def __init__(self, *args, **kwargs):
super(GymDiscreteProblemWithAgent, self).__init__(*args, **kwargs)
self._env = None
self.debug_dump_frames_path = "debug_frames_env"
self.make_extra_debug_info = True
# defaults
self.environment_spec = lambda: gym.make(self.env_name)
self.real_env = self.environment_spec()
self.in_graph_wrappers = []
self.collect_hparams = rl.ppo_atari_base()
self.settable_num_steps = 20000
self.simulated_environment = None
self.warm_up = 10 # TODO(piotrm): This should be probably removed.
def __init__(self, *args, **kwargs):
super(GymDiscreteProblemWithAgent, self).__init__(*args, **kwargs)
self._env = None
self.debug_dump_frames_path = "debug_frames_env"
self.make_extra_debug_info = True
self.autoencoder_model = None
# Defaults.
self.environment_spec = lambda: gym.make(self.env_name)
self.real_env = self.environment_spec()
self.in_graph_wrappers = []
self.collect_hparams = rl.ppo_atari_base()
self.settable_num_steps = 20000
self.simulated_environment = None
self.warm_up = 10 # TODO(piotrm): This should be probably removed.
# Debug info.
self.dones = 0
self.real_reward = 0
self.real_env.reset()
self.total_sim_reward, self.total_real_reward = 0.0, 0.0
self.sum_of_rewards = 0.0
self.successful_episode_reward_predictions = 0