Run tests and build Python wheels for aarch64 architecture #3948
Conversation
This file contains 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
|
Also, I tried to run tests on but then tests were failing with segfaults: |
StrikerRUS
commented
Feb 12, 2021
Comment on lines
+1048
to
+1051
| if platform.machine() == 'aarch64': | ||
| np.testing.assert_allclose(contribs_csr.toarray(), contribs_dense, rtol=1, atol=1e-12) | ||
| else: | ||
| np.testing.assert_allclose(contribs_csr.toarray(), contribs_dense) |
There was a problem hiding this comment.
Sometimes contribs_sparse tests are failing, so need to relax assertion tolerance
> np.testing.assert_allclose(contribs_csr_arr_re, contribs_dense)
E AssertionError:
E Not equal to tolerance rtol=1e-07, atol=0
E
E Mismatched elements: 20 / 840 (2.38%)
E Max absolute difference: 2.22044605e-16
E Max relative difference: 1.
E x: array([[ 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00,
E -3.840704e-02, -1.261461e-01, -3.519505e-02, 1.271551e-04,
E 0.000000e+00, -2.690382e-04, -1.218832e-01, 1.221108e-04,...
E y: array([[ 0.000000e+00, 0.000000e+00, -4.847735e-17, 0.000000e+00,
E -3.840704e-02, -1.261461e-01, -3.519505e-02, 1.271551e-04,
E 0.000000e+00, -2.690382e-04, -1.218832e-01, 1.221108e-04,...
StrikerRUS
requested review from
chivee,
henry0312,
jameslamb,
Laurae2 and
wxchan
as code owners
|
Prefetch and malloc tests are failing on @guolinke Maybe you would like to investigate this. FYI, |
jameslamb
reviewed
Feb 13, 2021
| @@ -42,7 +43,8 @@ | |||
|
|
|||
| pytestmark = [ | |||
| pytest.mark.skipif(getenv('TASK', '') == 'mpi', reason='Fails to run with MPI interface'), | |||
| pytest.mark.skipif(getenv('TASK', '') == 'gpu', reason='Fails to run with GPU interface') | |||
| pytest.mark.skipif(getenv('TASK', '') == 'gpu', reason='Fails to run with GPU interface'), | |||
| pytest.mark.skipif(machine() != 'x86_64', reason='Fails to run with non-x86_64 architecture') | |||
There was a problem hiding this comment.
In what way did the Dask tests fail?
There was a problem hiding this comment.
I don't remember, sorry! I marked Dask tests to be skipped at the very begging of enabling QEMU and cannot find logs for runs where they were enabled. I think we can investigate Dask support for aarch64 in a follow-up PR. WDYT?
There was a problem hiding this comment.
yep it's ok, can be a follow-up
jameslamb
approved these changes
Feb 13, 2021
|
@jameslamb Thanks a lot for your review! I'd like to ask @guolinke give it another look before merging as this PR is quite major. |
|
@guolinke Just want to make sure you have not missed this my comment: #3948 (comment). |
|
This pull request has been automatically locked since there has not been any recent activity since it was closed. To start a new related discussion, open a new issue at https://github.com/microsoft/LightGBM/issues including a reference to this. |
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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.
Contributes to #1129.
Fixed #3456.
Fixed #3517.
(Superceded) closed #3421.
I run 10x10 consecutive runs and all of them were successful, so I'm quite sure in stability of new QEMU job.
It can be seen from the picture above that sometimes QEMU job takes ~30min and sometimes ~1h. I thought that maybe there is any bottlenecking test that freezes whole job. But unfortunately, long runs are related to QEMU itself (it is common practice to set
timeout=360for QEMU builds) because in those jobs all tests contribute to slowness. Here are timings frompytest.~30min job:
~1h job:
Although this job can be quite long and slow down our CI checks, I believe that it should be run with every PR and commit because ARM64 support becomes more and more important nowadays and as you can see there were a lot of requests for ARM wheels.