init

VLMEvalKit/vlmeval/dataset/utils/__init__.py

+from .judge_util import build_judge, DEBUG_MESSAGE
+from .multiple_choice import extract_answer_from_item, prefetch_answer
+from .vqa_eval import levenshtein_distance
+
+
+__all__ = [
+    'build_judge', 'extract_answer_from_item', 'prefetch_answer',
+    'levenshtein_distance', 'DEBUG_MESSAGE'
+]

VLMEvalKit/vlmeval/dataset/utils/crpe.py

+import json
+import argparse
+from collections import defaultdict
+
+
+def is_correct(predict, answer):
+    # predict是标准答案 answer是预测
+    if len(answer) == 1:
+        return answer[0] == predict[0]
+    elif len(answer) != 1 and answer[0] in ['A', 'B', 'C', 'D']:
+        return answer[0] == predict[0]
+    elif len(answer) != 1 and answer[0] not in ['A', 'B', 'C', 'D']:
+        return predict[4:].lower() in answer.lower()