def __init__(self, rand_seed, display=False, no_op_max=7): self.env = gym.make(visual=VISUAL, game=GAME) self.state_size = self.env.observation_space_shape self.action_size = len(self.env.action_space) self._no_op_max = no_op_max if display: self._setup_display() self.reset()
import numpy as np from game import mygym as gym from training.ddqn import DQNAgent import constants import cv2 EPISODES = 5000 if __name__ == "__main__": visual = constants.VISUAL verbose = False env = gym.make(visual=visual, game='CDGame') state_size = env.observation_space_shape action_size = len(env.action_space) agent = DQNAgent(state_size, action_size) # agent.load("training/save/cd-ddqn.h5") done = False batch_size = 32 max_step = 500 for e in range(EPISODES): print('Episode {}/{}'.format(e + 1, EPISODES)) state = env.reset() state = np.reshape(state, [1, state_size[0], state_size[1], state_size[2]]) for time in range(max_step): action = agent.act(state) next_state, reward, done, total_reward = env.step(action) if visual:
import numpy as np from game import mygym as gym from training.ddqn_gw import DQNAgent import cv2 EPISODES = 5000 if __name__ == "__main__": visual = True verbose = False env = gym.make(visual=visual, game='GridWorld') state_size = env.observation_space_shape action_size = len(env.action_space) agent = DQNAgent(state_size, action_size) # agent.load("training/save/gw-ddqn.h5") done = False batch_size = 32 max_step = 500 for e in range(EPISODES): print('Episode {}/{}'.format(e + 1, EPISODES)) state = env.reset() state = np.reshape(state, [1, state_size]) for time in range(max_step): action = agent.act(state) next_state, reward, done, total_reward = env.step(action, agent) next_state = np.reshape(next_state, [1, state_size]) if visual: cv2.imshow('state', next_state) cv2.waitKey(10)
import game.mygym as gym import pygame import cv2 if __name__ == '__main__': env = gym.make() state = env.reset() carryOn = True while carryOn: next_state, reward, done, _ = env.step(2) pygame.quit()