Search results

Non-blocking DtH/HtD not ordered properly under compile

bot-triagedThis is a label only to be used by the auto triage botThis is a label only to be used by the auto triage bot

high priority

module: correctness (silent)issue that returns an incorrect result silentlyissue that returns an incorrect result silently

module: cudaRelated to torch.cuda, and CUDA support in generalRelated to torch.cuda, and CUDA support in general

module: dynamo

module: inductor

oncall: pt2

triage review

Status: Open.

#182190 In pytorch/pytorch;

· xmfan opened on May 1, 2026

[Dynamo/Inductor] torch.compile changes output shape of x.norm(dim=-1)

bot-triagedThis is a label only to be used by the auto triage botThis is a label only to be used by the auto triage bot

module: correctness (silent)issue that returns an incorrect result silentlyissue that returns an incorrect result silently

module: inductor

module: norms and normalization

oncall: pt2

triage review

Status: Open.

#182181 In pytorch/pytorch;

· michael-960 opened on May 1, 2026

DISABLED test_with_mode_created_separately (__main__.TestTorchFunctionMode)

module: __torch_function__

module: flaky-testsProblem is a flaky test in CIProblem is a flaky test in CI

oncall: pt2

skippedDenotes a (flaky) test currently skipped in CI.Denotes a (flaky) test currently skipped in CI.

Status: Open.

#182175 In pytorch/pytorch;

· pytorch-bot opened on May 1, 2026

DISABLED test_torchvision_models_efficientnet_b0 (__main__.TestVisionTracing)

module: flaky-testsProblem is a flaky test in CIProblem is a flaky test in CI

module: fx

skippedDenotes a (flaky) test currently skipped in CI.Denotes a (flaky) test currently skipped in CI.

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#182174 In pytorch/pytorch;

· pytorch-bot opened on May 1, 2026

Add torch.utils.nan_guard for forward-pass NaN/Inf debugging

bot-triagedThis is a label only to be used by the auto triage botThis is a label only to be used by the auto triage bot

featureA request for a proper, new feature.A request for a proper, new feature.

module: NaNs and InfsProblems related to NaN and Inf handling in floating pointProblems related to NaN and Inf handling in floating point

module: nnRelated to torch.nnRelated to torch.nn

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#182156 In pytorch/pytorch;

· Barosinec opened on May 1, 2026

UNSTABLE pull / dynamo-cpython-test / test-osdc (dynamo_cpython)

module: ciRelated to continuous integrationRelated to continuous integration

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

unstable

Status: Open.

#182144 In pytorch/pytorch;

· huydhn opened on May 1, 2026

torch.compile/Inductor produces different result for float16 cast before elementwise add

bot-triagedThis is a label only to be used by the auto triage botThis is a label only to be used by the auto triage bot

module: correctness (silent)issue that returns an incorrect result silentlyissue that returns an incorrect result silently

module: halfRelated to float16 half-precision floatsRelated to float16 half-precision floats

module: inductor

oncall: pt2

triage review

Status: Open.

#182131 In pytorch/pytorch;

· rookieLiu2018 opened on May 1, 2026

[vllm] [2.12 regression] PyTorch Fullgraph Smoke Test: full-cudagraph LLM diverges from piecewise-cudagraph LLM at batch=123

bot-triagedThis is a label only to be used by the auto triage botThis is a label only to be used by the auto triage bot

module: correctness (silent)issue that returns an incorrect result silentlyissue that returns an incorrect result silently

module: cuda graphsAbility to capture and then replay streams of CUDA kernelsAbility to capture and then replay streams of CUDA kernels

module: inductor

module: regressionIt used to work, and now it doesn'tIt used to work, and now it doesn't

module: vllm

oncall: pt2

triage review

Status: Open.

#182125 In pytorch/pytorch;

· atalman opened on May 1, 2026

[vllm] [2.12 regression] AsyncTP TP=2 (mp backend) on Llama-3.2-1B-Instruct: outputs differ between fuse_gemm_comms=True/False

bot-triagedThis is a label only to be used by the auto triage botThis is a label only to be used by the auto triage bot

module: correctness (silent)issue that returns an incorrect result silentlyissue that returns an incorrect result silently

module: inductor

module: multi-gpuProblem is related to running on multiple GPUsProblem is related to running on multiple GPUs

module: regressionIt used to work, and now it doesn'tIt used to work, and now it doesn't

module: vllm

oncall: distributedAdd this issue/PR to distributed oncall triage queueAdd this issue/PR to distributed oncall triage queue

oncall: distributed parallelismsAdd this issue/PR to the distributed parallelisms oncall triage queueAdd this issue/PR to the distributed parallelisms oncall triage queue

oncall: pt2

triage review

Status: Open.

#182124 In pytorch/pytorch;

· atalman opened on May 1, 2026

Perf regression: nn.Module init ~3.4× slower in PT 2.12 due to trunc_normal_ rewrite (BltModel example)

bot-triagedThis is a label only to be used by the auto triage botThis is a label only to be used by the auto triage bot

module: halfRelated to float16 half-precision floatsRelated to float16 half-precision floats

module: initializationRelated to weight initialization on operatorsRelated to weight initialization on operators

module: nnRelated to torch.nnRelated to torch.nn

module: performanceIssues related to performance, either of kernel code or framework glueIssues related to performance, either of kernel code or framework glue

module: randomRelated to random number generation in PyTorch (rng generator)Related to random number generation in PyTorch (rng generator)

module: regressionIt used to work, and now it doesn'tIt used to work, and now it doesn't

triage review

Status: Open.

#182116 In pytorch/pytorch;

· penguinwu opened on May 1, 2026

DISABLED test_ifft_norm_forward (__main__.TestFFT1D)

module: dynamo

module: flaky-testsProblem is a flaky test in CIProblem is a flaky test in CI

oncall: pt2

skippedDenotes a (flaky) test currently skipped in CI.Denotes a (flaky) test currently skipped in CI.

Status: Open.

#182112 In pytorch/pytorch;

· pytorch-bot opened on May 1, 2026

DISABLED test_max_cached_size (__main__.TestCachingHostAllocatorConfig)

module: cudaRelated to torch.cuda, and CUDA support in generalRelated to torch.cuda, and CUDA support in general

module: flaky-testsProblem is a flaky test in CIProblem is a flaky test in CI

skippedDenotes a (flaky) test currently skipped in CI.Denotes a (flaky) test currently skipped in CI.

triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module

Status: Open.

#182111 In pytorch/pytorch;

· pytorch-bot opened on May 1, 2026