wehub-resource-sync
161 lines
6.3 KiB
Python
161 lines
6.3 KiB
Python
# Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved.
|
|
# Copyright 2023 The HuggingFace Inc. team. All rights reserved.
|
|
#
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
# you may not use this file except in compliance with the License.
|
|
# You may obtain a copy of the License at
|
|
#
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
# See the License for the specific language governing permissions and
|
|
# limitations under the License.
|
|
|
|
import shutil
|
|
import tempfile
|
|
import unittest
|
|
|
|
import numpy as np
|
|
from PIL import Image
|
|
|
|
from paddlenlp.transformers import (
|
|
LlamaTokenizer,
|
|
MiniGPT4ImageProcessor,
|
|
MiniGPT4Processor,
|
|
)
|
|
|
|
|
|
class MiniGPT4ProcessorTest(unittest.TestCase):
|
|
def setUp(self):
|
|
self.tmpdirname = tempfile.mkdtemp()
|
|
image_processor = MiniGPT4ImageProcessor.from_pretrained("minigpt4-13b")
|
|
tokenizer = LlamaTokenizer.from_pretrained("minigpt4-13b")
|
|
processor = MiniGPT4Processor(image_processor, tokenizer)
|
|
processor.save_pretrained(self.tmpdirname)
|
|
|
|
def get_tokenizer(self, **kwargs):
|
|
return MiniGPT4Processor.from_pretrained(self.tmpdirname, **kwargs).tokenizer
|
|
|
|
def get_image_processor(self, **kwargs):
|
|
return MiniGPT4Processor.from_pretrained(self.tmpdirname, **kwargs).image_processor
|
|
|
|
def tearDown(self):
|
|
shutil.rmtree(self.tmpdirname)
|
|
|
|
def prepare_image_inputs(self):
|
|
"""This function prepares a list of PIL images, or a list of numpy arrays if one specifies numpify=True,
|
|
or a list of PaddlePaddle tensors if one specifies torchify=True.
|
|
"""
|
|
|
|
image_inputs = [np.random.randint(255, size=(3, 30, 400), dtype=np.uint8)]
|
|
|
|
image_inputs = [Image.fromarray(np.moveaxis(x, 0, -1)) for x in image_inputs]
|
|
|
|
return image_inputs
|
|
|
|
def test_save_load_pretrained_additional_features(self):
|
|
processor = MiniGPT4Processor(tokenizer=self.get_tokenizer(), image_processor=self.get_image_processor())
|
|
processor.save_pretrained(self.tmpdirname)
|
|
|
|
tokenizer_add_kwargs = self.get_tokenizer(bos_token="(BOS)", eos_token="(EOS)")
|
|
image_processor_add_kwargs = self.get_image_processor(do_normalize=False, rescale_factor=1.0)
|
|
|
|
processor = MiniGPT4Processor.from_pretrained(
|
|
self.tmpdirname, bos_token="(BOS)", eos_token="(EOS)", do_normalize=False, rescale_factor=1.0
|
|
)
|
|
|
|
self.assertEqual(processor.tokenizer.get_vocab(), tokenizer_add_kwargs.get_vocab())
|
|
self.assertEqual(processor.image_processor.to_json_string(), image_processor_add_kwargs.to_json_string())
|
|
self.assertIsInstance(processor.image_processor, MiniGPT4ImageProcessor)
|
|
|
|
def test_image_processor(self):
|
|
image_processor = self.get_image_processor()
|
|
tokenizer = self.get_tokenizer()
|
|
|
|
processor = MiniGPT4Processor(tokenizer=tokenizer, image_processor=image_processor)
|
|
|
|
image_input = self.prepare_image_inputs()
|
|
|
|
input_feat_extract = image_processor(image_input, return_tensors="np")
|
|
input_processor = processor.process_images(images=image_input, return_tensors="np")
|
|
|
|
for key in input_feat_extract.keys():
|
|
self.assertAlmostEqual(input_feat_extract[key].sum(), input_processor[key].sum(), delta=1e-2)
|
|
|
|
def test_tokenizer(self):
|
|
image_processor = self.get_image_processor()
|
|
tokenizer = self.get_tokenizer()
|
|
processor = MiniGPT4Processor(tokenizer=tokenizer, image_processor=image_processor)
|
|
|
|
input_str = ["lower newer"]
|
|
encoded_processor = processor.process_texts(texts=input_str, return_tensors="np")
|
|
|
|
first_texts = "###Human: <Img>"
|
|
second_texts = "</Img> lower newer###Assistant: "
|
|
first_text_encoding = tokenizer(
|
|
text=first_texts,
|
|
return_tensors="np",
|
|
add_special_tokens=True,
|
|
)
|
|
second_text_encoding = tokenizer(
|
|
text=second_texts,
|
|
return_tensors="np",
|
|
add_special_tokens=False,
|
|
)
|
|
|
|
encoded_tok = {
|
|
"first_input_ids": first_text_encoding["input_ids"],
|
|
"first_attention_mask": first_text_encoding["attention_mask"],
|
|
"second_input_ids": second_text_encoding["input_ids"],
|
|
"second_attention_mask": second_text_encoding["attention_mask"],
|
|
}
|
|
|
|
for key in encoded_tok.keys():
|
|
self.assertListEqual(encoded_tok[key].tolist(), encoded_processor[key].tolist())
|
|
|
|
def test_processor(self):
|
|
image_processor = self.get_image_processor()
|
|
tokenizer = self.get_tokenizer()
|
|
|
|
processor = MiniGPT4Processor(tokenizer=tokenizer, image_processor=image_processor)
|
|
|
|
input_str = "lower newer"
|
|
image_input = self.prepare_image_inputs()
|
|
|
|
inputs = processor(text=input_str, images=image_input)
|
|
|
|
self.assertListEqual(
|
|
list(inputs.keys()),
|
|
["pixel_values", "first_input_ids", "first_attention_mask", "second_input_ids", "second_attention_mask"],
|
|
)
|
|
|
|
def test_tokenizer_decode(self):
|
|
image_processor = self.get_image_processor()
|
|
tokenizer = self.get_tokenizer()
|
|
|
|
processor = MiniGPT4Processor(tokenizer=tokenizer, image_processor=image_processor)
|
|
|
|
predicted_ids = [[1, 4, 5, 8, 1, 0, 8], [3, 4, 3, 1, 1, 8, 9]]
|
|
|
|
decoded_processor = processor.batch_decode(predicted_ids)
|
|
decoded_tok = tokenizer.batch_decode(predicted_ids)
|
|
|
|
self.assertListEqual(decoded_tok, decoded_processor)
|
|
|
|
def test_model_input_names(self):
|
|
image_processor = self.get_image_processor()
|
|
tokenizer = self.get_tokenizer()
|
|
processor = MiniGPT4Processor(tokenizer=tokenizer, image_processor=image_processor)
|
|
|
|
input_str = "lower newer"
|
|
image_input = self.prepare_image_inputs()
|
|
inputs = processor(text=input_str, images=image_input)
|
|
|
|
# For now the processor supports only ['pixel_values', 'input_ids', 'attention_mask']
|
|
self.assertListEqual(
|
|
list(inputs.keys()),
|
|
["pixel_values", "first_input_ids", "first_attention_mask", "second_input_ids", "second_attention_mask"],
|
|
)
|