Search results

AOTInductor serializes entire underlying storage for each constant tensor, causing severe bloat for shared-storage weights

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

has workaround

module: aotinductoraot inductoraot inductor

module: serializationIssues related to serialization (e.g., via pickle, or otherwise) of PyTorch objectsIssues related to serialization (e.g., via pickle, or otherwise) of PyTorch objects

oncall: export

oncall: pt2

triage review

Status: Open.

#186176 In pytorch/pytorch;

· lothifia opened on Jun 4, 2026

ROCm: Flash/mem-efficient SDPA hangs during batched autoregressive generation with left-padded sequences

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

has workaround

module: deadlockProblems related to deadlocks (hang without exiting)Problems related to deadlocks (hang without exiting)

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

module: rocmAMD GPU support for PytorchAMD GPU support for Pytorch

module: sdpaAll things related to torch.nn.functional.scaled_dot_product_attentiionAll things related to torch.nn.functional.scaled_dot_product_attentiion

triage review

Status: Open.

#186174 In pytorch/pytorch;

· Finrandojin opened on Jun 4, 2026

AOTInductor miscompiles int64 aten.floor_divide when the DIVISOR is a tensor.

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

has workaround

module: aotinductoraot inductoraot inductor

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

oncall: export

oncall: pt2

triage review

Status: Open.

#186164 In pytorch/pytorch;

· Gasoonjia opened on Jun 4, 2026

torch.matmul misses mm fold for [S, 1, H] @ [H, K] when the size-1 batch dim has a non-contiguous stride

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

enhancementNot as big of a feature, but technically not a bug. Should be easy to fixNot as big of a feature, but technically not a bug. Should be easy to fix

has workaround

module: linear algebraIssues related to specialized linear algebra operations in PyTorch; includes matrix multiply matmulIssues related to specialized linear algebra operations in PyTorch; includes matrix multiply matmul

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

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.

#186148 In pytorch/pytorch;

· cuichenx opened on Jun 3, 2026

torch.onnx.export(dynamo=True) produces numerically wrong ONNX for masked_scatter

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: onnxRelated to torch.onnxRelated to torch.onnx

triage review

Status: Open.

#186146 In pytorch/pytorch;

· gonultasbu opened on Jun 3, 2026

AOTInductor Lowering Backward Compatibility

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: aotinductoraot inductoraot inductor

module: docsRelated to our documentation, both in docs/ and docblocksRelated to our documentation, both in docs/ and docblocks

oncall: export

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.

#186143 In pytorch/pytorch;

· leimao opened on Jun 3, 2026

fix qualname for OpOverload/OpOverloadPacket

actionable

module: custom-operatorscustom operators, custom ops, custom-operators, custom-opscustom operators, custom ops, custom-operators, custom-ops

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

Status: Open.

#186140 In pytorch/pytorch;

· zou3519 opened on Jun 3, 2026

The added test TestCompileWorker.test_shutdown_terminates_sidecar_worker_pool (and the corresponding version in TestCompileWorkerWithTimer) are flaky internally due to hitting timeout.

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: multiprocessingRelated to torch.multiprocessingRelated to torch.multiprocessing

module: testsIssues related to tests (not the torch.testing module)Issues related to tests (not the torch.testing module)

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.

#186136 In pytorch/pytorch;

· jansel opened on Jun 3, 2026

torch.scan gaps

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: higher order operatorstorch.cond and similartorch.cond and similar

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.

#186107 In pytorch/pytorch;

· zou3519 opened on Jun 3, 2026

Triton linux nightly wheels fail to install on Python 3.15 (requires-python: <3.15)

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: binariesAnything related to official binaries that we release to usersAnything related to official binaries that we release to users

module: ciRelated to continuous integrationRelated to continuous integration

module: python versionIssues related to specific Python versionsIssues related to specific Python versions

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.

#186099 In pytorch/pytorch;

· chuanqi129 opened on Jun 3, 2026

[inductor] insert_overlap_deps wraps non-overlap-critical compute with control_deps, blocking fusion and causing silent bf16 accuracy regression

Status: Open.

#186084 In pytorch/pytorch;

· lauri9 opened on Jun 3, 2026

Different ordering when exporting PixelUnshuffle to ONNX with dynamo=True

Status: Open.

#186080 In pytorch/pytorch;

· koningt opened on Jun 3, 2026