List view

[distributed] NotImplementedError: The operator 'symm_mem::fused_matmul_reduce_scatter' is not currently implemented for the XPU device

module: distributedFor distributed feature issueFor distributed feature issue

Status: Open.

Feature

#1547 In intel/torch-xpu-ops;

· PenghuiCheng opened on Apr 7, 2025

[distributed] AssertionError: 'fused_all_gather_matmul' not found in '# AOT ID: [\'2_inference\']\n......'

module: distributedFor distributed feature issueFor distributed feature issue

Status: Open.

Feature

#1548 In intel/torch-xpu-ops;

· PenghuiCheng opened on Apr 7, 2025

[distributed] AssertionError: 'fused_all_gather_scaled_matmul' not found in 'graph():\n......'

module: distributedFor distributed feature issueFor distributed feature issue

Status: Open.

Feature

#1549 In intel/torch-xpu-ops;

· PenghuiCheng opened on Apr 7, 2025

[distributed] NotImplementedError: The operator 'symm_mem::fused_scaled_matmul_reduce_scatter' is not currently implemented for the XPU device.

module: distributedFor distributed feature issueFor distributed feature issue

Status: Open.

Feature

#1551 In intel/torch-xpu-ops;

· PenghuiCheng opened on Apr 7, 2025

gets nan with complex dtype

bug_fix_stage1

client

dependency component: oneAPI

dtype: complex

module: dependency bugProblem is not caused by us, but caused by the library we useProblem is not caused by us, but caused by the library we use

Status: Open.

Bug

#1195 In intel/torch-xpu-ops;

· Stonepia opened on Dec 23, 2024

[distributed] Accuracy gap in _composable/fsdp on Xelink

bugSomething isn't workingSomething isn't working

module: distributedFor distributed feature issueFor distributed feature issue

Status: Open.

Bug

#1661 In intel/torch-xpu-ops;

· PenghuiCheng opened on May 14, 2025

Models got fail accuracy on BMG but passed on PVC

Accuracy

bug_fix_stage2

client

E2E

hw: BMG

module: inductor

os: Ubuntu

Status: Open.

Bug

#1305 In intel/torch-xpu-ops;

· mengfei25 opened on Jan 20, 2025

[E2E] HF/Timm/Torchbench models got "eager_two_runs_differ"

bug_fix_stage2

client

E2E

hw: BMG

os: WindowsWindows PlatformWindows Platform

Status: Open.

Bug

#1256 In intel/torch-xpu-ops;

· libohao1201 opened on Jan 7, 2025

[Windows][E2E] Several models got fail_accuracy on BMG/LNL.

bug_fix_stage2

client

E2E

module: inductor

os: WindowsWindows PlatformWindows Platform

Status: Open.

#1506 In intel/torch-xpu-ops;

· libohao1201 opened on Mar 25, 2025

Torchbench phlippe_resnet bfloat16 training got fail_accuracy

Accuracy

bug_fix_stage1

hw: PVC

Status: Open.

#1799 In intel/torch-xpu-ops;

· mengfei25 opened on Jul 2, 2025

[BMG-Windows][PT2.8]Torch-xpu-ops UT got accuracy issue

bug_fix_stage2

hw: BMG

os: WindowsWindows PlatformWindows Platform

Status: Open.

#1818 In intel/torch-xpu-ops;

· libohao1201 opened on Jul 8, 2025

CPP extension, extend from Linux to windows

feature_stage1_PT2.9

Status: Open.

#1852 In intel/torch-xpu-ops;

· ZhaoqiongZ opened on Jul 16, 2025

[release 2.8]Torchbench vision_maskrcnn (amp_b16/amp_fp16 inference) got fail_accuracy

Accuracy

bug_fix_stage2

hw: BMG

os: WindowsWindows PlatformWindows Platform

Status: Open.

Bug

#1866 In intel/torch-xpu-ops;

· libohao1201 opened on Jul 17, 2025

Torchbench model squeezenet1_1 and functorch_dp_cifar10 got fail_accuracy

Accuracy

bug_fix_stage2

hw: BMG

hw: PVC

Status: Open.

Bug

#1877 In intel/torch-xpu-ops;

· libohao1201 opened on Jul 23, 2025

[distributed] c10d::allgather_ and c10d::allreduce_ is not currently implemented for the XPU device

module: distributedFor distributed feature issueFor distributed feature issue

Status: Open.

Feature

#986 In intel/torch-xpu-ops;

· zhiyuan1i opened on Oct 19, 2024

Update to recommended shortcut API for kernel specific max work group size

windows_ciOnly for Windows CI triggerOnly for Windows CI trigger

Status: Open (in progress).

intel/torch-xpu-opsnumber 1989

#1989 In intel/torch-xpu-ops;

· CuiYifeng opened on Aug 29, 2025