add encoding params(defaut set to utf-8) for train

[superjcd]

The initial _get_data function in trainer.py opens file without setting encoding, which may lead to a codec error.

[EricFillion]

Thanks for the PR! I'll take a closer look later this week. It seems related to #253

[EricFillion]

I agree, we should set utf-8 as the default encoding and allow users to modify which encoding is used. There are just several changes that are required before we merge these changes:

1. TCTrainer.train() is called from HappyTextClassification.train(). So, we need to add the "encoding" parameter to HappyTextClassification.train()
2. We need to add this functionality for HappyTextClassification.eval() and HappyTextClassification.test()
3. We also need to add this functionality for HappyGeneration, HappyQuestionAnswering, HappyTextToText and HappyWordPrediction.
4. We need to update the documentation

So, thanks again for creating this PR and feel free to help accomplish any of these tasks. I'll complete them if you're too busy, but this would be a great way to learn more about Happy Transformer and to contribute if you have time.

[superjcd]

OK , I will try to fix the encoding problems

[superjcd]

I added a encoding param for train, eval， andtest for both HappyTextClassification and HappyQuestionAnswering; But i left other classes unchanged you mentioned above, cuz I'm not quite sure about those classes

[EricFillion]

Thanks for contributing! I just created a new PR (#271) that will be merged and published soon.

I decided to go with your initial suggestion of hardcoding the encoding format. I explain why within the description for the new PR.

I included all of your commits to give you credit for your contributions, and thanks again for your help.

[EricFillion]

Version 4.4.0 is now live and contains these changes!