MXNet is a lightweight, portable, and flexible deep learning framework with support to many programming languages such as C++, python, scala, julia, javascript, and R.

Since MXNet installation guide does not provide a guideline to install MXNet on CentOS 7, I write an installation guideline for CentOS 7. Here are the steps to build MXNet on CentOS 7:

1. Install dependencies

yum groupinstall "Development Tools" 
yum install atlas-devel opencv-devel

Atlas is installed in /usr/lib64/atlas.

2. Clone MXNet project

git clone --recursive https://github.com/dmlc/mxnet

3. Modify MSHADOW_LDFLAGS in mshadow/make/mshadow.mk (line 67) from -lcblas to -L/usr/lib64/atlas -lsatlas

...
else ifeq ($(USE_BLAS), atlas)
        MSHADOW_LDFLAGS += -L/usr/lib64/atlas -lsatlas
...

4. Make

make -j$(nproc)

5. MXNet Python package installation

Follow official MXNet Python package installation guide