MXNet – 安装

基于Ubuntu14.04/16.04，Python，GPU，Build from Sources

Prerequisites

• CUDA8.0
• cuDNN v5 for CUDA8.0
  确保添加CUDA安装路径到LD_LIBRARY_PATH：

export LD_LIBRARY_PATH=/usr/local/cuda/lib64/:$LD_LIBRARY_PATH

编译 MXNnet核心库

从 C++ 源码编译 MXNet core shared library – libmxnet.so.

Minimum Requirements：
– GCC 4.8 or later to compile C++ 11
– GNU Make

# Step 1 Install build tools and git.
$ sudo apt-get update
$ sudo apt-get install -y build-essential git

# Step 2 Install OpenBLAS.
$ sudo apt-get install -y libopenblas-dev

# Step 3 Install OpenCV.
$ sudo apt-get install -y libopencv-dev

# Step 4 Download MXNet sources and build MXNet core shared library.
$ git clone --recursive https://github.com/dmlc/mxnet
$ cd mxnet
$ make -j $(nproc) USE_OPENCV=1 USE_BLAS=openblas USE_CUDA=1 USE_CUDA_PATH=/usr/local/cuda USE_CUDNN=1

编译 MXNet的Python API

# Step 1 Install prerequisites - python setup tools and numpy.
$ sudo apt-get install -y python-dev python-setuptools python-numpy

# Step 2 Build the MXNet Python binding.
$ cd python
$ sudo python setup.py install

# Step 3 Validate the installation by running simple MXNet code.

>>> import mxnet as mx
>>> a = mx.nd.ones((2,3),mx.gpu()) # 在GPU上创建 2×3 矩阵
>>> b = a * 2 + 1 # 矩阵a各元素 ×2 + 1
>>> b.asnumpy()
array([[ 3.,3.,3.],[ 3.,3.]],dtype=float32)