Skip to content

[BUG] autoset_settings() fails with MIG GPUΒ #156

@btravouillon

Description

@btravouillon

With MIG GPU, the value of CUDA_VISIBLE_DEVICES is a string.

This breaks the code at

devices = ast.literal_eval(os.environ["CUDA_VISIBLE_DEVICES"])

This is reproducible with an NVIDIA A100 GPU with a MIG:

$ echo $CUDA_VISIBLE_DEVICES 
MIG-d5385a41-e608-5c4f-92c5-0932c7b636cd
$ python3.10 -c 'import ast, os; devices = ast.literal_eval(os.environ["CUDA_VISIBLE_DEVICES"]);'
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/cvmfs/ai.mila.quebec/apps/arch/distro/python/3.10/lib/python3.10/ast.py", line 64, in literal_eval
    node_or_string = parse(node_or_string.lstrip(" \t"), mode='eval')
  File "/cvmfs/ai.mila.quebec/apps/arch/distro/python/3.10/lib/python3.10/ast.py", line 50, in parse
    return compile(source, filename, mode, flags,
  File "<unknown>", line 1
    MIG-d5385a41-e608-5c4f-92c5-0932c7b636cd
                      ^
SyntaxError: invalid decimal literal

Note that the string format may vary depending on the NVIDIA driver (=R470), see https://docs.nvidia.com/datacenter/tesla/mig-user-guide/index.html#cuda-visible-devices

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