Table Of Contents
Table Of Contents


Getting started

A 60-minute Gluon crash coursegetting-started/crash-course/index.html

A quick overview of the core concepts of MXNet using the Gluon API.

Moving from other frameworksgetting-started/to-mxnet/index.html

Guides that ease your transition to MXNet from other framework.

Packages & Modules


MXNet’s imperative interface for Python. If you’re new to MXNet, start here!

NDArray APIpackages/ndarray/index.html

How to use the NDArray API to manipulate data. A useful set of tutorials for beginners.

Symbol APIpackages/symbol/index.html

How to use MXNet’s Symbol API.

Autograd APIpackages/autograd/autograd.html

How to use Automatic Differentiation with the Autograd API.

Learning Ratepackages/lr_scheduler.html

How to use the Learning Rate Scheduler.


Improving Performanceperformance/perf.html

How to get the best performance from MXNet.


How to profile MXNet models.

Tuning Numpy Operationsperformance/numpy.html

Gotchas using NumPy in MXNet.

Compression: float16performance/float16.html

How to use float16 in your model to boost training speed.

Compression: int8performance/index.html

How to use int8 in your model to boost training speed.

Gradient Compressionperformance/gradient_compression.html

How to use gradient compression to reduce communication bandwidth and increase speed.


How to get the most from your CPU by using Intel’s MKL-DNN.


How to use NVIDIA’s TensorRT to boost inference performance.


How to use TVM to boost performance.


MXNet on EC2deploy/run-on-aws/use_ec2.html

How to deploy MXNet on an Amazon EC2 instance.

MXNet on SageMakerdeploy/run-on-aws/use_sagemaker.html

How to run MXNet using Amazon SageMaker.

Training with Data from S3deploy/run-on-aws/use_s3.html

How to train with data from Amazon S3 buckets.

ONNX Modelsdeploy/onnx.html

How to export an MXNet model to the ONNX model format.


Custom Layers for Gluonextend/custom_layer.html

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

Custom Operators Using Numpyextend/custom_op.html

How to use Numpy to create custom MXNet operators.

New Layer Creationextend/new_op.html

How to create new MXNet operators.

Next steps

  • To learn more about using MXNet to implement various deep learning algorithms from scratch, we recommend the Dive into Deep Learning book.

  • If you are interested in building your projects based on state-of-the-art deep learning algorithms and/or pre-trained models, please refer to the toolkits in the MXNet ecosystem.

  • Check out the API Reference docs.