lm-sys / lm-sys/FastChat

Ask about the usage of template

Open
#2,517 2 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
39.5k
Forks
4.8k
PR merge metrics
No merged PRs in 30d

Description

Thanks for the awesome work! I am not sure whether I have missed this.

I see from here, the `train.py` uses vicuna template.
https://github.com/lm-sys/FastChat/blob/c3ad73a854c912132683b0a6b3df06596040385c/fastchat/train/train.py#L87-L88

Do I need to change the template to "llama-2" if I would like to finetune llama2 model? I am a bit confused as if I set it to "llama-2", the training loss is always 0, but if I stick to "vicuna", the loss is okay.
I try to take a look at `train_baichuan.py`, it also uses "vicuna" instead of "baichuan..."
https://github.com/lm-sys/FastChat/blob/c3ad73a854c912132683b0a6b3df06596040385c/fastchat/train/train_baichuan.py#L80-L82
In that case, when should these templates be used?

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with fastchat/train/train.py and fastchat/train/train_baichuan.py at the referenced template settings, then trace how template selection affects fine-tuning. Done means documenting when each template should be used for the models discussed and explaining the observed zero-loss behavior.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Documentation
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.