Transformers: cannot import name 'RobertaForTokenClassification'

[dsindex]

🐛 Bug

Model I am using (Bert, XLNet....): BERT

Language I am using the model on (English, Chinese....): English

The problem arise when using:

• [x] the official example scripts: (give details) examples/run_ner.py
• [ ] my own modified scripts: (give details)

The tasks I am working on is:

• [x] an official GLUE/SQUaD task: (give the name) NER on CoNLL2003 ENG
• [ ] my own task or dataset: (give details)

To Reproduce

Steps to reproduce the behavior:

1. $ pip install transformers
2. $ python
3. > from transformers import RobertaConfig, RobertaForTokenClassification, RobertaTokenizer
   Traceback (most recent call last):
   File "", line 1, in
   ImportError: cannot import name 'RobertaForTokenClassification'

Expected behavior

Environment

• OS: Ubuntu 18.04.3
• Python version: 3.6
• PyTorch version: 1.2.0
• PyTorch Transformers version (or branch): pip install transformers
• Using GPU ? Yes, CUDA 10
• Distributed of parallel setup ? No
• Any other relevant information:

Additional context

[dsindex]

• other bug
  
  
  
  • with '--evaluate_during_training'
  
  

File "/path-to/run_ner.py", line 167, in train
    results, _ = evaluate(args, model, tokenizer, labels, pad_token_label_id)
TypeError: evaluate() missing 1 required positional argument: 'mode'

[dsindex]

• update
  
  
  
  • using pip3 install git+https://github.com/huggingface/transformers.git --upgrade command, the first bug got away.
  
  

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[mohitsewak]

I am also facing the same issue, but even installing from git (as stated by dsindex) did not help.

[jyoti-peakon]

It could be because torch is not installed. transformers doesn't install torch automatically but needs the same to load the models. try pip install torch and import again!