Skip to content

Conversation

psiddh
Copy link
Contributor

@psiddh psiddh commented Aug 29, 2025

  • Fix CI job: test-arm-backend (test_run_ethosu_fvp) / linux-job.

Add 'qdq_fusion_pass' flag to run.sh & aot_arm_compiler. This
allows certain CI jobs not to fail as we conditionally allow fusion op
to be enabled / disabled with this flag. Note that since the fusion op
passes are still WIP and not all ops are yet supported, it is good to
rely on this flag until we have full fledged support for fusion ops.

Test Plan:

  1. examples/arm/run.sh --et_build_root=arm_test/test_run --target=ethos-u55-128 --model_name=qops --bundleio --no_delegate --portable_kernels="aten::sub.out,aten::add.out,aten::mul.out" --qdq_fusion_op=true

// The above fails as 'QuantOpTest' has mixed binary operations.

  1. examples/arm/run.sh --et_build_root=arm_test/test_run
    --target=ethos-u55-128 --model_name=qops --bundleio --no_delegate
    --portable_kernels="aten::sub.out,aten::add.out,aten::mul.out" -->
    passes (as default flag value is set to false)

  2. examples/arm/run.sh --et_build_root=arm_test/test_run --target=ethos-u55-128 --model_name=qadd2 --no_delegate

Reviewers:

Subscribers:

Tasks:

Tags:

Summary

[PLEASE REMOVE] See CONTRIBUTING.md's Pull Requests for ExecuTorch PR guidelines.

[PLEASE REMOVE] If this PR closes an issue, please add a Fixes #<issue-id> line.

[PLEASE REMOVE] If this PR introduces a fix or feature that should be the upcoming release notes, please add a "Release notes: " label. For a list of available release notes labels, check out CONTRIBUTING.md's Pull Requests.

Test plan

[PLEASE REMOVE] How did you test this PR? Please write down any manual commands you used and note down tests that you have written if applicable.

@psiddh psiddh requested a review from digantdesai August 29, 2025 21:23
Copy link

pytorch-bot bot commented Aug 29, 2025

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/13824

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:

❌ 1 New Failure, 2 Pending

As of commit da132c3 with merge base 2107afa (image):

NEW FAILURE - The following job has failed:

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

@meta-cla 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 Aug 29, 2025
Copy link

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.

Test Plan: examples/arm/run.sh --et_build_root=arm_test/test_run --target=ethos-u55-128 --model_name=qadd2 --no_delegate

Reviewers:

Subscribers:

Tasks:

Tags:
shoumikhin and others added 4 commits August 29, 2025 16:02
Add 'qdq_fusion_pass' flag to run.sh & aot_arm_compiler. This
allows certain CI jobs not to fail as we conditionally allow fusion op
to be enabled / disabled with this flag. Note that since the fusion op
passes are still WIP and not all ops are yet supported, it is good to
rely on this flag until we have full fledged support for fusion ops.

Test Plan:
1. examples/arm/run.sh --et_build_root=arm_test/test_run --target=ethos-u55-128 --model_name=qops --bundleio --no_delegate --portable_kernels="aten::sub.out,aten::add.out,aten::mul.out" --qdq_fusion_op=true

// The above fails as 'QuantOpTest' has mixed binary operations.

2. examples/arm/run.sh --et_build_root=arm_test/test_run
   --target=ethos-u55-128 --model_name=qops --bundleio --no_delegate
--portable_kernels="aten::sub.out,aten::add.out,aten::mul.out"  -->
passes (as default flag value is set to false)

Reviewers:

Subscribers:

Tasks:

Tags:
@psiddh psiddh merged commit 9de4c16 into pytorch:main Aug 30, 2025
118 of 119 checks passed
@psiddh
Copy link
Contributor Author

psiddh commented Aug 30, 2025

@shoumikhin Looks I somehow your PRs were referred to as part of this change (not sure how this happened, but seems ok).

@digantdesai I accidentally pushed few more changes on top of your approved PR. my bad! Can you please review it ?
Here are the changes that got committed eventually : 9de4c16
(you reviewed .h & .cpp only as part of your approval aot_arm_compiler.py & run.sh changes which are needed to address the broken CI job.)

The extra changes I made to run.sh & aot_arm_compiler.py is to address the broken CI job on the trunk (as a result of recent quantized_add op) Let me know if I need to revert this PR ?

cc @jackzhxng

@zingo
Copy link
Collaborator

zingo commented Aug 30, 2025

https://ossci-raw-job-status.s3.amazonaws.com/log/pytorch/executorch/49224733347

@digantdesai I accidentally pushed few more changes on top of your approved PR. my bad! Can you please review it ?
Here are the changes that got committed eventually : 9de4c16
(you reviewed .h & .cpp only as part of your approval aot_arm_compiler.py & run.sh changes which are needed to address the broken CI job.)

The extra changes I made to run.sh & aot_arm_compiler.py is to address the broken CI job on the trunk (as a result of recent quantized_add op) Let me know if I need to revert this PR ?

Hi, thanks to helping making sure it works again :)
Do you have links to your PR fixes for the broken CI jobs maybe I can help you get it on over the weekend.

The link in your msg above 9de4c16 seem to point ot this change/PR.

@psiddh
Copy link
Contributor Author

psiddh commented Aug 31, 2025

https://ossci-raw-job-status.s3.amazonaws.com/log/pytorch/executorch/49224733347

@digantdesai I accidentally pushed few more changes on top of your approved PR. my bad! Can you please review it ?
Here are the changes that got committed eventually : 9de4c16
(you reviewed .h & .cpp only as part of your approval aot_arm_compiler.py & run.sh changes which are needed to address the broken CI job.)
The extra changes I made to run.sh & aot_arm_compiler.py is to address the broken CI job on the trunk (as a result of recent quantized_add op) Let me know if I need to revert this PR ?

Hi, thanks to helping making sure it works again :) Do you have links to your PR fixes for the broken CI jobs maybe I can help you get it on over the weekend.

The link in your msg above 9de4c16 seem to point ot this change/PR.

Thanks @zingo for chiming ! Here is the job that was broken when the initial quantized add PR landed, But as soon as this PR landed, the job is ok again.
Reason: This particular testcase in the job was failing, as the test 'QuantOpTest' has mixed ops (add/mul/sub) in its testcase. But so far we added support only to quantized_add only. Therefore I added a 'flag' in aot _arm_compiler/py & run.sh to skip the newly added pass until we have full op support with cmsis integ (mul/sub). With this fix, the job/CI : test-arm-backend no longer fails . Hope this clarifies /helps.

@jackzhxng
Copy link
Contributor

@psiddh Thank you! Approving "Summary: Minor cleanup post quantized_add op" to fix trunk

@zingo
Copy link
Collaborator

zingo commented Sep 1, 2025

Hi @psiddh I think we might mean different fails, I was talking about that this PR seem to have started to fail this jobs:

trunk / test-arm-backend (test_models_ethos-u55) / linux-job
trunk / test-arm-backend (test_models_ethos-u85) / linux-job
trunk / test-arm-backend (test_models_tosa) / linux-job

and read you msg above that you had a fix. I now understand, you mean this PR fixes:

trunk / test-arm-backend (test_run_ethosu_fvp) / linux-job

See:
https://hud.pytorch.org/hud/pytorch/executorch/main/1?per_page=50&name_filter=test-arm-backend&mergeEphemeralLF=true

Thanks for clarifying :)

As a note you can add the ciflow/trunk flag to trigger all this jobs to be tested on github before submitting.

@zingo
Copy link
Collaborator

zingo commented Sep 1, 2025

Possible fix for problems here #13843

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ciflow/trunk CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants