'str' object has no attribute 'detach'

[sucong2]

제가 A6000 GPU를 사용하여 트레이닝을 진행하였는데, 몇 가지 에러가 발생하였습니다.
"python3 run_squad.py --task korquad --config_file xlm-roberta.json"를 실행하였을 때 진행되다 1/3 지점에서 발생한 문제인데 해결하는 방법에 대해 여쭤보고자 합니다.

좋은 트레이닝 모델을 제공해주셔서 감사합니다.

05/12/2022 09:45:19 - INFO - main - ***** Running training *****
05/12/2022 09:45:19 - INFO - main - Num examples = 65811
05/12/2022 09:45:19 - INFO - main - Num Epochs = 7
05/12/2022 09:45:19 - INFO - main - Train batch size per GPU = 4
05/12/2022 09:45:19 - INFO - main - Total train batch size (w. parallel, distributed & accumulation) = 4
05/12/2022 09:45:19 - INFO - main - Gradient Accumulation steps = 1
05/12/2022 09:45:19 - INFO - main - Total optimization steps = 115171
05/12/2022 09:52:44 - INFO - main - Loading features from cached file data/cached_dev_xlm-roberta-base_512
05/12/2022 09:52:48 - INFO - main - ***** Running evaluation 4000 *****
05/12/2022 09:52:48 - INFO - main - Num examples = 6742
05/12/2022 09:52:48 - INFO - main - Batch size = 16
Traceback (most recent call last):-------------------------------| 0.00% [0/422 00:00<00:00]
File "run_squad.py", line 483, in
main(cli_args)
File "run_squad.py", line 435, in main
global_step, tr_loss = train(args, train_dataset, model, tokenizer)
File "run_squad.py", line 184, in train
results = evaluate(args, model, tokenizer, global_step=global_step)
File "run_squad.py", line 259, in evaluate
output = [to_list(output[i]) for output in outputs]
File "run_squad.py", line 259, in
output = [to_list(output[i]) for output in outputs]
File "run_squad.py", line 58, in to_list
return tensor.detach().cpu().tolist()