Add support for reducing across the middle dimension for 3D matrices using the sum Triton kernel #2297
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
|
This pull request was exported from Phabricator. Differential Revision: D58307854 |
jananisriram
had a problem deploying
to
docker-s3-upload
GitHub Actions
Error
jananisriram
had a problem deploying
to
docker-s3-upload
GitHub Actions
Error
|
This pull request was exported from Phabricator. Differential Revision: D58307854 |
jananisriram
added a commit
to jananisriram/benchmark
that referenced
this pull request
Jun 12, 2024
…using the sum Triton kernel (pytorch#2297) Summary: Pull Request resolved: pytorch#2297 Support reducing 3-dimensional matrices across the middle dimension (`dim == 1`) such that the result is of dimensions `(M, K)`. This kernel assumes that `BLOCK_SIZE_M == 1`, as Triton is currently unable to perform reductions on a middle dimension, and that that the entire reduction dimension of the tensor fits in a thread block (`BLOCK_SIZE_N >= N`). Reviewed By: davidberard98 Differential Revision: D58307854
jananisriram
force-pushed
the
export-D58307854
branch
from
0c4c863 to
b1b470e
Compare
jananisriram
had a problem deploying
to
docker-s3-upload
GitHub Actions
Error
jananisriram
had a problem deploying
to
docker-s3-upload
GitHub Actions
Error
…using the sum Triton kernel (pytorch#2297) Summary: Pull Request resolved: pytorch#2297 Support reducing 3-dimensional matrices across the middle dimension (`dim == 1`) such that the result is of dimensions `(M, K)`. This kernel assumes that `BLOCK_SIZE_M == 1`, as Triton is currently unable to perform reductions on a middle dimension, and that that the entire reduction dimension of the tensor fits in a thread block (`BLOCK_SIZE_N >= N`). Reviewed By: davidberard98 Differential Revision: D58307854
|
This pull request was exported from Phabricator. Differential Revision: D58307854 |
jananisriram
force-pushed
the
export-D58307854
branch
from
b1b470e to
f08f969
Compare
jananisriram
had a problem deploying
to
docker-s3-upload
GitHub Actions
Failure
jananisriram
had a problem deploying
to
docker-s3-upload
GitHub Actions
Failure
|
This pull request has been merged in 55c975e. |
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.
Summary: Support reducing 3-dimensional matrices across the middle dimension (
dim == 1) such that the result is of dimensions(M, K). This kernel assumes thatBLOCK_SIZE_M == 1, as Triton is currently unable to perform reductions on a middle dimension, and that that the entire reduction dimension of the tensor fits in a thread block (BLOCK_SIZE_N >= N).Reviewed By: davidberard98
Differential Revision: D58307854