[LLVMGPUVectorDistribute] Fix batch dimensions extraction for attention-like ops #19040
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
manupak
requested review from
hanhanW,
MaheshRavishankar,
qedawkins,
kuhar and
Groverkss
as code owners
manupak
force-pushed
the
split-k2-online-attention
branch
2 times, most recently
from
a98864a to
ce7d7c1
Compare
Groverkss
approved these changes
Nov 7, 2024
attention-like ops Currently, the batch dimensions are extracted as the union of dimensions present across Q & K & V. This is not correct if one of the dims in inputs (Q,K and V) could be seen as broadcasting. Therefore, this commit changes this to be: B = Union ( Q & K & O , K & V & O ) where if parallel dimensions common between both matmuls will be treated as batching dimensions. Signed-off-by: Manupa Karunaratne <[email protected]>
manupak
force-pushed
the
split-k2-online-attention
branch
from
ce7d7c1 to
2c6d399
Compare
JamesMBartlett
pushed a commit
to gimletlabs/iree
that referenced
this pull request
Nov 8, 2024
…on-like ops (iree-org#19040) Currently, the batch dimensions are extracted as the union of dimensions present across Q & K & V. This is not correct if one of the dims in inputs (Q,K and V) could be seen as broadcasting. Therefore, this commit changes this to be: B = Union ( Q & K & O , K & V & O ) where if parallel dimensions common between both matmuls will be treated as batching dimensions. Signed-off-by: Manupa Karunaratne <[email protected]>
Groverkss
pushed a commit
to Groverkss/iree
that referenced
this pull request
Dec 1, 2024
…on-like ops (iree-org#19040) Currently, the batch dimensions are extracted as the union of dimensions present across Q & K & V. This is not correct if one of the dims in inputs (Q,K and V) could be seen as broadcasting. Therefore, this commit changes this to be: B = Union ( Q & K & O , K & V & O ) where if parallel dimensions common between both matmuls will be treated as batching dimensions. Signed-off-by: Manupa Karunaratne <[email protected]>
giacs-epic
pushed a commit
to giacs-epic/iree
that referenced
this pull request
Dec 4, 2024
…on-like ops (iree-org#19040) Currently, the batch dimensions are extracted as the union of dimensions present across Q & K & V. This is not correct if one of the dims in inputs (Q,K and V) could be seen as broadcasting. Therefore, this commit changes this to be: B = Union ( Q & K & O , K & V & O ) where if parallel dimensions common between both matmuls will be treated as batching dimensions. Signed-off-by: Manupa Karunaratne <[email protected]> Signed-off-by: Giacomo Serafini <[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.
Depends on: #19042
Currently, the batch dimensions are extracted as the union of dimensions present across Q & K & V. This is not correct if one of the dims in inputs (Q,K and V) could be seen as broadcasting.
Therefore, this commit changes this to be:
B = Union ( Q & K & O , K & V & O )
where if parallel dimensions common between both matmuls will be treated as batching dimensions.