1dda194
1
2
3
4
5
6
7
8
9
# evaluate.py and analysis.py use only the Python standard library. # Model encoding, retrieval and reranking additionally need: numpy torch transformers>=4.57 pillow pillow-heif qwen_vl_utils