OSError: file bert-base-uncased/config.json not found

[pkuzengqi]

Environment info

• transformers version: 4.4.2
• Python version: 3.6
• PyTorch version (GPU?): 1.8.0 (Tesla V100)

Information

The problem arises when using:

from transformers import BertModel
model = BertModel.from_pretrained('bert-base-uncased')

Error Info (Some personal info has been replaced by ---)

file bert-base-uncased/config.json not found
Traceback (most recent call last):
  File "---/anaconda3/envs/attn/lib/python3.6/site-packages/transformers-4.2.2-py3.8.egg/transformers/configuration_utils.py", line 420, in get_config_dict
  File "---/anaconda3/envs/attn/lib/python3.6/site-packages/transformers-4.2.2-py3.8.egg/transformers/file_utils.py", line 1063, in cached_path
OSError: file bert-base-uncased/config.json not found

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "---.py", line 107, in <module>
    from_pretrained_input()
  File "---.py", line 96, in from_pretrained_input
    model = BertModel.from_pretrained('bert-base-uncased')
  File "---/anaconda3/envs/attn/lib/python3.6/site-packages/transformers-4.2.2-py3.8.egg/transformers/modeling_utils.py", line 962, in from_pretrained
  File "---/anaconda3/envs/attn/lib/python3.6/site-packages/transformers-4.2.2-py3.8.egg/transformers/configuration_utils.py", line 372, in from_pretrained
  File "---/anaconda3/envs/attn/lib/python3.6/site-packages/transformers-4.2.2-py3.8.egg/transformers/configuration_utils.py", line 432, in get_config_dict
OSError: Can't load config for 'bert-base-uncased'. Make sure that:

- 'bert-base-uncased' is a correct model identifier listed on 'https://huggingface.co/models'

- or 'bert-base-uncased' is the correct path to a directory containing a config.json file

what I have read:

https://github.com/huggingface/transformers/issues/353

what I have tried:

1. loading from a downloaded model file works well

wget https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-uncased.tar.gz

unzip the file and rename bert_config.json as config.json, then

model = BertModel.from_pretrained(BERT_BASE_UNCASED_CACHE)

2. enough disk space, enough memory, free GPU

3. open internet connection, no proxy

4. import pytorch_pretrained_bert as ppb
   assert 'bert-large-cased' in ppb.modeling.PRETRAINED_MODEL_ARCHIVE_MAP

5. The following models work well

   
   model = BertModel.from_pretrained('bert-base-cased')

model = RobertaModel.from_pretrained('roberta-base')

6. working well in server cmd but not in local pycharm (remote deployment to server)

Observation: 
- Pycharm can found the ```transfromers``` installed with pip, but that will trigger this problem 
- Pycharm cannot find the current  ```transformers```  installed with conda 
```conda install transformers=4.4 -n env -c huggingface```

[theainerd]

I'm also facing the same issue. Did you find any fix yet . ??

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.

[stepbystep88]

I'm also facing the same issue. Did you guys find any fix yet . ??

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.

[VlaDanilov]

Same problem here, please write if someone found a valid solution.

[thak123]

Facing same error

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.

[vildhet]

Hi, I've had the same error but with roberta-base. It appeared that I had an empty folder named roberta-base in my working directory. Removing it solved the issue.

[leoliu0]

I found this issue is caused by setting cache directory using checkpoint name TrainingArguments(checkpoint,evaluation_strategy='steps')

change checkpoint to something else resolve the issue

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.

[vikas95]

353

Got the same issue, thanks for reporting it here. Was able to fix it following after going through your comment.

[3ricchen]

Hi, I've had the same error but with roberta-base. It appeared that I had an empty folder named roberta-base in my working directory. Removing it solved the issue.

You are literally an angel.