[Feature]: Try to load model from cache without checking HF revision first #575

What feature would you like to request?

There is a certain inconvenience about model loading process

After model was downloaded, each time its instance is created, hf_hub do a request to hf to check whether a new revision is available
However, it might not be an expected behaviour

What we'd like to do instead:
Try to load model from cache if it exists, do not make any calls to HF
If model is not available: run the full hf_hub pipeline to download the model

Is there any additional information you would like to provide?

No response