Search results

FR: wheels that support optional modules

Status: Open.

#67 In meta-pytorch/torchcomms;

· kiwi3shark opened on Nov 25, 2025

How to build with RCCLX?

Status: Open.

#66 In meta-pytorch/torchcomms;

· south-ocean opened on Nov 22, 2025

How to use C++ API with NCCLX?

Status: Open.

#61 In meta-pytorch/torchcomms;

· osayamenja opened on Nov 18, 2025

Extremely high HBM usage

Status: Open.

#59 In meta-pytorch/torchcomms;

· Gin-Sin opened on Nov 14, 2025

[Question] Multiple Vendor GPU Support

Status: Open.

#56 In meta-pytorch/torchcomms;

· jhlee508 opened on Nov 13, 2025

support dynamo/compile in torchcomms

Status: Open.

#55 In meta-pytorch/torchcomms;

· d4l3k opened on Nov 12, 2025

[Draft] RFC: Add XCCL Backend Support for Intel GPU in TorchComms

Status: Open.

#51 In meta-pytorch/torchcomms;

· zhangxiaoli73 opened on Nov 10, 2025

Named Symbol not Found Error on B200

Status: Open.

#50 In meta-pytorch/torchcomms;

· osayamenja opened on Nov 9, 2025

How to use rcclx

Status: Open.

#44 In meta-pytorch/torchcomms;

· yuguo-Jack opened on Nov 5, 2025

Question: Is it possible to get rid of <code>conda</code> when compiling?

Status: Open.

#26 In meta-pytorch/torchcomms;

· Gin-Sin opened on Oct 30, 2025

Request for reproduction script TP+GEMM overlapping in "Collective Communication for 100k+ GPUs"

Status: Open.

#24 In meta-pytorch/torchcomms;

· hanxdmech-ship-it opened on Oct 29, 2025

libglog and libfmt linkage fails due to looking for static, not shared, libs

Status: Open.

#21 In meta-pytorch/torchcomms;

· jeffhammond opened on Oct 28, 2025