Installing with Docker

To simplify and containerize Animius, we created a docker image that create virtual environments to isolate Animius installation. Animius is run within this virtual environment that can share resources with its host machine (access directories, use the GPU (on linux only), connect to the Internet, etc.).

Docker is the easiest way to enable GPU support on Linux since only the NVIDIA® GPU driver is required on the host machine (the NVIDIA® CUDA® Toolkit does not need to be installed).

Animius Docker requirements

  1. Install Docker on your local host machine.
  2. For GPU support on Linux, install nvidia-docker.

Download an Animius Docker image

The official TensorFlow Docker images are located in the gundammc/animius Docker Hub repository. Image releases are tagged using the following tags:

Each base tag has variants that add or change functionality:

For example, the following commands download Animius release images to your machine:

docker pull gundammc/animius                      # latest release
docker pull gundammc/animius:latest-gpu           # latest release w/ GPU support

Start an Animius Docker container

To start an Animius-configured Docker container, use the following:

docker run [-it] [--rm] [-p hostPort:containerPort] gundammc/animius[:tag] [command]

For details, see the docker run reference.