Haoye17
commented
3 months ago Hi @XiaoLei2123 !
We appreciate your interest in our work~
As for the error message, in the previous version of our codebase, we used this information for debugging. To fix it, you can simply comment out line 207 in /muffin/eval/muffin_inference_logp.py as indicated by the error message, or update the codebase to our latest one, we have commented it out in our latest code~
If you have any other questions, do not hesitate to ask, we are willing to help!
XiaoLei2123
Traceback (most recent call last): File "/running_package/code_package/./muffin/train/train_llava15.py", line 338, in
train(attn_implementation="flash_attention_2")
File "/running_package/code_package/./muffin/train/train_llava15.py", line 313, in train
model, data_module, tokenizer = init_model(
File "/running_package/code_package/./muffin/train/train_llava15.py", line 279, in init_model
data_module = make_dpo_data_module(tokenizer, data_args=data_args, reference_model=copy.deepcopy(model).cuda())
File "/running_package/code_package/./muffin/train/train_llava15.py", line 149, in make_dpo_data_module
train_dataset = DPODataset(tokenizer=tokenizer,
File "/running_package/code_package/./muffin/train/train_llava15.py", line 133, in init
self.list_data_dict = RLAIFVDataset(data_dir, reference_model, tokenizer,multimodal_cfg['image_token_len'], multimodal_cfg['image_processor'], multimodal_cfg['use_im_start_end'], is_llava15=True)
File "/running_package/code_package/muffin/data/datasets.py", line 40, in init
inference_logp(reference_model, tokenizer, hf_data, self.data_path,
File "/running_package/code_package/muffin/eval/muffin_inference_logp.py", line 326, in inference_logp
outputs = get_multimodal_sample_logps(model, dataloader, tokenizer, is_llava15=is_llava15) # win_logp_list, win_avg_logp_list, win_per_token_logp_list, rej_logp_list, rej_avg_logp_list, rej_per_token_logp_list
File "/running_package/code_package/muffin/eval/muffin_inference_logp.py", line 226, in get_multimodal_sample_logps
for batch in tqdm.tqdm(dataloader):
File "/miniconda3/envs/llava_cu122/lib/python3.10/site-packages/tqdm/std.py", line 1181, in iter
for obj in iterable:
File "/miniconda3/envs/llava_cu122/lib/python3.10/site-packages/torch/utils/data/dataloader.py", line 630, in next
data = self._next_data()
File "/miniconda3/envs/llava_cu122/lib/python3.10/site-packages/torch/utils/data/dataloader.py", line 1345, in _next_data
return self._process_data(data)
File "/miniconda3/envs/llava_cu122/lib/python3.10/site-packages/torch/utils/data/dataloader.py", line 1371, in _process_data
data.reraise()
File "/miniconda3/envs/llava_cu122/lib/python3.10/site-packages/torch/_utils.py", line 694, in reraise
raise exception
KeyError: Caught KeyError in DataLoader worker process 0.
Original Traceback (most recent call last):
File "/miniconda3/envs/llava_cu122/lib/python3.10/site-packages/torch/utils/data/_utils/worker.py", line 308, in _worker_loop
data = fetcher.fetch(index)
File "/miniconda3/envs/llava_cu122/lib/python3.10/site-packages/torch/utils/data/_utils/fetch.py", line 54, in fetch
return self.collate_fn(data)
File "/running_package/code_package/muffin/eval/muffin_inference_logp.py", line 207, in preference_collator_fn
idx=win_batch['idx']
KeyError: 'idx'
0%| | 0/13752 [00:00<?, ?it/s] Traceback (most recent call last): File "/running_package/code_package/./muffin/train/train_llava15.py", line 338, in
train(attn_implementation="flash_attention_2")
File "/running_package/code_package/./muffin/train/train_llava15.py", line 313, in train
model, data_module, tokenizer = init_model(
File "/running_package/code_package/./muffin/train/train_llava15.py", line 279, in init_model
data_module = make_dpo_data_module(tokenizer, data_args=data_args, reference_model=copy.deepcopy(model).cuda())
File "/running_package/code_package/./muffin/train/train_llava15.py", line 149, in make_dpo_data_module
train_dataset = DPODataset(tokenizer=tokenizer,
File "/running_package/code_package/./muffin/train/train_llava15.py", line 133, in init
self.list_data_dict = RLAIFVDataset(data_dir, reference_model, tokenizer,multimodal_cfg['image_token_len'], multimodal_cfg['image_processor'], multimodal_cfg['use_im_start_end'], is_llava15=True)
File "/running_package/code_package/muffin/data/datasets.py", line 40, in init
inference_logp(reference_model, tokenizer, hf_data, self.data_path,
File "/running_package/code_package/muffin/eval/muffin_inference_logp.py", line 326, in inference_logp
outputs = get_multimodal_sample_logps(model, dataloader, tokenizer, is_llava15=is_llava15) # win_logp_list, win_avg_logp_list, win_per_token_logp_list, rej_logp_list, rej_avg_logp_list, rej_per_token_logp_list
File "/running_package/code_package/muffin/eval/muffin_inference_logp.py", line 226, in get_multimodal_sample_logps
for batch in tqdm.tqdm(dataloader):
File "/miniconda3/envs/llava_cu122/lib/python3.10/site-packages/tqdm/std.py", line 1181, in iter
for obj in iterable:
File "/miniconda3/envs/llava_cu122/lib/python3.10/site-packages/torch/utils/data/dataloader.py", line 630, in next
data = self._next_data()
File "/miniconda3/envs/llava_cu122/lib/python3.10/site-packages/torch/utils/data/dataloader.py", line 1345, in _next_data
return self._process_data(data)
File "/miniconda3/envs/llava_cu122/lib/python3.10/site-packages/torch/utils/data/dataloader.py", line 1371, in _process_data
data.reraise()
File "/miniconda3/envs/llava_cu122/lib/python3.10/site-packages/torch/_utils.py", line 694, in reraise
raise exception
KeyError: Caught KeyError in DataLoader worker process 0.
Original Traceback (most recent call last):
File "/miniconda3/envs/llava_cu122/lib/python3.10/site-packages/torch/utils/data/_utils/worker.py", line 308, in _worker_loop
data = fetcher.fetch(index)
File "/miniconda3/envs/llava_cu122/lib/python3.10/site-packages/torch/utils/data/_utils/fetch.py", line 54, in fetch
return self.collate_fn(data)
File "/running_package/code_package/muffin/eval/muffin_inference_logp.py", line 207, in preference_collator_fn
idx=win_batch['idx']
KeyError: 'idx'