[Gasoonjia]

Stack from ghstack (oldest at bottom):

• [executorch][cuda] Run transposed convolutions on the implicit-GEMM template #25091
• -> [executorch][cuda] Test 3-D convolutions on the implicit-GEMM template #25090
• [executorch][cuda] Run 1-D convolutions on the implicit-GEMM template #25089
• [executorch][cuda] Converted convolutions follow the output layout the graph recorded #25088
• [executorch][cuda] Route 3-input-channel window and kernel-equals-stride convolutions to the implicit-GEMM template #25087
• [executorch][cuda] Add an implicit-GEMM template for dense triton::convolution #25086
• [executorch][cuda] Route 2-D depthwise and few-channel grouped convolutions to a Triton direct-stencil kernel #25085
• [executorch][cuda] Add triton::convolution and the conv routing framework #25084
• [executorch][cuda] Add planned replacement to the Triton replacement pass #25083
• [executorch][cuda] Deprecate triton_kernel_mode=OFF #25082

The implicit-GEMM template accepts 3-D convolutions, but they were neither tested nor measured at 3-D shapes, including convolutions after an explicit pad (padding 0, the pad left as its own node, as causal time padding is written).

No template code changes. A benchmark case can carry an explicit pad, which both paths run before a padding-0 convolution as in the exported graph, and the shapes suite gains 3-D convolutions: causal 3x3x3 after an explicit pad, 1x1x1, and (3,1,1) with stride (2,1,1), at 96 to 640 channels. None is significantly faster than Inductor at all its shapes in bf16, so worth_routing is unchanged and Inductor keeps them; the 1x1x1 case compiles and matches eager through the Triton-only lowering, which uses Inductor's 3-D conv template.

test_conv_implicit_gemm.py runs every schedule on 3-D geometries and lowers a 3-D block by default (Inductor) and with every template accepting (all three convolutions routed, the pad kept), against eager. test_conv_routing.py checks each 3-D case through both APIs and that the causal pad stays a separate node.

This diff was authored with Claude Code.

Differential Revision: D129802595

NOTE FOR REVIEWERS: This PR has internal Meta-specific changes or comments, please review them on Phabricator!

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/25090

• 📄 Preview Python docs built from this PR

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

❌ 1 Cancelled Job, 1 Pending, 1 Unrelated Failure

As of commit 40ad43a with merge base dea39f0 ():

CANCELLED JOB - The following job was cancelled. Please retry:

• Test CUDA Builds / unittest-cuda (system) / linux-job (gh)
  ##[error]received SIGINT

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

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

• Test ROCm AOTI / test-voxtral-realtime-rocm-gfx950-rocm7.2 / linux-job (gh) (trunk failure)
  ##[error]The operation was canceled.

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

[github-actions]

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.