EFF Speed-up MiniBatchDictionaryLearning by avoiding multiple validation #25490
Conversation
This file contains hidden or 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
|
I opened an alternative in #25493, which involves a lot less refactoring. |
|
Well now that it's done :) |
jjerphan
approved these changes
Mar 6, 2023
There was a problem hiding this comment.
Nice observation, @jeremiedbb! This LGTM.
I just have one remark. I also think we can merge this PR independently from #25493 as @ogrisel mentioned in #25490 (review) once a changelog entry is added. What do you think?
|
I added a what's new entry. Yes, let's merge this one first. |
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Outdated suggestions cannot be applied.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.
MinibatchDictionaryLearning calls public functions that call public classes themselves. We end up validating the parameters and the input/dict twice per minibatch. When the batch size is large it has barely no impact but for small batch sizes it can be very detrimental.
For instance, here's a profiling result in the extreme case batch_size=1
This PR removes the first one. It's a param validation coming from
sparse_encode. The profiling now givesthe first block is gone. I intend to deal with the other ones in follow up PRs