Open-source machine learning frameworks on Azure

Build and deploy models faster with an open ecosystem

Build and deploy machine learning models quickly on Azure using your favourite open-source frameworks. Azure provides an open and interoperable ecosystem to use the frameworks of your choice without getting locked in, accelerate every phase of the machine learning lifecycle and run your models anywhere from the cloud to the edge.

Build machine learning models in the framework of your choice

Azure supports all popular machine learning frameworks. Whether you’re developing models in deep learning frameworks such as PyTorch or TensorFlow, taking advantage of Azure automated machine learning capabilities or training traditional machine learning models in scikit-learn, you’ll be able to support your workloads on Azure.

Inference across any operating system and hardware platform

Optimise inferencing on a wide variety of hardware platforms using the open-source ONNX Runtime. ONNX Runtime works with popular frameworks such as PyTorch, TensorFlow, Keras, SciKit-Learn and more to deliver up to 17-times faster inferencing and up to 1.4-times faster training. Use ONNX Runtime to inference your ML models on Linux, Windows, Mac and even mobile devices. ONNX Runtime integrates latest accelerator software and hardware libraries from partners such as Intel and NVIDIA to help you maximise performance, whether you are running in the cloud or on the edge.

Accelerate the end-to-end machine learning life cycle

Accelerate your productivity with automated machine learning. Quickly identify suitable algorithms and tune hyperparameters, and manage the complete machine learning life cycle easily with simple deployment from the cloud to the edge. Access all of these capabilities from a tool-agnostic Python SDK.

Simplify and accelerate machine learning with Azure