Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

test_graph_unit_dnnl_convolution_cpu fails on AArch64 CI after recent quantization refactor #2572

Open
michalowski-arm opened this issue Jan 31, 2025 · 4 comments · May be fixed by #2599
Open
Assignees
Labels
bug A confirmed library bug platform:cpu-aarch64 Codeowner: @oneapi-src/onednn-cpu-aarch64

Comments

@michalowski-arm
Copy link
Contributor

michalowski-arm commented Jan 31, 2025

Summary

After a recent quantization refactor (#2270) we see a failure in Nightly AArch64 pipeline (https://github.com/oneapi-src/oneDNN/actions/runs/13066726551/job/36460347340). Upon bisecting, the first bad commit appears to be af1410c.

Version

ACL version v24.11.1

Environment

OneDNN GitHub CI for AArch64, c7g AWS instance

Steps to reproduce

ctest -R "test_graph_unit_dnnl_convolution_cpu" --rerun-failed --output-on-failure

Observed behavior

[ RUN ] test_convolution_execute.ConvBiasAddEltwise
/opt/actions-runner/_work/oneDNN/oneDNN/oneDNN/tests/gtests/graph/unit/backend/dnnl/test_convolution.cpp:2557: Failure
Expected equality of these values:
dst[i]
Which is: -0.072591752
param.ref_dst[I]
Which is: -0.072591871
[ FAILED ] test_convolution_execute.ConvBiasAddEltwise (23 ms)

Expected behavior

Test OK

@michalowski-arm michalowski-arm added the sighting Suspicious library behavior. Should be promoted to a bug when confirmed label Jan 31, 2025
@michalowski-arm
Copy link
Contributor Author

michalowski-arm commented Jan 31, 2025

@dzarukin, this appears to be your change

@vpirogov vpirogov added bug A confirmed library bug platform:cpu-aarch64 Codeowner: @oneapi-src/onednn-cpu-aarch64 and removed sighting Suspicious library behavior. Should be promoted to a bug when confirmed labels Jan 31, 2025
@dzarukin
Copy link
Contributor

Hi @michalowski-arm. Though it may be triggered by my changes, the problem seems to be in the way the test compares results. I'll check how to update it.

@Ryo-not-rio
Copy link
Contributor

@dzarukin this seems to be affecting our nightly and precommit pipelines as well: https://github.com/oneapi-src/oneDNN/actions/runs/13128702961

@vpirogov
Copy link
Member

vpirogov commented Feb 4, 2025

+@TaoLv who was looking into this.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
bug A confirmed library bug platform:cpu-aarch64 Codeowner: @oneapi-src/onednn-cpu-aarch64
Projects
None yet
Development

Successfully merging a pull request may close this issue.

4 participants