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.

Use your favourite open-source frameworks

Azure supports all popular machine learning frameworks, including PyTorch, TensorFlow, scikit-learn, MXNet, Chainer and Keras.

Easily move between frameworks

Develop in your preferred framework without worrying about downstream inferencing implications. Train models in one framework and transfer them to another for inferencing using ONNX, an open-source model format co-developed by Microsoft and other AI companies.

Run your models efficiently across platforms

Optimise inferencing on a wide variety of hardware platforms using the open-source ONNX Runtime. Train a model with any popular framework, transfer it to ONNX format and inference up to ten times faster. For optimal performance, ONNX Runtime integrates the latest CPU and GPU hardware acceleration from partners such as Intel and NVIDIA.

Accelerate machine learning

Accelerate your productivity with Azure Machine Learning, which supports popular frameworks and tools. Use automated machine learning to quickly identify suitable algorithms and tune hyperparameters. Manage the complete machine learning lifecycle with MLOps – DevOps for machine learning – including 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