Implementation of selected Reinforcement Learning Algorithms using Pytorch on environments from OpenAI gym.
Algorithms:
- REINFORCE algorithm
- Simple Actor-Critic with shared weights
- Actor-Critic without shared weights & GAE
- PPO without shared weights & GAE & early stopping with KL
- PPO without shared weights & GAE & mini batches
- PPO with continuous action space on Lunar Lander environment
- DDPG on the Pendulum environment
- PPO with continous action space & RNN & vectorized environments
Helpful stuff:
- Demonstration of saving, loading and creating a video of Pytorch network (save_load_and_video.ipynb)
- Creating, Saving and Loading Checkpoints for RL training loops
