obs = env.reset()
agent = DQNAgent(actions=actions,
memory=memory,
update_interval=500,
train_interval=1,
batch_size=32,
memory_interval=1,
observation=obs,
input_shape=[len(obs)],
training=True,
policy=policy)
agent.compile()
result = []
for episode in range(500): # 1000エピソード回す
agent.reset()
observation = env.reset() # 環境の初期化
# observation, _, _, _ = env.step(env.action_space.sample())
observation = deepcopy(observation)
agent.observe(observation)
for t in range(250): # n回試行する
# env.render() # 表示
action = agent.act()
observation, reward, done, info = env.step(
action) # アクションを実行した結果の状態、報酬、ゲームをクリアしたかどうか、その他の情報を返す
observation = deepcopy(observation)
agent.observe(observation, reward, done)
if done:
break
# test
