Record and emit scaledDotProductAttention for IREE (#543)#544
Merged
michalharakal merged 1 commit intodevelopfrom Apr 23, 2026
Merged
Record and emit scaledDotProductAttention for IREE (#543)#544michalharakal merged 1 commit intodevelopfrom
michalharakal merged 1 commit intodevelopfrom
Conversation
Three changes to enable SDPA in the SKaiNET → StableHLO → IREE path: 1. RecordingExecution: record SDPA calls with operation + params (was delegating without recording, like conv1d before PR #532) 2. TensorOperations: add ScaledDotProductAttentionOperation with inferOutputs (output shape = query shape) 3. NeuralNetOperationsConverter: decompose SDPA into StableHLO: - dot_general Q @ K.T (batching_dims=[0,1], contracting_dims=[3]x[3]) - scale + optional mask - softmax (max-subtract-exp-sum-div decomposition) - dot_general weights @ V (contracting_dims=[3]x[2]) Also includes: - SdpaHloExportTest: verifies tape → graph → MLIR with dot_general - TapeAttentionPermuteBugTest: proves raw array permute creates zero constants - ShapeOperationsConverter: concatenate input type annotation fix - ISSUE-SDPA-recording-and-hlo.md: issue documentation Fixes #543 Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
|
📖 Documentation Preview The documentation has been built successfully for this PR. Generated Files:
Artifacts:
This comment will be updated automatically when the PR is updated. |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Three changes to enable SDPA in the SKaiNET → StableHLO → IREE path:
RecordingExecution: record SDPA calls with operation + params (was delegating without recording, like conv1d before PR RecordingTensorOpsDecorator drops conv1d/conv3d traces silently #532)
TensorOperations: add ScaledDotProductAttentionOperation with inferOutputs (output shape = query shape)
NeuralNetOperationsConverter: decompose SDPA into StableHLO:
Also includes:
Fixes #543