Skip to content

Latest commit

 

History

History
77 lines (62 loc) · 6.46 KB

README_en.md

File metadata and controls

77 lines (62 loc) · 6.46 KB

中文|EN

JoyRL-Offline

Install

Currently JoyRL support Python3.7 and Gym0.25.2

conda create -n joyrl python=3.7
conda activate joyrl
pip install -r requirements.txt

Torch:

# CPU
conda install pytorch==1.10.0 torchvision==0.11.0 torchaudio==0.10.0 cpuonly -c pytorch
# GPU
conda install pytorch==1.10.0 torchvision==0.11.0 torchaudio==0.10.0 cudatoolkit=11.3 -c pytorch -c conda-forge
# GPU with mirrors
pip install torch==1.10.0+cu113 torchvision==0.11.0+cu113 torchaudio==0.10.0 --extra-index-url https://download.pytorch.org/whl/cu113

Install Mujoco

pip install mujoco==2.3.2
pip install mujoco-py==2.1.2.14

click mujoco-releases and download mujoco 2.1.0, unzip and move to C:/[username]/.mujoco/, it should be like C:/[username]/.mujoco/mujoco210, then add C:/[username]/.mujoco/mujoco210/bin to SYSTEM PATH.

Note that if you are using Windows, you'd better to download visual-cpp-build-tools.

Usage

you can simply change the parameters (like env_name, algo_name) in config.config.GeneralConfig() and run:

python main.py

then it will a new folder named tasks to save results and models.

Or you can custom parameters with a yaml file as you can seen in config/custom_config_Train.yaml and run:

python main.py --yaml config/custom_config_Train.yaml

And there are presets yaml files in the presets folder and well trained results in the benchmarks folder.

Docs

please click docs

Environments

Please click envs to read environments instruments.

Algorithms

Name Policy Reference Author Notes
Monte Carlo RL introduction johnjim0816
Value Iteration RL introduction guoshicheng
Q-learning RL introduction johnjim0816
Sarsa RL introduction johnjim0816
DQN DQN Paper johnjim0816, guoshicheng (CNN)
DoubleDQN DoubleDQN Paper johnjim0816
PER_DQN PER_DQN Paper wangzhongren
NoisyDQN NoisyDQN Paper wangzhongren
DRQN DRQN Paper johnjim0816 understand LSTM
C51 C51 Paper also called Categorical DQN
REINFORCE REINFORCE Paper johnjim0816
A2C A2C blog johnjim0816
A3C On A3C paper johnjim0816, Ariel Chen
PPO PPO Paper johnjim0816, Wen Qiu PPO-clip, PPO-kl
DDPG DDPG Paper johnjim0816
TD3 TD3 Paper johnjim0816
SoftQ SoftQ Paper johnjim0816
GAIL GAIL Paper Yi Zhang
Dyna Q Dyna Q Paper guoshicheng