[DispatchCreation] Collapse iree_linalg_ext.attention
#19012
Merged
Conversation
This file contains 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
IanWood1
changed the title
iree_linalg_ext.attention
Add support for attention in `CollapseDimensionsPass` so that the attention op gets collapsed as much as possible. This is motivated by reducing the different variants of attention that the sdxl attention spec has to handle. Signed-off-by: Ian Wood <[email protected]>
Since this pass now handles more than just `linalg.generic` ops. Fix up the comments and drop references to linalg.generic ops. Signed-off-by: Ian Wood <[email protected]>
MaheshRavishankar
requested changes
Nov 7, 2024
|
|
||
| /// Map from iteration domain index in the original op to the iteration domain | ||
| /// index in the collapsed op. | ||
| SmallVector<std::pair<int64_t, unsigned>> origOpToCollapsedOpIterationDim; |
There was a problem hiding this comment.
Why int64_t and unsigned ?
compiler/src/iree/compiler/Dialect/LinalgExt/Transforms/ReshapeFusion.cpp
Show resolved
Hide resolved
compiler/src/iree/compiler/DispatchCreation/test/collapse_dimensions.mlir
Show resolved
Hide resolved
MaheshRavishankar
approved these changes
Nov 7, 2024
Signed-off-by: Ian Wood <[email protected]>
Signed-off-by: Ian Wood <[email protected]>
Groverkss
pushed a commit
to Groverkss/iree
that referenced
this pull request
Dec 1, 2024
This change adds support for attention in `CollapseDimensionsPass` so that the attention op will be collapsed as much as possible. This is motivated by reducing the different variants of attention that the sdxl attention spec has to handle. Changes to LinalgExt/Transforms/ReshapeFusion.cpp are mostly taken directly from https://github.com/llvm/llvm-project/blob/002a0a27bc4702d6f34434c1838cb1698a0b0098/mlir/lib/Dialect/Linalg/Transforms/ElementwiseOpFusion.cpp (attributed at the top of the file). I attempted to keep not modify the original logic as much as possible to keep it general in case it needs to be reused for other `LinalgExt` ops. --------- Signed-off-by: Ian Wood <[email protected]>
giacs-epic
pushed a commit
to giacs-epic/iree
that referenced
this pull request
Dec 4, 2024
This change adds support for attention in `CollapseDimensionsPass` so that the attention op will be collapsed as much as possible. This is motivated by reducing the different variants of attention that the sdxl attention spec has to handle. Changes to LinalgExt/Transforms/ReshapeFusion.cpp are mostly taken directly from https://github.com/llvm/llvm-project/blob/002a0a27bc4702d6f34434c1838cb1698a0b0098/mlir/lib/Dialect/Linalg/Transforms/ElementwiseOpFusion.cpp (attributed at the top of the file). I attempted to keep not modify the original logic as much as possible to keep it general in case it needs to be reused for other `LinalgExt` ops. --------- Signed-off-by: Ian Wood <[email protected]> Signed-off-by: Giacomo Serafini <[email protected]>
IanWood1
added a commit
that referenced
this pull request
Jan 8, 2025
Reland after fixing sdxl int8 regressions via #19012. Running CI revealed further performance regressions that have pending patches: #19325 and #19326. This reverts commit 8d3faf8. --------- Signed-off-by: Ian Wood <[email protected]>
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.
This change adds support for attention in
CollapseDimensionsPassso that the attention op will be collapsed as much as possible. This is motivated by reducing the different variants of attention that the sdxl attention spec has to handle.Changes to LinalgExt/Transforms/ReshapeFusion.cpp are mostly taken directly from https://github.com/llvm/llvm-project/blob/002a0a27bc4702d6f34434c1838cb1698a0b0098/mlir/lib/Dialect/Linalg/Transforms/ElementwiseOpFusion.cpp (attributed at the top of the file). I attempted to keep not modify the original logic as much as possible to keep it general in case it needs to be reused for other
LinalgExtops.