Search results

Compiled HOPS do not support compilation when run under a custom DispatchMode

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

module: __torch_dispatch__

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.

#163013 In pytorch/pytorch;

· ydwu4 opened on Sep 15, 2025

sac(compile(f)): torch compile not work properly with selective activation checkpoint

high priority

module: __torch_dispatch__

module: activation checkpointingRelated to activation checkpointingRelated to activation checkpointing

module: dynamo

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.

#161889 In pytorch/pytorch;

· 984881878 opened on Sep 1, 2025

cudagraphs: <code>static_input_indices</code> incorrectly including SymInt graph args when using tensor subclasses + dynamic shapes

module: __torch_dispatch__

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

module: dynamic shapes

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.

#152343 In pytorch/pytorch;

· bdhirsh opened on Apr 28, 2025

TorchDispatchMode cann't capture the operator which name is aten::_index_put_ impl_

module: __torch_dispatch__

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.

#145041 In pytorch/pytorch;

· yangrudan opened on Jan 17, 2025

torch.compile() within TorchDispatchMode always causes an unknown guard failure.

dynamo-triage-jan2025

module: __torch_dispatch__

module: dynamo

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.

#144787 In pytorch/pytorch;

· galv opened on Jan 14, 2025

It is apparently very difficult to use decompose correctly in TorchDispatchMode

module: __torch_dispatch__

module: dispatchDispatchStub, Type, void pointer table, c10 dispatchDispatchStub, Type, void pointer table, c10 dispatch

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.

#141446 In pytorch/pytorch;

· ezyang opened on Nov 24, 2024

[2.5.0+] Running a compiled model under FlopCounterMode regresses its performance

module: __torch_dispatch__

module: dynamo

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.

#140909 In pytorch/pytorch;

· Birch-san opened on Nov 17, 2024

fix unnecessary re-trace in AOTDispatcher in subclass code path

module: __torch_dispatch__

module: aotdispatchumbrella label for AOTAutograd issuesumbrella label for AOTAutograd issues

module: compile-timeCompilation mechanism or time spent in (re)compilation, tracing, startupCompilation mechanism or time spent in (re)compilation, tracing, startup

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

tensor subclassRelated to tensor subclassesRelated to tensor subclasses

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.

#132651 In pytorch/pytorch;

· bdhirsh opened on Aug 5, 2024

Default argument removal for python dispatch fails for pin_memory on factory functions

module: __torch_dispatch__

module: libraryRelated to torch.library (for registering ops from Python)Related to torch.library (for registering ops from Python)

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.

#131772 In pytorch/pytorch;

· albanD opened on Jul 25, 2024

PyTorch complains <code>A leaf node was passed to _will_engine_execute_node</code> when calculating the gradient of a model warpped by <code>FlopCounterMode</code>.

actionable

module: __torch_dispatch__

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

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.

#131753 In pytorch/pytorch;

· undes1red opened on Jul 25, 2024

Cannot fakeify a tensor who's .grad field is a tensor subclass

module: __torch_dispatch__

module: fakeTensor

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.

#127470 In pytorch/pytorch;

· bdhirsh opened on May 29, 2024

torch.compile + dynamic shapes + tensor subclass graph output is broken

module: __torch_dispatch__

module: dynamic shapes

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.

#124619 In pytorch/pytorch;

· bdhirsh opened on Apr 22, 2024