Closed toffeecat closed 2 weeks ago
here's the command python eval_utils/chair_eval.py --data-path /datapath/val2014 --model llava-1.5 --use-cd --use-fast-v --sample True --sample-greedy True #SID_greedy
Done! ╭─────────────────────────────── Traceback (most recent call last) ────────────────────────────────╮ SID/transformers/src/transformers/models/llama/modeling_ │ │ llama.py:770 in forward │ │ │ │ 767 │ │ │ │ │ │ 768 │ │ │ │ # FastV Token Rerank, Attention Mask Implementation │ │ 769 │ │ │ │ elif USE_FAST_V: │ │ ❱ 770 │ │ │ │ │ SYS_LENGTH= key_position['image_start'] │ │ 771 │ │ │ │ │ IMAGE_TOKEN_LENGTH = key_position['image_end'] - key_position['image │ │ 772 │ │ │ │ │ ATTENTION_RANK = self.fast_v_attention_rank │ │ 773 │ │ │ │ │ ATTENTION_RANK_ADD = self.fast_v_attention_rank_add │ ╰──────────────────────────────────────────────────────────────────────────────────────────────────╯ TypeError: 'NoneType' object is not subscriptable
Please set --beams to 1
huofushuo
commented
2 weeks ago huofushuo
commented
2 weeks ago
here's the command python eval_utils/chair_eval.py --data-path /datapath/val2014 --model llava-1.5 --use-cd --use-fast-v --sample True --sample-greedy True #SID_greedy
Done! ╭─────────────────────────────── Traceback (most recent call last) ────────────────────────────────╮ SID/transformers/src/transformers/models/llama/modeling_ │ │ llama.py:770 in forward │ │ │ │ 767 │ │ │ │ │ │ 768 │ │ │ │ # FastV Token Rerank, Attention Mask Implementation │ │ 769 │ │ │ │ elif USE_FAST_V: │ │ ❱ 770 │ │ │ │ │ SYS_LENGTH= key_position['image_start'] │ │ 771 │ │ │ │ │ IMAGE_TOKEN_LENGTH = key_position['image_end'] - key_position['image │ │ 772 │ │ │ │ │ ATTENTION_RANK = self.fast_v_attention_rank │ │ 773 │ │ │ │ │ ATTENTION_RANK_ADD = self.fast_v_attention_rank_add │ ╰──────────────────────────────────────────────────────────────────────────────────────────────────╯ TypeError: 'NoneType' object is not subscriptable