SuperBenchmark

Azure Pipelines	Build Status
cpu-unit-test
gpu-unit-test

SuperBench is a validation and profiling tool for AI infrastructure, which supports:

• AI infrastructure validation and diagnosis
  • Distributed validation tools to validate hundreds or thousands of servers automatically
  • Consider both raw hardware and E2E model performance with ML workload patterns
  • Build a contract to identify hardware issues
  • Provide infrastructural-oriented criteria as Performance/Quality Gates for hardware and system release
  • Provide detailed performance report and advanced analysis tool
• AI workload benchmarking and profiling
  • Provide comprehensive performance comparison between different existing hardware
  • Provide insights for hardware and software co-design

It includes micro-benchmark for primitive computation and communication benchmarking, and model-benchmark to measure domain-aware end-to-end deep learning workloads.

🔴 Note: SuperBench is in the early pre-alpha stage for open source, and not ready for general public yet. If you want to jump in early, you can try building latest code yourself.

SuperBench capabilities, workflow and benchmarking metrics

The following graphic shows the capabilities provide by SuperBench core framework and its extension.

Benchmarking metrics provided by SuperBench are listed as below.

	Micro Benchmark	Model Benchmark
Metrics	Computation Benchmark Kernel Performance GFLOPS TensorCore cuBLAS cuDNN Kernel Launch Time Kernel_Launch_Event_Time Kernel_Launch_Wall_Time Operator Performance MatMul Sharding_MatMul Memory H2D_Mem_BW_<GPU ID> H2D_Mem_BW_<GPU ID> Communication Benchmark Device P2P Bandwidth P2P_BW_Max P2P_BW_Min P2P_BW_Avg RDMA RDMA_Peak RDMA_Avg NCCL NCCL_AllReduce NCCL_AllGather NCCL_broadcast NCCL_reduce NCCL_reduce_scatter Computation-Communication Benchmark Mul_During_NCCL MatMul_During_NCCL Storage Benchmark Disk Read/Write Rand_Read/Rand_Write R/W_Read R/W_Write Rand_R/W_Read Rand_R/W_Write	CNN models ResNet ResNet-50 ResNet-101 ResNet-152 DenseNet DenseNet-169 DenseNet-201 VGG VGG-11 VGG-13 VGG-16 VGG-19 Other CNN models ... BERT models BERT BERT_LARGE LSTM GPT-2

Installation

Using Docker (Preferred)

System Requirements

• Platform: Ubuntu 18.04 or later (64-bit)
• Docker: Docker CE 19.03 or later

Install SuperBench

• Using Pre-Build Images

  docker pull superbench/superbench:dev-cuda11.1.1
  docker run -it --rm \
      --privileged --net=host --ipc=host --gpus=all \
      superbench/superbench:dev-cuda11.1.1 bash
  

• Building the Image

  docker build -f dockerfile/cuda11.1.1.dockerfile -t superbench/superbench:dev .
  

Using Python

System Requirements

• Platform: Ubuntu 18.04 or later (64-bit); Windows 10 (64-bit) with WSL2

• Python: Python 3.6 or later, pip 18.0 or later

  Check whether Python environment is already configured:

  # check Python version
  python3 --version
  # check pip version
  python3 -m pip --version
  

  If not, install the followings:

  • Python
  • pip
  • venv

  It's recommended to use a virtual environment (optional):

  # create a new virtual environment
  python3 -m venv --system-site-packages ./venv
  # activate the virtual environment
  source ./venv/bin/activate
  
  # exit the virtual environment later
  # after you finish running superbench
  deactivate
  

Install SuperBench

• PyPI Binary

  # not available yet
  

• From Source

  # get source code
  git clone https://github.com/microsoft/superbenchmark
  cd superbenchmark
  
  # install superbench
  python3 -m pip install .
  make postinstall
  

Usage

Run SuperBench

# run benchmarks in default settings
sb exec

# use a custom config
sb exec --config-file ./superbench/config/default.yaml

Benchmark Gallary

Please find more benchmark examples here.

Developer Guide

If you want to develop new feature, please follow below steps to set up development environment.

Check Environment

Follow System Requirements.

Set Up

# get latest code
git clone https://github.com/microsoft/superbenchmark
cd superbenchmark

# install superbench
python3 -m pip install -e .[dev,test]

Lint and Test

# format code using yapf
python3 setup.py format

# check code style with mypy and flake8
python3 setup.py lint

# run all unit tests
python3 setup.py test

Submit a Pull Request

Please install pre-commit before git commit to run all pre-checks.

pre-commit install

Open a pull request to main branch on GitHub.

Contributing

Contributor License Agreement

This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.opensource.microsoft.com.

When you submit a pull request, a CLA bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., status check, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.

This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact opencode@microsoft.com with any additional questions or comments.

Contributing principles

SuperBenchmark is an open-source project. Your participation and contribution are highly appreciated. There are several important things you need know before contributing to this project:

What content can be added to SuperBenchmark

1. Bug fixes for existing features.

2. New features for benchmark module (micro-benchmark, model-benchmark, etc.)

   If you would like to contribute a new feature on SuperBenchmark, please submit your proposal first. In GitHub Issues module, choose Enhancement Request to finish the submission. If the proposal is accepted, you can submit pull requests to origin main branch.

Contribution steps

If you would like to contribute to the project, please follow below steps of joint development on GitHub.

1. Fork the repo first to your personal GitHub account.
2. Checkout from main branch for feature development.
3. When you finish the feature, please fetch the latest code from origin repo, merge to your branch and resolve conflict.
4. Submit pull requests to origin main branch.
5. Please note that there might be comments or questions from reviewers. It will need your help to update the pull request.

Trademarks

This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft's Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos are subject to those third-party's policies.