tensorforce tqdm progress callback

tensorforce_progressbar.py

from tqdm import tqdm_notebook, tqdm
class StepsProgressBarNotebook(object):
    """
    HTML5 logger for tensorforce using tqdm_notebook for jupyter-notebook.
    
    Usage: 
    `runner.run(episodes=np.inf, episode_finished=StepsProgressBar(steps=1e9, print_every=1000))`
    """
    
    def __init__(self, steps, print_every=None, mean_of=100):
        """
        steps - stop after this many total steps
        print_every - print the mean metrics every log_intv episodes
        mean_of - progress bar shows the reward as mean of the last n episodes
        """
        self.print_every = print_every
        self.mean_of = mean_of
        self.steps = steps
        self.progbar = tqdm_notebook(desc='', 
                                     total=steps, 
                                     leave=True, 
#                                      mininterval=1,
                                     unit='steps'
                                     
                                    )
    
    def __call__(self, r):
        desc = "reward: {reward: 2.4f} [{rewards_min: 2.0f}, {rewards_max: 2.0f}] episodes: {episodes}".format(
            reward=np.mean(r.episode_rewards[-self.mean_of:]),
            rewards_min=np.min(r.episode_rewards[-self.mean_of:]),
            rewards_max=np.max(r.episode_rewards[-self.mean_of:]),
            episodes=r.episode,
        )
        self.progbar.desc = desc
        self.progbar.update(r.episode_lengths[-1])
        
        if self.print_every and r.total_timesteps%self.print_every==0:
            print(self.progbar.desc)
        
        return r.total_timesteps<self.steps
        

class StepsProgressBar(object):
    """
    Progress bar for tensorforce that stop after N steps.
    
    Usage: 
    `runner.run(episodes=np.inf, episode_finished=StepsProgressBar(steps=1e9, print_every=1000))`
    """
    
    def __init__(self, steps, print_every=None, mean_of=100):
        """
        steps - stop after this many total steps
        print_every - print the mean metrics every log_intv episodes
        mean_of - progress bar shows the reward as mean of the last n episodes
        """
        self.print_every = print_every
        self.mean_of = mean_of
        self.steps = steps
        self.progbar = tqdm(desc='', 
                                     total=steps, 
                                     leave=True, 
#                                      mininterval=1,
                                     unit='steps'
                                     
                                    )
    
    def __call__(self, r):
        desc = "reward: {reward: 2.4f} [{rewards_min: 2.0f}, {rewards_max: 2.0f}] episodes: {episodes}".format(
            reward=np.mean(r.episode_rewards[-self.mean_of:]),
            rewards_min=np.min(r.episode_rewards[-self.mean_of:]),
            rewards_max=np.max(r.episode_rewards[-self.mean_of:]),
            episodes=r.episode,
        )
        self.progbar.desc = desc
        self.progbar.update(r.episode_lengths[-1])
        
        if self.print_every and r.total_timesteps%self.print_every==0:
            print(self.progbar.desc)
        
        return r.total_timesteps<self.steps