Hybrid ce loss changes #364
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
DomInvivo
requested changes
Jun 23, 2023
| regression_input = torch.inner(input, self.brackets.to(input.device)) | ||
| # regression loss needs normalized logits to probability as input to do inner product with self.brackets | ||
| # we apply softmax on the raw logits first | ||
| softmax_input = self.softmax(input) |
There was a problem hiding this comment.
I thought we discussed that the softmax should go as the last_activation since we want the classification metrics to work
There was a problem hiding this comment.
I mentioned about this on slack about the auroc and average precision metrics. It looks like they are happy with both probabilities or logits. In that case we don't need a custom CE loss? Or do you think there would be other classification metrics which only accept probabilities as the input?
` Accepts the following input tensors:
- ``preds`` (float tensor): ``(N, C, ...)``. Preds should be a tensor containing probabilities or logits for each
observation. If preds has values outside [0,1] range we consider the input to be logits and will auto apply
softmax per sample.`
There was a problem hiding this comment.
Thanks, just saw your Slack message!
I checked also Accuracy and f1. I think this is consistent across metrics 🎉
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Outdated suggestions cannot be applied.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.
Changelogs
Checklist:
feature,fixortest(or ask a maintainer to do it for you).discussion related to that PR
Fix to hybridceloss, but it is not actually helping the hybridceloss for val and test using the debug config.