Error messages for kernel selection #90783
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/90783
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit e522f45: This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
This pull request was exported from Phabricator. Differential Revision: D42008661 |
Summary: Pull Request resolved: pytorch#90783 Error messages fro kernel selection Test Plan: sandcastle & github Differential Revision: D42008661 fbshipit-source-id: 435eba80b509b4bc84143bad7ee9603820b8a152
mikekgfb
force-pushed
the
export-D42008661
branch
from
8201b88
to
8f57294
Compare
|
This pull request was exported from Phabricator. Differential Revision: D42008661 |
Summary: Pull Request resolved: pytorch#90783 Error messages fro kernel selection Test Plan: sandcastle & github Differential Revision: D42008661 fbshipit-source-id: f9639ab716601d31429a8f2d94b848980f3ea08f
mikekgfb
force-pushed
the
export-D42008661
branch
from
8f57294
to
50414e6
Compare
|
This pull request was exported from Phabricator. Differential Revision: D42008661 |
Summary: Pull Request resolved: pytorch#90783 Error messages fro kernel selection Test Plan: sandcastle & github Differential Revision: D42008661 fbshipit-source-id: 787f044eb09815eae16060ac5b6bcfec8d312976
mikekgfb
force-pushed
the
export-D42008661
branch
from
50414e6
to
2a75ed3
Compare
|
This pull request was exported from Phabricator. Differential Revision: D42008661 |
Summary: Pull Request resolved: pytorch#90783 Error messages fro kernel selection Test Plan: sandcastle & github Differential Revision: D42008661 fbshipit-source-id: d71eb0526713f9095990feff75e4548a89094f70
mikekgfb
force-pushed
the
export-D42008661
branch
from
2a75ed3
to
e522f45
Compare
|
This pull request was exported from Phabricator. Differential Revision: D42008661 |
|
@pytorchbot merge |
Merge failedReason: PR #90783 has not been reviewed yet (Rule superuser) Details for Dev Infra teamRaised by workflow job |
Because we're reporting the first error for the first kernel today. So, if you disable mem_efficient, and Flash predictaes do not work, you still get an answer why memory efficient kernel does not work. That's straight up wrong because it's not actionable, and it's also misleading. So, for the case when there's not kernel, I changed it to dumping all the reasons for all kerbnels, and then error out "and hence you lose, and don't have an executable kernel". This gives people a way to actually impact. @drisspg had previously mentioned he could fix the current issue with some if-spaghetti logic to make sure we give the error for the enabled kernel, but also pointed out that it's just not scalable if we ever add more kernels coz combinatorial explosion. |
|
@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 |
Merge failedReason: 2 additional jobs have failed, first few of them are: trunk ,trunk / linux-focal-rocm5.2-py3.8 / test (default, 1, 2, linux.rocm.gpu) Details for Dev Infra teamRaised by workflow job |
|
@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 |
Summary: Error messages fro kernel selection
Test Plan: sandcastle & github
Differential Revision: D42008661