/Mujoco-Pytorch

PPO, DDPG, SAC implementation on mujoco environment

Primary LanguagePythonMIT LicenseMIT

Mujoco-Pytorch

  • Reinforcement learning algorithm Application on Mujoco

Implemented Algorithm

  • PPO
  • SAC
  • DDPG

Performance

PPO

hopper

reacher

swimmer

walker2d

humanoid

ant

halfcheetah

Requirements

  • python 3.6. >=
  • gym 0.17. >=
  • mujoco_py 2.0. >=
  • pytorch 1.6. >=

RUN

python main.py

  • if you want to change any options, check "python main.py --help"

  • you train and test agent using main.py

  • '--env_name', type=str, default = 'Hopper-v2', help = "'Ant-v2','HalfCheetah-v2','Hopper-v2','Humanoid-v2','HumanoidStandup-v2',
    'InvertedDoublePendulum-v2', 'InvertedPendulum-v2','Walker2d-v2','Swimmer-v2','Reacher-v2'(default : Hopper-v2)"

  • '--train', type=bool, default=True, help="(default: True)"

  • '--render', type=bool, default=False, help="(default: False)"

  • '--epochs', type=int, default=1000, help='number of epochs, (default: 1000)'

  • '--entropy_coef', type=float, default=1e-2, help='entropy coef (default : 0.01)'

  • '--critic_coef', type=float, default=0.5, help='critic coef (default : 0.5)'

  • '--learning_rate', type=float, default=3e-4, help='learning rate (default : 0.0003)'

  • '--gamma', type=float, default=0.99, help='gamma (default : 0.99)'

  • '--lmbda', type=float, default=0.95, help='lambda using GAE(default : 0.95)'

  • '--eps_clip', type=float, default=0.2, help='actor and critic clip range (default : 0.2)'

  • '--K_epoch', type=int, default=64, help='train epoch number(default : 10)'

  • '--T_horizon', type=int, default=2048, help='one generation before training(default : 2048)'

  • '--hidden_dim', type=int, default=64, help='actor and critic network hidden dimension(default : 64)'

  • '--minibatch_size', type=int, default=64, help='minibatch size(default : 64)'

  • '--tensorboard', type=bool, default=False, help='use_tensorboard, (default: False)'

  • '--load', type=str, default = 'no', help = 'load network name in ./model_weights'

  • '--save_interval', type=int, default = 100, help = 'save interval(default: 100)'

  • '--print_interval', type=int, default = 20, help = 'print interval(default : 20)'

  • '--use_cuda', type=bool, default = True, help = 'cuda usage(default : True)'