Fix test_linear_transformation for ROCm #3099
Merged
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
vfdev-5
pushed a commit
to Quansight/vision
that referenced
this pull request
Dec 4, 2020
facebook-github-bot
pushed a commit
that referenced
this pull request
Dec 8, 2020
Reviewed By: datumbox Differential Revision: D25371946 fbshipit-source-id: cbf748a1e82638f936a21d12bcfea91b4779c9a5
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Outdated suggestions cannot be applied.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.
For test_linear_transformation in test_transforms_tensor.py running on ROCm for batched tensors uses a BLAS config which does not give EXACT same output every time, which is why this test would result in a failure on ROCm. rocBLAS in pyTorch could be made to use deterministic algorithms by the setting
torch.set_deterministic(True)implemented in PyTorch PR#48654This PR adds that setting in the test.
@fmassa - Would you suggest using
torch.set_deterministic(True)setting only for ROCm, and leave it as is for other platforms? In which case I can protect that using a conditional statementcc: @jeffdaily @sunway513