def log_diagnostics(self, paths, logger=default_logger): lms = get_stat_in_paths(paths, 'agent_infos', 'lagrange_multiplier') for key, value in create_stats_ordered_dict( "TDM LBFGS Lagrange Multiplier", lms, ).items(): logger.record_tabular(key, value)
def log_diagnostics(self, paths, logger=default_logger): statistics = OrderedDict() for name_in_env_infos, name_to_log in [ ('distance_to_target', 'Distance to Target'), ('speed', 'Speed'), ('distance_reward', 'Distance Reward'), ('action_reward', 'Action Reward'), ]: stats = get_stat_in_paths(paths, 'env_infos', name_in_env_infos) statistics.update(create_stats_ordered_dict( name_to_log, stats, )) final_stats = [s[-1] for s in stats] statistics.update( create_stats_ordered_dict( "Final " + name_to_log, final_stats, always_show_all_stats=True, )) statistics.update( create_stats_ordered_dict( "Path Lengths", get_path_lengths(paths), )) for key, value in statistics.items(): logger.record_tabular(key, value)
def log_diagnostics(self, paths, logger=default_logger): statistics = OrderedDict() for name_in_env_infos, name_to_log in [ ('distance_to_target', 'Distance to Target'), ('reward_ctrl', 'Action Reward'), ]: stat = get_stat_in_paths(paths, 'env_infos', name_in_env_infos) statistics.update(create_stats_ordered_dict( name_to_log, stat, )) distances = get_stat_in_paths(paths, 'env_infos', 'distance_to_target') statistics.update(create_stats_ordered_dict( "Final Distance to Target", [ds[-1] for ds in distances], )) for key, value in statistics.items(): logger.record_tabular(key, value)
def log_diagnostics(self, paths): reward_fwd = get_stat_in_paths(paths, 'env_infos', 'reward_fwd') reward_ctrl = get_stat_in_paths(paths, 'env_infos', 'reward_ctrl') logger.record_tabular('AvgRewardDist', np.mean(reward_fwd)) logger.record_tabular('AvgRewardCtrl', np.mean(reward_ctrl)) if len(paths) > 0: progs = [ path["observations"][-1][-3] - path["observations"][0][-3] for path in paths ] logger.record_tabular('AverageForwardProgress', np.mean(progs)) logger.record_tabular('MaxForwardProgress', np.max(progs)) logger.record_tabular('MinForwardProgress', np.min(progs)) logger.record_tabular('StdForwardProgress', np.std(progs)) else: logger.record_tabular('AverageForwardProgress', np.nan) logger.record_tabular('MaxForwardProgress', np.nan) logger.record_tabular('MinForwardProgress', np.nan) logger.record_tabular('StdForwardProgress', np.nan)
def log_diagnostics(self, paths): statistics = OrderedDict() for stat_name in [ 'arm to object distance', 'object to goal distance', 'arm to goal distance', ]: stat = get_stat_in_paths(paths, 'env_infos', stat_name) statistics.update(create_stats_ordered_dict(stat_name, stat)) for key, value in statistics.items(): logger.record_tabular(key, value)
def log_diagnostics(self, paths, logger=default_logger): statistics = OrderedDict() for name_in_env_infos, name_to_log in [ ('posafter', 'Position'), ('height', 'Height'), ('angle', 'Angle'), ]: stats = get_stat_in_paths(paths, 'env_infos', name_in_env_infos) statistics.update(create_stats_ordered_dict( name_to_log, stats, )) statistics.update( create_stats_ordered_dict( "Final " + name_to_log, [s[-1] for s in stats], )) for key, value in statistics.items(): logger.record_tabular(key, value)
def get_diagnostics(self, paths): statistics = OrderedDict() for stat_name_in_paths, stat_name_to_print in [ ('arm_object_distance', 'Distance hand to object'), ('arm_goal_distance', 'Distance hand to goal'), ]: stats = get_stat_in_paths(paths, 'env_infos', stat_name_in_paths) statistics.update(create_stats_ordered_dict( stat_name_to_print, stats, always_show_all_stats=True, )) final_stats = [s[-1] for s in stats] statistics.update(create_stats_ordered_dict( "Final " + stat_name_to_print, final_stats, always_show_all_stats=True, )) return statistics
def log_diagnostics(self, paths, logger=default_logger): statistics = OrderedDict() for stat_name_in_paths, stat_name_to_print in [ ('hand_to_object_distance', 'Distance hand to object'), ('object_to_goal_distance', 'Distance object to goal'), ('hand_to_hand_goal_distance', 'Distance hand to hand goal'), ('success', 'Success (within 0.1)'), ]: stats = get_stat_in_paths(paths, 'env_infos', stat_name_in_paths) statistics.update( create_stats_ordered_dict( stat_name_to_print, stats, always_show_all_stats=True, )) final_stats = [s[-1] for s in stats] statistics.update( create_stats_ordered_dict( "Final " + stat_name_to_print, final_stats, always_show_all_stats=True, )) for key, value in statistics.items(): logger.record_tabular(key, value)