Locking whisper model version

[AndyLogi]

I would like to lock the version of the base.en model so that if improvements are made in the future I'm still using the same model.
I can see that the model is downloaded from a specific URL, stored in the _MODELS dictionary.

If a new model is created, will:

1. the model at that URL be replaced; or
2. a new PyPI version will be released with an updated dictionary.

Essentially, my question boils down to if I lock to openai-whisper==20230314 will I always be using the same base.en model or is there potential that this will change in the future?

[jongwook]

Yes! the code even has a SHA1 check to make sure it exactly matches with the intended checkpoint.

whisper/whisper/__init__.py

Line 20 in b38a1f2

"base.en": "https://openaipublic.azureedge.net/main/whisper/models/25a8566e1d0c1e2231d1c762132cd20e0f96a85d16145c3a00adf5d1ac670ead/base.en.pt",