Aide python deep renforcement learning

Bonjour

J'aimerais mettre mon code de renforcement learning au deep renforcement learning facilement car les tutoriels que j'ai regardé n'étaient pas très clairs pour moi

Merci 

import numpy as np
from random import randint
import random

class EnvGrid(object):
    """
        docstring forEnvGrid.
    """
    def __init__(self):
        super(EnvGrid, self).__init__()

        self.grid = [
            [0, -1, 1],
            [0, 0, 0],
            [0, 0, -1]
        ]
        # Starting position
        self.y = 2
        self.x = 0

        self.actions = [
            [-1, 0], # Up
            [1, 0], #Down
            [0, -1], # Left
            [0, 1] # Right
        ]

    def reset(self):
        """
            Reset world
        """
        self.y = 2
        self.x = 0
        return (self.y*3+self.x+1)

    def step(self, action):
        """
            Action: 0, 1, 2, 3
        """
        self.y = max(0, min(self.y + self.actions[action][0],2))
        self.x = max(0, min(self.x + self.actions[action][1],2))

        return (self.y*3+self.x+1) , self.grid[self.y][self.x]

    def show(self):
        """
            Show the grid
        """
        print("---------------------")
        y = 0
        for line in self.grid:
            x = 0
            for pt in line:
                print("%s\t" % (pt if y != self.y or x != self.x else "X"), end="")
                x += 1
            y += 1
            print("")

    def is_finished(self):
        return self.grid[self.y][self.x] == 1

def take_action(st, Q, eps):
    # Take an action
    if random.uniform(0, 1) < eps:
        action = randint(0, 3)
    else: # Or greedy action
        action = np.argmax(Q[st])
    eps += 0.1
    return action

if __name__ == '__main__':
    env = EnvGrid()
    st = env.reset()

    Q = [
        [0, 0, 0, 0],
        [0, 0, 0, 0],
        [0, 0, 0, 0],
        [0, 0, 0, 0],
        [0, 0, 0, 0],
        [0, 0, 0, 0],
        [0, 0, 0, 0],
        [0, 0, 0, 0],
        [0, 0, 0, 0],
        [0, 0, 0, 0]
    ]

    for _ in range(500):
        # Reset the game
        st = env.reset()
        while not env.is_finished():
            env.show()
            #at = int(input("$>"))
            at = take_action(st, Q, 0.4)

            stp1, r = env.step(at)
            #print("s", stp1)
            #print("r", r)

            # Update Q function
            atp1 = take_action(stp1, Q, 0.1)
            Q[st][at] = Q[st][at] + 0.1*(r + 0.9*Q[stp1][atp1] - Q[st][at])
            st = stp1

    for s in range(1, 10):
        print(s, Q[s])