adding a parallelisation test for DeepEnsemble #898
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
uri-granta
reviewed
Jan 23, 2025
There was a problem hiding this comment.
Good idea for a test! Couple of questions:
- how long does it to take to run? If it's slow you could mark it with
@pytest.mark.slow. - why does the smaller ensemble have more units per layer? wouldn't that increase the run time and make the comparison less fair?
- just to be certain, have manually confirmed that the training is happening in parallel (e.g. by setting a breakpoint? or disabling parallelisation and seeing a change?)
about a minute on CPU, probably similar on GPU - I'll do that
that's specifically to account for the number of parameters in each model, so this is the fair comparison
its not possible to do that I think, disabling parllelisation would require a completely different DE architecture |
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Outdated suggestions cannot be applied.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.
we explicitly test that increasing number of networks in the ensemble, once accounted for difference in number of parameters, will lead to similar training time