This docker image provides an easy way to rollout an Ubuntu 16.04 enviroment that is suitable for RL. It contains many preinstalled libraries, and allows VNC and IPython Notebook access. The following tools are preinstalled in the image:
- Anaconda Python 3.6
- PyTorch 0.3
- CUDA 8.0
- OpenAI Gym 0.9.4
- MuJoCo 1.31 (in order to use MuJoCo you need to put your key
mjket.txtinto the root folder) - pysc2 1.2
- VNC
- Visdom
Make sure you install nvidia-docker to work with this image. Then, you can create a container by running:
nvidia-docker build -t rlenv .
And start the container via:
nvidia-docker run -v {LOCAL_CODE_DIR}:/workspace/code -it -p 5902:5900 -p 8888:8888 -p 8097:8097 --rm rlenv
This will run the container, open up VNC port (5902), ipython notebook port (8888), Visdom port (8097), as well as mount a local folder of your choice (need to replace {LOCAL_CODE_DIR}) to the container's /workspace/code. You will be directed to a shell with Anaconda enviroment (Python 3.6) enabled.
To access the graphical interface, launch your favourite VNC client (for example Screen Sharing in MacOS) and use localhost:5902 for the host, and 123456 for the password.