[inductor] [bug fix] Fix `conv` on processing uint #145136

shaoyuyoung · 2025-01-18T02:55:12Z

ut

pytest -s -v test/inductor/test_torchinductor.py -k test_conv_errors_with_uint

cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @yf225 @chenyang78 @kadeng @muchulee8 @ColinPeppler @amjames @desertfire @chauhang @aakhundov

pytorch-bot · 2025-01-18T02:55:16Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/145136

📄 Preview Python docs built from this PR
📄 Preview C++ docs built from this PR
❓ Need help or want to give feedback on the CI? Visit the bot commands wiki or our office hours

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

❌ 30 New Failures

As of commit f5c60f2 with merge base 425804d ():

NEW FAILURES - The following jobs have failed:

Lint / lintrunner-noclang / linux-job (gh)
>>> Lint for torch/_meta_registrations.py:
linux-aarch64 / linux-jammy-aarch64-py3.10 / test (default, 1, 3, linux.arm64.m7g.4xlarge) (gh)
inductor/test_cpu_cpp_wrapper.py::TestCppWrapper::test_qconv2d_dequant_promotion_cpu_cpp_wrapper
linux-aarch64 / linux-jammy-aarch64-py3.10 / test (default, 1, 4, linux.arm64.2xlarge) (gh)
inductor/test_cpu_cpp_wrapper.py::TestCppWrapper::test_qconv2d_add_relu_cpu_cpp_wrapper
linux-aarch64 / linux-jammy-aarch64-py3.10 / test (default, 2, 3, linux.arm64.m7g.4xlarge) (gh)
inductor/test_cpu_cpp_wrapper.py::TestCppWrapper::test_qconv2d_add_relu_cpu_cpp_wrapper
linux-aarch64 / linux-jammy-aarch64-py3.10 / test (default, 2, 4, linux.arm64.2xlarge) (gh)
inductor/test_cpu_cpp_wrapper.py::TestCppWrapper::test_qconv2d_add_cpu_cpp_wrapper
linux-aarch64 / linux-jammy-aarch64-py3.10 / test (default, 3, 3, linux.arm64.m7g.4xlarge) (gh)
inductor/test_cpu_cpp_wrapper.py::TestCppWrapper::test_qconv2d_with_concat_cpu_cpp_wrapper
linux-aarch64 / linux-jammy-aarch64-py3.10 / test (default, 3, 4, linux.arm64.2xlarge) (gh)
inductor/test_cpu_cpp_wrapper.py::TestCppWrapper::test_qconv2d_with_concat_cpu_cpp_wrapper
linux-aarch64 / linux-jammy-aarch64-py3.10 / test (default, 4, 4, linux.arm64.2xlarge) (gh)
inductor/test_cpu_cpp_wrapper.py::DynamicShapesCppWrapperCpuTests::test_qconv2d_cpu_dynamic_shapes_cpp_wrapper
pull / linux-focal-cuda12.4-py3.10-gcc9 / test (default, 2, 5, lf.linux.4xlarge.nvidia.gpu) (gh)
inductor/test_mkldnn_pattern_matcher.py::TestPatternMatcher::test_qat_qconv2d_relu
pull / linux-focal-cuda12.4-py3.10-gcc9-sm89 / test (default, 2, 5, lf.linux.g6.4xlarge.experimental.nvidia.gpu) (gh)
inductor/test_mkldnn_pattern_matcher.py::TestPatternMatcher::test_qat_qconv2d_relu
pull / linux-focal-cuda12.4-py3.10-gcc9-sm89 / test (default, 3, 5, lf.linux.g6.4xlarge.experimental.nvidia.gpu) (gh)
inductor/test_mkldnn_pattern_matcher.py::TestPatternMatcher::test_qat_qconv2d
pull / linux-focal-py3.13-clang10 / test (default, 1, 5, lf.linux.4xlarge) (gh)
inductor/test_cpu_cpp_wrapper.py::TestCppWrapper::test_qconv2d_add_cpu_cpp_wrapper
pull / linux-focal-py3.13-clang10 / test (default, 2, 5, lf.linux.4xlarge) (gh)
inductor/test_cpu_cpp_wrapper.py::TestCppWrapper::test_qconv2d_maxpool2d_linear_dynamic_cpu_cpp_wrapper
pull / linux-focal-py3.13-clang10 / test (default, 3, 5, lf.linux.4xlarge) (gh)
inductor/test_cpu_cpp_wrapper.py::DynamicShapesCppWrapperCpuTests::test_qconv2d_add_relu_cpu_dynamic_shapes_cpp_wrapper
pull / linux-focal-py3.13-clang10 / test (default, 4, 5, lf.linux.4xlarge) (gh)
inductor/test_mkldnn_pattern_matcher.py::TestPatternMatcher::test_qat_qconv2d_relu
pull / linux-focal-py3.13-clang10 / test (default, 5, 5, lf.linux.4xlarge) (gh)
inductor/test_cpu_cpp_wrapper.py::TestCppWrapper::test_qconv2d_with_concat_cpu_cpp_wrapper
pull / linux-focal-py3.9-clang10 / test (default, 1, 5, lf.linux.4xlarge) (gh)
inductor/test_cpu_cpp_wrapper.py::TestCppWrapper::test_qconv2d_maxpool2d_linear_dynamic_cpu_cpp_wrapper
pull / linux-focal-py3.9-clang10 / test (default, 2, 5, lf.linux.4xlarge) (gh)
inductor/test_cpu_cpp_wrapper.py::TestCppWrapper::test_qconv2d_add_relu_cpu_cpp_wrapper
pull / linux-focal-py3.9-clang10 / test (default, 3, 5, lf.linux.4xlarge) (gh)
inductor/test_cpu_cpp_wrapper.py::TestCppWrapper::test_qconv2d_add_cpu_cpp_wrapper
pull / linux-focal-py3.9-clang10 / test (default, 4, 5, lf.linux.4xlarge) (gh)
inductor/test_mkldnn_pattern_matcher.py::TestPatternMatcher::test_qat_qconv2d_relu
pull / linux-focal-py3.9-clang10 / test (default, 5, 5, lf.linux.4xlarge) (gh)
inductor/test_cpu_cpp_wrapper.py::TestCppWrapper::test_qconv2d_with_concat_cpu_cpp_wrapper
pull / linux-jammy-py3.10-clang15-asan / test (default, 2, 6, lf.linux.4xlarge) (gh)
inductor/test_cpu_cpp_wrapper.py::DynamicShapesCppWrapperCpuTests::test_qconv2d_add_cpu_dynamic_shapes_cpp_wrapper
pull / linux-jammy-py3.10-clang15-asan / test (default, 3, 6, lf.linux.4xlarge) (gh)
inductor/test_cpu_cpp_wrapper.py::DynamicShapesCppWrapperCpuTests::test_qconv2d_add_relu_cpu_dynamic_shapes_cpp_wrapper
pull / linux-jammy-py3.10-clang15-asan / test (default, 4, 6, lf.linux.4xlarge) (gh)
inductor/test_mkldnn_pattern_matcher.py::TestPatternMatcher::test_qat_qconv2d_relu
pull / linux-jammy-py3.10-clang15-asan / test (default, 5, 6, lf.linux.4xlarge) (gh)
inductor/test_mkldnn_pattern_matcher.py::TestPatternMatcher::test_qat_qconv2d
pull / linux-jammy-py3.10-clang15-asan / test (default, 6, 6, lf.linux.4xlarge) (gh)
inductor/test_cpu_cpp_wrapper.py::TestCppWrapper::test_qconv2d_add_relu_cpu_cpp_wrapper
pull / linux-jammy-py3.9-gcc11 / test (default, 2, 5, lf.linux.2xlarge) (gh)
inductor/test_cpu_cpp_wrapper.py::TestCppWrapper::test_qconv2d_add_cpu_cpp_wrapper
pull / linux-jammy-py3.9-gcc11 / test (default, 3, 5, lf.linux.2xlarge) (gh)
inductor/test_cpu_cpp_wrapper.py::TestCppWrapper::test_qconv2d_add_relu_cpu_cpp_wrapper
pull / linux-jammy-py3.9-gcc11 / test (default, 4, 5, lf.linux.2xlarge) (gh)
inductor/test_mkldnn_pattern_matcher.py::TestPatternMatcher::test_qat_qconv2d_relu
pull / linux-jammy-py3.9-gcc11 / test (default, 5, 5, lf.linux.2xlarge) (gh)
inductor/test_cpu_cpp_wrapper.py::TestCppWrapper::test_qconv2d_with_concat_cpu_cpp_wrapper

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

shaoyuyoung · 2025-01-18T02:56:09Z

@pytorchbot label "topic: not user facing"

shaoyuyoung · 2025-01-18T02:56:41Z

I am very sorry that I deleted the previous PR because of my misoperation :(

shaoyuyoung · 2025-01-18T02:57:52Z

@jansel , mind retaking a look?
really sorry, please forgive me
It seems that I made a mess.... some third-party modules have been added :(

third_party/cutlass

…into fix_conv_uint

shaoyuyoung · 2025-01-20T04:57:20Z

this time, it should be fine.
I will take care next time. :)

jansel · 2025-01-20T21:10:53Z

@pytorchbot merge

pytorchmergebot · 2025-01-20T21:13:04Z

Merge started

Your change will be merged once all checks pass (ETA 0-4 Hours).

Learn more about merging in the wiki.

Questions? Feedback? Please reach out to the PyTorch DevX Team

Advanced Debugging

Check the merge workflow status
here

pytorchmergebot · 2025-01-20T21:34:49Z

Merge failed

Reason: 1 mandatory check(s) failed. The first few are:

Lint / lintrunner-noclang / linux-job

Dig deeper by viewing the failures on hud

Details for Dev Infra team

Raised by workflow job

Failing merge rule: Core Maintainers

shaoyuyoung · 2025-01-21T05:57:47Z

CI was broaken just now?

jansel · 2025-01-21T17:19:14Z

The lint errors are real. Not sure about the others, you could try rebasing.

shaoyuyoung · 2025-01-22T02:19:22Z

The lint errors are real. Not sure about the others, you could try rebasing.

got it, I will investigate this :)

test/inductor/test_torchinductor.py

shaoyuyoung · 2025-02-06T06:58:03Z

I have rewritten UT by testing this case on CPU.
But CI still fails on cpu_cpp_wrapper :(

Should I use @skip_if_cpp_wrapper on this UT?

shaoyuyoung · 2025-02-06T07:44:28Z

It seems that regression testing fails on here (?)

pytorch/test/inductor/test_cpu_cpp_wrapper.py

Lines 286 to 291 in aab7925

    
           BaseTest( 
        
               "test_qconv2d_dequant_promotion", 
        
               "cpu", 
        
               test_mkldnn_pattern_matcher.TestPatternMatcher(), 
        
               condition=torch.backends.mkldnn.is_available() and not IS_WINDOWS, 
        
           ),

forgive me I can't understand why. :(

jansel · 2025-02-07T02:56:48Z

What is the error from the cpu_cpp_wraper? Can you reproduce it locally?

shaoyuyoung · 2025-02-07T03:07:38Z

CI reminds me this ut fails.

I have run the ut as follows locally

pytest -s -v test/inductor/test_cpu_cpp_wrapper.py -k test_qconv2d_add_relu_cpu_cpp_wrapper

It throws the error

RuntimeError: "conv" not implemented for 'torch.uint8'

I think maybe test_cpu_cpp_wraper uses torch.uint8?

My change adds the error check for conv with uint. Subsequently, these UTs fails

jansel · 2025-02-07T17:05:43Z

Did this previously work? I'm assuming the "q" in qconv2d is quantized, so maybe this is breaking that.

shaoyuyoung · 2025-02-08T02:24:19Z

Did this previously work?

Sry, not sure about "this previously work". Do you mean whether this UT works in an older pytorch version or before my code change?

jansel · 2025-02-08T18:07:18Z

yes

shaoyuyoung · 2025-02-10T02:25:23Z

Do you mean whether this UT works in an older pytorch version or before my code

I have done this and found that some UTs in test_cpu_cpp_wraper which consist qconv2d breaking my code change. I guess that qconv2d in inductor uses uint8? (conflicted UTs' log is below)

torch._dynamo.exc.BackendCompilerFailed: backend='inductor' raised:
RuntimeError: "conv" not implemented for 'torch.uint8'

how should we address this conflict? I think a "dirty" solution is to bring this type check to some other places instead of _meta_registrations (e.g., decomposition or lowering). But I'm not sure where is the optimization pass of conv

jansel · 2025-02-10T04:30:42Z

We shouldn't break the existing behavior. Might need to look more into what the test is doing.

shaoyuyoung · 2025-02-10T04:33:21Z

Might need to look more into what the test is doing.

I'll try my best to investigate this.

jansel

Re-request review once I should take another look at this

github-actions · 2025-05-30T02:12:30Z

Looks like this PR hasn't been updated in a while so we're going to go ahead and mark this as Stale.
Feel free to remove the Stale label if you feel this was a mistake.
If you are unable to remove the Stale label please contact a maintainer in order to do so.
If you want the bot to never mark this PR stale again, add the no-stale label.
Stale pull requests will automatically be closed after 30 days of inactivity.

shaoyuyoung and others added 2 commits January 18, 2025 10:51

fix conv on processing uint

52c1cec

Merge branch 'pytorch:main' into fix_conv_uint

0819681

pytorch-bot bot added the module: inductor label Jan 18, 2025

pytorch-bot bot added the topic: not user facing topic category label Jan 18, 2025

pytorchbot added the open source label Jan 18, 2025

Merge branch 'pytorch:main' into fix_conv_uint

b7d5ccf

jansel requested changes Jan 18, 2025

View reviewed changes

third_party/cutlass Outdated Show resolved Hide resolved

shaoyuyoung and others added 4 commits January 20, 2025 12:51

fix the third-party modules

9b3f9d0

Merge branch 'pytorch:main' into fix_conv_uint

bfbeafd

fix 3rd-party

4896927

Merge branch 'fix_conv_uint' of https://github.com/shaoyuyoung/pytorch …

8d1034c

…into fix_conv_uint

shaoyuyoung requested a review from jansel January 20, 2025 04:57

jansel approved these changes Jan 20, 2025

View reviewed changes

pytorch-bot bot added the ciflow/trunk Trigger trunk jobs on your pull request label Jan 20, 2025

pytorchmergebot added the merging label Jan 20, 2025

pytorchmergebot removed the merging label Jan 20, 2025

pytorch-bot bot temporarily deployed to upload-benchmark-results January 20, 2025 21:40 Inactive

Merge branch 'pytorch:main' into fix_conv_uint

1697815

etaf reviewed Feb 5, 2025

View reviewed changes

test/inductor/test_torchinductor.py Outdated Show resolved Hide resolved

shaoyuyoung and others added 3 commits February 6, 2025 11:20

Merge branch 'pytorch:main' into fix_conv_uint

4b91656

fix: use

c64571e

fix: remove .to(GPU_TYPE)

f5c60f2

jansel self-requested a review February 6, 2025 21:32

jansel requested changes Feb 18, 2025

View reviewed changes

soulitzer added the triaged This issue has been looked at a team member, and triaged and prioritized into an appropriate module label Feb 24, 2025

shaoyuyoung mentioned this pull request Mar 2, 2025

[inductor] torch.slice_scatter throws AssertionError when meeting internal float32 #147842

Open

shaoyuyoung marked this pull request as draft March 31, 2025 01:44

github-actions bot added the Stale label May 30, 2025

github-actions bot closed this Jun 29, 2025

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[inductor] [bug fix] Fix `conv` on processing uint #145136

[inductor] [bug fix] Fix `conv` on processing uint #145136

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

[inductor] [bug fix] Fix conv on processing uint #145136

[inductor] [bug fix] Fix conv on processing uint #145136

Uh oh!

Conversation

Uh oh!

Uh oh!

Uh oh!

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/145136

❌ 30 New Failures

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Merge started

Uh oh!

Merge failed

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

[inductor] [bug fix] Fix `conv` on processing uint #145136

[inductor] [bug fix] Fix `conv` on processing uint #145136