Case 1: sub-optimal RL agent as expert knowledge
Case 2: low-dimensional dynamics as expert knowledge
Offline demonstration from Walker2d-v2 environment of Mujoco
mkdir demon && cd demon/
wget https://rail.eecs.berkeley.edu/datasets/offline_rl/gym_mujoco_v2/walker2d_expert-v2.hdf5
wget https://rail.eecs.berkeley.edu/datasets/offline_rl/gym_mujoco_v2/walker2d_medium-v2.hdf5
wget https://rail.eecs.berkeley.edu/datasets/offline_rl/gym_mujoco_v2/walker2d_random-v2.hdf5Offline demonstration from domain name 'walker' and task name 'walk' of DeepMind Control Suite
- access the link from V-D4RL
- download the folder
main/walker_walk/medium/64px - the overall file structure becomes:
ExpertRL
└───64px
│ └───20220109T013713-488d0464873a40c99446c53b1468e1c9-501.npz
│ └───20220109T013713-681bef276feb456c8096851ae536d2a2-501.npz
│ │ ...
│ └───20220109T013743-e3b1c9be2e534355a7427ffcceb9d042-501.npz
└───demon
│ └───walker2d_expert-v2.hdf5
│ └───walker2d_medium-v2.hdf5
│ └───walker2d_random-v2.hdf5
└───dmc2gym
└───LowDynamics
│ └───README.md
│ │ ...
└───SubOptimal
│ └───README.md
│ │ ...
└───README.md
docker pull pytorch/pytorch:1.13.1-cuda11.6-cudnn8-devel
docker run -it --shm-size=8G \
--gpus=all --restart=always \
pytorch/pytorch:1.13.1-cuda11.6-cudnn8-devel \
/bin/bashpip install h5pyMujoco: We followed the instruction in mujoco-py.
wget https://mujoco.org/download/mujoco210-linux-x86_64.tar.gz
tar -zxvf mujoco210-linux-x86_64.tar.gz
mkdir ~/.mujoco && mv mujoco210 ~/.mujoco/
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/root/.mujoco/mujoco210/bin
pip3 install -U 'mujoco-py<2.2,>=2.1'
pip install mujoco
pip install gym[mujoco]
pip install 'cython<3'
apt install libosmesa6-dev libgl1-mesa-glx libglfw3
ln -s /usr/lib/x86_64-linux-gnu/libGL.so.1 /usr/lib/x86_64-linux-gnu/libGL.soDeepmind Control Suite: We used the image wrapper in dmc2gym with a few modifications using local install, which is necessary for the updated mujoco version.
cd dmc2gym/ && pip install -e .
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent