Todo
- increase number of steps in episode - changed to 5k from 2k
- max_steps in each episode too small
- HER says that it improves performance on sparse/binary rewards, rewards here are dense however
- need to inherit from gym.GoalEnv to use HER, that requires a change in how observation is structured
Installation
Create virtualenv and install dependencies from requirements.txt
pip install -r requirements.txt
To install the environment
pip install -e .