Search results

[C10] Incorrect channels-last contiguity flag handling in c10/core/SymbolicShapeMeta.h

actionable

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: dynamic shapes

module: internalsRelated to internal abstractions in c10 and ATenRelated to internal abstractions in c10 and ATen

oncall: pt2

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.

#178102 In pytorch/pytorch;

· benediktjohannes opened on Mar 22, 2026

torch.compile(dynamic=True) on CUDA: large eager vs compiled mismatch for BatchNorm2d + Conv2d

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: convolutionProblems related to convolutions (THNN, THCUNN, CuDNN)Problems related to convolutions (THNN, THCUNN, CuDNN)

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

module: inductor

oncall: pt2

triage review

Status: Open.

#178096 In pytorch/pytorch;

· hiahu329 opened on Mar 22, 2026

[Inductor CPU] Compilation fails when converting bool to float8_e4m3fn dtype

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

oncall: cpu inductorCPU Inductor issues for Intel team to triageCPU Inductor issues for Intel team to triage

Status: Open.

#178095 In pytorch/pytorch;

· yoheiueda opened on Mar 22, 2026

convolution_backward meta function uses grad_output instead of input to determine grad_input memory format

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: autogradRelated to torch.autograd, and the autograd engine in generalRelated to torch.autograd, and the autograd engine in general

module: nnRelated to torch.nnRelated to torch.nn

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

oncall: pt2

triage review

Status: Open.

#178092 In pytorch/pytorch;

· imperatormk opened on Mar 22, 2026

propagate_single_input_strategy copies input tensor_meta to output, producing wrong dtype for ops like convert_element_type

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

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

Status: Open.

#178091 In pytorch/pytorch;

· fmassa opened on Mar 22, 2026

bad-free in torch.sparse.spdiags due to missing bounds check on diagonal offsets

bot-mislabeledIf you notice that the claude-bot mislabeled an issueIf you notice that the claude-bot mislabeled an issue

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: crashProblem manifests as a hard crash, as opposed to a RuntimeErrorProblem manifests as a hard crash, as opposed to a RuntimeError

module: error checkingBugs related to incorrect/lacking error checkingBugs related to incorrect/lacking error checking

module: sparseRelated to torch.sparseRelated to torch.sparse

triage review

Status: Open.

#178089 In pytorch/pytorch;

· yyds1233 opened on Mar 22, 2026

[Profiler] Expose moduleHierarchy, activityType, and other KinetoEvent methods to Python

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

oncall: profilerprofiler-related issues (cpu, gpu, kineto)profiler-related issues (cpu, gpu, kineto)

Status: Open.

#178087 In pytorch/pytorch;

· aMayzner opened on Mar 22, 2026

DDP mode: CUDA error: an illegal memory access was encountered

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

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

Status: Open.

#178085 In pytorch/pytorch;

· rs-lsl opened on Mar 22, 2026

torch.compile introduces NaN in LayerNorm on valid float32 boundary inputs while eager remains finite

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

needs reproductionEnsure you have actionable steps to reproduce the issue. Someone else needs to confirm the repro.Ensure you have actionable steps to reproduce the issue. Someone else needs to confirm the repro.

Status: Open.

#178084 In pytorch/pytorch;

· hiahu329 opened on Mar 22, 2026

CompiledFunctionBackward RuntimeError, involving "invalid gradient" shape with a norm-like operation

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: aotdispatchumbrella label for AOTAutograd issuesumbrella label for AOTAutograd issues

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

module: pt2-dispatcherPT2 dispatcher-related issues (e.g., aotdispatch, functionalization, faketensor, custom-op,PT2 dispatcher-related issues (e.g., aotdispatch, functionalization, faketensor, custom-op,

oncall: pt2

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.

#178080 In pytorch/pytorch;

· lightvector opened on Mar 22, 2026

[MPS] Intermittent indexing out of bounds AcceleratorError

bot-mislabeledIf you notice that the claude-bot mislabeled an issueIf you notice that the claude-bot mislabeled an issue

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: advanced indexingRelated to x[i] = y, index functionsRelated to x[i] = y, index functions

module: macosMac OS related issuesMac OS related issues

module: mpsRelated to Apple Metal Performance Shaders frameworkRelated to Apple Metal Performance Shaders framework

triage review

Status: Open.

#178079 In pytorch/pytorch;

· hvaara opened on Mar 21, 2026

DISABLED test_serialize_export_inline_asm_elementwise_simple_cuda_float32 (__main__.TestHOPCUDA)

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

oncall: export

oncall: pt2

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

Status: Open.

#178077 In pytorch/pytorch;

· yangw-dev opened on Mar 21, 2026