Bert-score

- Being compatible with Huggingface's transformers version >=4.0.0. Thanks to public contributers ([84](https://github.com/Tiiiger/bert_score/pull/84), [#85](https://github.com/Tiiiger/bert_score/issues/85), [#86](https://github.com/Tiiiger/bert_score/pull/86)).

- Support custom baseline files [74](https://github.com/Tiiiger/bert_score/pull/74)
- The option `--rescale-with-baseline` is changed to `--rescale_with_baseline` so that it is consistent with other options.

- Being compatible with Huggingface's transformers >=v3.0.0 and minor fixes ([58](https://github.com/Tiiiger/bert_score/pull/58), [#66](https://github.com/Tiiiger/bert_score/pull/66), [#68](https://github.com/Tiiiger/bert_score/pull/68))
- Several improvements related to efficency ([67](https://github.com/Tiiiger/bert_score/pull/67), [#69](https://github.com/Tiiiger/bert_score/pull/69))

- Upgrade to be compatible with transformers v2.11.0 (58)
- Bug Fixd: support CPU inference (50)

- Fixing the bug with empty strings [issue 47](https://github.com/Tiiiger/bert_score/issues/47).
- Supporting 6 [ELECTRA](https://github.com/google-research/electra) models and 24 smaller [BERT](https://github.com/google-research/bert) models.
- A new [Google sheet](https://docs.google.com/spreadsheets/d/1RKOVpselB98Nnh_EOC4A2BYn8_201tmPODpNWu4w7xI/edit?usp=sharing) for keeping the performance (i.e., pearson correlation with human judgment) of different models on WMT16 to-English.
- Including the script for tuning the best number of layers of an English pre-trained model on WMT16 to-English data (See the [details](tune_layers)).

- **Bug fixed**: fixing the bug in v0.3.1 when having multiple reference sentences.
- Supporting multiple reference sentences with our command-line tool