Function play_one

rl2/atari/dqn_tf.py:278–341

(
  env,
  sess,
  total_t,
  experience_replay_buffer,
  model,
  target_model,
  image_transformer,
  gamma,
  batch_size,
  epsilon,
  epsilon_change,
  epsilon_min)

def play_one(
  env,
  sess,
  total_t,
  experience_replay_buffer,
  model,
  target_model,
  image_transformer,
  gamma,
  batch_size,
  epsilon,
  epsilon_change,
  epsilon_min):

  t0 = datetime.now()

  # Reset the environment
  obs = env.reset()
  obs_small = image_transformer.transform(obs, sess)
  state = np.stack([obs_small] * 4, axis=2)
  loss = None


  total_time_training = 0
  num_steps_in_episode = 0
  episode_reward = 0

  done = False
  while not done:

    # Update target network
    if total_t % TARGET_UPDATE_PERIOD == 0:
      target_model.copy_from(model)
      print("Copied model parameters to target network. total_t = %s, period = %s" % (total_t, TARGET_UPDATE_PERIOD))


    # Take action
    action = model.sample_action(state, epsilon)
    obs, reward, done, _ = env.step(action)
    obs_small = image_transformer.transform(obs, sess)
    next_state = update_state(state, obs_small)

    # Compute total reward
    episode_reward += reward

    # Save the latest experience
    experience_replay_buffer.add_experience(action, obs_small, reward, done)    

    # Train the model, keep track of time
    t0_2 = datetime.now()
    loss = learn(model, target_model, experience_replay_buffer, gamma, batch_size)
    dt = datetime.now() - t0_2

    # More debugging info
    total_time_training += dt.total_seconds()
    num_steps_in_episode += 1