Add sum reduction operator to TritonBench #2282
Closed
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: D58048782 |
jananisriram
had a problem deploying
to
docker-s3-upload
GitHub Actions
Error
jananisriram
had a problem deploying
to
docker-s3-upload
GitHub Actions
Error
jananisriram
force-pushed
the
export-D58048782
branch
from
5b1d48f to
605f419
Compare
|
This pull request was exported from Phabricator. Differential Revision: D58048782 |
jananisriram
had a problem deploying
to
docker-s3-upload
GitHub Actions
Error
jananisriram
had a problem deploying
to
docker-s3-upload
GitHub Actions
Error
jananisriram
added a commit
to jananisriram/benchmark
that referenced
this pull request
Jun 6, 2024
Summary: Add a Triton reduction kernel for the `sum` operator where `dim=None` to TritonBench, following the [TritonBench guide](https://fb.workplace.com/notes/953949486404240). This implementation works for all matrices being reduced to a scalar value. To measure accuracy of Triton reduction kernel, add accuracy metric to sum kernel in TritonBench in order to test accuracy of Triton implementation against baseline PyTorch implementation, referencing [`torchbenchmark/operators/gemm/operator.py`](https://www.internalfb.com/code/fbsource/[767bb6faa353685b84f08a39f36fdcf6ca170c85]/fbcode/pytorch/benchmark/torchbenchmark/operators/gemm/operator.py?lines=236). Reset output registers per run of the Triton kernel for accurate Triton output. Referenced the existing [vector_add](https://www.internalfb.com/code/fbsource/fbcode/pytorch/benchmark/torchbenchmark/operators/vector_add/) and [grouped_gemm](https://www.internalfb.com/code/fbsource/fbcode/pytorch/benchmark/torchbenchmark/operators/grouped_gemm/) TritonBench operators as frameworks for implementation. See the [TritonBench Operator Coverage Tracker](https://docs.google.com/spreadsheets/d/1091POOPSPsUnlNVEKaz2X_DQXdIwFv-fGOH_g9by-Zo/edit#gid=0) for current operator coverage in TritonBench. Reviewed By: xuzhao9, davidberard98 Differential Revision: D58048782
jananisriram
force-pushed
the
export-D58048782
branch
from
605f419 to
c331f9c
Compare
|
This pull request was exported from Phabricator. Differential Revision: D58048782 |
jananisriram
had a problem deploying
to
docker-s3-upload
GitHub Actions
Error
jananisriram
had a problem deploying
to
docker-s3-upload
GitHub Actions
Error
jananisriram
force-pushed
the
export-D58048782
branch
from
c331f9c to
6d44f49
Compare
|
This pull request was exported from Phabricator. Differential Revision: D58048782 |
jananisriram
had a problem deploying
to
docker-s3-upload
GitHub Actions
Error
jananisriram
had a problem deploying
to
docker-s3-upload
GitHub Actions
Error
Summary: Add a Triton reduction kernel for the `sum` operator where `dim=None` to TritonBench, following the [TritonBench guide](https://fb.workplace.com/notes/953949486404240). This implementation works for all matrices being reduced to a scalar value. To measure accuracy of Triton reduction kernel, add accuracy metric to sum kernel in TritonBench in order to test accuracy of Triton implementation against baseline PyTorch implementation, referencing [`torchbenchmark/operators/gemm/operator.py`](https://www.internalfb.com/code/fbsource/[767bb6faa353685b84f08a39f36fdcf6ca170c85]/fbcode/pytorch/benchmark/torchbenchmark/operators/gemm/operator.py?lines=236). Reset output registers per run of the Triton kernel for accurate Triton output. To measure performance of the Triton reduction kernel against PyTorch, add gbps metric, referencing [`torchbenchmark/operators/vector_add/operator.py`](https://www.internalfb.com/code/fbsource/[858eda681c7618f9427ba55cef8d4aba712cb26e]/fbcode/pytorch/benchmark/torchbenchmark/operators/vector_add/operator.py?lines=19). Referenced the existing [vector_add](https://www.internalfb.com/code/fbsource/fbcode/pytorch/benchmark/torchbenchmark/operators/vector_add/) and [grouped_gemm](https://www.internalfb.com/code/fbsource/fbcode/pytorch/benchmark/torchbenchmark/operators/grouped_gemm/) TritonBench operators as frameworks for implementation. See the [TritonBench Operator Coverage Tracker](https://docs.google.com/spreadsheets/d/1091POOPSPsUnlNVEKaz2X_DQXdIwFv-fGOH_g9by-Zo/edit#gid=0) for current operator coverage in TritonBench. Reviewed By: xuzhao9, davidberard98 Differential Revision: D58048782
jananisriram
force-pushed
the
export-D58048782
branch
from
6d44f49 to
1f8df93
Compare
|
This pull request was exported from Phabricator. Differential Revision: D58048782 |
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 2d8999b. |
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:
Add a Triton reduction kernel for the sum operator where dim=None to TritonBench, following the TritonBench guide. This implementation works for all matrices being reduced to a scalar value.
To measure accuracy of Triton reduction kernel, add accuracy metric to sum kernel in TritonBench in order to test accuracy of Triton implementation against baseline PyTorch implementation, referencing torchbenchmark/operators/gemm/operator.py. Reset output registers per run of the Triton kernel for accurate Triton output.
To measure performance of the Triton reduction kernel against PyTorch, add gbps metric, referencing torchbenchmark/operators/vector_add/operator.py.
Referenced the existing vector_add and grouped_gemm TritonBench operators as frameworks for implementation.
See the TritonBench Operator Coverage Tracker for current operator coverage in TritonBench.
Reviewed By: xuzhao9, davidberard98
Differential Revision: D58048782