MujocoTestbed

MujocoTestbed of AIRLAB

Requirements

Note that: pytorch should folllow CUDA version of your GPU (If not, you may not be able to use pytorch-cuda)

• mujoco-py
• PyTorch
• Python >=3.8
• Pytorch >=1.5
• CUDA enabled computing device
• gym==0.14.0

Build

first intall mujoco-py in your conda env click this link mujoco-py

git clone https://github.com/AIRLABkhu/MujocoTestbed.git
pip install -r requirements.txt

Algorithm

DDPG,SAC,TD3 is available.

Default Arguments and Usage

---

For SAC

python SAC/main.py --env-name Humanoid-v2 --alpha 0.05

For DDPG

python DDPG/main.py --env-name Huamnoid-v2

For TD3

python TD3/main.py --env-name Humanoid-v2

Visualization

First make csv_file

python tools.py --root-dir run/{ENV_NAME}/{Algo}

and make plot

python plotter.py --task {ENV_NAME} --algo {algo1},{algo2},{algo3} (for comparison of algorithms)

Example (I run experiments on two algoritms (sac,ddpg) in Ant-v2) then

python tools.py --root-dir run/Ant-v2/sac
python tools.py --root-dir run/Ant-v2/ddpg
python plotter.py --task Ant-v2 --algo sac,ddpg