This repository has been archived by the owner on Aug 15, 2019. It is now read-only.
Add gradients to oneHot and argMax
#1218
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
`oneHot` op is not differentiable and thus should have zero gradient with the same shape and dtype as its input.
`argMax` op is not differentiable and thus should have zero gradient with the same shape and dtype as its input.
`tf.split()`, `tf.unstack()` modify `begin` argument in a loop and thus change its value for the backpropagation. Make a copy of `begin` in `slice` to prevent such changes.
|
Sorry, added one more commit to the branch. I'm reusing this branch to fix my code, thus it is easier for to just pile up things here. |
`argMin` op is not differentiable and thus should have zero gradient with the same shape and dtype as its input.
|
Would like to see this land as well 👍 |
`greaterEqual` op is not differentiable and thus should have zero gradient with the same shape and dtype as its input.
|
Any chance this could be reviewed? |
|
Sorry for the delay, nice work! I filed a bug to track the "correct" fix for this. For now, zeros are fine. |
|
👍 Thanks for review! |
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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Outdated suggestions cannot be applied.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.
Description
Add gradients to
oneHotandargMax. Both are not differentiable, and thus their gradients must be zero.For repository owners only:
Please remember to apply all applicable tags to your pull request.
Tags: FEATURE
For more info see: https://github.com/tensorflow/tfjs/blob/master/DEVELOPMENT.md
This change is