You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Questions about running the cola dataset script #176
hello,An error occurs when I run the debert cola.sh script
Traceback (most recent call last):
File "examples/text-classification/run_glue.py", line 626, in
main()
File "examples/text-classification/run_glue.py", line 558, in main
train_result = trainer.train(resume_from_checkpoint=checkpoint)
File "/home/dengxingzhi/LoRA/examples/NLU/src/transformers/trainer.py", line 881, in train
self._load_state_dict_in_model(state_dict)
File "/home/dengxingzhi/LoRA/examples/NLU/src/transformers/trainer.py", line 1987, in _load_state_dict_in_model
if set(load_result.missing_keys) == set(self.model._keys_to_ignore_on_save):
TypeError: 'NoneType' object is not iterable
so,Is it a code problem,because running other scripts works fine
I met the same problem, and I added the following in the _load_state_dict_in_model() function in transformers/trainer.py
if self.model._keys_to_ignore_on_save == None:
self.model._keys_to_ignore_on_save = []
Then I could run the cola.sh script.
Activity
Zhoutong-Wu commentedon May 30, 2024
I met the same problem, and I added the following in the _load_state_dict_in_model() function in transformers/trainer.py
if self.model._keys_to_ignore_on_save == None:
self.model._keys_to_ignore_on_save = []
Then I could run the cola.sh script.