epoche_actions.append(action_onehot)
if done:
episode_rewards_sum = sum(epoche_rewards)
max_reward = max(episode_rewards_sum, max_reward)
if episode_rewards_sum > 0:
suc_count += 1
print("-----------------------")
print("Episode: ", epoch)
print("Reward: ", episode_rewards_sum)
print("Max reward during train: ", max_reward)
print("-----------------------")
epoche_rewards = model.calc_reward(epoche_rewards)
replBuffer.append(epoche_observations, epoche_actions,
epoche_rewards)
model.fit(epoche_observations, epoche_actions, epoche_rewards,
replBuffer)
epoche_observations = []
epoche_actions = []
epoche_rewards = []
training_version = load_version + (
epochs_count - current_epoch) // save_period
save_path = "res/{}/{}/LunarLander-v2.ckpt".format(
train_model_name, training_version)
