Skip to content

minGPT example does not verify the GPU count across all nodes #1395

@Michaelvll

Description

@Michaelvll

if not verify_min_gpu_count(min_gpus=_min_gpu_count):

The above line check the GPU count for the current process, which makes 2 node with 1 GPU each node fail to run, i.e. the slurm launcher script to fail:

https://github.com/pytorch/examples/blob/main/distributed/minGPT-ddp/mingpt/slurm/sbatch_run.sh#L19

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions