lm-sys / FastChat

An open platform for training, serving, and evaluating large language models. Release repo for Vicuna and Chatbot Arena.

Apache License 2.0

36.97k stars 4.56k forks source link

Can support the new model about Llama3 #3263

Open 1737686924 opened 7 months ago

1737686924 commented 7 months ago

Can support the new model about Llama3

namespace-Pt commented 7 months ago

+1 here. Especially the chat template

March-7 commented 7 months ago

Guys, is it the same as llama2?

1737686924 commented 7 months ago

伙计们，它和 llama2 一样吗？

I have not tested it myself, but I saw the evaluation results on meta, which are better and more advanced than llama2, and I heard that the open source Llama for 400B is being prepared, which should be better than GPT4

sohelzerdoumi commented 7 months ago

They are 3 pending pull request

3257 #3256 #3259

March-7 commented 7 months ago

Guys, when will the conversation template for new models like llama3 be updated to main？

hongyinjie commented 6 months ago

Change the file tokenizer_config.json: eos_token: end_of_text ==> eot_id

it work!

Oscarjia commented 6 months ago

@hongyinjie What change are you referring to? does this can fix llama3 can't stop problem? https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct/discussions/4