IWoL

---

Metadrive

1. Create conda environment for MetaDrive

   cd ~/IWoL/MetaDrive
   conda env create -f environment.yaml
   conda activate [your env name]

2. MetaDrive installation (Metadrive version: https://drive.google.com/file/d/13XeZ2h9h9g-To3phFurFBmADHujnkvIT/view?usp=sharing)

   cd ~/
   git clone https://github.com/metadriverse/metadrive.git
   cd metadrive
   pip install -e .

3. Install dependencies

   cd ~/IWoL/MetaDrive
   pip install -e .
   pip install -r requirements.txt
   pip install wandb setproctitle absl-py tensorboardX imageio wandb[media]

4. Test Metadrive running example

   cd onpolicy/scrips/train_metadrive_scripts/
   sh train_inter_small_ImIWoL.sh

[Trouble shooting: simplpbr version]

If you encounter issues related to panda3d-simplepbr, try downgrading it to version 0.9.x:

pip uninstall -y panda3d-simplepbr
pip install --no-cache-dir "panda3d-simplepbr==0.9.*"

If you encounter other version issues, we recommend reinstalling specific versions of numpy, scipy, matplotlib, and seaborn:

pip install --upgrade --force-reinstall "numpy==1.26.4" "scipy==1.10.1" "matplotlib<3.9" "seaborn<0.13"

---

Robotarium

1. Create conda environment for MetaDrive

   cd ~/IWoL/Robotarium
   conda env create -f environment.yaml
   conda activate [your env name]

2. Install dependencies

   pip install -e .
   pip install -r requirements.txt # ignore pip version conflict
   pip install wandb gym etproctitle absl-py tensorboardX tensorflow imageio wandb[media] matplotlib cvxopt scipy seaborn psutil

Trouble shooting

If you encounter issues with Generator from typing_extensions, try installing it via conda:

pip install --upgrade "typing_extensions>=4.7"

---

MQE

1. Create conda environment

   conda create -n mqe python=3.8

2. Install PyTorch and Isaac Gym.

   • Install appropriate PyTorch version from https://pytorch.org/.

     pip3 install torch==1.10.0+cu113 torchvision==0.11.1+cu113 torchaudio==0.10.0+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html

   • Download and install Isaac Gym Preview 4 from https://developer.nvidia.com/isaac-gym. Ubuntu20.04 and Python3.8 recommended.

     tar -xzvf IsaacGym_Preview_4_Package.tar.gz cd isaacgym/python && pip install -e .

3. Install dependencies

   cd ~/IWoL/MQE/
   pip install -e .
   pip install wandb gym etproctitle absl-py tensorboardX tensorflow imageio wandb[media] matplotlib cvxopt scipy seaborn psutil

4. Train MQE

   cd ~/IWoL/MQE/
   python train_imiwol.py --headless --sim_device cuda:0 --rl_device cuda:0

Trouble shooting

If you encounter issues with RuntimeError: Ninja is required to load C++ extension not found, install it via conda:

conda install -y ninja

---

Bidexhands

1. Create conda environment

   conda create -n bidexhands python=3.8

2. Install PyTorch and Isaac Gym.

   • Install appropriate PyTorch version from https://pytorch.org/.

     pip3 install torch==1.10.0+cu113 torchvision==0.11.1+cu113 torchaudio==0.10.0+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html

   • Download and install Isaac Gym Preview 4 from https://developer.nvidia.com/isaac-gym. Ubuntu20.04 and Python3.8 recommended.

     tar -xzvf IsaacGym_Preview_4_Package.tar.gz cd isaacgym/python && pip install -e .

3. Install dependencies

   cd ~/IWoL/BiDexHands/
   pip install -e .
   pip install wandb gym etproctitle absl-py tensorboardX tensorflow imageio wandb[media] matplotlib cvxopt scipy seaborn psutil

4. Train BiDexHands

   train.py --task=ShadowHandTwoCatchUnderarm --algo=imiwol --num_envs=250 --headless --compute_device_id 0 --sim_device cuda:0 --graphics_device_id 0

Trouble shooting

If you encounter issues with RuntimeError: Ninja is required to load C++ extension not found, install it via conda:

conda install -y ninja