lm-sys / lm-sys/FastChat

What is the fastest token generation speed for Vicuna-13B?

Open
#3,192 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

Hi, I was running the huggingface_api.py following its default setting, except the --temperature is reduced to 0.001.

I would like to reduce the inference time of Vicuna-13B-v1.3, however, I found that the fastest speed I can get is only 27 tokens/second. I am not sure if this is the fastest speed it is supposed to be and feel like it is too slow.

Could anyone please help to advice how can I improve the inference speed?
I am using Ubuntu 20.04.06, single NVIDIA A100 80GB GPU, model path lmsys/vicuna-7b-v1.3

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/serve/huggingface_api.py and reproduce the reported generation speed using the stated Ubuntu, A100, and model settings. First resolve whether the target is Vicuna-13B or the lmsys/vicuna-7b-v1.3 model; the issue does not define a specific code change or a measurable completion criterion.

Written by the indexing model from the issue text.

Assessment

Tech stack
linux, python, ubuntu
Domain
ai, performance
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
20/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.