Run MXNet on Docker¶
Docker is a system that lets you build self-contained versions of a Linux operating system that can run in isolation on your computer. On the self-contained version of Linux, you can run MXNet and other software packages without them interacting with the packages on your computer.
MXNet provides two Docker images for running MXNet:
- MXNet Docker (CPU) - https://hub.docker.com/r/kaixhin/mxnet/
- MXNet Docker (GPU) - https://hub.docker.com/r/kaixhin/cuda-mxnet/
These Docker images are updated weekly with the latest builds of MXNet. For CUDA support, you need the NVIDIA Docker image.
To run MXNet on Docker:
Install Docker on your computer. For more information, see the Docker documentation.
Run the command to start the MXNet Docker container.
For CPU containers, run this command:
sudo docker run -it kaixhin/mxnet
For GPU containers, run this command:
sudo nvidia-docker run -it kaixhin/cuda-mxnet:7.0
For more details on how to use the MXNet Docker images, see the MXNet Docker GitHub documentation.