Perfectly solves the problem of error reporting when executing env.render in the Jupyter notebook gym of the reinforcement learning server

1. Create a new virtual environment. My environment is python 3.8 and the name is: myrl

2. Find the virtual environment in jupyter notebook

In the Jupyter notebook cell, execute the following commands in sequence, as shown below:

!pip install gym==0.22

!pip install matplotlib

!pip install pygame

!pip install imageio-ffmpeg

Then execute:

#remove " > /dev/null 2>&1" to see what is going on under the hood

!pip install pyvirtualdisplay > /dev/null 2>&1

!apt-get install -y xvfb python-opengl ffmpeg > /dev/null 2>&1

Then execute:

!apt-get update > /dev/null 2>&1

!apt-get install cmake > /dev/null 2>&1

!pip install --upgrade setuptools 2>&1

!pip install ez_setup > /dev/null 2>&1

!pip install gym[atari] > /dev/null 2>&1

At this point, the environment is successfully installed.

3. Run the following code and a familiar screen will appear.

import gym

from gym import logger as gymlogger

from gym.wrappers import Monitor

gymlogger.set_level(40) #error only

import numpy as np

import random

import matplotlib

import matplotlib.pyplot as plt

%matplotlib inline

import math

import glob

import io

import base64

from IPython.display import HTML

from IPython import display as ipythondisplay

from pyvirtualdisplay import Display



display = Display(visible=0, size=(1400, 900))

display.start()



def show_video():

    mp4list = glob.glob('video/*.mp4')

    if len(mp4list) > 0:

        mp4 = mp4list[0]

        video = io.open(mp4, 'r+b').read()

        encoded = base64.b64encode(video)

        ipythondisplay.display(HTML(data='''<video alt="test" autoplay

                loop controls style="height: 400px;">

                <source src="data:video/mp4;base64,{0}" type="video/mp4" />

             </video>'''.format(encoded.decode('ascii'))))

    else:

        print("Could not find video")





def wrap_env(env):

    env = Monitor(env, './video', force=True)

    return env



# env = wrap_env(gym.make("CartPole-v0"))

env = wrap_env(gym.make("MountainCar-v0"))

observation = env.reset()

while True:

    env.render()

    #your agent goes here

    action = env.action_space.sample()

    observation, reward, done, info = env.step(action)

    if done:

        break;

env.close()

show_video()

4. The running results are as follows:

Other notes:

1. This article refers to the following links:

Reinforcement Learning (reinforcement learning)-Gym usage introduction | Literary Mathematics Jun

2. Solve the problem that gym server cannot display

https://www.twblogs.net/a/5e510a3bbd9eee21167ef3d6

3. Start jupyter as xvfb-run

xvfb-run -s "-screen 0 1400x900x24" jupyter notebook

If the startup fails and appears: xvfb-run: error: Xvfb failed to start, you can execute the following command:

pkill Xvfb, note that X must be capitalized

3. Gym can use version 0.22

pip install gym==0.22