def reset(self):
obs = CartPoleEnv.reset(self)
self.steps_beyond_done = 0
self.success_steps = 0
return obs
#instanciate agent
agent = PolicyGradientAgent(state_size, action_size)
done = True
last_render_time = time.time()
episode_count = -1
#forever
while True:
#if episode is done
if done:
episode_count += 1
logging.info("Episode: %i", episode_count)
#reset env
state_np = env.reset()
#set render flag based on time
render = (time.time() - last_render_time) > 5
#Get an action from our agent
action = agent.get_action(state_np)
#take a step in the environment
new_state_np, reward, done, info_dict = env.step(action)
#let the agent record the experience
agent.record_experience(state_np, action, reward, done)
#update the state
state_np = new_state_np
