"None" ensembling for classfication accuracy #290
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
|
|
a38e2a3
to
a63995c
Compare
black'd
a63995c
to
728cd42
Compare
lauritowal
approved these changes
Sep 3, 2023
AlexTMallen
approved these changes
Sep 5, 2023
There was a problem hiding this comment.
I am a bit confused why we're only evaluating the positive examples in our "none" ensembling case--it seems slightly arbitrary, but fine. I will also note that this function has always and continues to handle LM logits incorrectly. LM logits are log probs, so it does not make sense to apply a sigmoid to them and gives us slightly different ensembled results. Meanwhile, reporter logits are log odds, and applying a sigmoid to log odds gives probabilities.
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Outdated suggestions cannot be applied.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.
Closes NOT-372
ATM, accuracy for "none" ensembling == "partial" ensembling.
This PR implements a reasonable interpretation of what "No ensembling" would look like for classification accuracy: i.e. for accuracy and calibrated accuracy, use the positive hiddens for inference. I also added logging for
cal_thresh.