This repository was archived by the owner on Aug 21, 2025. It is now read-only.
Use nll_loss_backward decomposition for jvp transform #764
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
|
Before anyone asks: this is a pretty big hack. Ideally we would register the decomposition with PyTorch's forward-mode AD and since functorch leverages the PyTorch forward-mode AD implementation and rules for its jvp transform, then we would get them for free. Since this was relatively easy to do and I want to do some initial testing for this approach, I added this hack, but in parallel we should make progress toward being able to use decompositions with the PyTorch forward-mode AD subsystem. |
zou3519
added a commit
to zou3519/pytorch
that referenced
this pull request
Jul 20, 2022
bigfootjon
pushed a commit
to pytorch/pytorch
that referenced
this pull request
Jul 21, 2022
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
2 participants
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Outdated suggestions cannot be applied.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.
No description provided.