hugging-quants

/

Meta-Llama-3.1-70B-Instruct-AWQ-INT4

like 73

@alvarobartt Can you please sync the tokenizer to the offical llama3 tokenizer please

Confirm that all of the quants repos should use the updated chat template (it's the same template on every 3.1 model). The original template from Meta had several issues.

I can do this on Wednesday if no-one gets to it before then!

Updated the prompt template to the latest version of the llama models. Seems like they changed something about the tools:

tokenizer_config.json is 100% the same now as the original repo

Looks good to me, thanks a lot @SebastianBodza !

(We still need to update the others)

Edit: I was looking at the previous commit within this PR (and it's already merged!)