Skip to content

Latest commit

 

History

History
98 lines (67 loc) · 2.74 KB

README.md

File metadata and controls

98 lines (67 loc) · 2.74 KB
Raw

Docker image usage instructions

The docker image that can be built using Dockerfile allows automating setup and ensures reproducible work environment.

Pre-built images

See the existing list of docker images

Usage example:

docker pull ghcr.io/antoine-galataud/rl-testbed-for-energyplus:ray1.12.1_ep22.2.0

where image tag ray1.12.1_ep22.2.0 are Ray RLlib and EnergyPlus versions the image was built with.

Building

You need docker installed (ie sudo apt install docker.io).

Build image with:

cd rl-testbed-for-energyplus/
docker build . -f docker/Dockerfile -t rl-testbed-for-energyplus

The default EnergyPlus version used in the image is 22-2-0. The default RL framework is OpenAI baselines.

To use different options, follow example below:

docker build . \
  --build-arg EPLUS_VERSION=9-6-0 \
  --build-arg EPLUS_DL_URL="<EnergyPlus download URL>" \
  --build-arg RL_FRAMEWORK="ray" \
  -t rl-testbed-for-energyplus

Running

The image can be used to run a training as it contains all necessary dependencies. Run:

docker run -t -i rl-testbed-for-energyplus

to start the container and open a shell. Then:

cd /root/rl-testbed-for-energyplus
# launch (OpenAI baselines)
time python3 -m baselines_energyplus.trpo_mpi.run_energyplus --num-timesteps 1000000000
# launch (Ray RLlib - PPO)
time python3 -m ray_energyplus.ppo.run_energyplus --num-timesteps 1000000000

Another option is to use your own project sources to launch a training (e.g. you have forked present git repo). To do so, you can start docker image with your project sources mounted as a container volume:

docker run -t -i -v /path/to/project:/root/my-project rl-testbed-for-energyplus

Tip: simply use $(pwd) instead of /path/to/project if you want to mount the current directory.

Note: all modifications you will make to your sources located on your hard drive will be reflected in the running docker container.

Plotting

If you want to monitor progress using plots from inside the docker container, this can be done in 3 steps (tested on Ubuntu, probably not working on MacOS):

Make sure all clients can access your display:

sudo apt install x11-xserver-utils
xhost +

Run the container so it can access your display:

docker run -it \
    --env="QT_X11_NO_MITSHM=1" \
    --env="DISPLAY" \
    --volume="/tmp/.X11-unix:/tmp/.X11-unix:rw" \
    rl-testbed-for-energyplus

Run the experiment, then run monitoring tool as usual (inside the container):

python3 -m common.plot_energyplus

For more information, see also this tutorial