meta-pytorch/tritonbench: Tritonbench is a collection of PyTorch custom operators with example inputs to measure their performance.

TritonBench

TritonBench is a collection of PyTorch operators used to evaluation the performance of Triton, and its integration with PyTorch.

Installation

The benchmark suite should be self-contained of its dependencies. To install, follow the steps below.

Step 1: clone the repository and checkout all submodules

``bash git clone https://github.com/meta-pytorch/tritonbench.git cd tritonbench git submodule update --init --recursive

Step 2: run install.py

 bash
python install.py
By default, it will install the latest PyTorch nightly release and use the Triton version bundled with it.
Basic Usage
To benchmark an operator, run the following command: bash
python run.py --op gemm
## Install as a library
To install as a library: bash
pip install -e .
In your own benchmark script: python
import tritonbench
from tritonbench.utils import parser
op_args = parser.parse_args()
addmm_bench = tritonbench.load_opbench_by_name("addmm")(op_args)
addmm_bench.run()

Submodules

We depend on the following projects as a source of customized Triton or CUTLASS kernels:

(CUDA, HIP) generative-recommenders
(CUDA, HIP) Liger-Kernel
(CUDA, HIP) tilelang
(CUDA) flash-attention
(CUDA) FBGEMM
(CUDA) ThunderKittens
(HIP) AITer

License

TritonBench is BSD 3-Clause licensed, as found in the LICENSE file.

meta-pytorch/tritonbench: Tritonbench is a collection of PyTorch custom operators with example inputs to measure their performance.

TritonBench

Installation

Basic Usage

Submodules

License

Reading Mode

Related Stories

How Hackers Weaponized DEV.to's Mention System to Spread a Phishing Scam

Patch Your Rails: 21 Security Fixes in 8.0.5.2 and 8.1.4.1

Your Login Works. Now Prove Who Did What

Your Agent Has a Side Door: Testing MCP Policy Beyond the Tool List

Last week in Agent Security Update 5: The Pentagon Prompts

Related Stories