Repository navigation
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/25032
Note: Links to docs will display an error until the docs builds have been completed. ⏳ 2 Pending, 1 Unrelated FailureAs of commit 6dd073e with merge base 8c77fe1 ( BROKEN TRUNK - The following job failed but were present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
Bot
added
the
CLA Signed
Collaborator
Author
|
@pytorchbot label "partner: arm" |
pytorch-bot
Bot
added
the
partner: arm
Collaborator
Author
|
@pytorchbot label "release notes: arm" |
Tighten the LLM custom SDPA export rewrite so it can be exercised in an export-only environment before any Arm backend lowering is added. The pass now accepts GQA decode shapes only when the query and KV head layout is compatible, and it cleanly falls back to the Python llama op registration when the compiled custom-op library is unavailable. Add export-pass coverage for a representative decode-time GQA shape and for the unsupported head-mismatch case without enable_gqa. The fallback registration also provides a Python reference body for llama.custom_sdpa so the rewritten graph remains executable in local tests. Test: PYTHONPATH=/private/tmp/executorch_patch1_sdpa/src \ PYTEST_DISABLE_PLUGIN_AUTOLOAD=1 \ /Users/bardem01/devkit7/executorch/.venv-fvp/bin/pytest -q \ extension/llm/export/test_export_passes.py Authored with assistance from OpenAI Codex. Signed-off-by: Baris Demir <baris.demir@arm.com> Change-Id: I880b2dbd201218467254db2eb5a981f00254623b
bdemirb
force-pushed
the
tmp-1366005-rebase2
branch
from
November 13, 2026 11:27
07f0c96 to
154f112
Compare
bdemirb
changed the title
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Tighten the LLM custom SDPA export rewrite so it can be exercised in an export-only environment before any Arm backend lowering is added. The pass now accepts GQA decode shapes only when the query and KV head layout is compatible, and it cleanly falls back to the Python llama op registration when the compiled custom-op library is unavailable.
Add export-pass coverage for a representative decode-time GQA shape and for the unsupported head-mismatch case without enable_gqa. The fallback registration also provides a Python reference body for llama.custom_sdpa so the rewritten graph remains executable in local tests.
cc @digantdesai @freddan80 @per @zingo @oscarandersson8218 @mansnils @Sebastian-Larsson @robell @rascani