adding no_cuda flag

[rahular]

The --no_cuda flag is missing from the flagset in extract_features.py. On running the current code, the following error occurs.

(py3.5) [rahul pytorch-pretrained-BERT]$ python extract_features.py \
>   --input_file=./input.txt \
>   --output_file=./output.jsonl \
>   --vocab_file=$BERT_BASE_DIR/vocab.txt \
>   --bert_config_file=$BERT_BASE_DIR/bert_config.json \
>   --init_checkpoint=$BERT_BASE_DIR/pytorch_model.bin \
>   --layers=-4 \
>   --max_seq_length=128 \
>   --batch_size=8
Traceback (most recent call last):
  File "extract_features.py", line 306, in <module>
    main()
  File "extract_features.py", line 223, in main
    device = torch.device("cuda" if torch.cuda.is_available() and not args.no_cuda else "cpu")
AttributeError: 'Namespace' object has no attribute 'no_cuda'

[thomwolf]

Thanks, I've added that manually (the library organization has changed a bit with the first pip release).