README.md

Requirements:

The code is written to be run on Python 3.5.
The simulator is using MuJoCo 1.31 and/or Flex 2.
Most of the code is written to run using docker and/or nvidia-docker

Compile Docker Image

Flex

Make sure that you have a copy of FlexRobotics-master.zip inside the /docker/flex/ folder. Then run

cd docker/flex
docker build -t pytorchflex $(pwd)/docker/flex/.

MuJoCo

Make sure that you have a valid mjkey.txt inside the /docker/mujoco/ folder. Then run

cd docker/mujoco
docker build -t pytorchmujoco $(pwd)/docker/mujoco/.

Run Docker

docker run --runtime=nvidia --rm pytorchflex nvidia-smi

To run the docker image: docker run --runtime=nvidia --rm -ti -v $(pwd):/home/code -v $(pwd)/docker/mujoco/mjkey.txt:/.mujoco/mjkey.txt --ipc=host pytorchmujoco
docker run --runtime=nvidia --rm -e DISPLAY=$DISPLAY -ti -v $(pwd):/home/code -v $(pwd)/docker/mujoco/mjkey.txt:/.mujoco/mjkey.txt --ipc=host pytorchmujoco

Run demo:

demo_mujoco_pointmass