Skip to content

torch.AcceleratorError: CUDA error: invalid device ordinal #176

@FindHao

Description

@FindHao

when we try reproducing a multigpu run and the original tensor locates at GPU6, our script raises this error. we need a better device mapping

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions