Add test for consistency between meta and CPU devices. #138515
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/138515
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 8716a9b with merge base 4dd04db ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
It's kind of up in the air whether or not CPU is a good default, but I guess it's easy to test and if you really care you use FakeTensor, so sure why no |
|
@pytorchbot merge |
Merge startedYour 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 |
|
The merge job was canceled or timed out. This most often happen if two merge requests were issued for the same PR, or if merge job was waiting for more than 6 hours for tests to finish. In later case, please do not hesitate to reissue the merge command |
|
@pytorchbot merge |
Merge startedYour 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 |
|
The merge job was canceled or timed out. This most often happen if two merge requests were issued for the same PR, or if merge job was waiting for more than 6 hours for tests to finish. In later case, please do not hesitate to reissue the merge command |
|
I've just realized that the CI failure I was getting in #138520 wasn't being run in this PR. I believe this PR also fails on |
|
Moving this to "draft", in order to debug said CI failure. |
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
|
Apparently the reason for the CI failure was that the test was being run under inductor. Added a skip for inductor (and dynamo) runs. |
|
The CI failure looks not related, so merging anyway. |
|
@pytorchbot merge -i |
Merge startedYour change will be merged while ignoring the following 1 checks: Lint / lintrunner-noclang / linux-job Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
|
@pytorchbot revert -m 'Sorry for reverting your change, but the test is failing in trunk, maybe a landrace' -c landrace test_ops.py::TestCommonCPU::test_meta_consistency_out_dtype_mismatch_permute_copy_cpu_float32 GH job link HUD commit link |
|
@pytorchbot successfully started a revert job. Check the current status here. |
…)" This reverts commit 006130d. Reverted #138515 on behalf of https://github.com/huydhn due to Sorry for reverting your change, but the test is failing in trunk, maybe a landrace ([comment](#138515 (comment)))
|
@ysiraichi your PR has been successfully reverted. |
Pull Request resolved: #138520 Approved by: https://github.com/ezyang ghstack dependencies: #138515
Reference: pytorch#138399 This PR introduces an `OpInfo` test that checks whether running each `out=` operation using meta inputs is consistent with using concrete (e.g. CPU) inputs. More specifically, it tests the case where the output tensors are not of the expected data type. According to the `out=` specification, some operations should error. I have added XFAIL to the set of operations that are currently failing. Pull Request resolved: pytorch#138515 Approved by: https://github.com/ezyang
Pull Request resolved: pytorch#138520 Approved by: https://github.com/ezyang ghstack dependencies: pytorch#138515
5 participants
Stack from ghstack (oldest at bottom):
addmm: error on output dtype mismatch. #138520Reference: #138399
This PR introduces an
OpInfotest that checks whether running eachout=operationusing meta inputs is consistent with using concrete (e.g. CPU) inputs. More specifically,
it tests the case where the output tensors are not of the expected data type. According to
the
out=specification, some operations should error.I have added XFAIL to the set of operations that are currently failing.