PPO与其他强化学习算法的比较:DQNA3C和A2C

输入代码“ agent = DQNAgent(state_size, action_size) # train agent batch_size = 32 episodes = 1000 for e in range(episodes): state = np.zeros(state_size) for bus in data['bus']: state[bus[0] - 1] = bus[3] state = np.reshape(state, [1, state_size]) for t in range(24): action = agent.act(state) next_state, reward, done, _ = data.step(action) next_state = np.zeros(state_size) for bus in data['bus']: next_state[bus[0] - 1] = bus[3] next_state = np.reshape(next_state, [1, state_size]) agent.remember(state, action, reward, next_state, done) state = next_state if done: break if len(agent.memory) > batch_size: agent.replay(batch_size) print("episode: {}/{}, score: {}".format(e+1, episodes, t)) # test agent state = np.zeros(state_size) for bus in data['bus']: state[bus[0] - 1] = bus[3] state = np.reshape(state, [1, state_size]) action = agent.act(state) print("key node: {}".format(action))”显示错误：'gbk' codec can't decode byte 0x93 in position 119265: illegal multibyte sequence