shouvikcirca / pongusinga2c

Training an agent to play pong using Pytorch with A2C which is a RL algorithm

Training_Sequential.ipynb contains the latest code for Sequential training.

Installation

pip3 install torch
pip3 install gym
pip3 install gym[atari]

Frame Dimensions

env.observation_space

Box(210, 160, 3)

Actions

 env.unwrapped.get_action_meanings()

['NOOP', 'FIRE', 'RIGHT', 'LEFT', 'RIGHTFIRE', 'LEFTFIRE']

Sample Run 1

python3 pong.py 1

Score obtained: 0.0

Sample Run 2

python3 pong.py 0

Enter Number of games
10

Score        0	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	20	21	
Frequency    10	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0