如何渲染丹瑟力的环境？

Question

如何使用Tensorforce库呈现环境？

我尝试过调用environment.render，但是它说函数不存在。这是我的密码：

from tensorforce.agents import Agent
from tensorforce.environments import Environment
from tensorforce.execution import Runner

# Create an OpenAI-Gym environment
environment = Environment.create(environment='gym', level='MountainCarContinuous-v0')

agent = Agent.create(agent='random', environment=environment)   

# runner = Runner(agent=agent, environment=environment) # Initialize the runner
# runner.run(num_episodes=NUM_EPISODES) # Start the runner
# runner.close()

# Train
for ep in range(NUM_EPISODES): # Number of episodes

    print('********Episode ' + str(ep) + '********')

    # Initialize episode
    states = environment.reset()
    done = False
    step = 0

    while not done: # Episode timestep
        actions = agent.act(states=states)
        states, done, reward = environment.execute(actions=actions)
        agent.observe(terminal=done, reward=reward)
        environment.render() # Gives error

environment.close()
agent.close()

这就是我遇到的错误：

Traceback (most recent call last):

  File "c:\users\user\reinforcement learning\rl.py", line 179, in <module>
    environment.render()

AttributeError: 'OpenAIGym' object has no attribute 'render'

Answer 1

如果你使用https://github.com/tensorforce/tensorforce/blob/master/examples/act_观察_interface.py，

以下修改有效。

进口健身房：

import gym
from gym import wrappers

然后定义一个具有健身房环境的Tensorforce环境：

    env = gym.make('CartPole-v1')
    env = wrappers.Monitor(env, 'tmp', force=True)

    environment = Environment.create(environment=env, max_episode_timesteps=500)
    # environment = Environment.create(environment='/Users/rondelion/git/tensorforce/benchmarks/configs/cartpole.json')

并渲染健身房环境：

            ....
            num_updates += agent.observe(terminal=terminal, reward=reward)
            env.render()
            sum_rewards += reward
            ....