opponent = possible_opponents[int(selected_opponent)]()
mcts_agent = GreedyAgent()
tree = Mcts(game, simulation_agent=mcts_agent)
# tree = Mcts(game)
count = 0
while not game.done:
count = count + 1
start = time()
tree.run(30, verbose=True)
stop = time()
ob = game._get_obs()
# if ob is not None:
# game.ale.saveScreenPNG('images/' + str(count) + '-state.png')
# print(count, end=" ")
# for i, val in enumerate(ob):
# print(val, end=" ")
# print("")
print("total time: ", stop - start)
action1 = tree.predict()
action2 = opponent.act(ob, player=1)
game.act(action1)
tree.move_root(action1)
game.act(action2)
tree.move_root(action2)
game.render()
from pong.pong_game import PongGame
from pong.gym_agents import RandomAgent
if __name__ == '__main__':
# You can set the level to logger.DEBUG or logger.WARN if you
# want to change the amount of output.
logger.set_level(logger.INFO)
env = PongGame()
env.seed(0)
agent1 = RandomAgent()
agent2 = RandomAgent()
episode_count = 1
reward = 0
done = False
print(env.action_space.n)
for i in range(episode_count):
ob = env.reset()
while True:
action1 = agent1.act(ob, player=0)
action2 = agent2.act(ob, player=1)
ob, reward = env.step(action1, a2=action2)
if done:
break
env.render()
# Close the env and write monitor result info to disk
env.close()
if __name__ == '__main__':
# You can set the level to logger.DEBUG or logger.WARN if you
# want to change the amount of output.
logger.set_level(logger.INFO)
env = PongGame()
outdir = '/tmp/random-agent-results'
ACTIONS = env.action_space.n
SKIP_CONTROL = 0 # Use previous control decision SKIP_CONTROL times, that's how you
# can test what skip is still usable.
env = wrappers.Monitor(env, directory=outdir, force=True)
env.seed(0)
env.render()
env.unwrapped.viewer.window.on_key_press = key_press
env.unwrapped.viewer.window.on_key_release = key_release
episode_count = 100
reward = 0
done = False
print(env.action_space.n)
for i in range(episode_count):
ob = env.reset()
skip = 0
total_reward = 0
total_timesteps = 0
while True:
if not skip: