watersink / tutorials

Tutorials for using ONNX

ONNX Tutorials

Open Neural Network Exchange (ONNX) is an open standard format for representing machine learning models. ONNX is supported by a community of partners who have implemented it in many frameworks and tools.

Getting ONNX models

• Choose a pre-trained ONNX model from the ONNX Model Zoo. A lot of pre-trained ONNX models are provided for common scenarios.
• Convert models from mainstream frameworks. More tutorials are below.

Framework / tool	Installation	Exporting to ONNX (frontend)	Importing ONNX models (backend)
Caffe	apple/coremltools and onnx/onnxmltools	Exporting	n/a
Caffe2	part of caffe2 package	Exporting	Importing
Chainer	chainer/onnx-chainer	Exporting	coming soon
Cognitive Toolkit (CNTK)	built-in	Exporting	Importing
Apple CoreML	onnx/onnx-coreml and onnx/onnxmltools	Exporting	Importing
Keras	onnx/keras-onnx	Exporting	n/a
LibSVM	onnx/onnxmltools	Exporting	n/a
LightGBM	onnx/onnxmltools	Exporting	n/a
MATLAB	onnx converter on matlab central file exchange	Exporting	Importing
Menoh	pfnet-research/menoh	n/a	Importing
ML.NET	built-in	Exporting	Importing
Apache MXNet	part of mxnet package docs github	Exporting	Importing
PyTorch	part of pytorch package	Exporting, Exporting - Different ONNX Opsets, Extending support	coming soon
SciKit-Learn	onnx/sklearn-onnx	Exporting	n/a
Apache SINGA	built-in	Exporting [experimental]	Importing [experimental]
TensorFlow	onnx/onnx-tensorflow and onnx/tensorflow-onnx	Exporting	Importing
TensorRT	onnx/onnx-tensorrt	n/a	Importing

• Use services like Azure Custom Vision service that generate customized ONNX models for your data

End-to-end tutorials

• For preparation

  • Docker image for Caffe2/PyTorch/ONNX tutorials
  • Docker image for ONNX, ONNX Runtime, and Converters

• For serving

  • Serving ONNX models with MXNet Model Server
  • ONNX model hosting with AWS SageMaker and MXNet
  • Serving ONNX models with ONNX Runtime on Azure ML - FER Facial Expression Recognition, MNIST Handwritten Digits, Resnet50 Image Classification
  • Inferencing ONNX models using ONNX Runtime Python API

• For conversion

  • Convert a PyTorch model to Tensorflow using ONNX

  For direct conversion to/from ONNX format, see the "Exporting" and "Importing" columns in the table under Getting ONNX Models

• From conversion to deployment

  • Converting SuperResolution model from PyTorch to Caffe2 with ONNX and deploying on mobile device
  • Transferring SqueezeNet from PyTorch to Caffe2 with ONNX and to Android app
  • Converting Style Transfer model from PyTorch to CoreML with ONNX and deploying to an iPhone
  • Serving PyTorch Models on AWS Lambda with Caffe2 & ONNX
  • MXNet to ONNX to ML.NET with SageMaker, ECS and ECR - external link
  • Convert CoreML YOLO model to ONNX, score with ONNX Runtime, and deploy in Azure

ONNX tools

• Verifying correctness and comparing performance
• Visualizing an ONNX model (useful for debugging)
• Netron: a viewer for ONNX models
• Example of operating on ONNX protobuf
• Float16 <-> Float32 converter

Contributing

We welcome improvements to the convertor tools and contributions of new ONNX bindings. Check out contributor guide to get started.

Use ONNX for something cool? Send the tutorial to this repo by submitting a PR.