Implementation of reinforcement learning algorithms that is applied RAY in PyTorch.
In this repo, we implemented this RL_with_RAY code only use RAY CORE from scratch, didn't use RLLib.
- Python 3.6.x
- PyTorch 1.8.1
- RAY 1.3.0
- pybullet-gym
- numpy
pip install -r requirements.txt
- n_cpu = n_workers = 5
- hdims = [256,256]
- actv = nn.Tanh()
- out_actv = nn.Tanh()
- n_worker
- total_steps
- ep_len_rollout
- hdims
- https://docs.ray.io/en/master/index.html, RAY Documentation
- https://arxiv.org/abs/1707.06347, PPO Paper
- https://arxiv.org/abs/1801.01290, SAC Paper
- https://arxiv.org/abs/1803.07055, ARS Paper