lm-sys / lm-sys/FastChat

--gpus 1,2,3 does not work

Open
#2,054 10 comments 1 reaction 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

Hello,

I installed FastChat locally with commit "f60b66093ca08c44701af00114df92d3ab66f3ac".
I have a server with 4 GPUS. GPU 0 is busy and I would like to use the other 3 GPUs to run fastchat.
However, the `--gpus 1,2,3` seems to have no effect and FastChat tries to use GPU 0 as well.
I was able to circumvent this by setting CUDA_VISIBLE_DEVICES directly in the shell.

```bash
CUDA_VISIBLE_DEVICES=1,2,3 python3 -m fastchat.serve.cli --model-path lmsys/vicuna-13b-v1.3 --num-gpus 3
```

I can not verify this on other machines, so I wonder if this a local problem with my server or if there is a problem with how the env variable is set in the code.

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

Reproduce the issue with `python3 -m fastchat.serve.cli --model-path lmsys/vicuna-13b-v1.3 --num-gpus 3` and the `--gpus 1,2,3` option, then inspect the `fastchat.serve.cli` entry point and its environment-variable handling. Done means the option prevents GPU 0 from being used and preserves the documented workaround behavior.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
ai
Issue type
Bug
Difficulty
3/5
Estimated time
1-2 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
45/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.