Pad token

#4
by Teja-Gollapudi - opened

Hi what was the pad token and eos_token used to train this model?

For eos_token_id, i see it mentioned as 2 in the model config but 50256 in the custom tokenizer py file?

What values should i use for both?
Thank you

Hi what was the pad token and eos_token used to train this model?

For eos_token_id, i see it mentioned as 2 in the model config but 50256 in the custom tokenizer py file?

What values should i use for both?
Thank you

The eos_token_id is 50256. No pad token used in training.

rooa changed discussion status to closed

Sign up or log in to comment