Skip to content
Implementation of the Deep Deterministic Policy Gradient(DDPG) in bullet Gym using pytorch
Branch: master
Clone or download
hzwer
hzwer hard data
Latest commit 6c9223e Mar 23, 2018
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
DDPG_from_baselines implement baselines via pytorch Feb 6, 2018
PPO_example PPO Feb 8, 2018
TRPO_example Merge branch 'master' of https://github.com/megvii-rl/pytorch-gym Feb 8, 2018
base.clean
base.resolve
base
ensemble new version of ensemble Feb 3, 2018
jupyter_plot
rl-painter hard data Mar 23, 2018
.gitignore
README.md

README.md

DDPG in bullet Gym using pytorch

Overview

This is an implementation of Deep Deterministic Policy Gradient (DDPG) in bullet Gym using PyTorch.

Dependencies

  • Python 3.6.2
  • pytorch 0.2.0
  • gym
  • tensorboardX-1.0
  • pybullet (if you want to train agents for bullet env)

Run

  • here is a simple example to train CartPole with high efficiency:

$ cd base

$ python main.py --debug --discrete --env=CartPole-v0 --vis

  • you can use this to understand usage of each argument:

$ python main.py --help

  • some explanation of important arguments:

--debug: print the reward and some other information

--discrete: if the actions are discrete rather than continuous

--vis: render each action (but it would slow down your training speed)

--cuda: train this task using GPU

--test: testing mode

--resume : load model from the path

DDPG from baselines

# run HalfCheetah-v1 with default parameters
cd ./baselines
python main.py 

Contributors

You can’t perform that action at this time.