mlcommons / mlcommons/inference

Multi-GPU run for Llama2-70b inference

Open
#2,165 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
1.6k
Forks
650
Avg merge
1d 22h
Merged PRs (30d)
6

Description

Hi,
Is it possible to run Llama2-70b inference on multiple GPUs? When I specify --device "cuda:0,1", I get this error:

main.py: error: argument --device: invalid choice: 'cuda:0,1' (choose from 'cpu', 'cuda:0')

I even tried without quotes. It seems that only two options are available: cpu or 1 gpu device. Is that right?

Contributor guide

Open the contributing guide

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 main.py and the --device argument, then reproduce the shown command-line validation error for cuda:0,1. Determine whether multi-GPU Llama2-70b inference is intended to be supported; done should be a clearly defined supported behavior or an explicit documented limitation, with validation for the resulting device configuration.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.