All issues

Search results

[OOB Performance] dynamo models regression on 2.15.0.dev20260911+xpu

Status: Open.

#5448 In intel/torch-xpu-ops;

· RUIJIEZHONG66166 opened on Sep 20, 2026

test_snode_runtime.py ComputeBoundedTests return 0 estimated runtime on XPU (roofline flops fallback swallowed)

Status: Open.

#5444 In intel/torch-xpu-ops;

· daisyden opened on Sep 18, 2026

[Inductor][XPU] test_lookup_table.py TestLookupTableE2E fails: choice filtering not applied on XPU

Status: Open.

#5443 In intel/torch-xpu-ops;

· daisyden opened on Sep 18, 2026

[Inductor][XPU] test_lookup_table.py TestLookupTable unit tests fail: lookup returns empty on XPU

Status: Open.

#5442 In intel/torch-xpu-ops;

· daisyden opened on Sep 18, 2026

[upstream ut] AssertionError: RuntimeError not raised in test/test_spectral_ops.py

Status: Open.

#5441 In intel/torch-xpu-ops;

· zxd1997066 opened on Sep 18, 2026

[upstream ut] AssertionError: Tensor-likes are not close! in test/test_spectral_ops.py

Status: Open.

#5440 In intel/torch-xpu-ops;

· zxd1997066 opened on Sep 18, 2026

Upgrade xpu support package to 2026.1.3

Status: Open.

#5436 In intel/torch-xpu-ops;

· chuanqi129 opened on Sep 18, 2026

·  

PT2.15

test_onednn_attention_gqa_vs_math_float16_batch_size_2_n_head_64_n_head_kv_16_q_size_9216_kv_size_77_head_dim_64_is_causal_False_xpu_float16 - AssertionError: Tensor-likes are not close!

hw: LNLLunar LakeLunar Lake

os: WindowsWindows PlatformWindows Platform

test: utUnit-test failure surfaceUnit-test failure surface

Status: Open.

Bug

#5434 In intel/torch-xpu-ops;

· mazurjax opened on Sep 18, 2026

The test enters an unrecoverable "Level Zero DEVICE_LOST" or resource failure state and may hang, requiring a machine reboot

functionalityFunctional defect: error/assert/hang/timeout (non-numeric)Functional defect: error/assert/hang/timeout (non-numeric)

randomFlaky/nondeterministic failure (intermittent)Flaky/nondeterministic failure (intermittent)

Status: Open.

#5432 In intel/torch-xpu-ops;

· mengfei25 opened on Sep 18, 2026

[xpu-alignment] [26-09-17] [26-09-17] abs_xpu has no Float8_e5m2 case, so isinf/isfinite raise NotImplementedError on XPU float8_e5m2 while CPU works

agent:doneAgent lifecycle: fix verified, patch ready for a human to applyAgent lifecycle: fix verified, patch ready for a human to apply

ai_generatedContent generated by an AI assistantContent generated by an AI assistant

Status: Open.

#5431 In intel/torch-xpu-ops;

· torchxpubot opened on Sep 18, 2026

[xpu-alignment] [26-09-17] [26-09-17] max_unpool2d XPU kernel narrows the int64 index to int32 before its bounds check and silently writes the wrong slot

agent:doneAgent lifecycle: fix verified, patch ready for a human to applyAgent lifecycle: fix verified, patch ready for a human to apply

ai_generatedContent generated by an AI assistantContent generated by an AI assistant

Status: Open.

#5430 In intel/torch-xpu-ops;

· torchxpubot opened on Sep 18, 2026

Feature request: scatter float8 support

skippedUsed for temp UT failure to parallel fixUsed for temp UT failure to parallel fix

Status: Open.

#5429 In intel/torch-xpu-ops;

· daisyden opened on Sep 18, 2026

·  

UT Backlog