yzha/ctc_eval · Awesome job!

Hi @yzha

Awesome job adding a new metric! Let me know if I can help in anyway. I found a few minor things that could improve the current state a bit.

1. The example in the README should be:

results = ctc_score.compute(references=['hello world'], predictions=['hi world'])

(compute always expects lists for the inputs)

2. Install

We want to avoid automatically installing things in a users environment. It's better to add the dependencies to requirements.txt which will be automatically install on Spaces and the user will get a warning too.

3. Imports

You can put all imports at the beginning. evaluate will check automatically if those packages are available and throw an error telling the user which packages to install.

4. Docs

I think it would be great to document the possible configs in the README.

These are just minor things - great job! Let me know if you'd like my help!