Add option for loading transformer model from file path

[oskarvanderwal]

It is currently not possible to pass a file path for loading a transformer model, only a transformer_type. I would like to run the syntax.py script for finetuned BERT models that I have saved locally. You could for example add another option for passing a file path here.

[oskarvanderwal]

No changes are required after all, it seems! The current code works fine with a file path to the fine-tuned transformer model as input (transformer_type), but requires the following:

• the model is saved using save_pretrained(); and
• the tokenizer is passed explicitly in the configuration.