Telugu performance?

#1
by n8duo - opened

I don't see any discussion about the model's performance on Telugu here. So I am slightly wondering if the inclusion of Telugu in the supported languages was a mistake. Is there any comment on this? Does the model indeed support Telugu?

Follow up: it seems the model does not support Telugu, per my initial attempts. It seems unable to distinguish between benign and highly offensive content. And the tokenizer is unable to tokenize Telugu.

Sign up or log in to comment