table) + self.assertTrue(result, explanation) + + def test_diffs(self): + table = """| CATEGORY | POINTS EARNED | +|------------------------------|------------------| +| Sustainable Sites | 9 | +| Water Efficiency | 3 | +| Energy & Atmosphere | 12 | +| Materials & Resources | 6 | +| Indoor Environmental Quality | 11 | +| Innovation & Design Process | 5 | +| TOTAL | 46 |""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="9", left="Sustl Sie", max_diffs=2) + result, explanation = test.run(table) + self.assertFalse(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="9", left="Sustainable Site", max_diffs=2) + result, explanation = test.run(table) + self.assertTrue(result, explanation) + + def test_markdown_marker2(self): + table = """| Concentration +level | [CO] | [SO2] | [NOx] | +|------------------------|-----------|-------|----------| +| Control | 0 μM | 0 μM | 0 nM | +| Low | 250 +μM | 8 μM | 0.002 nM | +| Medium | 625 μM | 20 μM | 0.005 nM | +| High | 1250 μM | 40 μM | 0.01 nM |""" + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="20 μM", up=".002 nM") + result, explanation = test.run(table) + self.assertFalse(result, explanation) + + def test_marker3(self): + table = """| | N | Minimum | Maximum | Gemiddelde | Sd | +|-----------------------------------------------|-------|---------|---------|------------|-----| +| Slaapkwaliteit tijdens +gewone nachten | 2017 | 1,0 | 6,0 | 3,9 | 1,0 | +| Slaapkwaliteit tijdens +consignatiediensten | 19816 | 1,0 | 6,0 | 2,8 | 1,2 | +""" + test = TableTest( + pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="2,8", left_heading="Slaapkwaliteit tijdens\nconsignatiediensten" + ) + result, explanation = test.run(table) + self.assertFalse(result, explanation) + + def test_big_table(self): + table = """ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
Base: Resident respondents (n=1,315) and Business respondents (n=397)
Question – Satisfaction on scale of 10 Response Resident Sample Business Sample
Planning for and managing residential, commercial and industrial development Rating of 8, 9 or 10 13% 11%
Average rating 6.4 5.7
Don't know responses 11% 6%
Environmental protection, support for green projects (e.g. green grants, building retrofits programs, zero waste) Rating of 8, 9 or 10 35% 34%
Average rating 8.0 7.5
Don't know responses 8% 6%
Providing and maintaining parks and green spaces Rating of 8, 9 or 10 42% 41%
Average rating 7.7 7.3
Don't know responses 1% 1%
+""" + test = TableTest( + pdf="test.pdf", + page=1, + id="test_id", + type=TestType.TABLE.value, + max_diffs=5, + cell="Planning for and managing residential, commercial and industrial development", + down="Environmental protection,\nsupport for green projects\n(e.g. green grants,\nbuilding retrofits programs,\nzero waste)", + ) + result, explanation = test.run(table) + self.assertTrue(result, explanation) + + def test_html_rowspans_colspans(self): + table = """ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
Product Category Product Subcategory Quarterly Sales ($000s) Annual Total
Q1 Q2 Q3 Q4
Electronics Smartphones 245 278 312 389 1,224
Laptops 187 192 243 297 919
Tablets 95 123 135 156 509
Accessories 64 72 87 105 328
Home Appliances Refrigerators 132 145 151 162 590
Washing Machines 98 112 127 143 480
Microwaves 54 67 72 84 277
Furniture Sofas 112 128 134 142 516
Tables 87 95 103 124 409
Chairs 76 84 92 110 362
Quarterly Totals 1,150 1,296 1,456 1,712 5,614
""" + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Refrigerators", left="Home Appliances") + result, explanation = test.run(table) + self.assertTrue(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Washing Machines", left="Home Appliances") + result, explanation = test.run(table) + self.assertTrue(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Microwaves", left="Home Appliances") + result, explanation = test.run(table) + self.assertTrue(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Sofas", top_heading="Product Subcategory") + result, explanation = test.run(table) + self.assertTrue(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="135", top_heading="Q3") + result, explanation = test.run(table) + self.assertTrue(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="135", top_heading="Quarterly Sales ($000s)") + result, explanation = test.run(table) + self.assertTrue(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="1,712", top_heading="Quarterly Sales ($000s)") + result, explanation = test.run(table) + self.assertTrue(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="135", top_heading="Q2") + result, explanation = test.run(table) + self.assertFalse(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="135", top_heading="Q1") + result, explanation = test.run(table) + self.assertFalse(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="135", top_heading="Q4") + result, explanation = test.run(table) + self.assertFalse(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Home Appliances", top_heading="Product Category") + result, explanation = test.run(table) + self.assertTrue(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Washing Machines", top_heading="Product Category") + result, explanation = test.run(table) + self.assertFalse(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Washing Machines", top_heading="Q3") + result, explanation = test.run(table) + self.assertFalse(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Washing Machines", top_heading="Quarterly Sales ($000s)") + result, explanation = test.run(table) + self.assertFalse(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Electronics", right="Laptops") + result, explanation = test.run(table) + self.assertTrue(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Electronics", right="Accessories") + result, explanation = test.run(table) + self.assertTrue(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Quarterly Sales ($000s)", down="Q2") + result, explanation = test.run(table) + self.assertTrue(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Q2", up="Quarterly Sales ($000s)") + result, explanation = test.run(table) + self.assertTrue(result, explanation) + + def test_multiple_markdown_tables(self): + """Test that we can find and verify cells in multiple markdown tables in one document""" + content = """ +# First Table + +| Name | Age | Role | +| ---- | --- | ---- | +| John | 28 | Developer | +| Jane | 32 | Designer | +| Bob | 45 | Manager | + +Some text between tables... + +# Second Table + +| Department | Budget | Employees | +| ---------- | ------ | --------- | +| Engineering | 1.2M | 15 | +| Design | 0.8M | 8 | +| Marketing | 1.5M | 12 | +| HR | 0.5M | 5 | +""" + + # Test cells in the first table + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="John", right="28") + result, explanation = test.run(content) + self.assertTrue(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="32", left="Jane") + result, explanation = test.run(content) + self.assertTrue(result, explanation) + + # Test cells in the second table + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Engineering", right="1.2M") + result, explanation = test.run(content) + self.assertTrue(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="12", left="1.5M") + result, explanation = test.run(content) + self.assertTrue(result, explanation) + + # Verify top headings work correctly across tables + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Bob", top_heading="Name") + result, explanation = test.run(content) + self.assertTrue(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="HR", top_heading="Department") + result, explanation = test.run(content) + self.assertTrue(result, explanation) + + def test_multiple_html_tables(self): + """Test that we can find and verify cells in multiple HTML tables in one document""" + content = """ +
First Table
+ + + + + + + + + + + + + + + + + + + + + + + + + + +
Country Capital Population
USA Washington DC 331M
France Paris 67M
Japan Tokyo 126M
+ +
Some text between tables...
+ +
Second Table
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
Company Industry Revenue Employees
ABC Corp Technology $5B 10,000
XYZ Inc Healthcare $2.5B 8,500
Acme Co Manufacturing $1.8B 15,000
Global LLC Finance $3.2B 6,200
+""" + + # Test cells in the first table + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="USA", right="Washington DC") + result, explanation = test.run(content) + self.assertTrue(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="126M", left="Tokyo") + result, explanation = test.run(content) + self.assertTrue(result, explanation) + + # Test cells in the second table + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="XYZ Inc", right="Healthcare") + result, explanation = test.run(content) + self.assertTrue(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="15,000", left="$1.8B") + result, explanation = test.run(content) + self.assertTrue(result, explanation) + + # Verify top headings work correctly across tables + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Tokyo", top_heading="Capital") + result, explanation = test.run(content) + self.assertTrue(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Finance", top_heading="Industry") + result, explanation = test.run(content) + self.assertTrue(result, explanation) + + def test_mixed_markdown_and_html_tables(self): + """Test that we can find and verify cells in mixed markdown and HTML tables in one document""" + content = """ +# Markdown Table + +| Product | Price | Quantity | +| ------- | ----- | -------- | +| Apple | $1.20 | 100 | +| Orange | $0.80 | 150 | +| Banana | $0.60 | 200 | + +
HTML Table
+ + + + + + + + + + + + + + + + + + + + + + + + + + +
Month Income Expenses Profit
January $10,000 $8,000 $2,000
February $12,000 $9,500 $2,500
March $15,000 $10,200 $4,800
+""" + + # Test cells in the markdown table + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Orange", right="$0.80") + result, explanation = test.run(content) + self.assertTrue(result, explanation) + + # Test cells in the HTML table + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="February", right="$12,000") + result, explanation = test.run(content) + self.assertTrue(result, explanation) + + # Verify we can find cells with specific top headings + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="100", top_heading="Quantity") + result, explanation = test.run(content) + self.assertTrue(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="$4,800", top_heading="Profit") + result, explanation = test.run(content) + self.assertTrue(result, explanation) + + def test_br_tags_replacement(self): + """Test that
and
tags are correctly replaced with newlines""" + table = """ + + + + + + + + +
Header 1 Header 2
Line 1
Line 2
Line 3 Single line
""" + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="Line 1 Line 2 Line 3") + result, explanation = test.run(table) + self.assertTrue(result, explanation) + + def test_real_complicated_table(self): + table = """ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
Table 1    Differences in diagnoses, gender and family status for participants with a suicide attempt and those without a suicide attempt within the 12-month follow-up interval
Participants with no
suicide attempt
(n = 132)^a Participants with a
suicide attempt
(n = 43)^b
n % n % χ² d.f. P
ICD-10 diagnoses
  F0 1 0.76 0 0.00 0.00 1 1.00
  F1 17 12.88 12 27.91 4.39 1 0.04
  F2 1 0.76 0 0.00 0.00 1 1.00
  F3 106 80.30 31 72.09 0.74 1 0.39
  F4 42 31.82 17 39.53 0.61 1 0.43
  F5 5 3.79 5 11.63 2.44 1 0.12
  F6 20 15.15 19 44.19 14.48 1 0.00
  F7 0 0.00 0 0.00 — — —
  F8 1 0.76 0 0.00 0.00 1 1.00
  F9 2 1.52 1 2.33 0.00 1 1.00
Gender 3.09 2 0.21
  Female 75 56.8 24 55.8
  Male 57 43.2 18 41.9
  Diverse 0 0 1 2.3
Family status 4.87 4 0.30
  Single 55 41.7 14 32.6
  Partnership 25 18.9 9 20.9
  Married 27 20.5 5 11.6
  Divorced 20 15.2 11 25.6
  Widowed 1 0.8 1 2.3
+ F0: Organic, including symptomatic, mental disorders; F1: Mental and behavioural disorders due to psychoactive substance use; F2: Schizophrenia, schizotypal and delusional disorders; F3: affective disorders; F4: Neurotic, stress-related and somatoform disorders; F5: Behavioural syndromes associated with physiological disturbances and physical factors; F6: Disorders of adult personality and behaviour; F7: Mental retardation; F8: Disorders of psychological development; F9: Behavioural and emotional disorders with onset usually occurring in childhood and adolescence.
+ a. 75.43% of the total sample with full information on suicide reattempts within the entire 12-month follow-up interval.
+ b. 24.57% of the total sample with full information on suicide reattempts within the entire 12-month follow-up interval. +
""" + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="4.39", top_heading="χ2") + result, explanation = test.run(table) + self.assertTrue(result, explanation) + + test = TableTest(pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="12.88", top_heading="%") + result, explanation = test.run(table) + self.assertTrue(result, explanation) + + # Account for the superscript in the header + test = TableTest( + pdf="test.pdf", page=1, id="test_id", type=TestType.TABLE.value, cell="12.88", top_heading="Participants with no suicide attempt (n = 132)a" + ) + result, explanation = test.run(table) + self.assertTrue(result, explanation) + + test = TableTest( + pdf="test.pdf", + page=1, + id="test_id", + type=TestType.TABLE.value, + cell="12.88", + top_heading="Table 1 Differences in diagnoses, gender and family status for participants with a suicide attempt and those without a suicide attempt within the 12-month follow-up interval", + ) + result, explanation = test.run(table) + self.assertTrue(result, explanation) + + +class TestBaselineTest(unittest.TestCase): + """Test the BaselineTest class""" + + def test_valid_initialization(self): + """Test that valid initialization works""" + test = BaselineTest(pdf="test.pdf", page=1, id="test_id", type=TestType.BASELINE.value, max_repeats=50) + self.assertEqual(test.max_repeats, 50) + + def test_non_empty_content(self): + """Test that non-empty content passes""" + test = BaselineTest(pdf="test.pdf", page=1, id="test_id", type=TestType.BASELINE.value) + result, _ = test.run("This is some normal content") + self.assertTrue(result) + + def test_empty_content(self): + """Test that empty content fails""" + test = BaselineTest(pdf="test.pdf", page=1, id="test_id", type=TestType.BASELINE.value) + result, explanation = test.run(" \n\t ") + self.assertFalse(result) + self.assertIn("no alpha numeric characters", explanation) + + def test_repeating_content(self): + """Test that highly repeating content fails""" + test = BaselineTest(pdf="test.pdf", page=1, id="test_id", type=TestType.BASELINE.value, max_repeats=2) + # Create highly repeating content - repeat "abc" many times + repeating_content = "abc" * 10 + result, explanation = test.run(repeating_content) + self.assertFalse(result) + self.assertIn("repeating", explanation) + + def test_content_with_disallowed_characters(self): + """Test that content with disallowed characters fails""" + test = BaselineTest(pdf="test.pdf", page=1, id="test_id", type=TestType.BASELINE.value) + result, explanation = test.run("This has Chinese characters: 你好") + self.assertFalse(result) + self.assertIn("disallowed characters", explanation) + + def test_content_with_emoji(self): + """Test that content with emoji fails""" + test = BaselineTest(pdf="test.pdf", page=1, id="test_id", type=TestType.BASELINE.value) + result, explanation = test.run("This has emoji: 😊") + self.assertFalse(result) + self.assertIn("disallowed characters", explanation) + self.assertIn("😊", explanation) + + def test_content_with_mandarin(self): + test = BaselineTest(pdf="test.pdf", page=1, id="test_id", type=TestType.BASELINE.value) + result, explanation = test.run("asdfasdfas維基百科/中文asdfw") + self.assertFalse(result) + self.assertIn("disallowed characters", explanation) + + def test_valid_content(self): + """Test that valid content passes all checks""" + test = BaselineTest(pdf="test.pdf", page=1, id="test_id", type=TestType.BASELINE.value) + content = "This is some normal content with proper English letters and no suspicious repetition." + result, _ = test.run(content) + self.assertTrue(result) + + def test_max_length_with_image_tags_skipped(self): + """Test that image tags are properly removed when max_length_skips_image_alt_tags is True""" + # Test with max_length_skips_image_alt_tags=True + test = BaselineTest(pdf="test.pdf", page=1, id="test_id", type=TestType.BASELINE.value, max_length=10, max_length_skips_image_alt_tags=True) + + # Content with image tag that would exceed max_length if counted + content = "Hello ![Diagram showing labeled components 100, 101, 102, 103, 104, 105, 102A, 103A, 104A, 105A, 220, 130, and 140 within a rectangular frame.](page_370_682_1012_1012.png) World" + result, _ = test.run(content) + # Should pass because "HelloWorld" has 10 alphanumeric characters after removing image tag + self.assertTrue(result) + + def test_max_length_with_image_tags_not_skipped(self): + """Test that image tags are counted when max_length_skips_image_alt_tags is False""" + # Test with max_length_skips_image_alt_tags=False (default) + test = BaselineTest(pdf="test.pdf", page=1, id="test_id", type=TestType.BASELINE.value, max_length=10) + + # Same content with image tag + content = "Hello ![Diagram showing labeled components 100, 101, 102, 103, 104, 105, 102A, 103A, 104A, 105A, 220, 130, and 140 within a rectangular frame.](page_370_682_1012_1012.png) World" + result, explanation = test.run(content) + # Should fail because full content has way more than 10 alphanumeric characters + self.assertFalse(result) + self.assertIn("characters were output for a page we expected to be blank", explanation) + + def test_multiple_image_tags_removed(self): + """Test that multiple image tags are removed correctly""" + test = BaselineTest(pdf="test.pdf", page=1, id="test_id", type=TestType.BASELINE.value, max_length=15, max_length_skips_image_alt_tags=True) + + content = "Start ![First image](img1.png) middle ![Second image](img2.png) end" + result, _ = test.run(content) + # Should pass because "Start middle end" = "Startmiddleend" has 14 alphanumeric characters + self.assertTrue(result) + + def test_nested_brackets_in_image_tags(self): + """Test that image tags with nested brackets are handled correctly""" + test = BaselineTest(pdf="test.pdf", page=1, id="test_id", type=TestType.BASELINE.value, max_length=8, max_length_skips_image_alt_tags=True) + + content = "Text ![Complex [nested] description](image.png) here" + result, _ = test.run(content) + # Should pass because "Text here" = "Texthere" has 8 alphanumeric characters + self.assertTrue(result) + + +class TestMathTest(unittest.TestCase): + """Test the MathTest class""" + + def test_valid_initialization(self): + """Test that valid initialization works""" + try: + test = MathTest(pdf="test.pdf", page=1, id="test_id", type=TestType.MATH.value, math="a + b = c") + self.assertEqual(test.math, "a + b = c") + except Exception as e: + self.fail(f"Valid initialization failed with: {e}") + + def test_invalid_test_type(self): + """Test that invalid test type raises ValidationError""" + with self.assertRaises(ValidationError): + MathTest(pdf="test.pdf", page=1, id="test_id", type=TestType.PRESENT.value, math="a + b = c") + + def test_empty_math(self): + """Test that empty math raises ValidationError""" + with self.assertRaises(ValidationError): + MathTest(pdf="test.pdf", page=1, id="test_id", type=TestType.MATH.value, math="") + + def test_exact_math_match(self): + """Test exact match of math equation""" + try: + test = MathTest(pdf="test.pdf", page=1, id="test_id", type=TestType.MATH.value, math="a + b = c") + + # Test content with exact math match + content = "Here is an equation: $$a + b = c$$" + result, _ = test.run(content) + self.assertTrue(result) + except Exception as e: + self.fail(f"Test failed with: {e}") + + def test_rendered_math_match(self): + """Test rendered match of math equation""" + try: + test = MathTest(pdf="test.pdf", page=1, id="test_id", type=TestType.MATH.value, math="a + b = c") + + # Test content with different but equivalent math + content = "Here is an equation: $$a+b=c$$" + result, _ = test.run(content) + self.assertTrue(result) + except Exception as e: + self.fail(f"Test failed with: {e}") + + def test_no_math_match(self): + """Test no match of math equation""" + try: + test = MathTest(pdf="test.pdf", page=1, id="test_id", type=TestType.MATH.value, math="a + b = c") + + # Test content with no matching math + content = "Here is an equation: $$x + y = z$$" + result, explanation = test.run(content) + self.assertFalse(result) + self.assertIn("No match found", explanation) + except Exception as e: + self.fail(f"Test failed with: {e}") + + def test_different_math_delimiters(self): + """Test different math delimiters""" + try: + test = MathTest(pdf="test.pdf", page=1, id="test_id", type=TestType.MATH.value, math="a + b = c") + + # Test different delimiters + delimiters = [ + "$$a + b = c$$", # $$...$$ + "$a + b = c$", # $...$ + "\$a + b = c\$", # $...$ + "\\[a + b = c\\]", # \[...\] + ] + + for delim in delimiters: + content = f"Here is an equation: {delim}" + result, _ = test.run(content) + self.assertTrue(result) + except Exception as e: + self.fail(f"Test failed with: {e}") + + +class TestFormatTest(unittest.TestCase): + """Test the FormatTest class""" + + def test_valid_initialization(self): + """Test that valid initialization works""" + test = FormatTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FORMAT.value, text="Important Title", format="heading") + self.assertEqual(test.text, "Important Title") + self.assertEqual(test.format, "heading") + self.assertTrue(test.case_sensitive) + + def test_invalid_test_type(self): + """Test that invalid test type raises ValidationError""" + with self.assertRaises(ValidationError): + FormatTest(pdf="test.pdf", page=1, id="test_id", type=TestType.PRESENT.value, text="test text", format="bold") + + def test_empty_text(self): + """Test that empty text raises ValidationError""" + with self.assertRaises(ValidationError): + FormatTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FORMAT.value, text="", format="heading") + + def test_invalid_format_type(self): + """Test that invalid format type raises ValidationError""" + with self.assertRaises(ValidationError): + FormatTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FORMAT.value, text="test text", format="underline") # Not supported + + def test_markdown_heading(self): + """Test detection of markdown headings""" + test = FormatTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FORMAT.value, text="Chapter One", format="heading") + + # Test various heading levels + content = "# Chapter One\nSome text here" + result, _ = test.run(content) + self.assertTrue(result) + + content = "## Chapter One\nSome text here" + result, _ = test.run(content) + self.assertTrue(result) + + content = "### Chapter One\nSome text here" + result, _ = test.run(content) + self.assertTrue(result) + + def test_html_heading(self): + """Test detection of HTML headings""" + test = FormatTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FORMAT.value, text="Chapter One", format="heading") + + # Test various HTML heading tags + content = "
Chapter One
\nSome text here" + result, _ = test.run(content) + self.assertTrue(result) + + content = "
Chapter One
\nSome text here" + result, _ = test.run(content) + self.assertTrue(result) + + def test_markdown_bold(self): + """Test detection of markdown bold text""" + test = FormatTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FORMAT.value, text="important", format="bold") + + # Test ** syntax + content = "This is **important** text" + result, _ = test.run(content) + self.assertTrue(result) + + # Test __ syntax + content = "This is __important__ text" + result, _ = test.run(content) + self.assertTrue(result) + + # Should not match if not bold + content = "This is important text" + result, explanation = test.run(content) + self.assertFalse(result) + self.assertIn("not found with bold formatting", explanation) + + def test_html_bold(self): + """Test detection of HTML bold text""" + test = FormatTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FORMAT.value, text="important", format="bold") + + # Test tag + content = "This is important text" + result, _ = test.run(content) + self.assertTrue(result) + + # Test tag + content = "This is important text" + result, _ = test.run(content) + self.assertTrue(result) + + def test_markdown_italic(self): + """Test detection of markdown italic text""" + test = FormatTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FORMAT.value, text="emphasis", format="italic") + + # Test * syntax + content = "This needs *emphasis* here" + result, _ = test.run(content) + self.assertTrue(result) + + # Test _ syntax + content = "This needs _emphasis_ here" + result, _ = test.run(content) + self.assertTrue(result) + + # Should not match bold + content = "This needs **emphasis** here" + result, explanation = test.run(content) + self.assertFalse(result) + self.assertIn("not found with italic formatting", explanation) + + def test_html_italic(self): + """Test detection of HTML italic text""" + test = FormatTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FORMAT.value, text="emphasis", format="italic") + + # Test tag + content = "This needs emphasis here" + result, _ = test.run(content) + self.assertTrue(result) + + # Test tag + content = "This needs emphasis here" + result, _ = test.run(content) + self.assertTrue(result) + + def test_partial_match(self): + """Test that partial matches work""" + test = FormatTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FORMAT.value, text="Chapter", format="heading") + + content = "# Chapter One: Introduction\nSome text here" + result, _ = test.run(content) + self.assertTrue(result) + + def test_case_insensitive(self): + """Test case insensitive matching""" + test = FormatTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FORMAT.value, text="IMPORTANT", format="bold", case_sensitive=False) + + content = "This is **important** text" + result, _ = test.run(content) + self.assertTrue(result) + + def test_fuzzy_matching(self): + """Test fuzzy matching with max_diffs""" + test = FormatTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FORMAT.value, text="Chapter One", format="heading", max_diffs=2) + + # Slightly misspelled heading + content = "# Chaptre Oen\nSome text here" + result, _ = test.run(content) + self.assertTrue(result) + + def test_multiple_formatted_sections(self): + """Test when multiple sections have the same format""" + test = FormatTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FORMAT.value, text="Section Two", format="heading") + + content = """ +# Section One +Some text + +## Section Two +More text + +### Section Three +Even more text +""" + result, _ = test.run(content) + self.assertTrue(result) + + def test_nested_formatting(self): + """Test nested formatting scenarios""" + test = FormatTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FORMAT.value, text="bold text inside", format="bold") + + # Bold text inside a heading + content = "# This has **bold text inside** the heading" + result, _ = test.run(content) + self.assertTrue(result) + + def test_mixed_markdown_html(self): + """Test content with both markdown and HTML formatting""" + test = FormatTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FORMAT.value, text="Important", format="bold") + + content = """ +Some **Important** markdown bold text. +And some Important HTML bold text. +""" + result, _ = test.run(content) + self.assertTrue(result) + + def test_italic_not_matching_bold(self): + """Test that italic patterns don't incorrectly match bold""" + test = FormatTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FORMAT.value, text="text", format="italic") + + # Should not match double asterisk bold + content = "This is **text** here" + result, _ = test.run(content) + self.assertFalse(result) + + # Should match single asterisk italic + content = "This is *text* here" + result, _ = test.run(content) + self.assertTrue(result) + + def test_normalization_in_formatted_text(self): + """Test that text normalization works within formatted sections""" + test = FormatTest( + pdf="test.pdf", + page=1, + id="test_id", + type=TestType.FORMAT.value, + text="\"fancy\" 'quotes'", # This is what it looks like after normalization + format="bold", + ) + + # Content has fancy quotes that should be normalized + content = "This is **\u201cfancy\u201d \u2018quotes\u2019** here" + result, _ = test.run(content) + self.assertTrue(result) + + +class TestFootnoteTest(unittest.TestCase): + """Test the FootnoteTest class""" + + def test_valid_initialization(self): + """Test that valid initialization works""" + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="1") + self.assertEqual(test.marker, "1") + self.assertIsNone(test.appears_before_marker) + self.assertIsNone(test.appears_after_marker) + + def test_invalid_test_type(self): + """Test that invalid test type raises ValidationError""" + with self.assertRaises(ValidationError): + FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.PRESENT.value, marker="1") + + def test_marker_required(self): + """Test that marker field is required""" + with self.assertRaises(TypeError): + FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value) + + def test_marker_only(self): + """Test footnote with only marker specified""" + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="2") + + # Test with markdown footnote reference + content = "This is some text[^2] with a footnote.\n\n[^2]: This is the footnote text." + result, _ = test.run(content) + self.assertTrue(result) + + # Test with HTML superscript + content = "This is some text² with a footnote." + result, _ = test.run(content) + self.assertTrue(result) + + # Test with Unicode superscript + content = "This is some text² with a footnote." + result, _ = test.run(content) + self.assertTrue(result) + + # Test when marker not found + content = "This text has no footnote markers." + result, explanation = test.run(content) + self.assertFalse(result) + self.assertIn("not found", explanation) + + def test_appears_before_marker(self): + """Test footnote with appears_before_marker specified""" + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="1", appears_before_marker="important fact") + + # Test when text appears before marker + content = "This is an important fact[^1] in the document." + result, _ = test.run(content) + self.assertTrue(result) + + content = "This is an important fact [^1] in the document." + result, _ = test.run(content) + self.assertTrue(result) + + content = "This is an important fact\t[^1] in the document." + result, _ = test.run(content) + self.assertTrue(result) + + # Test when text doesn't appear before marker + content = "This is something else[^1] important fact comes later." + result, explanation = test.run(content) + self.assertFalse(result) + self.assertIn("not found before", explanation) + + def test_appears_after_marker(self): + """Test footnote with appears_after_marker specified""" + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="3", appears_after_marker="that needs citation") + + # Test when text appears after marker + content = """This is a statement[^3] that needs citation from the study.""" + result, _ = test.run(content) + self.assertTrue(result) + + content = """This is a statement[^3]that needs citation from the study.""" + result, _ = test.run(content) + self.assertTrue(result) + + content = """This is a statement ³that needs citation from the study.""" + result, _ = test.run(content) + self.assertTrue(result) + + # Test when text doesn't appear after marker + content = """This is a statement that needs citation before [^3] the marker.""" + result, explanation = test.run(content) + self.assertFalse(result) + self.assertIn("not found after", explanation) + + def test_both_before_and_after(self): + """Test footnote with both appears_before_marker and appears_after_marker specified""" + test = FootnoteTest( + pdf="test.pdf", + page=1, + id="test_id", + type=TestType.FOOTNOTE.value, + marker="1", + appears_before_marker="important fact", + appears_after_marker="documented here", + ) + + # Test when both conditions are met (can be at different occurrences) + content = "This is an important fact[^1] and also[^1] documented here." + result, _ = test.run(content) + self.assertTrue(result) + + # Test when both conditions are met at same occurrence + content = "This is an important fact[^1] documented here in the text." + result, _ = test.run(content) + self.assertTrue(result) + + # Test when one condition is not met + content = "This is something else[^1] documented here." + result, explanation = test.run(content) + self.assertFalse(result) + + content = "Thiis is something else [^1] and also an important fact [^1] documented here." + result, _ = test.run(content) + self.assertTrue(result) + + def test_fuzzy_matching_for_appears_before(self): + """Test fuzzy matching with max_diffs for appears_before_marker""" + test = FootnoteTest( + pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="1", appears_before_marker="Example Reference", max_diffs=3 + ) + + # Slightly misspelled text before marker + content = """The Exampl Referenc[^1] is cited here.""" + result, _ = test.run(content) + self.assertTrue(result) + + def test_fuzzy_matching_for_appears_after(self): + """Test fuzzy matching with max_diffs for appears_after_marker""" + test = FootnoteTest( + pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="2", appears_after_marker="statistical analysis", max_diffs=2 + ) + + # Slightly misspelled text after marker + content = "The reference[^2] statistcal analysys shows significant results." + result, _ = test.run(content) + self.assertTrue(result) + + def test_multiple_footnotes(self): + """Test handling of multiple footnotes in document""" + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="2") + + content = """First claim[^1] and second claim[^2]. + +[^1]: First reference text. +[^2]: Second reference with more details.""" + result, _ = test.run(content) + self.assertTrue(result) + + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="1") + + result, _ = test.run(content) + self.assertTrue(result) + + def test_html_superscript_variations(self): + """Test various HTML superscript formats""" + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="4") + + # Test with attributes in sup tag + content = 'Text with footnote⁴.' + result, _ = test.run(content) + self.assertTrue(result) + + # Test case insensitive HTML + content = "Text with footnote⁴." + result, _ = test.run(content) + self.assertTrue(result) + + def test_unicode_superscripts(self): + """Test all Unicode superscript digits""" + superscripts = {"0": "⁰", "1": "¹", "2": "²", "3": "³", "4": "⁴", "5": "⁵", "6": "⁶", "7": "⁷", "8": "⁸", "9": "⁹"} + + for digit, superscript in superscripts.items(): + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker=digit) + content = f"Text with footnote{superscript}." + result, _ = test.run(content) + self.assertTrue(result, f"Failed for superscript {digit}") + + def test_multi_digit_marker(self): + """Test multi-digit footnote markers""" + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="12") + + # Test markdown reference + content = "Text with footnote[^12].\n\n[^12]: Reference text." + result, _ = test.run(content) + self.assertTrue(result) + + # Test Unicode superscript for multi-digit + content = "Text with footnote¹²." + result, _ = test.run(content) + self.assertTrue(result) + + def test_appears_before_ignores_whitespace(self): + """Test that appears_before_marker ignores whitespace and non-alpha characters""" + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="1", appears_before_marker="claim") + + # Should match even with punctuation and whitespace between + content = "This is a claim, ... !![^1] here." + result, _ = test.run(content) + self.assertTrue(result) + + # Should match with just spaces + content = "This is a claim [^1] here." + result, _ = test.run(content) + self.assertTrue(result) + + def test_empty_appears_before_validation(self): + """Test that empty appears_before_marker field raises ValidationError""" + with self.assertRaises(ValidationError): + FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="1", appears_before_marker=" ") # Whitespace only + + def test_empty_appears_after_validation(self): + """Test that empty appears_after_marker field raises ValidationError""" + with self.assertRaises(ValidationError): + FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="1", appears_after_marker=" ") # Whitespace only + + def test_normalized_text_in_footnotes(self): + """Test that text normalization works in appears_before/after_marker""" + test = FootnoteTest( + pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="1", appears_before_marker='"fancy" quotes and-dashes' # Normalized form + ) + + # Content with fancy characters that should be normalized + content = """"fancy" quotes and—dashes[^1] here.""" + result, _ = test.run(content) + self.assertTrue(result) + + def test_exact_marker_match(self): + """Test that marker matching is exact""" + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="1") + + # Should match "1" but not "10", "11", etc. + content = "Text with footnote[^10]." + result, explanation = test.run(content) + self.assertFalse(result) + self.assertIn("not found", explanation) + + # Should match exactly "1" + content = "Text with footnote[^1]." + result, _ = test.run(content) + self.assertTrue(result) + + def test_special_symbol_markers(self): + """Test footnote markers with special symbols""" + # Test common special symbols used as footnote markers + special_symbols = ["*", "†", "‡", "§", "¶", "∥", "**", "††", "‡‡"] + + for symbol in special_symbols: + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker=symbol) + + # Test with markdown footnote reference + content = f"Text with footnote[^{symbol}].\n\n[^{symbol}]: Footnote text." + result, _ = test.run(content) + self.assertTrue(result, f"Failed for symbol marker '{symbol}' in markdown") + + # Test with HTML superscript + content = f"Text with footnote^{symbol}." + result, _ = test.run(content) + self.assertTrue(result, f"Failed for symbol marker '{symbol}' in HTML") + + # Test that the symbol as plain text doesn't match + # (it needs to be in proper footnote format) + content = f"Text with {symbol} in the middle." + result, _ = test.run(content) + self.assertFalse(result, f"Plain text '{symbol}' should not match as footnote marker") + + def test_word_markers(self): + """Test footnote markers that are words""" + word_markers = ["apple", "note", "ref", "A", "B", "foo"] + + for word in word_markers: + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker=word) + + # Test with markdown footnote reference + content = f"Text with footnote[^{word}].\n\n[^{word}]: Footnote text." + result, _ = test.run(content) + self.assertTrue(result, f"Failed for word marker '{word}' in markdown") + + # Test with HTML superscript + content = f"Text with footnote^{word}." + result, _ = test.run(content) + self.assertTrue(result, f"Failed for word marker '{word}' in HTML") + + # Test that the word as plain text doesn't match + content = f"Text with {word} in the middle." + result, _ = test.run(content) + self.assertFalse(result, f"Plain text '{word}' should not match as footnote marker") + + def test_marker_with_whitespace_validation(self): + """Test that markers with whitespace are rejected""" + with self.assertRaises(ValidationError) as cm: + FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="has space") + self.assertIn("cannot contain whitespace", str(cm.exception)) + + def test_special_symbols_with_appears_before(self): + """Test special symbol markers combined with appears_before_marker""" + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="†", appears_before_marker="important") + + # Use proper footnote format + content = """This is important[^†] here.""" + result, _ = test.run(content) + self.assertTrue(result) + + # Test when text doesn't appear before + content = """This is something[^†] else important.""" + result, explanation = test.run(content) + self.assertFalse(result) + self.assertIn("not found before", explanation) + + def test_special_symbols_with_appears_after(self): + """Test special symbol markers with position checking""" + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="‡", appears_after_marker="critical finding") + + # Test when text appears after the marker + content = "The reference^‡ critical finding was unexpected." + result, _ = test.run(content) + self.assertTrue(result) + + # Test when text doesn't appear after the marker + content = "The critical finding was before ^‡ the marker." + result, explanation = test.run(content) + self.assertFalse(result) + self.assertIn("not found after", explanation) + + def test_escaped_special_characters_in_regex(self): + """Test that special regex characters in markers are properly escaped""" + # These characters have special meaning in regex and need escaping + regex_special_markers = ["*", "**", ".", "^", "$", "+", "?", "|", "\\", "(", ")", "[", "]", "{", "}"] + + for marker in regex_special_markers: + try: + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker=marker) + + # Test with markdown format + content = f"Text[^{marker}].\n\n[^{marker}]: Note." + result, _ = test.run(content) + # We expect this to work for valid markers like "*" and "**" + # But some like "\" might not be valid in markdown + + # Test with HTML format which should work for all + content = f"Text^{marker}." + result, _ = test.run(content) + + except Exception as e: + # Some markers might not be valid, that's okay + pass + + def test_mixed_formats_same_document(self): + """Test document with mixed footnote formats""" + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="†") + + # Document with both markdown and HTML footnotes + content = """ +Some text[^†] with markdown format. +Other text^† with HTML format. + +[^†]: This is the footnote definition. +""" + result, _ = test.run(content) + self.assertTrue(result) + + def test_double_dagger_and_similar(self): + """Test markers that are repeated symbols""" + repeated_markers = ["**", "††", "‡‡", "§§", "***"] + + for marker in repeated_markers: + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker=marker) + + # Test markdown format + content = f"Text[^{marker}].\n\n[^{marker}]: Note text." + result, _ = test.run(content) + self.assertTrue(result, f"Failed for repeated marker '{marker}'") + + def test_case_sensitive_word_markers(self): + """Test that word markers are case-sensitive""" + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="Note") + + # Should match exact case + content = "Text[^Note].\n\n[^Note]: Reference." + result, _ = test.run(content) + self.assertTrue(result) + + # Should not match different case + content = "Text[^note].\n\n[^note]: Reference." + result, _ = test.run(content) + self.assertFalse(result) + + def test_html_footnote_definitions(self): + """Test HTML-style footnote definitions""" + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="1") + + # HTML style: ¹ marker + content = """ +Some text with a reference¹ in the middle. + +At the bottom: +¹This is the footnote text that appears right after the marker. +""" + result, _ = test.run(content) + self.assertTrue(result) + + def test_html_footnote_with_special_symbols(self): + """Test HTML-style footnotes with special symbol markers""" + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="†") + + # HTML style with special symbol + content = """ +Important finding^† in the research. + +^†See appendix for details and supplementary materials. +""" + result, _ = test.run(content) + self.assertTrue(result) + + def test_mixed_footnote_styles(self): + """Test document with both markdown and HTML footnote definitions""" + # Test for first footnote (markdown style) + test1 = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="1") + + content = """ +First point[^1] and second point². + +[^1]: First reference in markdown style. + +²Second reference in HTML style. +""" + result, _ = test1.run(content) + self.assertTrue(result) + + # Test for second footnote (HTML style) + test2 = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="2") + result, _ = test2.run(content) + self.assertTrue(result) + + def test_html_footnote_multiline(self): + """Test HTML footnote that spans multiple lines""" + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="3") + + content = """ +Reference here³. + +³This footnote text +spans multiple lines +with more information. +""" + result, _ = test.run(content) + self.assertTrue(result) + + def test_real_world_case(self): + content = """## Abstract + +Automated scientific discovery promises to accelerate progress across scientific domains. However, developing and evaluating an AI agent's capacity for end-to-end scientific reasoning is challenging as running real-world experiments is often prohibitively expensive or infeasible. In this work we introduce DISCOVERYWORLD, the first virtual environment for developing and benchmarking an agent's ability to perform complete cycles of novel scientific discovery. DISCOVERYWORLD contains a variety of different challenges, covering topics as diverse as radioisotope dating, rocket science, and proteomics, to encourage development of *general* discovery skills rather than task-specific solutions. DISCOVERYWORLD itself is an inexpensive, simulated, text-based environment (with optional 2D visual overlay). It includes 120 different challenge tasks, spanning eight topics each with three levels of difficulty and several parametric variations. Each task requires an agent to form hypotheses, design and run experiments, analyze results, and act on conclusions. DISCOVERYWORLD further provides three automatic metrics for evaluating performance, based on (a) task completion, (b) task-relevant actions taken, and (c) the discovered explanatory knowledge. We find that strong baseline agents, that perform well in prior published environments, struggle on most DISCOVERYWORLD tasks, suggesting that DISCOVERYWORLD captures some of the novel challenges of discovery, and thus that DISCOVERYWORLD may help accelerate near-term development and assessment of scientific discovery competency in agents. Code available at github.com/allenai/discoveryworld.¹ + +## 1 Introduction + +A long-standing dream of AI has been to build systems that can perform scientific discovery, potentially leading to new breakthroughs for the benefit of humanity [13]. Recently, with the rise of neural techniques, there have been several successful discovery systems developed for specialized problems such as protein folding [10, 15], mathematics [22], and material science [24]. However, while the results have been impressive, these systems (deliberately) bypass the full discovery process of ideation, hypothesis formation, experiment design, etc., and instead (expertly) perform systematic searches over a pre-defined hypothesis space, with pre-defined goals. This raises the question: how much more can be achieved if AI is applied to the broader scientific process? Some works have indeed developed early systems for this, for example, in chemistry [1], and genetics [12]. These systems can also generate hypotheses, design experiments, and execute them (via robotics) in real environments. However, operating in real environments is expensive and complex, creating a barrier + +¹Released under Apache-2.0 license.""" + + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="1") + + result, _ = test.run(content) + self.assertTrue(result) + + test = FootnoteTest( + pdf="test.pdf", + page=1, + id="test_id", + type=TestType.FOOTNOTE.value, + marker="1", + appears_before_marker="Code available at github.com/allenai/discoveryworld.", + ) + + result, _ = test.run(content) + self.assertTrue(result) + + test = FootnoteTest(pdf="test.pdf", page=1, id="test_id", type=TestType.FOOTNOTE.value, marker="1", appears_after_marker="Released under Apache") + + result, _ = test.run(content) + self.assertTrue(result) + + +if __name__ == "__main__": + unittest.main()

PDF: {{ pdf_name }}

Equations ({{ tests|length }})

All Done! 🎉

PDF Model Comparison

Differences

All Tests Reviewed!

All Done! 🎉

PDF Test Review: {{ pdf_name }} ({{ pdf_index + 1 }}/{{ total_pdfs }})

Tests ({{ tests|length }})

PDF: {{ pdf_name }}

Equations ({{ tests|length }})

through

Text Evaluation Review

Numeric Attribute Distributions

{attr_name}

Summary

PII Rule Comparison Results

Summary

Result Categories

True Positives: {tp}

True Negatives: {tn}

False Positives: {fp}

False Negatives: {fn}

Task Overview

How to Annotate

How to Identify PII

Identifiers for PII

PII that must co-occur with an Identifier

PII that occurs even without an Identifier

Generated On

Workspace

Sample Size

Identifiers for PII (Select these if found)

PII that must co-occur with an Identifier

PII that occurs even without an Identifier

Identifiers (Select these if found)

PII that requires an identifier above

PII that is always sensitive (even without an identifier)

Hello World

Hello World

Continent and country or area Continent et pays ou zone	Population estimates (in thousands) - Estimations de population (en milliers)¹
Continent and country or area Continent et pays ou zone	2003	2004	2005	2006	2007	2008	2009	2010	2011	2012
AFRICA - AFRIQUE
Algeria - Algérie	33 003	33 461	33 961	34 507	35 097	35 725	36 383	37 063	37 763	38 482
Angola	15 421	15 977	16 544	17 122	17 713	18 314	18 927	19 549	20 180	20 821
Benin - Bénin	7 606	7 823	8 042	8 264	8 487	8 712	8 942	9 180	9 420	9 661
Botswana	1 833	1 855	1 876	1 898	1 919	1 934	1 952	1 969	1 987	2 004
Burkina Faso	12 829	13 034	13 222	13 622	14 235	14 680	15 095	15 540	15 995	16 460
Burundi	7 264	7 511	7 770	8 043	8 328	8 624	8 927	9 233	9 540	9 850
Cabo Verde	467	474	479	482	484	485	486	488	491	494
Cameroon - Cameroun	17 075	17 523	17 981	18 450	18 929	19 419	19 919	20 430	20 950	21 700
Central African Republic - République centrafricaine	3 830	3 894	3 961	4 032	4 107	4 185	4 266	4 350	4 436	4 525
Chad - Tchad	9 311	9 665	10 014	10 357	10 694	11 031	11 371	11 721	12 080	12 448
Comoros - Comores	570	585	601	617	633	649	666	683	700	718
Congo	3 383	3 449	3 543	3 647	3 759	3 876	3 995	4 112	4 225	4 337
Côte d'Ivoire	18 103	18 440	18 804	19 192	19 600	20 025	20 461	20 902	21 346	21 795
Democratic Republic of the Congo - République démocratique du Congo	50 972	52 487	54 028	55 591	57 188	58 819	60 486	62 191	63 932	65 705
Djibouti	755	766	777	788	799	810	822	834	847	860
Egypt - Égypte	69 432	70 501	71 774	72 901	74 230	75 492	76 775	78 076	79 302	80 722
Equatorial Guinea - Guinée équatoriale	682	698	715	732	749	766	783	801	818	836
Eritrea - Érythrée	4 473	4 666	4 854	5 035	5 210	5 382	5 558	5 741	5 933	6 131
Ethiopia - Éthiopie	71 090	74 066	76 167	78 201	80 441	82 621	84 838	87 095	89 393	91 729
Gabon	1 301	1 330	1 359	1 388	1 417	1 447	1 477	1 508	1 539	1 672
Gambia - Gambie	1 349	1 392	1 437	1 482	1 529	1 578	1 628	1 681	1 735	1 791
Ghana	20 302	20 836	21 384	21 948	22 526	23 110	23 692	24 263	24 821	25 366
Guinea - Guinée	9 205	9 380	9 576	9 799	10 047	10 315	10 593	10 876	11 162	11 451
Guinea-Bissau - Guinée-Bissau	1 361	1 391	1 422	1 453	1 484	1 517	1 551	1 587	1 624	1 664
Kenya	33 913	34 701	35 516	36 360	37 236	38 140	39 070	40 028	41 010	42 027
Lesotho	1 899	1 912	1 926	1 940	1 956	1 972	1 990	2 009	2 030	2 052
Liberia - Libéria	3 124	3 185	3 270	3 385	3 522	3 673	3 821	3 958	4 080	4 190
Libya - Libye	5 423	5 507	5 594	5 686	5 782	5 877	5 964	6 041	6 103	6 155
Madagascar	17 245	17 763	18 290	18 826	19 371	19 927	20 496	21 080	21 679	22 294
Malawi	12 126	12 433	12 744	13 060	13 384	13 717	14 061	14 416	14 783	15 164
Mali	11 220	11 573	11 941	12 326	12 726	13 138	13 559	13 986	14 417	14 854
Mauritania - Mauritanie	2 966	3 055	3 146	3 238	3 330	3 423	3 516	3 609	3 703	3 796
Mauritius - Maurice^a	1 204	1 209	1 213	1 217	1 220	1 223	1 227	1 231	1 235	1 240
Mayotte	164	170	175	181	186	192	198	204	210	216
Morocco - Maroc^b	29 897	30 160	30 425	30 693	30 967	31 255	31 577	31 942	32 309	32 521
Mozambique	19 873	20 439	21 010	21 587	22 171	22 763	23 361	23 967	24 581	25 203
Namibia - Namibie	1 981	2 003	2 027	2 053	2 081	2 111	2 143	2 179	2 218	2 259
Niger	12 661	13 076	13 499	13 929	14 368	14 817	15 278	15 751	16 239	16 744
Nigeria - Nigéria	132 550	135 999	139 586	143 315	147 187	151 208	155 381	159 708	164 193	168 834
Republic of South Sudan - République du Soudan du Sud	7 490	7 730	8 039	8 377	8 737	9 118	9 521	9 941	10 381	10 838
Réunion	770	780	791	802	813	823	834	845	855	865
Rwanda	9 126	9 254	9 429	9 651	9 928	10 223	10 530	10 837	11 144	11 458
Sao Tome and Principe - Sao Tomé-et-Principe	145	148	151	154	157	160	163	166	169	172
Senegal - Sénégal	10 674	10 988	11 271	11 583	11 905	12 239	12 587	12 951	13 331	13 726
Seychelles	84	86	87	88	89	90	91	91	92	92
Sierra Leone	4 713	4 928	5 120	5 281	5 416	5 532	5 641	5 752	5 865	5 979
Somalia - Somalie	8 038	8 250	8 487	8 698	8 913	9 140	9 381	9 636	9 908	10 195
South Africa - Afrique du Sud	46 889	47 643	48 282	48 869	49 401	49 877	50 280	50 632	50 946	51 216
Sudan - Soudan	29 974	30 779	31 586	32 398	33 218	34 040	34 853	35 652	36 431	37 195
Swaziland	1 088	1 095	1 105	1 118	1 135	1 154	1 174	1 193	1 212	1 231
Togo	5 429	5 604	5 780	5 956	6 135	6 307	6 474	6 632	6 773	6 643
Tunisia - Tunisie	9 847	9 947	10 051	10 170	10 274	10 379	10 541	10 632	10 753	10 875
Uganda - Ouganda	26 938	27 807	28 725	29 691	30 722	31 791	32 884	33 987	35 148	36 346
United Republic of Tanzania - République Unie de Tanzanie	36 761	37 755	38 824	39 942	41 120	42 354	43 640	44 973	46 355	47 783
Western Sahara - Sahara occidental	379	403	428	448	475	482	498	515	532	549
Zambia - Zambie	10 895	11 175	11 470	11 782	12 110	12 457	12 825	13 217	13 634	14 075
Zimbabwe	12 673	12 693	12 711	12 724	12 740	12 784	12 889	13 077	13 359	13 724

AMERICA, NORTH - AMÉRIQUE DU NORD
Anguilla	12	12	13	13	13	13	14	14	14	14
Antigua and Barbuda - Antigua-et-Barbuda	81	82	83	83	84	85	86	87	88	89
Aruba	97	99	100	101	101	101	101	102	102	102
Bahamas	316	322	329	336	342	348	354	360	366	372
Barbados - Barbade	271	272	274	275	276	278	279	280	282	283

Outcome	0-12 weeks OR (95% CI)		>12 weeks OR (95% CI)		All weeks OR (95% CI)
Outcome	Unweighted	Weighted	Unweighted model	Weighted model	Unweighted model	Weighted model
Death	0.84	1.03	0.53	0.93	Hidden value 1	Hidden value 2