Community contributions to MXNet have added many new valuable features and functionality to support use cases such as model serving & portability, easy and flexible APIs, and educational material like crash courses and online books. This ecosystem page lists the projects that use MXNet, teach MXNet, or augment MXNet in some way.
Learning MXNet and other Educational Resources¶
- Clojure API - use MXNet with Clojure.
- C++ API - not be confused with the C++ backend, this API allows C++ programmers to train networks in C++.
- Gluon Python Interface - train complex models imperatively and then deploy with a symbolic graph.
- Julia API (Community Supported) - train models with multiple GPUs using Julia.
- Keras-MXNet - design with Keras2 and train with MXNet as the backend for 2x or more speed improvement.
- MinPy - Pure numpy practice with third party operator integration and MXNet as backend for GPU computing
- Module Python API - backed by the Symbol API, you can define your network in a declarative fashion.
- ONNX-MXnet API - train and use Open Neural Network eXchange (ONNX) model files.
- Perl API (Community Supported) - train models with multiple GPUs using Perl.
- R API (Community Supported) - train models with multiple GPUs using R.
- Scala Infer API - model loading and inference functionality.
- TensorFuse - Common interface for Theano, CGT, TensorFlow, and MXNet (experimental) by dementrock
Toolkits to Extend MXNet¶
Debugging and Visualization¶
- MXBoard - lets you to visually inspect and interpret your MXNet runs and graphs using the TensorBoard software.
Do you know of a project or resource in the MXNet ecosystem that should be listed here? Or would you like to get involved by providing your own contribution? Check out the guide for contributing to MXNet, and browse the design proposals to see what others are working on. You might find something you would like to help with or use those design docs as a template for your own proposal. Use one of the developer communication channels if you would like to know more, or create a GitHub issue if you would like to propose something for the MXNet ecosystem.