Extend

The following tutorials will help you learn how to customize MXNet.

Custom Layers for Gluon../packages/gluon/blocks/custom-layer.html

How to add new layer functionality to MXNet’s imperative interface.

Custom Loss../packages/gluon/loss/custom-loss.html

A guide to implementing custom losses.

Custom Operators Using Numpycustomop.html

How to use Numpy to create custom MXNet operators.

New Operator Creation/api/faq/new_op

How to create new MXNet operators using CustomOp (Python) or NNVM (C++).

A Beginner’s Guide to Implementing Operators in MXNet Backend/api/faq/add_op_in_backend

How to create new MXNet operators in MXNet’s backend using C++. An example custom quadratic function op.

Using runtime compilation (RTC) to write CUDA kernels in MXNet/api/faq/using_rtc

How to write CUDA kernels in MXNet using runtime compilation.