Add Phonon (Gradium 31M TTS) example with multi-backend export (#24810) by psiddh · Pull Request #24810 · pytorch/executorch

psiddh · 2026-11-09T20:35:50Z

Summary:

Adds examples/models/phonon: Gradium's Apache-2.0 Phonon 31M English text-to-speech model exported to one .pte, with a backend-agnostic C++ runner (phonon_runner) and a Python runner (run_phonon.py).
It follows nemotron3_diarization: one export script, CMake presets with make phonon-{cpu,cuda,mlx,vulkan}, CI jobs in pull.yml, mlx.yml and cuda.yml, and a short README in the same shape. CUDA writes aoti_cuda_blob.ptd beside the .pte.
Validated against the official runtime on XNNPACK, Vulkan, WebGPU, CUDA and MLX. Exports default to FP32 on every backend; --dtype bf16 is available on CUDA and MLX.
WebGPU runs in the browser: the README gives the --backend webgpu --webgpu-compat-ops export and links the Text to Speech demo of the ExecuTorch WebGPU Playground.
Also fixes an XNNPACK delegate use-after-free that Phonon hits: XNNCompiler::compileModel frees the named-data buffer of every packed filter/bias right after xnn_create_runtime_v4, even when another node still reads that constant. Phonon's StreamingConvTranspose1d feeds the transposed-conv bias to the conv and to the held-back tail subtraction, so a program loaded from a file (C++ runner, pybindings) segfaulted in the first mimi_decode call. A packed value that another node also reads now stays alive; an unknown node type keeps every constant alive. Same class of bug as the PReLU fix (#21480).

Reviewed By: JakeStevens

Differential Revision: D128116963

pytorch-bot · 2026-11-09T20:35:55Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/24810

📄 Preview Python docs built from this PR

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

❌ 1 New Failure, 1 Pending, 1 Unrelated Failure

As of commit ab60983 with merge base a2ddd36 ():

NEW FAILURE - The following job has failed:

pull / unittest-editable / linux / linux-job (gh)
examples/models/muse-glimmer/tests/test_cuda_sequence_export_inputs.py::SequenceExportInputsTest::test_small_logits_example_preserves_token_and_embedding_capacity

BROKEN TRUNK - The following job failed but were present on the merge base:

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

pull / unittest / linux / linux-job (gh) (trunk failure)
examples/models/muse-glimmer/tests/test_cuda_sequence_export_inputs.py::SequenceExportInputsTest::test_small_logits_example_preserves_token_and_embedding_capacity

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

meta-codesync · 2026-11-09T20:36:00Z

@psiddh has exported this pull request. If you are a Meta employee, you can view the originating Diff in D128116963.

Copilot

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota. Try requesting a review at a lower effort level.

github-actions · 2026-11-09T20:36:43Z

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.

mergennachin · 2026-11-09T20:46:11Z

+    parser.add_argument(
+        "--qlinear",
+        choices=("8da4w", "8da8w", "4w", "8w"),
+        help="Quantize the flow LM's linear layers (default: fp32)",


can we not give users this option, just have an option do 4bit and it will figure out whether it would do weight-only or dynamic-activation-quantization based on the backend

Done. Replaced --qlinear with --quantize 4bit|8bit; the export picks the scheme per backend (XNNPACK: 8da4w/8da8w; CUDA: weight-only 4w/8w, with 4-bit switching to BF16 automatically; Vulkan/WebGPU: 4w; MLX: asymmetric 4w/8w). Unsupported combos fail upfront with a one-line message saying what to run instead.

Copilot

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota. Try requesting a review at a lower effort level.

…ch#24810) Summary: Adds `examples/models/phonon`: Gradium's Apache-2.0 Phonon 31M English text-to-speech model exported to one `.pte`, with a backend-agnostic C++ runner (`phonon_runner`) and a Python runner (`run_phonon.py`). It follows `nemotron3_diarization`: one export script, CMake presets with `make phonon-{cpu,cuda,mlx,vulkan}`, CI jobs in `pull.yml`, `mlx.yml` and `cuda.yml`, and a short README in the same shape. CUDA writes `aoti_cuda_blob.ptd` beside the `.pte`. Validated against the official runtime on XNNPACK, Vulkan, WebGPU, CUDA and MLX. Exports default to FP32 on every backend; `--dtype bf16` is available on CUDA and MLX. WebGPU runs in the browser: the README gives the `--backend webgpu --webgpu-compat-ops` export and links the Text to Speech demo of the ExecuTorch WebGPU Playground. Also fixes an XNNPACK delegate use-after-free that Phonon hits: `XNNCompiler::compileModel` frees the named-data buffer of every packed filter/bias right after `xnn_create_runtime_v4`, even when another node still reads that constant. Phonon's `StreamingConvTranspose1d` feeds the transposed-conv bias to the conv and to the held-back tail subtraction, so a program loaded from a file (C++ runner, pybindings) segfaulted in the first `mimi_decode` call. A packed value that another node also reads now stays alive; an unknown node type keeps every constant alive. Same class of bug as the PReLU fix (pytorch#21480). Differential Revision: D128116963

Copilot

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota. Try requesting a review at a lower effort level.

…ch#24810) Summary: Adds `examples/models/phonon`: Gradium's Apache-2.0 Phonon 31M English text-to-speech model exported to one `.pte`, with a backend-agnostic C++ runner (`phonon_runner`) and a Python runner (`run_phonon.py`). It follows `nemotron3_diarization`: one export script, CMake presets with `make phonon-{cpu,cuda,mlx,vulkan}`, CI jobs in `pull.yml`, `mlx.yml` and `cuda.yml`, and a short README in the same shape. CUDA writes `aoti_cuda_blob.ptd` beside the `.pte`. Validated against the official runtime on XNNPACK, Vulkan, WebGPU, CUDA and MLX. Exports default to FP32 on every backend; `--dtype bf16` is available on CUDA and MLX. WebGPU runs in the browser: the README gives the `--backend webgpu --webgpu-compat-ops` export and links the Text to Speech demo of the ExecuTorch WebGPU Playground. Also fixes an XNNPACK delegate use-after-free that Phonon hits: `XNNCompiler::compileModel` frees the named-data buffer of every packed filter/bias right after `xnn_create_runtime_v4`, even when another node still reads that constant. Phonon's `StreamingConvTranspose1d` feeds the transposed-conv bias to the conv and to the held-back tail subtraction, so a program loaded from a file (C++ runner, pybindings) segfaulted in the first `mimi_decode` call. A packed value that another node also reads now stays alive; an unknown node type keeps every constant alive. Same class of bug as the PReLU fix (pytorch#21480). Differential Revision: D128116963

Copilot

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota. Try requesting a review at a lower effort level.

…ch#24810) Summary: Adds `examples/models/phonon`: Gradium's Apache-2.0 Phonon 31M English text-to-speech model exported to one `.pte`, with a backend-agnostic C++ runner (`phonon_runner`) and a Python runner (`run_phonon.py`). It follows `nemotron3_diarization`: one export script, CMake presets with `make phonon-{cpu,cuda,mlx,vulkan}`, CI jobs in `pull.yml`, `mlx.yml` and `cuda.yml`, and a short README in the same shape. CUDA writes `aoti_cuda_blob.ptd` beside the `.pte`. Validated against the official runtime on XNNPACK, Vulkan, WebGPU, CUDA and MLX. Exports default to FP32 on every backend; `--dtype bf16` is available on CUDA and MLX. WebGPU runs in the browser: the README gives the `--backend webgpu --webgpu-compat-ops` export and links the Text to Speech demo of the ExecuTorch WebGPU Playground. Also fixes an XNNPACK delegate use-after-free that Phonon hits: `XNNCompiler::compileModel` frees the named-data buffer of every packed filter/bias right after `xnn_create_runtime_v4`, even when another node still reads that constant. Phonon's `StreamingConvTranspose1d` feeds the transposed-conv bias to the conv and to the held-back tail subtraction, so a program loaded from a file (C++ runner, pybindings) segfaulted in the first `mimi_decode` call. A packed value that another node also reads now stays alive; an unknown node type keeps every constant alive. Same class of bug as the PReLU fix (pytorch#21480). Differential Revision: D128116963

Copilot

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota. Try requesting a review at a lower effort level.

…ch#24810) Summary: Adds `examples/models/phonon`: Gradium's Apache-2.0 Phonon 31M English text-to-speech model exported to one `.pte`, with a backend-agnostic C++ runner (`phonon_runner`) and a Python runner (`run_phonon.py`). It follows `nemotron3_diarization`: one export script, CMake presets with `make phonon-{cpu,cuda,mlx,vulkan}`, CI jobs in `pull.yml`, `mlx.yml` and `cuda.yml`, and a short README in the same shape. CUDA writes `aoti_cuda_blob.ptd` beside the `.pte`. Validated against the official runtime on XNNPACK, Vulkan, WebGPU, CUDA and MLX. Exports default to FP32 on every backend; `--dtype bf16` is available on CUDA and MLX. WebGPU runs in the browser: the README gives the `--backend webgpu --webgpu-compat-ops` export and links the Text to Speech demo of the ExecuTorch WebGPU Playground. Also fixes an XNNPACK delegate use-after-free that Phonon hits: `XNNCompiler::compileModel` frees the named-data buffer of every packed filter/bias right after `xnn_create_runtime_v4`, even when another node still reads that constant. Phonon's `StreamingConvTranspose1d` feeds the transposed-conv bias to the conv and to the held-back tail subtraction, so a program loaded from a file (C++ runner, pybindings) segfaulted in the first `mimi_decode` call. A packed value that another node also reads now stays alive; an unknown node type keeps every constant alive. Same class of bug as the PReLU fix (pytorch#21480). Differential Revision: D128116963

…ch#24810) Summary: Pull Request resolved: pytorch#24810 Adds `examples/models/phonon`: Gradium's Apache-2.0 Phonon 31M English text-to-speech model exported to one `.pte`, with a backend-agnostic C++ runner (`phonon_runner`) and a Python runner (`run_phonon.py`). It follows `nemotron3_diarization`: one export script, CMake presets with `make phonon-{cpu,cuda,mlx,vulkan}`, CI jobs in `pull.yml`, `mlx.yml` and `cuda.yml`, and a short README in the same shape. CUDA writes `aoti_cuda_blob.ptd` beside the `.pte`. Validated against the official runtime on XNNPACK, Vulkan, WebGPU, CUDA and MLX. Exports default to FP32 on every backend; `--dtype bf16` is available on CUDA and MLX. WebGPU runs in the browser: the README gives the `--backend webgpu --webgpu-compat-ops` export and links the Text to Speech demo of the ExecuTorch WebGPU Playground. Also fixes an XNNPACK delegate use-after-free that Phonon hits: `XNNCompiler::compileModel` frees the named-data buffer of every packed filter/bias right after `xnn_create_runtime_v4`, even when another node still reads that constant. Phonon's `StreamingConvTranspose1d` feeds the transposed-conv bias to the conv and to the held-back tail subtraction, so a program loaded from a file (C++ runner, pybindings) segfaulted in the first `mimi_decode` call. A packed value that another node also reads now stays alive; an unknown node type keeps every constant alive. Same class of bug as the PReLU fix (pytorch#21480). Differential Revision: D128116963

mergennachin · 2026-11-12T15:23:49Z

@JakeStevens - can you review this please? It touches Phonon model (TTS model) and also XNNPACK.

…ch#24810) Summary: Pull Request resolved: pytorch#24810 Adds `examples/models/phonon`: Gradium's Apache-2.0 Phonon 31M English text-to-speech model exported to one `.pte`, with a backend-agnostic C++ runner (`phonon_runner`) and a Python runner (`run_phonon.py`). It follows `nemotron3_diarization`: one export script, CMake presets with `make phonon-{cpu,cuda,mlx,vulkan}`, CI jobs in `pull.yml`, `mlx.yml` and `cuda.yml`, and a short README in the same shape. CUDA writes `aoti_cuda_blob.ptd` beside the `.pte`. Validated against the official runtime on XNNPACK, Vulkan, WebGPU, CUDA and MLX. Exports default to FP32 on every backend; `--dtype bf16` is available on CUDA and MLX. WebGPU runs in the browser: the README gives the `--backend webgpu --webgpu-compat-ops` export and links the Text to Speech demo of the ExecuTorch WebGPU Playground. Also fixes an XNNPACK delegate use-after-free that Phonon hits: `XNNCompiler::compileModel` frees the named-data buffer of every packed filter/bias right after `xnn_create_runtime_v4`, even when another node still reads that constant. Phonon's `StreamingConvTranspose1d` feeds the transposed-conv bias to the conv and to the held-back tail subtraction, so a program loaded from a file (C++ runner, pybindings) segfaulted in the first `mimi_decode` call. A packed value that another node also reads now stays alive; an unknown node type keeps every constant alive. Same class of bug as the PReLU fix (pytorch#21480). Differential Revision: D128116963

…ch#24810) Summary: Adds `examples/models/phonon`: Gradium's Apache-2.0 Phonon 31M English text-to-speech model exported to one `.pte`, with a backend-agnostic C++ runner (`phonon_runner`) and a Python runner (`run_phonon.py`). It follows `nemotron3_diarization`: one export script, CMake presets with `make phonon-{cpu,cuda,mlx,vulkan}`, CI jobs in `pull.yml`, `mlx.yml` and `cuda.yml`, and a short README in the same shape. CUDA writes `aoti_cuda_blob.ptd` beside the `.pte`. Validated against the official runtime on XNNPACK, Vulkan, WebGPU, CUDA and MLX. Exports default to FP32 on every backend; `--dtype bf16` is available on CUDA and MLX. WebGPU runs in the browser: the README gives the `--backend webgpu --webgpu-compat-ops` export and links the Text to Speech demo of the ExecuTorch WebGPU Playground. Also fixes an XNNPACK delegate use-after-free that Phonon hits: `XNNCompiler::compileModel` frees the named-data buffer of every packed filter/bias right after `xnn_create_runtime_v4`, even when another node still reads that constant. Phonon's `StreamingConvTranspose1d` feeds the transposed-conv bias to the conv and to the held-back tail subtraction, so a program loaded from a file (C++ runner, pybindings) segfaulted in the first `mimi_decode` call. A packed value that another node also reads now stays alive; an unknown node type keeps every constant alive. Same class of bug as the PReLU fix (pytorch#21480). Differential Revision: D128116963

Copilot

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota. Try requesting a review at a lower effort level.

…ch#24810) Summary: Pull Request resolved: pytorch#24810 Adds `examples/models/phonon`: Gradium's Apache-2.0 Phonon 31M English text-to-speech model exported to one `.pte`, with a backend-agnostic C++ runner (`phonon_runner`) and a Python runner (`run_phonon.py`). It follows `nemotron3_diarization`: one export script, CMake presets with `make phonon-{cpu,cuda,mlx,vulkan}`, CI jobs in `pull.yml`, `mlx.yml` and `cuda.yml`, and a short README in the same shape. CUDA writes `aoti_cuda_blob.ptd` beside the `.pte`. Validated against the official runtime on XNNPACK, Vulkan, WebGPU, CUDA and MLX. Exports default to FP32 on every backend; `--dtype bf16` is available on CUDA and MLX. WebGPU runs in the browser: the README gives the `--backend webgpu --webgpu-compat-ops` export and links the Text to Speech demo of the ExecuTorch WebGPU Playground. Also fixes an XNNPACK delegate use-after-free that Phonon hits: `XNNCompiler::compileModel` frees the named-data buffer of every packed filter/bias right after `xnn_create_runtime_v4`, even when another node still reads that constant. Phonon's `StreamingConvTranspose1d` feeds the transposed-conv bias to the conv and to the held-back tail subtraction, so a program loaded from a file (C++ runner, pybindings) segfaulted in the first `mimi_decode` call. A packed value that another node also reads now stays alive; an unknown node type keeps every constant alive. Same class of bug as the PReLU fix (pytorch#21480). Differential Revision: D128116963

JakeStevens

Review automatically exported from Phabricator review in Meta.

Summary: `XNNCompiler::compileModel` frees the named-data buffer of every packed filter and bias right after `xnn_create_runtime_v4`, even when another node still reads that constant by raw pointer. Phonon's `StreamingConvTranspose1d` feeds the transposed-conv bias to the convolution and to a subtraction, so a program loaded from a file (C++ runner, pybindings) crashed in its first `mimi_decode` call. Same class of bug as the PReLU fix (pytorch#21480). `getPackedValueIds` now removes from the packed set every value that a node reads as a plain input. It is a `switch` over `XNodeUnion` with no `default`, using the typed `xnode_union_as_*()` accessors, plus a `static_assert` on `XNodeUnion::MAX`, so a new node type is a build error rather than a silent fallback. A node it cannot read (no union value) logs an error and keeps every constant alive. Reviewed By: JakeStevens Differential Revision: D131816678

…ch#24810) Summary: Adds `examples/models/phonon`: Gradium's Apache-2.0 Phonon 31M English text-to-speech model exported to one `.pte`, with a backend-agnostic C++ runner (`phonon_runner`) and a Python runner (`run_phonon.py`). It follows `nemotron3_diarization`: one export script, CMake presets with `make phonon-{cpu,cuda,mlx,vulkan}`, CI jobs in `pull.yml`, `mlx.yml` and `cuda.yml`, and a short README in the same shape. CUDA writes `aoti_cuda_blob.ptd` beside the `.pte`. Validated against the official runtime on XNNPACK, Vulkan, WebGPU, CUDA and MLX. Exports default to FP32 on every backend; `--dtype bf16` is available on CUDA and MLX. WebGPU runs in the browser: the README gives the `--backend webgpu --webgpu-compat-ops` export and links the Text to Speech demo of the ExecuTorch WebGPU Playground. Also fixes an XNNPACK delegate use-after-free that Phonon hits: `XNNCompiler::compileModel` frees the named-data buffer of every packed filter/bias right after `xnn_create_runtime_v4`, even when another node still reads that constant. Phonon's `StreamingConvTranspose1d` feeds the transposed-conv bias to the conv and to the held-back tail subtraction, so a program loaded from a file (C++ runner, pybindings) segfaulted in the first `mimi_decode` call. A packed value that another node also reads now stays alive; an unknown node type keeps every constant alive. Same class of bug as the PReLU fix (pytorch#21480). Reviewed By: JakeStevens Differential Revision: D128116963

Copilot

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota. Try requesting a review at a lower effort level.

…ch#24810) Summary: Pull Request resolved: pytorch#24810 Adds `examples/models/phonon`: Gradium's Apache-2.0 Phonon 31M English text-to-speech model exported to one `.pte`, with a backend-agnostic C++ runner (`phonon_runner`) and a Python runner (`run_phonon.py`). It follows `nemotron3_diarization`: one export script, CMake presets with `make phonon-{cpu,cuda,mlx,vulkan}`, CI jobs in `pull.yml`, `mlx.yml` and `cuda.yml`, and a short README in the same shape. CUDA writes `aoti_cuda_blob.ptd` beside the `.pte`. Validated against the official runtime on XNNPACK, Vulkan, WebGPU, CUDA and MLX. Exports default to FP32 on every backend; `--dtype bf16` is available on CUDA and MLX. WebGPU runs in the browser: the README gives the `--backend webgpu --webgpu-compat-ops` export and links the Text to Speech demo of the ExecuTorch WebGPU Playground. Also fixes an XNNPACK delegate use-after-free that Phonon hits: `XNNCompiler::compileModel` frees the named-data buffer of every packed filter/bias right after `xnn_create_runtime_v4`, even when another node still reads that constant. Phonon's `StreamingConvTranspose1d` feeds the transposed-conv bias to the conv and to the held-back tail subtraction, so a program loaded from a file (C++ runner, pybindings) segfaulted in the first `mimi_decode` call. A packed value that another node also reads now stays alive; an unknown node type keeps every constant alive. Same class of bug as the PReLU fix (pytorch#21480). Reviewed By: JakeStevens Differential Revision: D128116963

psiddh requested a review from larryliu0820 as a code owner November 9, 2026 20:35

Copilot AI balanced review requested due to automatic review settings November 9, 2026 20:35

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 9, 2026

Copilot AI reviewed Nov 9, 2026

meta-codesync Bot added the meta-exported label Nov 9, 2026

mergennachin reviewed Nov 9, 2026

View reviewed changes

psiddh force-pushed the export-D128116963 branch from c5b0b09 to 0ad8bcf Compare November 11, 2026 21:34

psiddh requested a review from digantdesai as a code owner November 11, 2026 21:34

Copilot AI balanced review requested due to automatic review settings November 11, 2026 21:34

Copilot AI reviewed Nov 11, 2026

meta-codesync Bot changed the title ~~Add Phonon (Gradium 31M TTS) example with multi-backend export~~ Add Phonon (Gradium 31M TTS) example with multi-backend export (#24810) Nov 12, 2026

psiddh force-pushed the export-D128116963 branch from 0ad8bcf to 3c77de4 Compare November 12, 2026 00:52

Copilot AI balanced review requested due to automatic review settings November 12, 2026 00:52

Copilot AI reviewed Nov 12, 2026

psiddh force-pushed the export-D128116963 branch from 3c77de4 to 9fa5c89 Compare November 12, 2026 01:00

Copilot AI balanced review requested due to automatic review settings November 12, 2026 01:00

Copilot AI reviewed Nov 12, 2026

psiddh force-pushed the export-D128116963 branch from 9fa5c89 to 6716fb8 Compare November 12, 2026 01:28

Copilot AI balanced review requested due to automatic review settings November 12, 2026 01:28

Copilot AI reviewed Nov 12, 2026

mergennachin requested a review from JakeStevens November 12, 2026 15:23

psiddh force-pushed the export-D128116963 branch from 6716fb8 to 78694bb Compare November 13, 2026 17:27

Copilot AI balanced review requested due to automatic review settings November 13, 2026 17:27

Copilot AI reviewed Nov 13, 2026

JakeStevens approved these changes Nov 13, 2026

View reviewed changes

psiddh added 2 commits November 13, 2026 11:04

psiddh force-pushed the export-D128116963 branch from 78694bb to ab60983 Compare November 13, 2026 19:04

Copilot AI balanced review requested due to automatic review settings November 13, 2026 19:04

Copilot AI reviewed Nov 13, 2026

psiddh merged commit aa46257 into pytorch:main Nov 13, 2026
437 of 442 checks passed

Conversation

psiddh commented Nov 9, 2026 • edited by meta-codesync Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-bot Bot commented Nov 9, 2026 • 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/24810

❌ 1 New Failure, 1 Pending, 1 Unrelated Failure

Uh oh!

meta-codesync Bot commented Nov 9, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

github-actions Bot commented Nov 9, 2026

This PR needs a release notes: label

Uh oh!

mergennachin Nov 9, 2026

Choose a reason for hiding this comment

Uh oh!

psiddh Nov 12, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

mergennachin commented Nov 12, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

JakeStevens left a comment

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

psiddh commented Nov 9, 2026 •

edited by meta-codesync Bot

Loading

pytorch-bot Bot commented Nov 9, 2026 •

edited

Loading

This PR needs a `release notes:` label