A2C with intrinsic reward

Prerequisite

The following packages should be installed:

torch
numpy
gym
gym[atari]
opencv

Additionally, CUDA should be available.

To run the agents

# For simple policy-gradient:
python pg.py
# For simple policy-gradient with intrinsic reward:
python pg_r_in.py
# For A2C:
python original_actor_critic.py
# For A2C with intrinsic reward:
python intrinsic_reward_actor_critic.py