Conversation
Nanmur
force-pushed
the
codex/fix-fold-batchnorm-iohw
branch
from
October 3, 2026 14:14
427afae to
5530511
Compare
Contributor
Author
|
The CPU failure is unrelated to this PR. The BatchNorm tests, including the IOHW regression case, passed in Jenkins; the only failure is the same WebGPU test failure present on the tested main commit (main CPU build #4125). I opened #20538 with the isolated upstream TIRx test fix. I will rerun this PR after that fix lands on main. |
tqchen
pushed a commit
that referenced
this pull request
Oct 3, 2026
## Description The grid-pack guard regression test added in #20424 uses TIRx syntax that had already been replaced on `main`: - #20437 removed the `s_tir=True` arguments from TIRx decorators. - #20519 renamed the TIRx parameter annotation from `T.Buffer` to `T.Tensor`. As a result, the current CPU CI fails during test collection with: ```text AttributeError: No script namespace 'Buffer' ``` This test-only change updates the new test to the current TIRx syntax. The WebGPU test body and generated-code assertion are unchanged. The same upstream failure is currently blocking unrelated PRs #20529 and #20535. ## Testing - `python -m py_compile tests/python/codegen/test_target_codegen_webgpu.py` - `python -m pre_commit run --files tests/python/codegen/test_target_codegen_webgpu.py` A matching current-`main` native library was not available locally, so the targeted WebGPU runtime test is left to CI.
Contributor
Author
|
@tvm-bot rerun |
1 similar comment
Contributor
Author
|
@tvm-bot rerun |
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Fixes #20534.
FoldBatchnormToConv2Dalready builds the OIHW broadcast reshape withrelax.op.reshape, but the IOHW branch calls.reshape()directly on a RelaxCall. This changes the IOHW branch to use the Relax reshape operator as well, preserving its(1, O, 1, 1)broadcast shape.The regression test uses a valid IOHW Conv2D followed by BatchNorm and verifies that the pass completes, returns a well-formed module, and removes the BatchNorm call.
Testing
python -m pytest tests/python/relax/test_transform_fold_batch_norm_to_conv2d.py::test_fold_batchnorm_iohw_kernel_layout -qpre-commit run --files python/tvm/relax/transform/fold_batch_norm_to_conv2d_for_inference.py tests/python/relax/test_transform_fold_batch_norm_to_conv2d.py