Add a decomposition for _weight_norm_interface. #112193
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/112193
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit fbf7080 with merge base 9bfebf7 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
albanD
added
the
triaged
| @unittest.skipIf(TEST_WITH_ASAN, "Skipped under ASAN") | ||
| @onlyNativeDeviceTypes | ||
| @skipIfCrossRef | ||
| def test_weight_norm_interface(self, device): | ||
| g = torch.randn((3, 10, 10), device=device) | ||
| v = torch.randn((1, 1, 10), device=device) | ||
|
|
||
| ref = torch.ops.aten._weight_norm_interface(g, v, 2) | ||
| res = torch._decomp.decompositions._weight_norm_interface(g, v, 2) | ||
| self.assertTrue(torch.allclose(ref[0], res[0])) | ||
| self.assertTrue(torch.allclose(ref[1], res[1])) | ||
|
|
There was a problem hiding this comment.
Rather than a one-off test, please add relevant samples that exercise the _wegith_norm_interface path.
There was a problem hiding this comment.
Hi,
What do you mean by "relevant samples" exactly? Just more random test cases of different shapes or you have specific test cases in mind?
Thanks!
There was a problem hiding this comment.
Add samples to WeightNorm's OpInfo so that they exercise this path. No need to add any new test
There was a problem hiding this comment.
@qihqi will you do this in a follow up PR?
|
test failure is legit |
Thanks, taking a look |
|
@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 |
Follow up for pytorch#112193
Follow up for pytorch#112193
Fixes pytorch#112086 Pull Request resolved: pytorch#112193 Approved by: https://github.com/ezyang
Follow up for pytorch#112193
Fixes pytorch#112086 Pull Request resolved: pytorch#112193 Approved by: https://github.com/ezyang
Fixes pytorch#112086 Pull Request resolved: pytorch#112193 Approved by: https://github.com/ezyang
Fixes #112086