[Feature Request] Cache also the tokenizer.json when running the model

[felladrin]

Using Phi-3, I noticed it's re-downloading the tokenizer.json every time model.run() is called. Could it be cached, the same way the GGUF is being cached?

[FL33TW00D]

@felladrin Thanks for the reminder.

The infra is there i just haven't set it up yet: https://github.com/huggingface/ratchet/blob/master/crates/ratchet-web/src/db.rs#L158