huggingface / huggingface/picotron

Trying to install and run Picotron on CPU

Open
#36 1 comment 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
2.3k
Forks
203
PR merge metrics
No merged PRs in 30d

Description

I generated the configs with --use-cpu flag and still when I run it using torchrun locally end up with failing on flash_attn not being deployed.

I could not get flash-attn==2.5.0 installed in CPU. When going through the requirements for flash-attn, looks like it is not compatible with CPU.

Is CPU no longer supported for Picotron or am missing something to get flash-attn==2.5.0 deployed on CPUs?

Contributor guide

No contributing guide indexed for this repository

Research direction

Start by reproducing the local torchrun failure with configs generated using --use-cpu, then inspect the requirements and flash_attn==2.5.0 installation path. Determine whether CPU execution is supported and document the required configuration or the incompatibility; the issue is done when CPU support or its absence is clear.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
distributed-systems, machine-learning
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Quiet
Clarity
Needs clarification
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.