Add epsilon for NaN prevention + test case to RetinaNet #1684
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
|
/gcbrun |
ianstenbit
approved these changes
Apr 7, 2023
|
/gcbrun |
|
/gcbrun |
|
Looks like we need to upgrade tensorflow |
|
/gcbrun |
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Outdated suggestions cannot be applied.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.
fixes #1683
I actually took a
tf.cond()based approach here to zero out sample weights instead. I do this because I found that when we divide by normalizer, and normalizer is zero, the background classes all get absolutely massive values for class weight - causing massive gradients.An alternative is perhaps to use a small value, i.e. 0.3~ for the class weights? I think the ideal mask is something like:
tf.where(positive_mask, 0.1, 0.0)where 0.1 is any tiny value. Though, using all zeros is probably safer and easier.