Create and lower broadcasted_reduce to get softmax e2e working - #3269
Open
copybara-service[bot] wants to merge 1 commit into
Open
Create and lower broadcasted_reduce to get softmax e2e working#3269copybara-service[bot] wants to merge 1 commit into
copybara-service[bot] wants to merge 1 commit into
Conversation
This change introduces tensor_ext.broadcasted_reduce, which reduces along an axis of a tensor and then immediately broadcasts back along that axis. I found the default behavior (linalg.reduce + linalg.broadcast) introduced too much complexity and usually triggered an expensive layout conversion. A layout conversion is required, but it is usually simple enough and it seems easier to output the more efficient kernel when we handle the fused case directly. A future change will support identifying opportunities to fuse linalg.reduce with linalg.broadcast, but for now the softmax-to-cgf-softmax kernel can output it directly. PiperOrigin-RevId: 952401826
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Create and lower broadcasted_reduce to get softmax e2e working
This change introduces tensor_ext.broadcasted_reduce, which reduces along an axis of a tensor and then immediately broadcasts back along that axis. I found the default behavior (linalg.reduce + linalg.broadcast) introduced too much complexity and usually triggered an expensive layout conversion. A layout conversion is required, but it is usually simple enough and it seems easier to output the more efficient kernel when we handle the fused case directly.
A future change will support identifying opportunities to fuse linalg.reduce with linalg.broadcast, but for now the softmax-to-cgf-softmax kernel can output it directly.