Update chat template and eos_token configs

#9
No description provided.
ykhwang changed pull request status to open

This PR needs to be made to fix issues with special token handling when using the chat completion endpoints. Similar needs to be done with the other instruct models.

Hugging Quants org

Thanks for the fixes @ykhwang , would you mind applying those to the rest of the affected quants? Otherwise we can do that on our end, so whatever is more suitable for you! Thanks again 🤗

alvarobartt changed pull request status to merged

Hello @alvarobartt , sorry for the late reply. I found that additional updates on chat templates for tool usage (https://huggingface.co/meta-llama/Meta-Llama-3.1-405B-Instruct/commit/cfe126a16d4108374a6f9cda6d117c0d08b99e23) and seems already applied by the latest commit. Thanks a lot!

Sign up or log in to comment