Fix LayerNorm gradient flow issue #3012
Open
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
- Fix LayerNorm.forward() to use tensor operations instead of scalar operations - Replace sum_keepdim()/size with mean_keepdim() to preserve gradients - Use broadcast_add() with epsilon tensor instead of scalar addition - Fix ops::layer_norm_slow() with same gradient-preserving changes - Update ops::layer_norm() to use slow implementation for proper gradients - Add comprehensive gradient flow test (now passes with 100% gradient flow) - Add numerical equivalence test to ensure accuracy is maintained - Fixes training issues where LayerNorm parameters weren't being updated Resolves gradient propagation bug where only 33% of parameters received gradients during backpropagation, preventing proper model training.
|
great. it would be awesome to have more training code examples and workflows with candle |
|
Hey! Thanks for this :) I think we'll have to implement this in the optimized kernels as well before we can merge. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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Outdated suggestions cannot be applied.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.
Resolves gradient propagation bug where only 33% of parameters received gradients during backpropagation, preventing proper model training. #3011