the tokenizer loading code usually loads from a folder
it looks like the tokenizer files don't clash with the model files,
and so the two folders are used for both
that's helpful

Reply via email to