Skip to content

reduce_scatter raises a RuntimeError #66

New issue
New issue
Open
Open
reduce_scatter raises a RuntimeError#66
@garrett361

Description

@garrett361
garrett361
opened on Jun 5, 2024

Hello, cross posting from ipex #647: torch-ccl does not support torch.distributed.reduce_scatter, despite the claims in the docs.

For instance, in 2.1.300+xpu we have:

torch-ccl/src/ProcessGroupCCL.cpp

Lines 871 to 877 in b9ce713

c10::intrusive_ptr<C10D_Work> ProcessGroupCCL::reduce_scatter(
std::vector<at::Tensor>& /* unused */,
std::vector<std::vector<at::Tensor>>& /* unused */,
const ReduceScatterOptions& /* unused */)
{
TORCH_CHECK(false, "ProcessGroupCCL does not support reduce_scatter");
}

where the TORCH_CHECK line raises a RuntimeError.

See the ipex ticket for more details and code to reproduce the error.

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