[ARM] Enable some additional Aarch64 unit tests #146895
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/146895
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 04822a5 with merge base b004228 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
@pytorchbot label "topic: not user facing" |
|
@malfet this PR enables We urgently need this PR in order to expose some important unit test regressions with oneDNN 3.7 upgrade #138889 |
|
@pytorchbot label "module: arm" |
Adds inductor/test_fused_attention inductor/test_cpu_select_algorithm and nn/test_convolution. Skip/Xfail some tests with issues linked..
d257e23 to
f94426b
Compare
|
I have forced pushed in order to fix the merge conflict with a rebase |
|
@yanbing-j can you approve again. Not sure if my force push cancelled the CI approval. |
|
@pytorchbot label "ciflow/linux-aarch64" |
|
To add the ciflow label This helps ensure we don't trigger CI on this PR until it is actually authorized to do so. Please ping one of the reviewers if you do not have access to approve and run workflows. |
|
@pytorchbot label "ciflow/linux-aarch64" |
|
There's still a failing test in It could be a simple fix, if it's not obvious I will mark the test as skip and raise an issue so we don't hold up this PR too long
|
|
@fadara01 @yanbing-j could one of you approve the workflow again? Thanks |
Done! |
|
@pytorchbot merge |
|
This PR needs to be approved by an authorized maintainer before merge. |
|
@malfet approval needed on this when you get a moment |
|
@pytorchbot label "arm priority" |
|
Let's hold back on merging this, since the Arm Compute Library (ACL) version used for the CI jammy docker image here is v24.04 while the one used in manylinux here is v24.09. |
|
Hello @robert-hardwick , I believe we can avoid merging this for now. This needs to be addressed internally first. I will mark the PR as draft for now. |
Agreed. Let's get the ACL versions in the CI synchronised first.. and since #144992 has been reverted this is a lower priority now. Thanks @nikhil-arm |
|
Looks like this PR hasn't been updated in a while so we're going to go ahead and mark this as |
This PR adds some tests to Aarch64 ci. Notably
nn/test_convolutionandinductor/test_fused_attentionThe reason for this is that there are some additional regression test failures in the oneDNN 3.7 upgrade #138889 which do not have visibility because they are not enabled.
I have marked
test_ConvTranspose2d_output_size_downsample_upsampleas skipped ( #146857 ) due to a segmentation fault. But priority is to get visibility on oneDNN 3.7 new test failures.cc @malfet @snadampal @milpuz01 @aditew01 @nikhil-arm @fadara01 @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @chenyang78 @kadeng @muchulee8 @amjames @chauhang @aakhundov @yf225