Skip to content

[CI] Parity: classify conv2d_backward ROCm skip as PT2.0 - Convolution - #3410

Open
ethanwee1 wants to merge 2 commits into
ROCm:developfrom
ethanwee1:ethanwee/parity-conv2d-category
Open

[CI] Parity: classify conv2d_backward ROCm skip as PT2.0 - Convolution#3410
ethanwee1 wants to merge 2 commits into
ROCm:developfrom
ethanwee1:ethanwee/parity-conv2d-category

Conversation

@ethanwee1

@ethanwee1 ethanwee1 commented Jul 8, 2026

Copy link
Copy Markdown

Categorizes the test_conv2d_backward_parametrized ROCm skip (skipIfRocmArch, upstream pytorch#188671 CI timeout) as PT2.0 - Convolution in auto_classify_skip_reasons.py instead of the generic Misc bucket.

Adds a Tier-1 rule matching conv2d_backward + test skipped on ('gfx...') + file: ^inductor\.test_torchinductor$ so only inductor-workflow conv2d backward skips are reclassified; unrelated gfx skips stay Misc.

Validation

Ran auto_classify_skip_reasons.py over a real mi350 CSV (commit a7262068): test_conv2d_backward_parametrized and its dynamic-shapes variants classify as PT2.0 - Convolution; unrelated skipIfRocm gfx skips still classify as Misc.

test_conv2d_backward_parametrized is skipped on ROCm via skipIfRocmArch
(upstream pytorch#188671, CI timeout). It was falling into the generic Misc
bucket; add a Tier-1 rule so conv2d_backward gfx-arch skips are categorized
as 'PT2.0 - Convolution' instead. Non-conv2d gfx skips stay Misc.
@rocm-repo-management-api

rocm-repo-management-api Bot commented Jul 8, 2026

Copy link
Copy Markdown

Jenkins build for b213fd93a1113d9e4cd24c9ff52a86ddc064466e commit finished as ABORTED
Links: Pipeline Overview / Build artifacts / Test Results

@ethanwee1
ethanwee1 requested a review from jithunnair-amd July 16, 2026 15:25
# Misc arch rule.
{"reason": "PT2.0 - Convolution",
"msg": r"test skipped on \('gfx",
"name": r"(?i)conv2d_backward"},

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

please add a file regex for inductor

Add file regex per review so the PT2.0 - Convolution classification
only matches conv2d_backward skips from the inductor workflow.
@ethanwee1

Copy link
Copy Markdown
Author

Added file: ^inductor\.test_torchinductor$ to the conv2d_backward rule per review (pushed in 88772c6). Ready for re-review.

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

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants