Table Of Contents
Table Of Contents

Install

Platform: Local Cloud

Provider: AWS

Instruction:

There are several options you can access MXNet at AWS:

  • Deep Learning AMI: Amazon machine images with MXNet pre-installed, available for Ubuntu, Amazon Linux, and Windows 2016.
  • Sagemaer: a fully-managed machine learning platform with MXNet integrated.
OS: Linux Macos Windows
Package: Pip Docker

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

Build-in backend for CPU.
Required to run on Nvidia GPUs.
Accelarate Intel CPU performacne.
Enable both Nvidia CPUs and Inter CPU accelaration.

Prerequisite:

  • Require docker is installed and it can be used by a non-root user.
  • Require pip >= 9. is installed. Both Python 2 and Python 3 are supported.
  • Hint: append the flag --pre at the end of the command will install the nightly build.
  • Require CUDA is installed. Supported versions include 8.0, 9.0, 9.1, and 9.2.
  • Hint: cuDNN is already included in the MXNet binary, you don’t need to install it.
  • Hint: MKL-DNN is already included in the MXNet binary, you don’t need to install it.

Command:

pip install mxnet
# Here we assume CUDA 9.2 is installed. You can change the number
# according to your own CUDA version.
pip install mxnet-cu92
pip install mxnet-mkl
# Here we assume CUDA 9.2 is installed. You can change the number
# according to your own CUDA version.
pip install mxnet-cu92mkl
docker pull mxnet/python
docker pull mxnet/python:gpu
docker pull mxnet/python:1.3.0_cpu_mkl
docker pull mxnet/python:1.3.0_gpu_cu90_mkl_py3

Next steps

  • For new users: Crash Course. It contains a 60-minutes crash course that teaches to train a handwritten digits classifier.
  • For experienced users: Guide. Check all MXNet guidelines.
  • For advanced users: API. Browser all MXNet classes and methods.

Table Of Contents