introduce triton sdpa kernel to cuda backend #15859

Gasoonjia · 2025-11-17T20:31:58Z

Stack from ghstack (oldest at bottom):

-> introduce triton sdpa kernel to cuda backend #15859

Introduce Triton SDPA Kernel to CUDA Backend

This diff introduces a Triton-optimized implementation of scaled dot-product attention (SDPA) kernel to the CUDA backend. The new kernel is designed to replace the default Edge SDPA operator during graph transformation to accelerate the model inference and get rid of sdpa decomposition.

Changes

Added a new file sdpa.py to fbcode/executorch/backends/cuda/triton/kernels and fbcode/executorch/backends/cuda/triton/kernels directories, which contains the Triton-optimized SDPA kernel implementation.
Added a new file __init__.py to fbcode/executorch/backends/cuda/triton/replacement_pass, which replaces the given existing edge ops with target triton kernels.
Added tests for sdpa exporting with triton kernel. Without the triton kernel, sdpa model can not be exported.

Purpose

The purpose of this diff is to provide a high-performance SDPA kernel for the CUDA backend, which can be used to accelerate attention-based models on NVIDIA GPUs.

Differential Revision: D87259044

**Introduce Triton SDPA Kernel to CUDA Backend** This diff introduces a Triton-optimized implementation of scaled dot-product attention (SDPA) kernel to the CUDA backend. The new kernel is designed to replace the default Edge SDPA operator during graph transformation to accelerate the model inference and get rid of sdpa decomposition. **Changes** * Added a new file `sdpa.py` to `fbcode/executorch/backends/cuda/triton/kernels` and `fbcode/executorch/backends/cuda/triton/kernels` directories, which contains the Triton-optimized SDPA kernel implementation. * Added a new file `__init__.py` to `fbcode/executorch/backends/cuda/triton/replacement_pass`, which replaces the given existing edge ops with target triton kernels. * Added tests for sdpa exporting with triton kernel. Without the triton kernel, sdpa model can not be exported. **Purpose** The purpose of this diff is to provide a high-performance SDPA kernel for the CUDA backend, which can be used to accelerate attention-based models on NVIDIA GPUs. Differential Revision: [D87259044](https://our.internmc.facebook.com/intern/diff/D87259044/) [ghstack-poisoned]

**Introduce Triton SDPA Kernel to CUDA Backend** This diff introduces a Triton-optimized implementation of scaled dot-product attention (SDPA) kernel to the CUDA backend. The new kernel is designed to replace the default Edge SDPA operator during graph transformation to accelerate the model inference and get rid of sdpa decomposition. **Changes** * Added a new file `sdpa.py` to `fbcode/executorch/backends/cuda/triton/kernels` and `fbcode/executorch/backends/cuda/triton/kernels` directories, which contains the Triton-optimized SDPA kernel implementation. * Added a new file `__init__.py` to `fbcode/executorch/backends/cuda/triton/replacement_pass`, which replaces the given existing edge ops with target triton kernels. * Added tests for sdpa exporting with triton kernel. Without the triton kernel, sdpa model can not be exported. **Purpose** The purpose of this diff is to provide a high-performance SDPA kernel for the CUDA backend, which can be used to accelerate attention-based models on NVIDIA GPUs. Differential Revision: [D87259044](https://our.internmc.facebook.com/intern/diff/D87259044/) ghstack-source-id: 323839094 Pull Request resolved: #15859

pytorch-bot · 2025-11-17T20:32:04Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/15859

📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

❗ 1 Active SEVs

There are 1 currently active SEVs. If your PR is affected, please view them below:

[SEV] Pytorch Trunk Test Failures due to Git operation failures

❌ 4 New Failures, 3 Unrelated Failures

As of commit d88e987 with merge base 179a155 ():

NEW FAILURES - The following jobs have failed:

pull / android / run-emulator (gh)
The process '/usr/bin/sh' failed with exit code 255
pull / unittest-editable / windows / windows-job (gh)
backends/xnnpack/test/recipes/test_xnnpack_recipes.py::TestXnnpackRecipes::test_all_models_with_recipes
Test CUDA Builds / test-model-cuda-e2e (google, gemma-3-4b-it, non-quantized) / linux-job (gh)
RuntimeError: Command docker exec -t 3c5bbe500ed0d381bb80605f657eb6d277a3883cd3fa877a18f8307472442b56 /exec failed with exit code 1
Test CUDA Builds / test-model-cuda-e2e (google, gemma-3-4b-it, quantized-int4-tile-packed) / linux-job (gh)
RuntimeError: Command docker exec -t 85d765a217406799fe8834d82e143ad64b1847f6c5d592e226a339a1de90cd2b /exec failed with exit code 1

FLAKY - The following job failed but was likely due to flakiness present on trunk:

Test CUDA Builds / test-model-cuda-e2e (openai, whisper-large-v3-turbo, non-quantized) / linux-job (gh) (matched linux rule in flaky-rules.json)
The process '/usr/bin/git' failed with exit code 128

BROKEN TRUNK - The following jobs failed but was present on the merge base:

👉 Rebase onto the `viable/strict` branch to avoid these failures

pull / unittest / linux / linux-job (gh) (trunk failure)
exir/backend/test/test_lowered_backend_module.py::TestBackendAPI::test_emit_nested_lowered_backend_module
pull / unittest-editable / linux / linux-job (gh) (trunk failure)
exir/backend/test/test_lowered_backend_module.py::TestBackendAPI::test_emit_nested_lowered_backend_module

This comment was automatically generated by Dr. CI and updates every 15 minutes.

github-actions · 2025-11-17T20:32:50Z

This PR needs a `release notes:` label

If your change should be included in the release notes (i.e. would users of this library care about this change?), please use a label starting with release notes:. This helps us keep track and include your important work in the next release notes.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "release notes: none"

For more information, see
https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.

**Introduce Triton SDPA Kernel to CUDA Backend** This diff introduces a Triton-optimized implementation of scaled dot-product attention (SDPA) kernel to the CUDA backend. The new kernel is designed to replace the default Edge SDPA operator during graph transformation to accelerate the model inference and get rid of sdpa decomposition. **Changes** * Added a new file `sdpa.py` to `fbcode/executorch/backends/cuda/triton/kernels` and `fbcode/executorch/backends/cuda/triton/kernels` directories, which contains the Triton-optimized SDPA kernel implementation. * Added a new file `__init__.py` to `fbcode/executorch/backends/cuda/triton/replacement_pass`, which replaces the given existing edge ops with target triton kernels. * Added tests for sdpa exporting with triton kernel. Without the triton kernel, sdpa model can not be exported. **Purpose** The purpose of this diff is to provide a high-performance SDPA kernel for the CUDA backend, which can be used to accelerate attention-based models on NVIDIA GPUs. Differential Revision: [D87259044](https://our.internmc.facebook.com/intern/diff/D87259044/) [ghstack-poisoned]

Pull Request resolved: #15859 **Introduce Triton SDPA Kernel to CUDA Backend** This diff introduces a Triton-optimized implementation of scaled dot-product attention (SDPA) kernel to the CUDA backend. The new kernel is designed to replace the default Edge SDPA operator during graph transformation to accelerate the model inference and get rid of sdpa decomposition. **Changes** * Added a new file `sdpa.py` to `fbcode/executorch/backends/cuda/triton/kernels` and `fbcode/executorch/backends/cuda/triton/kernels` directories, which contains the Triton-optimized SDPA kernel implementation. * Added a new file `__init__.py` to `fbcode/executorch/backends/cuda/triton/replacement_pass`, which replaces the given existing edge ops with target triton kernels. * Added tests for sdpa exporting with triton kernel. Without the triton kernel, sdpa model can not be exported. **Purpose** The purpose of this diff is to provide a high-performance SDPA kernel for the CUDA backend, which can be used to accelerate attention-based models on NVIDIA GPUs. ghstack-source-id: 323839094 @exported-using-ghexport Differential Revision: [D87259044](https://our.internmc.facebook.com/intern/diff/D87259044/)

**Introduce Triton SDPA Kernel to CUDA Backend** This diff introduces a Triton-optimized implementation of scaled dot-product attention (SDPA) kernel to the CUDA backend. The new kernel is designed to replace the default Edge SDPA operator during graph transformation to accelerate the model inference and get rid of sdpa decomposition. **Changes** * Added a new file `sdpa.py` to `fbcode/executorch/backends/cuda/triton/kernels` and `fbcode/executorch/backends/cuda/triton/kernels` directories, which contains the Triton-optimized SDPA kernel implementation. * Added a new file `__init__.py` to `fbcode/executorch/backends/cuda/triton/replacement_pass`, which replaces the given existing edge ops with target triton kernels. * Added tests for sdpa exporting with triton kernel. Without the triton kernel, sdpa model can not be exported. **Purpose** The purpose of this diff is to provide a high-performance SDPA kernel for the CUDA backend, which can be used to accelerate attention-based models on NVIDIA GPUs. Differential Revision: [D87259044](https://our.internmc.facebook.com/intern/diff/D87259044/) [ghstack-poisoned]

Pull Request resolved: #15859 **Introduce Triton SDPA Kernel to CUDA Backend** This diff introduces a kernel-generator (https://github.com/meta-pytorch/KernelAgent) driven, Triton-optimized implementation of scaled dot-product attention (SDPA) kernel to the CUDA backend. The new kernel is designed to replace the default Edge SDPA operator during graph transformation to accelerate the model inference and get rid of sdpa decomposition. **Changes** * Added a new file `sdpa.py` to `fbcode/executorch/backends/cuda/triton/kernels` and `fbcode/executorch/backends/cuda/triton/kernels` directories, which contains the Triton-optimized SDPA kernel implementation. * Added a new file `__init__.py` to `fbcode/executorch/backends/cuda/triton/replacement_pass`, which replaces the given existing edge ops with target triton kernels. * Added tests for sdpa exporting with triton kernel. Without the triton kernel, sdpa model can not be exported. **Purpose** The purpose of this diff is to provide a high-performance SDPA kernel for the CUDA backend, which can be used to accelerate attention-based models on NVIDIA GPUs. ghstack-source-id: 323849720 @exported-using-ghexport Differential Revision: [D87259044](https://our.internmc.facebook.com/intern/diff/D87259044/)

**Introduce Triton SDPA Kernel to CUDA Backend** This diff introduces a Triton-optimized implementation of scaled dot-product attention (SDPA) kernel to the CUDA backend. The new kernel is designed to replace the default Edge SDPA operator during graph transformation to accelerate the model inference and get rid of sdpa decomposition. **Changes** * Added a new file `sdpa.py` to `fbcode/executorch/backends/cuda/triton/kernels` and `fbcode/executorch/backends/cuda/triton/kernels` directories, which contains the Triton-optimized SDPA kernel implementation. * Added a new file `__init__.py` to `fbcode/executorch/backends/cuda/triton/replacement_pass`, which replaces the given existing edge ops with target triton kernels. * Added tests for sdpa exporting with triton kernel. Without the triton kernel, sdpa model can not be exported. **Purpose** The purpose of this diff is to provide a high-performance SDPA kernel for the CUDA backend, which can be used to accelerate attention-based models on NVIDIA GPUs. Differential Revision: [D87259044](https://our.internmc.facebook.com/intern/diff/D87259044/) [ghstack-poisoned]

Pull Request resolved: #15859 **Introduce Triton SDPA Kernel to CUDA Backend** This diff introduces a kernel-generator (https://github.com/meta-pytorch/KernelAgent) driven, Triton-optimized implementation of scaled dot-product attention (SDPA) kernel to the CUDA backend. The new kernel is designed to replace the default Edge SDPA operator during graph transformation to accelerate the model inference and get rid of sdpa decomposition. **Changes** * Added a new file `sdpa.py` to `fbcode/executorch/backends/cuda/triton/kernels` and `fbcode/executorch/backends/cuda/triton/kernels` directories, which contains the Triton-optimized SDPA kernel implementation. * Added a new `fbcode/executorch/backends/cuda/triton/replacement_pass`, which replaces the given existing edge ops with target triton kernels. * Added tests for sdpa exporting with triton kernel. Without the triton kernel, sdpa model can not be exported. **Purpose** The purpose of this diff is to provide a high-performance SDPA kernel for the CUDA backend, which can be used to accelerate attention-based models on NVIDIA GPUs. ghstack-source-id: 323864198 @exported-using-ghexport Differential Revision: [D87259044](https://our.internmc.facebook.com/intern/diff/D87259044/)

Copilot

Pull Request Overview

This PR introduces a Triton-optimized implementation of scaled dot-product attention (SDPA) kernel to the ExecuTorch CUDA backend, along with a graph transformation pass that replaces the default Edge SDPA operator with the optimized Triton kernel during compilation.

Key Changes:

Added Triton SDPA kernel with autotuning and support for different query/key-value sequence lengths
Implemented a replacement pass to swap Edge dialect SDPA operations with Triton kernels
Added SdpaModule toy model and test coverage for SDPA export

Reviewed Changes

Copilot reviewed 12 out of 12 changed files in this pull request and generated 7 comments.

Show a summary per file

File	Description
backends/cuda/triton/kernels/sdpa.py	Implements the Triton-optimized SDPA kernel with fused attention computation, autotuning configs, and fake implementation for torch.export
backends/cuda/triton/kernels/init.py	Exports the sdpa kernel from the kernels module
backends/cuda/triton/replacement_pass.py	Implements ReplaceEdgeOpWithTritonOpPass to replace edge dialect SDPA operations with Triton kernels during graph transformation
backends/cuda/triton/init.py	Module initialization that imports kernels and exports the replacement pass
backends/cuda/cuda_backend.py	Integrates the Triton replacement pass into the CUDA backend preprocessing pipeline and removes the SDPBackend.MATH constraint
backends/cuda/tests/test_cuda_export.py	Adds test_sdpa_single_kernel to verify SDPA export functionality
backends/cuda/tests/TARGETS	Adds dependency on toy_model for SDPA testing
backends/cuda/TARGETS	Adds triton_kernels and triton_replacement_pass build targets with appropriate dependencies
examples/models/toy_model/model.py	Adds SdpaModule toy model with example inputs using bfloat16 dtype
examples/models/toy_model/init.py	Exports SdpaModule from toy_model package
examples/models/init.py	Registers "sdpa" model in the Model enum and model registry
.github/workflows/cuda.yml	Adds "sdpa" to the CI test matrix

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

backends/cuda/triton/replacement_pass.py

backends/cuda/triton/kernels/sdpa.py

backends/cuda/triton/replacement_pass.py

Copilot · 2025-11-17T23:03:16Z

examples/models/toy_model/model.py

+class SdpaModule(torch.nn.Module, EagerModelBase):
+    def __init__(self):
+        super().__init__()
+


This class does not call EagerModelBase.init during initialization. (SdpaModule.init may be missing a call to a base class init)

Suggested change

EagerModelBase.__init__(self)

Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * #15859 * __->__ #15857 # aoti_torch_new_tensor_handle Shim for Cuda Backend ## summary This diff introduces the `aoti_torch_new_tensor_handle` shim for the CUDA backend. The shim is a wrapper around the `aoti_torch_new_tensor_handle` function, which creates a new tensor handle from an existing tensor, to unblock the custom triton kernel usage The diff affects the following files: * `fbcode/executorch/backends/aoti/common_shims.h`: The function prototype is declared. * `fbcode/executorch/backends/apple/metal/runtime/shims/memory.cpp`: A fallback implementation is provided for the Apple Metal backend to keep its behavior. * `fbcode/executorch/backends/cuda/runtime/shims/tests/test_aoti_torch_new_tensor_handle.cpp`: A test file is added for the CUDA backend. Differential Revision: [D87254968](https://our.internmc.facebook.com/intern/diff/D87254968/)

**Introduce Triton SDPA Kernel to CUDA Backend** This diff introduces a Triton-optimized implementation of scaled dot-product attention (SDPA) kernel to the CUDA backend. The new kernel is designed to replace the default Edge SDPA operator during graph transformation to accelerate the model inference and get rid of sdpa decomposition. **Changes** * Added a new file `sdpa.py` to `fbcode/executorch/backends/cuda/triton/kernels` and `fbcode/executorch/backends/cuda/triton/kernels` directories, which contains the Triton-optimized SDPA kernel implementation. * Added a new file `__init__.py` to `fbcode/executorch/backends/cuda/triton/replacement_pass`, which replaces the given existing edge ops with target triton kernels. * Added tests for sdpa exporting with triton kernel. Without the triton kernel, sdpa model can not be exported. **Purpose** The purpose of this diff is to provide a high-performance SDPA kernel for the CUDA backend, which can be used to accelerate attention-based models on NVIDIA GPUs. Differential Revision: [D87259044](https://our.internmc.facebook.com/intern/diff/D87259044/) [ghstack-poisoned]

Pull Request resolved: #15859 **Introduce Triton SDPA Kernel to CUDA Backend** This diff introduces a kernel-generator (https://github.com/meta-pytorch/KernelAgent) driven, Triton-optimized implementation of scaled dot-product attention (SDPA) kernel to the CUDA backend. The new kernel is designed to replace the default Edge SDPA operator during graph transformation to accelerate the model inference and get rid of sdpa decomposition. **Changes** * Added a new file `sdpa.py` to `fbcode/executorch/backends/cuda/triton/kernels` and `fbcode/executorch/backends/cuda/triton/kernels` directories, which contains the Triton-optimized SDPA kernel implementation. * Added a new `fbcode/executorch/backends/cuda/triton/replacement_pass`, which replaces the given existing edge ops with target triton kernels. * Added tests for sdpa exporting with triton kernel. Without the triton kernel, sdpa model can not be exported. **Purpose** The purpose of this diff is to provide a high-performance SDPA kernel for the CUDA backend, which can be used to accelerate attention-based models on NVIDIA GPUs. ghstack-source-id: 323941017 @exported-using-ghexport Differential Revision: [D87259044](https://our.internmc.facebook.com/intern/diff/D87259044/)

larryliu0820

Please take a look at Copilot comments as well

larryliu0820

Review automatically exported from Phabricator review in Meta.

**Introduce Triton SDPA Kernel to CUDA Backend** This diff introduces a Triton-optimized implementation of scaled dot-product attention (SDPA) kernel to the CUDA backend. The new kernel is designed to replace the default Edge SDPA operator during graph transformation to accelerate the model inference and get rid of sdpa decomposition. **Changes** * Added a new file `sdpa.py` to `fbcode/executorch/backends/cuda/triton/kernels` and `fbcode/executorch/backends/cuda/triton/kernels` directories, which contains the Triton-optimized SDPA kernel implementation. * Added a new file `__init__.py` to `fbcode/executorch/backends/cuda/triton/replacement_pass`, which replaces the given existing edge ops with target triton kernels. * Added tests for sdpa exporting with triton kernel. Without the triton kernel, sdpa model can not be exported. **Purpose** The purpose of this diff is to provide a high-performance SDPA kernel for the CUDA backend, which can be used to accelerate attention-based models on NVIDIA GPUs. Differential Revision: [D87259044](https://our.internmc.facebook.com/intern/diff/D87259044/) [ghstack-poisoned]

Pull Request resolved: #15859 **Introduce Triton SDPA Kernel to CUDA Backend** This diff introduces a kernel-generator (https://github.com/meta-pytorch/KernelAgent) driven, Triton-optimized implementation of scaled dot-product attention (SDPA) kernel to the CUDA backend. The new kernel is designed to replace the default Edge SDPA operator during graph transformation to accelerate the model inference and get rid of sdpa decomposition. **Changes** * Added a new file `sdpa.py` to `fbcode/executorch/backends/cuda/triton/kernels` and `fbcode/executorch/backends/cuda/triton/kernels` directories, which contains the Triton-optimized SDPA kernel implementation. * Added a new `fbcode/executorch/backends/cuda/triton/replacement_pass`, which replaces the given existing edge ops with target triton kernels. * Added tests for sdpa exporting with triton kernel. Without the triton kernel, sdpa model can not be exported. **Purpose** The purpose of this diff is to provide a high-performance SDPA kernel for the CUDA backend, which can be used to accelerate attention-based models on NVIDIA GPUs. ghstack-source-id: 323996159 @exported-using-ghexport Differential Revision: [D87259044](https://our.internmc.facebook.com/intern/diff/D87259044/)

**Introduce Triton SDPA Kernel to CUDA Backend** This diff introduces a Triton-optimized implementation of scaled dot-product attention (SDPA) kernel to the CUDA backend. The new kernel is designed to replace the default Edge SDPA operator during graph transformation to accelerate the model inference and get rid of sdpa decomposition. **Changes** * Added a new file `sdpa.py` to `fbcode/executorch/backends/cuda/triton/kernels` and `fbcode/executorch/backends/cuda/triton/kernels` directories, which contains the Triton-optimized SDPA kernel implementation. * Added a new file `__init__.py` to `fbcode/executorch/backends/cuda/triton/replacement_pass`, which replaces the given existing edge ops with target triton kernels. * Added tests for sdpa exporting with triton kernel. Without the triton kernel, sdpa model can not be exported. **Purpose** The purpose of this diff is to provide a high-performance SDPA kernel for the CUDA backend, which can be used to accelerate attention-based models on NVIDIA GPUs. Differential Revision: [D87259044](https://our.internmc.facebook.com/intern/diff/D87259044/) [ghstack-poisoned]

Pull Request resolved: #15859 **Introduce Triton SDPA Kernel to CUDA Backend** This diff introduces a kernel-generator (https://github.com/meta-pytorch/KernelAgent) driven, Triton-optimized implementation of scaled dot-product attention (SDPA) kernel to the CUDA backend. The new kernel is designed to replace the default Edge SDPA operator during graph transformation to accelerate the model inference and get rid of sdpa decomposition. **Changes** * Added a new file `sdpa.py` to `fbcode/executorch/backends/cuda/triton/kernels` and `fbcode/executorch/backends/cuda/triton/kernels` directories, which contains the Triton-optimized SDPA kernel implementation. * Added a new `fbcode/executorch/backends/cuda/triton/replacement_pass`, which replaces the given existing edge ops with target triton kernels. * Added tests for sdpa exporting with triton kernel. Without the triton kernel, sdpa model can not be exported. **Purpose** The purpose of this diff is to provide a high-performance SDPA kernel for the CUDA backend, which can be used to accelerate attention-based models on NVIDIA GPUs. ghstack-source-id: 324134561 @exported-using-ghexport Differential Revision: [D87259044](https://our.internmc.facebook.com/intern/diff/D87259044/)

Gasoonjia requested review from jackzhxng and lucylq as code owners November 17, 2025 20:31

Gasoonjia mentioned this pull request Nov 17, 2025

introduce aoti_torch_new_tensor_handle shim for cuda backend #15857

Merged

meta-cla bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Nov 17, 2025

meta-codesync bot added fb-exported meta-exported labels Nov 17, 2025

mergennachin requested a review from Copilot November 17, 2025 22:58

Copilot started reviewing on behalf of mergennachin November 17, 2025 22:59 View session

Copilot finished reviewing on behalf of mergennachin November 17, 2025 23:01

Copilot AI reviewed Nov 17, 2025

View reviewed changes

larryliu0820 approved these changes Nov 18, 2025

View reviewed changes

Gasoonjia closed this Nov 19, 2025

Gasoonjia had a problem deploying to cherry-pick-bot November 19, 2025 18:57 — with GitHub Actions Failure

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

introduce triton sdpa kernel to cuda backend #15859

introduce triton sdpa kernel to cuda backend #15859

Uh oh!

Gasoonjia commented Nov 17, 2025 •

edited

Loading

Uh oh!

pytorch-bot bot commented Nov 17, 2025 •

edited

Loading

Uh oh!

github-actions bot commented Nov 17, 2025

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI Nov 17, 2025

Uh oh!

larryliu0820 left a comment

Uh oh!

larryliu0820 left a comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

introduce triton sdpa kernel to cuda backend #15859

introduce triton sdpa kernel to cuda backend #15859

Uh oh!

Conversation

Gasoonjia commented Nov 17, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-bot bot commented Nov 17, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/15859

❗ 1 Active SEVs

❌ 4 New Failures, 3 Unrelated Failures

Uh oh!

github-actions bot commented Nov 17, 2025

This PR needs a release notes: label

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Reviewed Changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI Nov 17, 2025

Choose a reason for hiding this comment

Uh oh!

larryliu0820 left a comment

Choose a reason for hiding this comment

Uh oh!

larryliu0820 left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Gasoonjia commented Nov 17, 2025 •

edited

Loading

pytorch-bot bot commented Nov 17, 2025 •

edited

Loading

This PR needs a `release notes:` label