Repolode
Submit
microsoft

microsoft/onnxruntime

A cross-platform accelerator for running ML models exported to ONNX format, with training support for PyTorch scripts.

View on GitHubWebsite
Stars
22,137
Forks
4,347
Open issues
1,820
Last commit
2 hours ago

What is it?

ONNX Runtime is a cross-platform accelerator for machine learning inference and training. It runs models from frameworks such as PyTorch and TensorFlow/Keras, as well as classical libraries like scikit-learn, LightGBM and XGBoost, once they are expressed in the ONNX format.

For inference, it applies graph optimizations and transforms, and uses hardware accelerators where available, aiming to deliver faster response times and lower compute costs. It is designed to run across different hardware, drivers and operating systems.

For training, the project offers acceleration for transformer models on multi-node NVIDIA GPUs, and the project describes it as adding only a single line to an existing PyTorch training script.

Key features

  • 01Runs models exported from PyTorch, TensorFlow/Keras and classical ML libraries in ONNX format
  • 02Applies graph optimizations and transforms to improve inference performance
  • 03Uses hardware accelerators where applicable across different hardware and drivers
  • 04Accelerates transformer model training on multi-node NVIDIA GPUs

Use cases

  • Deploying trained deep learning models in production services with lower latency
  • Running one model across different hardware and operating systems
  • Speeding up transformer model training on NVIDIA GPU clusters

Quick start

  1. 1

    Get a release from the GitHub releases page

  2. 2

    Read the usage documentation and tutorials at onnxruntime.ai/docs

Things to know

  • The README describes training acceleration on NVIDIA GPUs.

Topics

#machine-learning#inference#onnx#deep-learning#hardware-acceleration#pytorch#ai-framework#neural-networks#scikit-learn#tensorflow

FAQ

What is onnxruntime?
A cross-platform accelerator for running ML models exported to ONNX format, with training support for PyTorch scripts. ONNX Runtime is a cross-platform accelerator for machine learning inference and training. It runs models from frameworks such as PyTorch and TensorFlow/Keras, as well as classical libraries like scikit-learn, LightGBM and XGBoost, once they are expressed in the ONNX format.
Is onnxruntime free to use? What license does it have?
onnxruntime is open source under the MIT license. Check the license terms before using it in commercial products.
Is onnxruntime still maintained?
Actively maintained. The last commit to onnxruntime was on Oct 22, 2026.
What can I use onnxruntime for?
Deploying trained deep learning models in production services with lower latency Running one model across different hardware and operating systems Speeding up transformer model training on NVIDIA GPU clusters
How do I install onnxruntime?
Get a release from the GitHub releases page Read the usage documentation and tutorials at onnxruntime.ai/docs
Where can I get onnxruntime?
The source code and installation instructions are on GitHub: https://github.com/microsoft/onnxruntime

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects