-
Notifications
You must be signed in to change notification settings - Fork 24.1k
Disable cudnn to avoid creating guards that denies exporting #148140
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
base: main
Are you sure you want to change the base?
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/148140
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (1 Unrelated Failure)As of commit da059f1 with merge base 9db9593 ( UNSTABLE - The following job is marked as unstable, possibly due to flakiness on trunk:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
This pull request was exported from Phabricator. Differential Revision: D70357703 |
4b90c32
to
98cd695
Compare
…#148140) Summary: Fixes pytorch#147623 This code https://github.com/pytorch/pytorch/blob/main/aten/src/ATen/native/Normalization.cpp#L504-L518 produces guards that raise ConstraintViolation error in batchnorm op. We disable cudnn in export tracing to avoid creating such guards. In the exported graph, the `cudnn_enabled` input to `torch.ops.aten.batch_norm.default` will be False instead of True now. Test Plan: ``` buck2 run mode/dev-nosan //caffe2/test:test_export -- -r bn_dynamic_shapes buck run fbcode//mode/dev-nosan //caffe2/test:test_export -- -r decomp_batch_norm ``` Differential Revision: D70357703
This pull request was exported from Phabricator. Differential Revision: D70357703 |
Fixes #2084 This is required to land pytorch/pytorch#148140 in torch.export(). cc @angelayi @@justinchuby Co-authored-by: Shangdi Yu <[email protected]>
Summary: title required to land pytorch#148140 to include microsoft/onnxscript#2085 Test Plan: CI Differential Revision: D70526777
Summary: title required to land pytorch#148140 to include microsoft/onnxscript#2085 Test Plan: CI Differential Revision: D70526777
Unblock #148140 Pull Request resolved: #148388 Approved by: https://github.com/malfet
Unblock #148140 Pull Request resolved: #148388 Approved by: https://github.com/malfet
Unblock pytorch#148140 Pull Request resolved: pytorch#148388 Approved by: https://github.com/malfet
98cd695
to
f9bde34
Compare
…#148140) Summary: Fixes pytorch#147623 This code https://github.com/pytorch/pytorch/blob/main/aten/src/ATen/native/Normalization.cpp#L504-L518 produces guards that raise ConstraintViolation error in batchnorm op. We disable cudnn in export tracing to avoid creating such guards. In the exported graph, the `cudnn_enabled` input to `torch.ops.aten.batch_norm.default` will be False instead of True now. Test Plan: ``` buck2 run mode/dev-nosan //caffe2/test:test_export -- -r bn_dynamic_shapes buck run fbcode//mode/dev-nosan //caffe2/test:test_export -- -r decomp_batch_norm ``` Differential Revision: D70357703
f9bde34
to
0e1cdf9
Compare
…#148140) Summary: Fixes pytorch#147623 This code https://github.com/pytorch/pytorch/blob/main/aten/src/ATen/native/Normalization.cpp#L504-L518 produces guards that raise ConstraintViolation error in batchnorm op. We disable cudnn in export tracing to avoid creating such guards. In the exported graph, the `cudnn_enabled` input to `torch.ops.aten.batch_norm.default` will be False instead of True now. Test Plan: ``` buck2 run mode/dev-nosan //caffe2/test:test_export -- -r bn_dynamic_shapes buck run fbcode//mode/dev-nosan //caffe2/test:test_export -- -r decomp_batch_norm ``` Differential Revision: D70357703
This pull request was exported from Phabricator. Differential Revision: D70357703 |
…#148140) Summary: Pull Request resolved: pytorch#148140 Fixes pytorch#147623 This code https://github.com/pytorch/pytorch/blob/main/aten/src/ATen/native/Normalization.cpp#L504-L518 produces guards that raise ConstraintViolation error in batchnorm op. We disable cudnn in export tracing to avoid creating such guards. In the exported graph, the `cudnn_enabled` input to `torch.ops.aten.batch_norm.default` will be False instead of True now. Test Plan: ``` buck2 run mode/dev-nosan //caffe2/test:test_export -- -r bn_dynamic_shapes buck run fbcode//mode/dev-nosan //caffe2/test:test_export -- -r decomp_batch_norm ``` Differential Revision: D70357703
This pull request was exported from Phabricator. Differential Revision: D70357703 |
0e1cdf9
to
da059f1
Compare
Looks like this PR hasn't been updated in a while so we're going to go ahead and mark this as |
Summary:
Fixes #147623
This code https://github.com/pytorch/pytorch/blob/main/aten/src/ATen/native/Normalization.cpp#L504-L518 produces guards that raise ConstraintViolation error in batchnorm op.
We disable cudnn in export tracing to avoid creating such guards
Dependency: We need to land microsoft/onnxscript#2085 first in onnxscript, and them bump the onnxscript version in #148388
Test Plan:
Differential Revision: D70357703