chore: import upstream snapshot with attribution

PFPO/post_processors/pattern/tags.py

@@ -0,0 +1,52 @@
+import re
+
+
+def think_tag_cleaner(completion: str):
+    # Regular expression to match content between [BEGIN] and [END]
+    pattern = r'<thinking>(.*?)</thinking>'
+
+    # Find all matches
+    matches = re.findall(pattern, completion, re.DOTALL)
+
+    # Check if there's exactly one pair of tags
+    if len(matches) == 1:
+        return matches[0].strip()  # Return the content of the single match
+    else:
+        return False  # Return False for 0 or more than 1 pair of tags
+
+
+def solution_tag_cleaner(completion: str):
+    # Regular expression to match content between [BEGIN] and [END]
+    pattern = r'<solution>(.*?)</solution>'
+
+    matches = re.findall(pattern, completion, re.DOTALL)
+
+    # Check if there's exactly one pair of tags
+    if len(matches) == 1:
+        return matches[0].strip()  # Return the content of the single match
+    else:
+        return False  # Return False for 0 or more than 1 pair of tags
+
+
+def output_tag_cleaner(completion: str):
+    # Regular expression to match content between [BEGIN] and [END]
+    pattern = r'<Output>(.*?)</Output>'
+
+    matches = re.findall(pattern, completion, re.DOTALL)
+
+    # Check if there's exactly one pair of tags
+    if len(matches) == 1:
+        return matches[0].strip()  # Return the content of the single match
+    else:
+        return False  # Return False for 0 or more than 1 pair of tags
+
+
+_code_cleaners = {
+    "think": think_tag_cleaner,
+    "solution": solution_tag_cleaner,
+    "output": output_tag_cleaner
+}
+
+
+def get(name: str):
+    return _code_cleaners[name]