SimToReal

Installation / Setup

Note: This repo has only been tested/used on Linux Ubuntu and may have limited/broken functionality on other operating systems

1. Clone SimToReal GitHub repository

git clone https://github.com/CreativeNick/SimToReal.git

2. Clone ManiSkill GitHub repository

Visit the ManiSkill repository and follow the installation instructions. Be sure to clone/install ManiSkill inside the SimToReal/ folder. It's recommened to do this in a virtual environment. \ In summary:

Create virtual environment
Run pip install --upgrade mani_skill to install ManiSkill package
Run pip install torch torchvision torchaudio to install Torch
Run sudo apt-get install libvulkan1 to install Vulkan
Run sudo apt-get install vulkan-utils followed by vulkaninfo to test Vulkan installation

Optional Setup / Installation

There are some libraries, assets, etc you may need to install when trouble-shooting. If the above installation works for you, then you may ignore this optional setup.

1. Download YCB Assets

This repo uses the Yale-CMU-Berkeley (YCB) Object and Model set, which you can learn more about here. To download the entire asset folder, follow the instructions found in the ycb-tools repository. Be sure to install the asset folder inside the SimToReal/ folder.

2. Download other libraries

There may be other libraries/programs you may need to install before running certain commands, such as tyro in ppo.py.

pip install tyro

FFmpeg ImageIO are both required to capture and render videos for both training and evaluation. In case you already pip installed both of these, make sure you have the right version (Only run the second line if you do not have both installed).

pip uninstall "imageio-ffmpeg" "imageio"
pip install "imageio-ffmpeg"=="0.4.9" imageio=="2.34.0"