def __init__(self):
game_mode, render, total_step_limit, total_run_limit, clip = self.get_args(
)
env_name = "BreakoutDeterministic-v4" # Handles frame skipping (4) at every iteration in deterministic env
env = MainGymWrapper.wrap(gym.make(env_name))
self.main_loop(self.game_model(game_mode, env.action_space.n), env,
render, total_step_limit, total_run_limit, clip)
def __init__(self):
game_name, game_mode, render, total_step_limit, total_run_limit, clip = self.gameargs(
)
env_name = game_name + "Deterministic-v4"
env = MainGymWrapper.wrap(gym.make(env_name))
self.main_loop(
self.game_models(game_mode, game_name, env.action_space.n), env,
render, total_step_limit, total_run_limit, clip)
def __init__(self):
game_name, game_mode, render, total_step_limit, total_run_limit, clip, shields = self._args()
print(shields)
env_name = game_name + "Deterministic-v4" # Handles frame skipping (4) at every iteration
env = MainGymWrapper.wrap(gym.make(env_name))
#env.set_shields(shields)
self._main_loop(self._game_model(game_mode, game_name, env.action_space.n), env, render, total_step_limit, total_run_limit, clip)
def run_game(n_rep=2000):
env_name = 'BreakoutDeterministic-v4'
print(env_name)
env = MainGymWrapper.wrap(gym.make(env_name))
# env = gym.make(env_name)
env.reset()
print("Metadata", env.metadata)
print("Actions:", env.action_space)
print("Observation:", env.observation_space)
actions = env.action_space
env.render()
for i in range(n_rep):
next_state, reward, terminal, info = env.step(np.random.randint(low=0, high=4))
env.render()
time.sleep(0.4)
print("Four consecutive frames")
print(info)
print(np.asarray(next_state).shape)
print(reward)
print(terminal)
plt.imshow(np.asarray(next_state)[0, ...])
plt.plot()
def __init__(self,
game_name: str = 'Alien',
game_mode: str = 'ddqn_training',
render: bool = False,
total_step_limit: int = 500000,
total_run_limit: int = None,
clip: bool = False):
"""Initialise Atari class
"""
self.game_name = game_name
self.game_mode = game_mode
self.render = render
self.total_step_limit = total_step_limit
self.total_run_limit = total_run_limit
self.clip = clip
self._args()
env_name = game_name + "Deterministic-v4" # Handles frame skipping (4) at every iteration
env = MainGymWrapper.wrap(gym.make(env_name))
self._main_loop(
self._game_model(env.action_space.n),
env) #, render, total_step_limit, total_run_limit, clip)
def _initialize_environmnt(self):
self.env = MainGymWrapper.wrap(gym.make(self.env_name))
def main():
game_name = "Breakout-v0"
env = MainGymWrapper.wrap(gym.make(game_name))
_main_loop(DDQNTrainer(game_name, INPUT_SHAPE, env.action_space.n), env)