Apache MXNet (incubating) documents

Date:Dec 09, 2019

Note

This is a preview of the Apache MXNet (incubating) new NumPy-like interface. It only contains a subset of documents. For more information, see the MXNet main website.

Installation

There is a preview pip package available. These packages haven’t uploaded into Pypi server yet, so you need to select various options manually.

OS: Linux macOS Windows

Backend: Native CUDA MKL-DNN CUDA + MKL-DNN

Build-in backend for CPU.
Required to run on Nvidia GPUs.
Accelerate Intel CPU performance.
Enable both Nvidia CPUs and Inter CPU acceleration.

Prerequisites:

  • Requires CUDA. Supported versions include 8.0, 9.0, and 9.2.
  • Hint: cuDNN is already included in the MXNet binary, so you don’t need to install it.
  • Hint: MKL-DNN is already included in the MXNet binary, so you don’t need to install it.
  • For detailed information on MKL and MKL-DNN, refer to the MKLDNN_README.

Command:

# For python 3.7
pip install https://apache-mxnet.s3-accelerate.amazonaws.com/dist/python/numpy/latest/mxnet-1.5.0-cp37-cp37m-macosx_10_11_x86_64.whl
# For python 3.6
pip install https://apache-mxnet.s3-accelerate.amazonaws.com/dist/python/numpy/latest/mxnet-1.5.0-cp36-cp36m-macosx_10_11_x86_64.whl

# For python 3.7
pip install https://apache-mxnet.s3-accelerate.amazonaws.com/dist/python/numpy/latest/mxnet_mkl-1.5.0-cp37-cp37m-macosx_10_11_x86_64.whl
# For python 3.6
pip install https://apache-mxnet.s3-accelerate.amazonaws.com/dist/python/numpy/latest/mxnet_mkl-1.5.0-cp36-cp36m-macosx_10_11_x86_64.whl

pip install https://apache-mxnet.s3-accelerate.amazonaws.com/dist/python/numpy/latest/mxnet-1.5.0-py2.py3-none-manylinux1_x86_64.whl

# You must have CUDA 10.0 is installed.
pip install https://apache-mxnet.s3-accelerate.amazonaws.com/dist/python/numpy/latest/mxnet_cu100-1.5.0-py2.py3-none-manylinux1_x86_64.whl

# You must have  CUDA 10.1 is installed.
pip install https://apache-mxnet.s3-accelerate.amazonaws.com/dist/python/numpy/latest/mxnet_cu101mkl-1.5.0-py2.py3-none-manylinux1_x86_64.whl
Not supported.

Guidelines

This guide section is intended as an introductory overview of MXNet modules for manipulating multi-dimensional arrays, loading data, constructing and training neural networks.

Crash Courseguide/crash-course/index.html

A crash course to train and predict with a convolutional neural network.

NP on MXNetguide/deepnumpy/index.html

Manipulate multi-dimensional arrays with the NumPy-like interface.

Gluonguide/gluon/index.html

Load data, construct and train neural networks with Gluon.

API Documents

This API section details functions, modules, and objects included in MXNet, describing what they are and what they do. The APIs are grouped into the following categories:

NP on MXNetapi/deepnumpy/index.html

API documents for mxnet.np and mxnet.npx.

Gluonapi/gluon/index.html

API documents for mxnet.gluon and other related modules.