[x] Abstract models & tokenizers out to allow loading transformers tokenizers via @xenova/transformers
[x] Fix error with case in the model selector ← I had to use a nasty hack that recreates the callbacks every render, but it works
[x] Fix error with loading model .json from transformers ← we need to auth against the HF api to load tokenizers for gated models. Maybe we should do a proxy, or republish just the tokenizers somewhere? thoughts
[x] Fix segment calculation—have either been missing whitespace characters or getting extra junk :/ Needs another pass.
transformers
tokenizers via@xenova/transformers