This is an implementation of a configurable square grid world environment (tasks can be defined based on changes in the reward structure, input space or transition dynamics).
For a 3x3 grid world, the states or input space are defined as MNIST images.
- Clone repository using
git clone <repo-url>.git - Install environment
pip install -e . - To use environment run:
>>> import dynamic_grid
>>> env = gym.make('DynamicGrid-v0')
or run
>>> import dynamic_grid
>>> env = dynamic_grid.DynamicGrid()