Docker images for MXNET

Pre-built docker images are available at https://hub.docker.com/r/dmlc/mxnet/

How to use

First pull the pre-built image
```
docker pull dmlc/mxnet
```

Then we can run the python shell in the docker

docker run -ti dmlc/mxnet python

For example

$ docker run -ti dmlc/mxnet python
Python 2.7.6 (default, Jun 22 2015, 17:58:13)
[GCC 4.8.2] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import mxnet as mx
import mxnet as mx
>>> quit()
quit()

Note: One may get the error message libdc1394 error: Failed to initialize libdc1394, which is due to opencv and can be ignored.

Train a model on MNIST to check everything works

docker run dmlc/mxnet python /mxnet/example/image-classification/train_mnist.py

If the host machine has Nvidia GPUs, we can use dmlc/mxnet:cuda, which has both CUDA and CUDNN installed. To launch the docker, we need to install nvidia-docker first.

Pull the image
```
docker pull dmlc/mxnet:cuda
```

Train MNIST on GPU 0

nvidia-docker run dmlc/mxnet:cuda python /mxnet/example/image-classification/train_mnist.py --gpus 0

How to build

docker build -t dmlc/mxnet:cpu cpu
docker build -t dmlc/mxnet:cuda cuda