Add plots variable along 1 input parameter to jagged_sum operator in TritonBench #2340
Conversation
|
This pull request was exported from Phabricator. Differential Revision: D59034792 |
|
This pull request was exported from Phabricator. Differential Revision: D59034792 |
…TritonBench (pytorch#2340) Summary: Pull Request resolved: pytorch#2340 Add plotting functionality to the `jagged_sum` operator in TritonBench, enabling the creation of line plots for any set of benchmarks variable along one of the following input parameters: `B`, `M`, `seqlen`, or `sparsity`. This diff makes it easier to visualize the differences in `latency` among the different benchmarks in the `jagged_sum` operator. Add a command-line argument to toggle the benchmarks displayed on the plots, making it easier to visualize just the 2 Triton benchmarks or just the 2 PyTorch benchmarks, if necessary. This modification helps more clearly visualize the `latency` differences between the simple fused and variable-length loop Triton kernels as well as the unpadded and padded PyTorch benchmarks. Reviewed By: davidberard98 Differential Revision: D59034792
baa16a1
to
eb929f6
Compare
…TritonBench Summary: Add plotting functionality to the `jagged_sum` operator in TritonBench, enabling the creation of line plots for any set of benchmarks variable along one of the following input parameters: `B`, `M`, `seqlen`, or `sparsity`. This diff makes it easier to visualize the differences in `latency` among the different benchmarks in the `jagged_sum` operator. Add a command-line argument to toggle the benchmarks displayed on the plots, making it easier to visualize just the 2 Triton benchmarks or just the 2 PyTorch benchmarks, if necessary. This modification helps more clearly visualize the `latency` differences between the simple fused and variable-length loop Triton kernels as well as the unpadded and padded PyTorch benchmarks. Note that for plots testing multiple values of `B` or `M`, the x-axis is on a log scale to more accurately depict trends in latency; whether or not the x-axis is on a log scale is noted in the plot name in the Test Plan. Reviewed By: davidberard98 Differential Revision: D59034792
|
This pull request was exported from Phabricator. Differential Revision: D59034792 |
eb929f6
to
3c94931
Compare
|
Hi @jananisriram! Thank you for your pull request. We require contributors to sign our Contributor License Agreement, and yours needs attention. You currently have a record in our system, but the CLA is no longer valid, and will need to be resubmitted. ProcessIn order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA. Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with If you have received this in error or have any questions, please contact us at [email protected]. Thanks! |
|
This is already in main as 847d38e, closing. |
Summary:
Add plotting functionality to the
jagged_sumoperator in TritonBench, enabling the creation of line plots for any set of benchmarks variable along one of the following input parameters:B,M,seqlen, orsparsity. This diff makes it easier to visualize the differences inlatencyamong the different benchmarks in thejagged_sumoperator.Add a command-line argument to toggle the benchmarks displayed on the plots, making it easier to visualize just the 2 Triton benchmarks or just the 2 PyTorch benchmarks, if necessary. This modification helps more clearly visualize the
latencydifferences between the simple fused and variable-length loop Triton kernels as well as the unpadded and padded PyTorch benchmarks.Reviewed By: davidberard98
Differential Revision: D59034792