layer_norm_grad for npu #10560
Draft
layer_norm_grad for npu #10560
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 SBP settings for LayerNormGradOp to ensure correct gradient aggregation for gamma_diff and beta_diff Changes - Updated SBP strategy in LayerNormGradOp: Set gamma_diff and beta_diff to use PartialSum instead of Split to avoid dimension mismatches during distributed training. - Added consistency check for begin_norm_axis and begin_params_axis: Enforce equality to ensure proper alignment of normalization and parameter dimensions.
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Outdated suggestions cannot be applied.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.
因为 NPU GPT2 测试场景下，layer_norm 都是 affine 的，为减少 CANN 调用，对 layer_norm_grad 进行重构。在一个 kernel 内完成 gamma，beta，dx 的梯度计算。