- This codebase is tested with Ubuntu 20.04.6 LTS, CUDA Version: 12.0, Cuda compilation tools, release 11.8, V11.8.89 and Nvidia driver version 525.125.06.
The following command will install some necessary dependencies.
sudo apt-get install build-essential libgl1-mesa-dev freeglut3-dev libglfw3 libgles2-mesa-dev
-
Create conda environment Create a conda environment and activate it:
conda env create -f environment.yml -
Using Docker to compile PyFlex.
- Install docker-ce
- Install nvidia-docker
- Install Anaconda
- Install Pybind11 using
conda install pybind11
We posted the Dockerfile. To generate the pre-built file, download the Dockerfile in this directory and run
docker build -t multiddpg .
in the directory that contains the Dockerfile.
Using the following command to run docker
nvidia-docker run \
-v PATH_TO_MultiDDPG:/workspace/multiddpg \
-v PATH_TO_CONDA:PATH_TO_CONDA \
-v /tmp/.X11-unix:/tmp/.X11-unix \
--gpus all \
-e DISPLAY=$DISPLAY \
-e QT_X11_NO_MITSHM=1 \
-it multiddpg bash
Now you are in the Docker environment. Go to the multiddpg directory and compile PyFlex
export PATH="PATH_TO_CONDA/bin:$PATH"
. ./prepare_1.0.sh && ./compile_1.0.sh
install softgym
cd softgym
pip install .
- Go to the root folder of MultiDDPG and run
. ./prepare_1.0.sh && ./compile_1.0.sh. Please see the example test scripts and the bottom ofbindings/pyflex.cppfor available APIs.
Test MultiDDPG for 10 tasks with different inital positions using the network in './model'
python examples/script.py --train 0
Train MultiDDPG for 50 episodes as an example stores in './model'
python examples/script.py --train 1
If the module 'softgym' can not be found. Please change the sys path in '.\examples\ddpg_env.py' and '.\exaples\ddpg_test.py'
- NVIDIA FleX - 1.2.0: https://github.com/NVIDIAGameWorks/FleX
- Our python interface builds on top of PyFleX: https://github.com/YunzhuLi/PyFleX
- Softgym: https://github.com/Xingyu-Lin/softgym