def train():
env = gym.make('CartPole-v0')
agent = DQNAgent(env=env)
num_episodes = 200
for i_episode in range(num_episodes):
state = env.reset()
total_reward = 0
while True:
action = agent.get_action(state)
next_state, reward, done, info = env.step(action)
total_reward += reward
update_array = [state, action, reward, next_state, done]
agent.update(update_array)
state = next_state
if done:
print("Episode ", i_episode, ": ", total_reward, " epsilon: ", agent.epsilon)
break
agent.save('myClassModel')
env.close()
env = gym.make(env_id)
# env = gym.wrappers.Monitor(env, osp.join(log_dir,"record"))
env = WrapPytorch(env)
agent = DQNAgent(env, log_dir=log_dir)
# agent.load(log_dir)
episode_rewards = []
ep = 0
obs = env.reset()
episode_reward = 0
for frame in range(Config.MAX_FRAMES):
# print("frame", frame)
# env.render()
epsilon = Config.epsilon_by_frame(frame)
action = agent.get_action(obs, epsilon)
prev_obs = obs
obs, reward, done, _ = env.step(action)
episode_reward += reward
agent.update(prev_obs, action, reward, obs, frame)
if done:
episode_rewards.append(episode_reward)
agent.writer.add_scalar("data/reward", episode_reward, ep)
print("episode", ep, "reward:", episode_reward)
ep += 1
obs = env.reset()
episode_reward = 0
if ep % 50 == 0:
agent.save(log_dir)
agent.save(log_dir)