lm-sys / lm-sys/FastChat

When using the new version of 'mtbench', I encountered an issue

Open
#2,453 0 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

When I was using the latest version of the code and running mt-bench within llm-judge, I encountered the following error:
```bash
llm_judge git:(main) ✗ python3 gen_model_answer.py --model-path /chenhui/zhangwuhan/model/baichuan2 --model-id baichuan2-base
[2023-09-20 15:49:04,731] [INFO] [real_accelerator.py:133:get_accelerator] Setting ds_accelerator to cuda (auto detect)
Output to data/mt_bench/model_answer/baichuan2-base.jsonl
Traceback (most recent call last):
File "/chenhui/zhangwuhan/evaluation/FastChat/fastchat/llm_judge/gen_model_answer.py", line 267, in
run_eval(
File "/chenhui/zhangwuhan/evaluation/FastChat/fastchat/llm_judge/gen_model_answer.py", line 54, in run_eval
get_answers_func(
File "/usr/local/lib/python3.10/dist-packages/torch/utils/_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "/chenhui/zhangwuhan/evaluation/FastChat/fastchat/llm_judge/gen_model_answer.py", line 83, in get_model_answers
model, tokenizer = load_model(
TypeError: load_model() got an unexpected keyword argument 'dtype'
```
I removed the 'dtype' parameter from the 'load_model' function. Will this affect the results? Also, will the results from the new version of 'mt-bench' differ significantly from the old version?

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/llm_judge/gen_model_answer.py, especially run_eval and get_model_answers, and inspect the load_model definition referenced by the traceback. Reproduce the shown mt-bench command, then verify that the call and function signature agree and determine whether removing dtype affects evaluation results.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
machine-learning, testing-qa
Issue type
Bug
Difficulty
3/5
Estimated time
1-2 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
30/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.