Dont mutate tensor stride in place in cudnn conv #126786
Closed
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
[ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/126786
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 0734941 with merge base 7e166e8 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
eellison
added a commit
that referenced
this pull request
May 21, 2024
ghstack-source-id: 5109bc89093fb2e0dca22675133c3ffbf9ee8633 Pull Request resolved: #126786
eqy
reviewed
May 21, 2024
|
What are the callsites of those apis ? Do they invoke |
|
We don't compile RNNs in torch.compile - would prefer separate issue/fix for the other legacy callsites. |
|
@eellison Yes I think overloads of pytorch/aten/src/ATen/cudnn/Descriptors.h Line 169 in b40fb2d fixSizeOneDimStride. The callsites are IIRC RNN, Conv v7, and maybe other legacy API places like batchnorm?
|
Fix for #126241. Within the cudnn convolution, we were in-place updating the strides of the tensor to disambiguate for size-1 dims and contiguous and channels last tensors. Instead of mutating the tensors stride, just use a temporary. Inside cudnn it is then copied: https://github.com/NVIDIA/cudnn-frontend/blob/d7ccb5b3c47b4de709604cce463ad66b775b7812/include/cudnn_frontend_Tensor.h#L201-L203. [ghstack-poisoned]
eellison
added a commit
that referenced
this pull request
May 21, 2024
ghstack-source-id: 3de7eb31768a05fe2e9686b1b2c183fa8aee31c6 Pull Request resolved: #126786
Fix for #126241. Within the cudnn convolution, we were in-place updating the strides of the tensor to disambiguate for size-1 dims and contiguous and channels last tensors. Instead of mutating the tensors stride, just use a temporary. Inside cudnn it is then copied: https://github.com/NVIDIA/cudnn-frontend/blob/d7ccb5b3c47b4de709604cce463ad66b775b7812/include/cudnn_frontend_Tensor.h#L201-L203. [ghstack-poisoned]
eellison
added a commit
that referenced
this pull request
May 21, 2024
ghstack-source-id: bf7ad7fcf19506b9387511d895e4ce412903a385 Pull Request resolved: #126786
ezyang
approved these changes
May 21, 2024
shunting314
approved these changes
May 21, 2024
eqy
approved these changes
May 21, 2024
|
@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 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Stack from ghstack (oldest at bottom):
Fix for #126241.
Within the cudnn convolution, we were in-place updating the strides of the tensor to disambiguate for size-1 dims and contiguous and channels last tensors. Instead of mutating the tensors stride, just use a temporary. Inside cudnn it is then copied: https://github.com/NVIDIA/cudnn-frontend/blob/d7ccb5b3c47b4de709604cce463ad66b775b7812/include/cudnn_frontend_Tensor.h#L201-L203.