Persist pairs to save large overhead when training model on large datasets - #304
Merged
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 larger datasets (say millions of spectra or > 100,000 compounds) the bias-compensating pair search is quite compute intensive. There would still be room for some more runtime optimization, but that will be a lot of additional work. So, for now, this PR will add one big helping step if it comes to repeated trainings on the same dataset/split, for instance for parameter searches etc.
--> The selected pairs/splits can now be stored in a dedicated folder
pair_data_folderand will be reused when possible.