强化学习

import gym, os

from itertools import count

import paddle

import paddle.nn as nn

import paddle.optimizer as optim

import paddle.nn.functional as F

from paddle.distribution import Categorical

 

device = paddle.get_device()

env = gym.make("CartPole-v0")  ### 或者 env = gym.make("CartPole-v0").unwrapped 开启无锁定环境训练

state_size = env.observation_space.shape[0]

action_size = env.action_space.n

lr = 0.001

class Actor(nn.Layer):

    def __init__(self, state_size, action_size):

        super(Actor, self).__init__()

        self.state_size = state_size

        self.action_size = action_size

        self.linear1 = nn.Linear(self.state_size, 128)

        self.linear2 = nn.Linear(128, 256)

        self.linear3 = nn.Linear(256, self.action_size)

<