Use SARSA to Play MoutainCar-v0¶

In [1]:
%matplotlib inline

import sys
import logging
import itertools

import numpy as np
np.random.seed(0)
import gym
import matplotlib.pyplot as plt

logging.basicConfig(level=logging.INFO,
        format='%(asctime)s [%(levelname)s] %(message)s',
        stream=sys.stdout, datefmt='%H:%M:%S')
In [2]:
env = gym.make('MountainCar-v0')
for key in vars(env):
    logging.info('%s: %s', key, vars(env)[key])
for key in vars(env.spec):
    logging.info('%s: %s', key, vars(env.spec)[key])

15:06:49 [INFO] env: <MountainCarEnv<MountainCar-v0>>
15:06:49 [INFO] action_space: Discrete(3)
15:06:49 [INFO] observation_space: Box(-1.2000000476837158, 0.6000000238418579, (2,), float32)
15:06:49 [INFO] reward_range: (-inf, inf)
15:06:49 [INFO] metadata: {'render.modes': ['human', 'rgb_array'], 'video.frames_per_second': 30}
15:06:49 [INFO] _max_episode_steps: 200
15:06:49 [INFO] _elapsed_steps: None
15:06:49 [INFO] id: MountainCar-v0
15:06:49 [INFO] entry_point: gym.envs.classic_control:MountainCarEnv
15:06:49 [INFO] reward_threshold: -110.0
15:06:49 [INFO] nondeterministic: False
15:06:49 [INFO] max_episode_steps: 200
15:06:49 [INFO] _kwargs: {}
15:06:49 [INFO] _env_name: MountainCar
In [3]:
class TileCoder:
    def __init__(self, layer_count, feature_count):
        self.layer_count = layer_count
        self.feature_count = feature_count
        self.codebook = {}

    def get_feature(self, codeword):
        if codeword in self.codebook:
            return self.codebook[codeword]
        count = len(self.codebook)
        if count >= self.feature_count:  # resolve conflicts
            return hash(codeword) % self.feature_count
        self.codebook[codeword] = count
        return count

    def __call__(self, floats=(), ints=()):
        dim = len(floats)
        scaled_floats = tuple(f * (self.layer_count ** 2) for f in floats)
        features = []
        for layer in range(self.layer_count):
            codeword = (layer,) + tuple(
                    int((f + (1 + dim * i) * layer) / self.layer_count)
                    for i, f in enumerate(scaled_floats)) + ints
            feature = self.get_feature(codeword)
            features.append(feature)
        return features
In [4]:
class SARSAAgent:
    def __init__(self, env):
        self.action_n = env.action_space.n
        self.obs_low = env.observation_space.low
        self.obs_scale = env.observation_space.high - \
                env.observation_space.low
        self.encoder = TileCoder(8, 1896)
        self.w = np.zeros(self.encoder.feature_count)
        self.gamma = 1.
        self.learning_rate = 0.03

    def encode(self, observation, action):
        states = tuple((observation - self.obs_low) / self.obs_scale)
        actions = (action,)
        return self.encoder(states, actions)

    def get_q(self, observation, action):  # action value
        features = self.encode(observation, action)
        return self.w[features].sum()

    def reset(self, mode=None):
        self.mode = mode
        if self.mode == 'train':
            self.trajectory = []

    def step(self, observation, reward, terminated):
        if self.mode == 'train' and np.random.rand() < 0.001:
            action = np.random.randint(self.action_n)
        else:
            qs = [self.get_q(observation, action) for action in
                    range(self.action_n)]
            action = np.argmax(qs)
        if self.mode == 'train':
            self.trajectory += [observation, reward, terminated, action]
            if len(self.trajectory) >= 8:
                self.learn()
        return action

    def close(self):
        pass

    def learn(self):
        observation, _, _, action, next_observation, reward, terminated, \
                next_action = self.trajectory[-8:]
        target = reward + (1. - terminated) * self.gamma * \
                self.get_q(next_observation, next_action)
        td_error = target - self.get_q(observation, action)
        features = self.encode(observation, action)
        self.w[features] += (self.learning_rate * td_error)


agent = SARSAAgent(env)
In [5]:
def play_episode(env, agent, seed=None, mode=None, render=False):
    observation, _ = env.reset(seed=seed)
    reward, terminated, truncated = 0., False, False
    agent.reset(mode=mode)
    episode_reward, elapsed_steps = 0., 0
    while True:
        action = agent.step(observation, reward, terminated)
        if render:
            env.render()
        if terminated or truncated:
            break
        observation, reward, terminated, truncated, _ = env.step(action)
        episode_reward += reward
        elapsed_steps += 1
    agent.close()
    return episode_reward, elapsed_steps


logging.info('==== train ====')
episode_rewards = []
for episode in itertools.count():
    episode_reward, elapsed_steps = play_episode(env, agent, seed=episode,
            mode='train')
    episode_rewards.append(episode_reward)
    logging.info('train episode %d: reward = %.2f, steps = %d',
            episode, episode_reward, elapsed_steps)
    if np.mean(episode_rewards[-10:]) > -110:
        break
plt.plot(episode_rewards)


logging.info('==== test ====')
episode_rewards = []
for episode in range(100):
    episode_reward, elapsed_steps = play_episode(env, agent)
    episode_rewards.append(episode_reward)
    logging.info('test episode %d: reward = %.2f, steps = %d',
            episode, episode_reward, elapsed_steps)
logging.info('average episode reward = %.2f ± %.2f',
        np.mean(episode_rewards), np.std(episode_rewards))

15:06:49 [INFO] ==== train ====
15:06:49 [INFO] train episode 0: reward = -200.00, steps = 200
15:06:50 [INFO] train episode 1: reward = -200.00, steps = 200
15:06:50 [INFO] train episode 2: reward = -200.00, steps = 200
15:06:50 [INFO] train episode 3: reward = -200.00, steps = 200
15:06:50 [INFO] train episode 4: reward = -200.00, steps = 200
15:06:50 [INFO] train episode 5: reward = -200.00, steps = 200
15:06:50 [INFO] train episode 6: reward = -200.00, steps = 200
15:06:51 [INFO] train episode 7: reward = -200.00, steps = 200
15:06:51 [INFO] train episode 8: reward = -200.00, steps = 200
15:06:51 [INFO] train episode 9: reward = -200.00, steps = 200
15:06:51 [INFO] train episode 10: reward = -200.00, steps = 200
15:06:51 [INFO] train episode 11: reward = -200.00, steps = 200
15:06:52 [INFO] train episode 12: reward = -200.00, steps = 200
15:06:52 [INFO] train episode 13: reward = -200.00, steps = 200
15:06:52 [INFO] train episode 14: reward = -200.00, steps = 200
15:06:52 [INFO] train episode 15: reward = -200.00, steps = 200
15:06:52 [INFO] train episode 16: reward = -200.00, steps = 200
15:06:53 [INFO] train episode 17: reward = -200.00, steps = 200
15:06:53 [INFO] train episode 18: reward = -200.00, steps = 200
15:06:53 [INFO] train episode 19: reward = -200.00, steps = 200
15:06:53 [INFO] train episode 20: reward = -200.00, steps = 200
15:06:53 [INFO] train episode 21: reward = -200.00, steps = 200
15:06:54 [INFO] train episode 22: reward = -200.00, steps = 200
15:06:54 [INFO] train episode 23: reward = -200.00, steps = 200
15:06:54 [INFO] train episode 24: reward = -200.00, steps = 200
15:06:54 [INFO] train episode 25: reward = -200.00, steps = 200
15:06:54 [INFO] train episode 26: reward = -200.00, steps = 200
15:06:55 [INFO] train episode 27: reward = -200.00, steps = 200
15:06:55 [INFO] train episode 28: reward = -200.00, steps = 200
15:06:55 [INFO] train episode 29: reward = -195.00, steps = 195
15:06:55 [INFO] train episode 30: reward = -200.00, steps = 200
15:06:55 [INFO] train episode 31: reward = -200.00, steps = 200
15:06:55 [INFO] train episode 32: reward = -200.00, steps = 200
15:06:56 [INFO] train episode 33: reward = -200.00, steps = 200
15:06:56 [INFO] train episode 34: reward = -200.00, steps = 200
15:06:56 [INFO] train episode 35: reward = -200.00, steps = 200
15:06:56 [INFO] train episode 36: reward = -200.00, steps = 200
15:06:56 [INFO] train episode 37: reward = -200.00, steps = 200
15:06:56 [INFO] train episode 38: reward = -200.00, steps = 200
15:06:56 [INFO] train episode 39: reward = -200.00, steps = 200
15:06:57 [INFO] train episode 40: reward = -200.00, steps = 200
15:06:57 [INFO] train episode 41: reward = -200.00, steps = 200
15:06:57 [INFO] train episode 42: reward = -200.00, steps = 200
15:06:57 [INFO] train episode 43: reward = -199.00, steps = 199
15:06:57 [INFO] train episode 44: reward = -200.00, steps = 200
15:06:57 [INFO] train episode 45: reward = -200.00, steps = 200
15:06:58 [INFO] train episode 46: reward = -172.00, steps = 172
15:06:58 [INFO] train episode 47: reward = -200.00, steps = 200
15:06:58 [INFO] train episode 48: reward = -161.00, steps = 161
15:06:58 [INFO] train episode 49: reward = -200.00, steps = 200
15:06:58 [INFO] train episode 50: reward = -200.00, steps = 200
15:06:58 [INFO] train episode 51: reward = -200.00, steps = 200
15:06:59 [INFO] train episode 52: reward = -200.00, steps = 200
15:06:59 [INFO] train episode 53: reward = -200.00, steps = 200
15:06:59 [INFO] train episode 54: reward = -200.00, steps = 200
15:06:59 [INFO] train episode 55: reward = -200.00, steps = 200
15:06:59 [INFO] train episode 56: reward = -170.00, steps = 170
15:06:59 [INFO] train episode 57: reward = -162.00, steps = 162
15:06:59 [INFO] train episode 58: reward = -166.00, steps = 166
15:07:00 [INFO] train episode 59: reward = -200.00, steps = 200
15:07:00 [INFO] train episode 60: reward = -200.00, steps = 200
15:07:00 [INFO] train episode 61: reward = -200.00, steps = 200
15:07:00 [INFO] train episode 62: reward = -200.00, steps = 200
15:07:00 [INFO] train episode 63: reward = -161.00, steps = 161
15:07:00 [INFO] train episode 64: reward = -200.00, steps = 200
15:07:01 [INFO] train episode 65: reward = -200.00, steps = 200
15:07:01 [INFO] train episode 66: reward = -161.00, steps = 161
15:07:01 [INFO] train episode 67: reward = -200.00, steps = 200
15:07:01 [INFO] train episode 68: reward = -170.00, steps = 170
15:07:01 [INFO] train episode 69: reward = -200.00, steps = 200
15:07:01 [INFO] train episode 70: reward = -199.00, steps = 199
15:07:01 [INFO] train episode 71: reward = -161.00, steps = 161
15:07:02 [INFO] train episode 72: reward = -200.00, steps = 200
15:07:02 [INFO] train episode 73: reward = -197.00, steps = 197
15:07:02 [INFO] train episode 74: reward = -200.00, steps = 200
15:07:02 [INFO] train episode 75: reward = -200.00, steps = 200
15:07:02 [INFO] train episode 76: reward = -200.00, steps = 200
15:07:03 [INFO] train episode 77: reward = -165.00, steps = 165
15:07:03 [INFO] train episode 78: reward = -200.00, steps = 200
15:07:03 [INFO] train episode 79: reward = -158.00, steps = 158
15:07:03 [INFO] train episode 80: reward = -200.00, steps = 200
15:07:03 [INFO] train episode 81: reward = -200.00, steps = 200
15:07:03 [INFO] train episode 82: reward = -200.00, steps = 200
15:07:04 [INFO] train episode 83: reward = -168.00, steps = 168
15:07:04 [INFO] train episode 84: reward = -200.00, steps = 200
15:07:04 [INFO] train episode 85: reward = -200.00, steps = 200
15:07:04 [INFO] train episode 86: reward = -200.00, steps = 200
15:07:04 [INFO] train episode 87: reward = -171.00, steps = 171
15:07:04 [INFO] train episode 88: reward = -162.00, steps = 162
15:07:04 [INFO] train episode 89: reward = -200.00, steps = 200
15:07:05 [INFO] train episode 90: reward = -200.00, steps = 200
15:07:05 [INFO] train episode 91: reward = -200.00, steps = 200
15:07:05 [INFO] train episode 92: reward = -192.00, steps = 192
15:07:05 [INFO] train episode 93: reward = -200.00, steps = 200
15:07:05 [INFO] train episode 94: reward = -200.00, steps = 200
15:07:05 [INFO] train episode 95: reward = -154.00, steps = 154
15:07:06 [INFO] train episode 96: reward = -200.00, steps = 200
15:07:06 [INFO] train episode 97: reward = -152.00, steps = 152
15:07:06 [INFO] train episode 98: reward = -199.00, steps = 199
15:07:06 [INFO] train episode 99: reward = -200.00, steps = 200
15:07:06 [INFO] train episode 100: reward = -200.00, steps = 200
15:07:06 [INFO] train episode 101: reward = -197.00, steps = 197
15:07:06 [INFO] train episode 102: reward = -163.00, steps = 163
15:07:07 [INFO] train episode 103: reward = -161.00, steps = 161
15:07:07 [INFO] train episode 104: reward = -200.00, steps = 200
15:07:07 [INFO] train episode 105: reward = -187.00, steps = 187
15:07:07 [INFO] train episode 106: reward = -166.00, steps = 166
15:07:07 [INFO] train episode 107: reward = -200.00, steps = 200
15:07:07 [INFO] train episode 108: reward = -198.00, steps = 198
15:07:08 [INFO] train episode 109: reward = -164.00, steps = 164
15:07:08 [INFO] train episode 110: reward = -193.00, steps = 193
15:07:08 [INFO] train episode 111: reward = -160.00, steps = 160
15:07:08 [INFO] train episode 112: reward = -156.00, steps = 156
15:07:08 [INFO] train episode 113: reward = -152.00, steps = 152
15:07:08 [INFO] train episode 114: reward = -200.00, steps = 200
15:07:08 [INFO] train episode 115: reward = -155.00, steps = 155
15:07:09 [INFO] train episode 116: reward = -187.00, steps = 187
15:07:09 [INFO] train episode 117: reward = -152.00, steps = 152
15:07:09 [INFO] train episode 118: reward = -156.00, steps = 156
15:07:09 [INFO] train episode 119: reward = -154.00, steps = 154
15:07:09 [INFO] train episode 120: reward = -154.00, steps = 154
15:07:09 [INFO] train episode 121: reward = -165.00, steps = 165
15:07:09 [INFO] train episode 122: reward = -155.00, steps = 155
15:07:10 [INFO] train episode 123: reward = -200.00, steps = 200
15:07:10 [INFO] train episode 124: reward = -158.00, steps = 158
15:07:10 [INFO] train episode 125: reward = -160.00, steps = 160
15:07:10 [INFO] train episode 126: reward = -190.00, steps = 190
15:07:10 [INFO] train episode 127: reward = -156.00, steps = 156
15:07:10 [INFO] train episode 128: reward = -124.00, steps = 124
15:07:10 [INFO] train episode 129: reward = -115.00, steps = 115
15:07:11 [INFO] train episode 130: reward = -196.00, steps = 196
15:07:11 [INFO] train episode 131: reward = -126.00, steps = 126
15:07:11 [INFO] train episode 132: reward = -156.00, steps = 156
15:07:11 [INFO] train episode 133: reward = -118.00, steps = 118
15:07:11 [INFO] train episode 134: reward = -200.00, steps = 200
15:07:11 [INFO] train episode 135: reward = -155.00, steps = 155
15:07:11 [INFO] train episode 136: reward = -200.00, steps = 200
15:07:11 [INFO] train episode 137: reward = -117.00, steps = 117
15:07:12 [INFO] train episode 138: reward = -198.00, steps = 198
15:07:12 [INFO] train episode 139: reward = -179.00, steps = 179
15:07:12 [INFO] train episode 140: reward = -156.00, steps = 156
15:07:12 [INFO] train episode 141: reward = -200.00, steps = 200
15:07:12 [INFO] train episode 142: reward = -153.00, steps = 153
15:07:12 [INFO] train episode 143: reward = -131.00, steps = 131
15:07:12 [INFO] train episode 144: reward = -123.00, steps = 123
15:07:13 [INFO] train episode 145: reward = -119.00, steps = 119
15:07:13 [INFO] train episode 146: reward = -127.00, steps = 127
15:07:13 [INFO] train episode 147: reward = -154.00, steps = 154
15:07:13 [INFO] train episode 148: reward = -149.00, steps = 149
15:07:13 [INFO] train episode 149: reward = -153.00, steps = 153
15:07:13 [INFO] train episode 150: reward = -149.00, steps = 149
15:07:13 [INFO] train episode 151: reward = -200.00, steps = 200
15:07:14 [INFO] train episode 152: reward = -143.00, steps = 143
15:07:14 [INFO] train episode 153: reward = -117.00, steps = 117
15:07:14 [INFO] train episode 154: reward = -160.00, steps = 160
15:07:14 [INFO] train episode 155: reward = -193.00, steps = 193
15:07:14 [INFO] train episode 156: reward = -156.00, steps = 156
15:07:14 [INFO] train episode 157: reward = -149.00, steps = 149
15:07:14 [INFO] train episode 158: reward = -151.00, steps = 151
15:07:15 [INFO] train episode 159: reward = -116.00, steps = 116
15:07:15 [INFO] train episode 160: reward = -123.00, steps = 123
15:07:15 [INFO] train episode 161: reward = -156.00, steps = 156
15:07:15 [INFO] train episode 162: reward = -177.00, steps = 177
15:07:15 [INFO] train episode 163: reward = -147.00, steps = 147
15:07:15 [INFO] train episode 164: reward = -118.00, steps = 118
15:07:15 [INFO] train episode 165: reward = -148.00, steps = 148
15:07:16 [INFO] train episode 166: reward = -120.00, steps = 120
15:07:16 [INFO] train episode 167: reward = -153.00, steps = 153
15:07:16 [INFO] train episode 168: reward = -151.00, steps = 151
15:07:16 [INFO] train episode 169: reward = -118.00, steps = 118
15:07:16 [INFO] train episode 170: reward = -116.00, steps = 116
15:07:16 [INFO] train episode 171: reward = -123.00, steps = 123
15:07:16 [INFO] train episode 172: reward = -163.00, steps = 163
15:07:16 [INFO] train episode 173: reward = -146.00, steps = 146
15:07:16 [INFO] train episode 174: reward = -151.00, steps = 151
15:07:17 [INFO] train episode 175: reward = -175.00, steps = 175
15:07:17 [INFO] train episode 176: reward = -111.00, steps = 111
15:07:17 [INFO] train episode 177: reward = -160.00, steps = 160
15:07:17 [INFO] train episode 178: reward = -112.00, steps = 112
15:07:17 [INFO] train episode 179: reward = -122.00, steps = 122
15:07:17 [INFO] train episode 180: reward = -159.00, steps = 159
15:07:17 [INFO] train episode 181: reward = -149.00, steps = 149
15:07:17 [INFO] train episode 182: reward = -144.00, steps = 144
15:07:18 [INFO] train episode 183: reward = -141.00, steps = 141
15:07:18 [INFO] train episode 184: reward = -115.00, steps = 115
15:07:18 [INFO] train episode 185: reward = -155.00, steps = 155
15:07:18 [INFO] train episode 186: reward = -144.00, steps = 144
15:07:18 [INFO] train episode 187: reward = -125.00, steps = 125
15:07:18 [INFO] train episode 188: reward = -156.00, steps = 156
15:07:18 [INFO] train episode 189: reward = -155.00, steps = 155
15:07:18 [INFO] train episode 190: reward = -148.00, steps = 148
15:07:18 [INFO] train episode 191: reward = -115.00, steps = 115
15:07:18 [INFO] train episode 192: reward = -152.00, steps = 152
15:07:19 [INFO] train episode 193: reward = -114.00, steps = 114
15:07:19 [INFO] train episode 194: reward = -112.00, steps = 112
15:07:19 [INFO] train episode 195: reward = -113.00, steps = 113
15:07:19 [INFO] train episode 196: reward = -121.00, steps = 121
15:07:19 [INFO] train episode 197: reward = -120.00, steps = 120
15:07:19 [INFO] train episode 198: reward = -90.00, steps = 90
15:07:19 [INFO] train episode 199: reward = -150.00, steps = 150
15:07:19 [INFO] train episode 200: reward = -113.00, steps = 113
15:07:19 [INFO] train episode 201: reward = -156.00, steps = 156
15:07:19 [INFO] train episode 202: reward = -146.00, steps = 146
15:07:20 [INFO] train episode 203: reward = -152.00, steps = 152
15:07:20 [INFO] train episode 204: reward = -116.00, steps = 116
15:07:20 [INFO] train episode 205: reward = -145.00, steps = 145
15:07:20 [INFO] train episode 206: reward = -151.00, steps = 151
15:07:20 [INFO] train episode 207: reward = -151.00, steps = 151
15:07:20 [INFO] train episode 208: reward = -123.00, steps = 123
15:07:20 [INFO] train episode 209: reward = -145.00, steps = 145
15:07:20 [INFO] train episode 210: reward = -141.00, steps = 141
15:07:21 [INFO] train episode 211: reward = -161.00, steps = 161
15:07:21 [INFO] train episode 212: reward = -113.00, steps = 113
15:07:21 [INFO] train episode 213: reward = -168.00, steps = 168
15:07:21 [INFO] train episode 214: reward = -155.00, steps = 155
15:07:21 [INFO] train episode 215: reward = -154.00, steps = 154
15:07:21 [INFO] train episode 216: reward = -149.00, steps = 149
15:07:21 [INFO] train episode 217: reward = -153.00, steps = 153
15:07:21 [INFO] train episode 218: reward = -168.00, steps = 168
15:07:22 [INFO] train episode 219: reward = -149.00, steps = 149
15:07:22 [INFO] train episode 220: reward = -151.00, steps = 151
15:07:22 [INFO] train episode 221: reward = -146.00, steps = 146
15:07:22 [INFO] train episode 222: reward = -151.00, steps = 151
15:07:22 [INFO] train episode 223: reward = -151.00, steps = 151
15:07:22 [INFO] train episode 224: reward = -154.00, steps = 154
15:07:22 [INFO] train episode 225: reward = -151.00, steps = 151
15:07:22 [INFO] train episode 226: reward = -112.00, steps = 112
15:07:23 [INFO] train episode 227: reward = -166.00, steps = 166
15:07:23 [INFO] train episode 228: reward = -147.00, steps = 147
15:07:23 [INFO] train episode 229: reward = -114.00, steps = 114
15:07:23 [INFO] train episode 230: reward = -121.00, steps = 121
15:07:23 [INFO] train episode 231: reward = -110.00, steps = 110
15:07:23 [INFO] train episode 232: reward = -153.00, steps = 153
15:07:23 [INFO] train episode 233: reward = -125.00, steps = 125
15:07:23 [INFO] train episode 234: reward = -155.00, steps = 155
15:07:23 [INFO] train episode 235: reward = -149.00, steps = 149
15:07:24 [INFO] train episode 236: reward = -145.00, steps = 145
15:07:24 [INFO] train episode 237: reward = -147.00, steps = 147
15:07:24 [INFO] train episode 238: reward = -151.00, steps = 151
15:07:24 [INFO] train episode 239: reward = -148.00, steps = 148
15:07:24 [INFO] train episode 240: reward = -173.00, steps = 173
15:07:24 [INFO] train episode 241: reward = -138.00, steps = 138
15:07:24 [INFO] train episode 242: reward = -120.00, steps = 120
15:07:24 [INFO] train episode 243: reward = -158.00, steps = 158
15:07:24 [INFO] train episode 244: reward = -144.00, steps = 144
15:07:25 [INFO] train episode 245: reward = -150.00, steps = 150
15:07:25 [INFO] train episode 246: reward = -153.00, steps = 153
15:07:25 [INFO] train episode 247: reward = -143.00, steps = 143
15:07:25 [INFO] train episode 248: reward = -144.00, steps = 144
15:07:25 [INFO] train episode 249: reward = -148.00, steps = 148
15:07:25 [INFO] train episode 250: reward = -132.00, steps = 132
15:07:25 [INFO] train episode 251: reward = -151.00, steps = 151
15:07:25 [INFO] train episode 252: reward = -150.00, steps = 150
15:07:26 [INFO] train episode 253: reward = -148.00, steps = 148
15:07:26 [INFO] train episode 254: reward = -144.00, steps = 144
15:07:26 [INFO] train episode 255: reward = -146.00, steps = 146
15:07:26 [INFO] train episode 256: reward = -143.00, steps = 143
15:07:26 [INFO] train episode 257: reward = -148.00, steps = 148
15:07:26 [INFO] train episode 258: reward = -136.00, steps = 136
15:07:26 [INFO] train episode 259: reward = -111.00, steps = 111
15:07:26 [INFO] train episode 260: reward = -152.00, steps = 152
15:07:26 [INFO] train episode 261: reward = -116.00, steps = 116
15:07:27 [INFO] train episode 262: reward = -164.00, steps = 164
15:07:27 [INFO] train episode 263: reward = -149.00, steps = 149
15:07:27 [INFO] train episode 264: reward = -142.00, steps = 142
15:07:27 [INFO] train episode 265: reward = -145.00, steps = 145
15:07:27 [INFO] train episode 266: reward = -153.00, steps = 153
15:07:27 [INFO] train episode 267: reward = -165.00, steps = 165
15:07:27 [INFO] train episode 268: reward = -101.00, steps = 101
15:07:27 [INFO] train episode 269: reward = -146.00, steps = 146
15:07:27 [INFO] train episode 270: reward = -92.00, steps = 92
15:07:27 [INFO] train episode 271: reward = -114.00, steps = 114
15:07:28 [INFO] train episode 272: reward = -144.00, steps = 144
15:07:28 [INFO] train episode 273: reward = -113.00, steps = 113
15:07:28 [INFO] train episode 274: reward = -89.00, steps = 89
15:07:28 [INFO] train episode 275: reward = -170.00, steps = 170
15:07:28 [INFO] train episode 276: reward = -109.00, steps = 109
15:07:28 [INFO] train episode 277: reward = -92.00, steps = 92
15:07:28 [INFO] train episode 278: reward = -87.00, steps = 87
15:07:28 [INFO] train episode 279: reward = -160.00, steps = 160
15:07:28 [INFO] train episode 280: reward = -107.00, steps = 107
15:07:28 [INFO] train episode 281: reward = -161.00, steps = 161
15:07:29 [INFO] train episode 282: reward = -131.00, steps = 131
15:07:29 [INFO] train episode 283: reward = -142.00, steps = 142
15:07:29 [INFO] train episode 284: reward = -159.00, steps = 159
15:07:29 [INFO] train episode 285: reward = -114.00, steps = 114
15:07:29 [INFO] train episode 286: reward = -111.00, steps = 111
15:07:29 [INFO] train episode 287: reward = -155.00, steps = 155
15:07:29 [INFO] train episode 288: reward = -146.00, steps = 146
15:07:29 [INFO] train episode 289: reward = -129.00, steps = 129
15:07:29 [INFO] train episode 290: reward = -154.00, steps = 154
15:07:30 [INFO] train episode 291: reward = -155.00, steps = 155
15:07:30 [INFO] train episode 292: reward = -147.00, steps = 147
15:07:30 [INFO] train episode 293: reward = -146.00, steps = 146
15:07:30 [INFO] train episode 294: reward = -154.00, steps = 154
15:07:30 [INFO] train episode 295: reward = -123.00, steps = 123
15:07:30 [INFO] train episode 296: reward = -149.00, steps = 149
15:07:30 [INFO] train episode 297: reward = -159.00, steps = 159
15:07:30 [INFO] train episode 298: reward = -159.00, steps = 159
15:07:31 [INFO] train episode 299: reward = -158.00, steps = 158
15:07:31 [INFO] train episode 300: reward = -164.00, steps = 164
15:07:31 [INFO] train episode 301: reward = -155.00, steps = 155
15:07:31 [INFO] train episode 302: reward = -129.00, steps = 129
15:07:31 [INFO] train episode 303: reward = -157.00, steps = 157
15:07:31 [INFO] train episode 304: reward = -160.00, steps = 160
15:07:31 [INFO] train episode 305: reward = -93.00, steps = 93
15:07:31 [INFO] train episode 306: reward = -145.00, steps = 145
15:07:31 [INFO] train episode 307: reward = -108.00, steps = 108
15:07:31 [INFO] train episode 308: reward = -151.00, steps = 151
15:07:32 [INFO] train episode 309: reward = -88.00, steps = 88
15:07:32 [INFO] train episode 310: reward = -109.00, steps = 109
15:07:32 [INFO] train episode 311: reward = -93.00, steps = 93
15:07:32 [INFO] train episode 312: reward = -169.00, steps = 169
15:07:32 [INFO] train episode 313: reward = -91.00, steps = 91
15:07:32 [INFO] train episode 314: reward = -156.00, steps = 156
15:07:32 [INFO] train episode 315: reward = -123.00, steps = 123
15:07:32 [INFO] train episode 316: reward = -138.00, steps = 138
15:07:32 [INFO] train episode 317: reward = -138.00, steps = 138
15:07:33 [INFO] train episode 318: reward = -142.00, steps = 142
15:07:33 [INFO] train episode 319: reward = -116.00, steps = 116
15:07:33 [INFO] train episode 320: reward = -119.00, steps = 119
15:07:33 [INFO] train episode 321: reward = -114.00, steps = 114
15:07:33 [INFO] train episode 322: reward = -115.00, steps = 115
15:07:33 [INFO] train episode 323: reward = -150.00, steps = 150
15:07:33 [INFO] train episode 324: reward = -119.00, steps = 119
15:07:33 [INFO] train episode 325: reward = -148.00, steps = 148
15:07:33 [INFO] train episode 326: reward = -148.00, steps = 148
15:07:33 [INFO] train episode 327: reward = -149.00, steps = 149
15:07:34 [INFO] train episode 328: reward = -149.00, steps = 149
15:07:34 [INFO] train episode 329: reward = -149.00, steps = 149
15:07:34 [INFO] train episode 330: reward = -146.00, steps = 146
15:07:34 [INFO] train episode 331: reward = -147.00, steps = 147
15:07:34 [INFO] train episode 332: reward = -143.00, steps = 143
15:07:34 [INFO] train episode 333: reward = -156.00, steps = 156
15:07:34 [INFO] train episode 334: reward = -114.00, steps = 114
15:07:34 [INFO] train episode 335: reward = -148.00, steps = 148
15:07:34 [INFO] train episode 336: reward = -113.00, steps = 113
15:07:35 [INFO] train episode 337: reward = -157.00, steps = 157
15:07:35 [INFO] train episode 338: reward = -148.00, steps = 148
15:07:35 [INFO] train episode 339: reward = -156.00, steps = 156
15:07:35 [INFO] train episode 340: reward = -156.00, steps = 156
15:07:35 [INFO] train episode 341: reward = -152.00, steps = 152
15:07:35 [INFO] train episode 342: reward = -138.00, steps = 138
15:07:35 [INFO] train episode 343: reward = -148.00, steps = 148
15:07:35 [INFO] train episode 344: reward = -154.00, steps = 154
15:07:35 [INFO] train episode 345: reward = -116.00, steps = 116
15:07:36 [INFO] train episode 346: reward = -114.00, steps = 114
15:07:36 [INFO] train episode 347: reward = -177.00, steps = 177
15:07:36 [INFO] train episode 348: reward = -97.00, steps = 97
15:07:36 [INFO] train episode 349: reward = -145.00, steps = 145
15:07:36 [INFO] train episode 350: reward = -115.00, steps = 115
15:07:36 [INFO] train episode 351: reward = -138.00, steps = 138
15:07:36 [INFO] train episode 352: reward = -146.00, steps = 146
15:07:36 [INFO] train episode 353: reward = -144.00, steps = 144
15:07:37 [INFO] train episode 354: reward = -156.00, steps = 156
15:07:37 [INFO] train episode 355: reward = -135.00, steps = 135
15:07:37 [INFO] train episode 356: reward = -138.00, steps = 138
15:07:37 [INFO] train episode 357: reward = -139.00, steps = 139
15:07:37 [INFO] train episode 358: reward = -150.00, steps = 150
15:07:37 [INFO] train episode 359: reward = -146.00, steps = 146
15:07:37 [INFO] train episode 360: reward = -156.00, steps = 156
15:07:37 [INFO] train episode 361: reward = -155.00, steps = 155
15:07:37 [INFO] train episode 362: reward = -127.00, steps = 127
15:07:38 [INFO] train episode 363: reward = -147.00, steps = 147
15:07:38 [INFO] train episode 364: reward = -148.00, steps = 148
15:07:38 [INFO] train episode 365: reward = -145.00, steps = 145
15:07:38 [INFO] train episode 366: reward = -142.00, steps = 142
15:07:38 [INFO] train episode 367: reward = -154.00, steps = 154
15:07:38 [INFO] train episode 368: reward = -151.00, steps = 151
15:07:38 [INFO] train episode 369: reward = -145.00, steps = 145
15:07:39 [INFO] train episode 370: reward = -150.00, steps = 150
15:07:39 [INFO] train episode 371: reward = -145.00, steps = 145
15:07:39 [INFO] train episode 372: reward = -141.00, steps = 141
15:07:39 [INFO] train episode 373: reward = -146.00, steps = 146
15:07:39 [INFO] train episode 374: reward = -148.00, steps = 148
15:07:39 [INFO] train episode 375: reward = -121.00, steps = 121
15:07:39 [INFO] train episode 376: reward = -157.00, steps = 157
15:07:39 [INFO] train episode 377: reward = -154.00, steps = 154
15:07:40 [INFO] train episode 378: reward = -157.00, steps = 157
15:07:40 [INFO] train episode 379: reward = -148.00, steps = 148
15:07:40 [INFO] train episode 380: reward = -131.00, steps = 131
15:07:40 [INFO] train episode 381: reward = -123.00, steps = 123
15:07:40 [INFO] train episode 382: reward = -168.00, steps = 168
15:07:40 [INFO] train episode 383: reward = -155.00, steps = 155
15:07:40 [INFO] train episode 384: reward = -154.00, steps = 154
15:07:40 [INFO] train episode 385: reward = -97.00, steps = 97
15:07:40 [INFO] train episode 386: reward = -112.00, steps = 112
15:07:40 [INFO] train episode 387: reward = -84.00, steps = 84
15:07:41 [INFO] train episode 388: reward = -159.00, steps = 159
15:07:41 [INFO] train episode 389: reward = -155.00, steps = 155
15:07:41 [INFO] train episode 390: reward = -163.00, steps = 163
15:07:41 [INFO] train episode 391: reward = -147.00, steps = 147
15:07:41 [INFO] train episode 392: reward = -162.00, steps = 162
15:07:41 [INFO] train episode 393: reward = -125.00, steps = 125
15:07:41 [INFO] train episode 394: reward = -98.00, steps = 98
15:07:41 [INFO] train episode 395: reward = -99.00, steps = 99
15:07:41 [INFO] train episode 396: reward = -161.00, steps = 161
15:07:42 [INFO] train episode 397: reward = -117.00, steps = 117
15:07:42 [INFO] train episode 398: reward = -157.00, steps = 157
15:07:42 [INFO] train episode 399: reward = -114.00, steps = 114
15:07:42 [INFO] train episode 400: reward = -99.00, steps = 99
15:07:42 [INFO] train episode 401: reward = -149.00, steps = 149
15:07:42 [INFO] train episode 402: reward = -102.00, steps = 102
15:07:42 [INFO] train episode 403: reward = -105.00, steps = 105
15:07:42 [INFO] train episode 404: reward = -102.00, steps = 102
15:07:42 [INFO] train episode 405: reward = -108.00, steps = 108
15:07:42 [INFO] train episode 406: reward = -145.00, steps = 145
15:07:43 [INFO] train episode 407: reward = -108.00, steps = 108
15:07:43 [INFO] train episode 408: reward = -108.00, steps = 108
15:07:43 [INFO] train episode 409: reward = -97.00, steps = 97
15:07:43 [INFO] train episode 410: reward = -162.00, steps = 162
15:07:43 [INFO] train episode 411: reward = -141.00, steps = 141
15:07:43 [INFO] train episode 412: reward = -141.00, steps = 141
15:07:43 [INFO] train episode 413: reward = -107.00, steps = 107
15:07:43 [INFO] train episode 414: reward = -153.00, steps = 153
15:07:44 [INFO] train episode 415: reward = -153.00, steps = 153
15:07:44 [INFO] train episode 416: reward = -139.00, steps = 139
15:07:44 [INFO] train episode 417: reward = -143.00, steps = 143
15:07:44 [INFO] train episode 418: reward = -164.00, steps = 164
15:07:44 [INFO] train episode 419: reward = -150.00, steps = 150
15:07:44 [INFO] train episode 420: reward = -143.00, steps = 143
15:07:44 [INFO] train episode 421: reward = -108.00, steps = 108
15:07:44 [INFO] train episode 422: reward = -148.00, steps = 148
15:07:44 [INFO] train episode 423: reward = -156.00, steps = 156
15:07:44 [INFO] train episode 424: reward = -106.00, steps = 106
15:07:45 [INFO] train episode 425: reward = -106.00, steps = 106
15:07:45 [INFO] train episode 426: reward = -107.00, steps = 107
15:07:45 [INFO] train episode 427: reward = -87.00, steps = 87
15:07:45 [INFO] train episode 428: reward = -87.00, steps = 87
15:07:45 [INFO] train episode 429: reward = -189.00, steps = 189
15:07:45 [INFO] train episode 430: reward = -155.00, steps = 155
15:07:45 [INFO] train episode 431: reward = -107.00, steps = 107
15:07:45 [INFO] train episode 432: reward = -150.00, steps = 150
15:07:45 [INFO] train episode 433: reward = -89.00, steps = 89
15:07:46 [INFO] train episode 434: reward = -141.00, steps = 141
15:07:46 [INFO] train episode 435: reward = -139.00, steps = 139
15:07:46 [INFO] train episode 436: reward = -87.00, steps = 87
15:07:46 [INFO] train episode 437: reward = -151.00, steps = 151
15:07:46 [INFO] train episode 438: reward = -96.00, steps = 96
15:07:46 [INFO] train episode 439: reward = -108.00, steps = 108
15:07:46 [INFO] train episode 440: reward = -136.00, steps = 136
15:07:46 [INFO] train episode 441: reward = -108.00, steps = 108
15:07:46 [INFO] train episode 442: reward = -101.00, steps = 101
15:07:46 [INFO] train episode 443: reward = -107.00, steps = 107
15:07:47 [INFO] train episode 444: reward = -106.00, steps = 106
15:07:47 [INFO] train episode 445: reward = -91.00, steps = 91
15:07:47 [INFO] ==== test ====
15:07:47 [INFO] test episode 0: reward = -106.00, steps = 106
15:07:47 [INFO] test episode 1: reward = -161.00, steps = 161
15:07:47 [INFO] test episode 2: reward = -86.00, steps = 86
15:07:47 [INFO] test episode 3: reward = -161.00, steps = 161
15:07:47 [INFO] test episode 4: reward = -200.00, steps = 200
15:07:47 [INFO] test episode 5: reward = -104.00, steps = 104
15:07:47 [INFO] test episode 6: reward = -83.00, steps = 83
15:07:47 [INFO] test episode 7: reward = -103.00, steps = 103
15:07:47 [INFO] test episode 8: reward = -91.00, steps = 91
15:07:47 [INFO] test episode 9: reward = -104.00, steps = 104
15:07:47 [INFO] test episode 10: reward = -84.00, steps = 84
15:07:47 [INFO] test episode 11: reward = -200.00, steps = 200
15:07:47 [INFO] test episode 12: reward = -173.00, steps = 173
15:07:48 [INFO] test episode 13: reward = -106.00, steps = 106
15:07:48 [INFO] test episode 14: reward = -90.00, steps = 90
15:07:48 [INFO] test episode 15: reward = -85.00, steps = 85
15:07:48 [INFO] test episode 16: reward = -88.00, steps = 88
15:07:48 [INFO] test episode 17: reward = -103.00, steps = 103
15:07:48 [INFO] test episode 18: reward = -103.00, steps = 103
15:07:48 [INFO] test episode 19: reward = -200.00, steps = 200
15:07:48 [INFO] test episode 20: reward = -200.00, steps = 200
15:07:48 [INFO] test episode 21: reward = -178.00, steps = 178
15:07:48 [INFO] test episode 22: reward = -104.00, steps = 104
15:07:48 [INFO] test episode 23: reward = -88.00, steps = 88
15:07:48 [INFO] test episode 24: reward = -107.00, steps = 107
15:07:48 [INFO] test episode 25: reward = -86.00, steps = 86
15:07:48 [INFO] test episode 26: reward = -147.00, steps = 147
15:07:48 [INFO] test episode 27: reward = -92.00, steps = 92
15:07:48 [INFO] test episode 28: reward = -170.00, steps = 170
15:07:48 [INFO] test episode 29: reward = -96.00, steps = 96
15:07:48 [INFO] test episode 30: reward = -104.00, steps = 104
15:07:48 [INFO] test episode 31: reward = -170.00, steps = 170
15:07:48 [INFO] test episode 32: reward = -90.00, steps = 90
15:07:48 [INFO] test episode 33: reward = -84.00, steps = 84
15:07:48 [INFO] test episode 34: reward = -134.00, steps = 134
15:07:49 [INFO] test episode 35: reward = -104.00, steps = 104
15:07:49 [INFO] test episode 36: reward = -107.00, steps = 107
15:07:49 [INFO] test episode 37: reward = -85.00, steps = 85
15:07:49 [INFO] test episode 38: reward = -175.00, steps = 175
15:07:49 [INFO] test episode 39: reward = -188.00, steps = 188
15:07:49 [INFO] test episode 40: reward = -104.00, steps = 104
15:07:49 [INFO] test episode 41: reward = -90.00, steps = 90
15:07:49 [INFO] test episode 42: reward = -83.00, steps = 83
15:07:49 [INFO] test episode 43: reward = -200.00, steps = 200
15:07:49 [INFO] test episode 44: reward = -200.00, steps = 200
15:07:49 [INFO] test episode 45: reward = -83.00, steps = 83
15:07:49 [INFO] test episode 46: reward = -84.00, steps = 84
15:07:49 [INFO] test episode 47: reward = -103.00, steps = 103
15:07:49 [INFO] test episode 48: reward = -84.00, steps = 84
15:07:49 [INFO] test episode 49: reward = -86.00, steps = 86
15:07:49 [INFO] test episode 50: reward = -107.00, steps = 107
15:07:49 [INFO] test episode 51: reward = -91.00, steps = 91
15:07:49 [INFO] test episode 52: reward = -183.00, steps = 183
15:07:49 [INFO] test episode 53: reward = -103.00, steps = 103
15:07:49 [INFO] test episode 54: reward = -103.00, steps = 103
15:07:50 [INFO] test episode 55: reward = -174.00, steps = 174
15:07:50 [INFO] test episode 56: reward = -87.00, steps = 87
15:07:50 [INFO] test episode 57: reward = -161.00, steps = 161
15:07:50 [INFO] test episode 58: reward = -200.00, steps = 200
15:07:50 [INFO] test episode 59: reward = -104.00, steps = 104
15:07:50 [INFO] test episode 60: reward = -85.00, steps = 85
15:07:50 [INFO] test episode 61: reward = -104.00, steps = 104
15:07:50 [INFO] test episode 62: reward = -190.00, steps = 190
15:07:50 [INFO] test episode 63: reward = -184.00, steps = 184
15:07:50 [INFO] test episode 64: reward = -104.00, steps = 104
15:07:50 [INFO] test episode 65: reward = -106.00, steps = 106
15:07:50 [INFO] test episode 66: reward = -104.00, steps = 104
15:07:50 [INFO] test episode 67: reward = -103.00, steps = 103
15:07:50 [INFO] test episode 68: reward = -153.00, steps = 153
15:07:50 [INFO] test episode 69: reward = -171.00, steps = 171
15:07:51 [INFO] test episode 70: reward = -155.00, steps = 155
15:07:51 [INFO] test episode 71: reward = -107.00, steps = 107
15:07:51 [INFO] test episode 72: reward = -87.00, steps = 87
15:07:51 [INFO] test episode 73: reward = -178.00, steps = 178
15:07:51 [INFO] test episode 74: reward = -103.00, steps = 103
15:07:51 [INFO] test episode 75: reward = -107.00, steps = 107
15:07:51 [INFO] test episode 76: reward = -84.00, steps = 84
15:07:51 [INFO] test episode 77: reward = -110.00, steps = 110
15:07:51 [INFO] test episode 78: reward = -181.00, steps = 181
15:07:51 [INFO] test episode 79: reward = -200.00, steps = 200
15:07:51 [INFO] test episode 80: reward = -157.00, steps = 157
15:07:51 [INFO] test episode 81: reward = -103.00, steps = 103
15:07:51 [INFO] test episode 82: reward = -93.00, steps = 93
15:07:51 [INFO] test episode 83: reward = -103.00, steps = 103
15:07:51 [INFO] test episode 84: reward = -104.00, steps = 104
15:07:51 [INFO] test episode 85: reward = -103.00, steps = 103
15:07:51 [INFO] test episode 86: reward = -86.00, steps = 86
15:07:52 [INFO] test episode 87: reward = -179.00, steps = 179
15:07:52 [INFO] test episode 88: reward = -103.00, steps = 103
15:07:52 [INFO] test episode 89: reward = -103.00, steps = 103
15:07:52 [INFO] test episode 90: reward = -188.00, steps = 188
15:07:52 [INFO] test episode 91: reward = -153.00, steps = 153
15:07:52 [INFO] test episode 92: reward = -103.00, steps = 103
15:07:52 [INFO] test episode 93: reward = -86.00, steps = 86
15:07:52 [INFO] test episode 94: reward = -108.00, steps = 108
15:07:52 [INFO] test episode 95: reward = -103.00, steps = 103
15:07:52 [INFO] test episode 96: reward = -200.00, steps = 200
15:07:52 [INFO] test episode 97: reward = -107.00, steps = 107
15:07:52 [INFO] test episode 98: reward = -103.00, steps = 103
15:07:52 [INFO] test episode 99: reward = -86.00, steps = 86
15:07:52 [INFO] average episode reward = -123.57 ± 40.09
In [6]:
env.close()