Feature request

It would be nice to add in the APIs something to know if a model is supported, given a hub, model version, id and revision and quantization level.

I guess this would consist of listing all the files that will get downloaded by Transformers.js before we actually start it.

I am thinking something along those lines:

• grab the config.json
• determine the model architecture
• compute the files required and their URL
• check if present in cache, if not do HEAD calls

Motivation

This is to avoid starting a session if we know it'll fail (missing quantization, or missing version etc).

Your contribution

I am happy to do the patch if I am mentored