Python result.logged_results_to_HBS_result方法代碼示例

# 需要導入模塊: from hpbandster.core import result [as 別名]
# 或者: from hpbandster.core.result import logged_results_to_HBS_result [as 別名]
def fit(self, pipeline_config, optimized_hyperparameter_config, budget, loss, info, refit=None):
        if refit or pipeline_config["ensemble_size"] == 0 or pipeline_config["task_id"] not in [-1, 1]:
            return {"optimized_hyperparameter_config": optimized_hyperparameter_config, "budget": budget}
        
        filename = os.path.join(pipeline_config["result_logger_dir"], 'predictions_for_ensemble.npy')
        optimize_metric = self.pipeline[MetricSelector.get_name()].metrics[pipeline_config["optimize_metric"]]
        y_transform = self.pipeline[OneHotEncoding.get_name()].complete_y_tranformation
        result = logged_results_to_HBS_result(pipeline_config["result_logger_dir"])

        all_predictions, labels, model_identifiers, _ = read_ensemble_prediction_file(filename=filename, y_transform=y_transform)
        ensemble_selection, ensemble_configs = build_ensemble(result=result,
            optimize_metric=optimize_metric, ensemble_size=pipeline_config["ensemble_size"],
            all_predictions=all_predictions, labels=labels, model_identifiers=model_identifiers,
            only_consider_n_best=pipeline_config["ensemble_only_consider_n_best"],
            sorted_initialization_n_best=pipeline_config["ensemble_sorted_initialization_n_best"])

        return {"optimized_hyperparameter_config": optimized_hyperparameter_config, "budget": budget,
            "ensemble": ensemble_selection,
            "ensemble_configs": ensemble_configs,
            "loss": loss,
            "info": info
            } 

# 需要導入模塊: from hpbandster.core import result [as 別名]
# 或者: from hpbandster.core.result import logged_results_to_HBS_result [as 別名]
def fit(self, pipeline_config, autonet, run_result_dir, optimize_metric, trajectories):
        ensemble_log_file = os.path.join(run_result_dir, "ensemble_log.json")
        test_log_file = os.path.join(run_result_dir, "test_result.json")
        if not pipeline_config["enable_ensemble"] or optimize_metric is None or \
            (not os.path.exists(ensemble_log_file) and not os.path.exists(test_log_file)):
            return {"trajectories": trajectories, "optimize_metric": optimize_metric}

        try:
            started = logged_results_to_HBS_result(run_result_dir).HB_config["time_ref"]
        except:
            return {"trajectories": trajectories, "optimize_metric": optimize_metric}
        
        metrics = autonet.pipeline[MetricSelector.get_name()].metrics
        ensemble_trajectories = dict()
        test_trajectories = dict()
        if os.path.exists(ensemble_log_file):
            ensemble_trajectories = get_ensemble_trajectories(ensemble_log_file, started, metrics)
        if os.path.exists(test_log_file):
            test_trajectories = get_ensemble_trajectories(test_log_file, started, metrics,  prefix="", only_test=True)
        
        return {"trajectories": dict(trajectories, **ensemble_trajectories, **test_trajectories), "optimize_metric": optimize_metric} 

# 需要導入模塊: from hpbandster.core import result [as 別名]
# 或者: from hpbandster.core.result import logged_results_to_HBS_result [as 別名]
def test_incumbent_trajectory(self):
        """ Load example result and check incumbent_trajectory generation for general errors (whitebox-test)"""
        result = logged_results_to_HBS_result(self.result_path)

        # All budgets
        traj = get_incumbent_trajectory(result, result.HB_config['budgets'], mode='racing')
        traj = get_incumbent_trajectory(result, result.HB_config['budgets'], mode='minimum')
        traj = get_incumbent_trajectory(result, result.HB_config['budgets'], mode='prefer_higher_budget')

        # Single budgets
        traj = get_incumbent_trajectory(result, [result.HB_config['budgets'][0]], mode='racing')
        traj = get_incumbent_trajectory(result, [result.HB_config['budgets'][0]], mode='minimum')
        traj = get_incumbent_trajectory(result, [result.HB_config['budgets'][0]], mode='prefer_higher_budget') 

# 需要導入模塊: from hpbandster.core import result [as 別名]
# 或者: from hpbandster.core.result import logged_results_to_HBS_result [as 別名]
def parse_results(self, pipeline_config):
        """Parse the results of the optimization run
        
        Arguments:
            pipeline_config {dict} -- The configuration of the pipeline.
        
        Raises:
            RuntimeError:  An Error occurred when parsing the results.
        
        Returns:
            dict -- Dictionary summarizing the results
        """
        try:
            res = logged_results_to_HBS_result(pipeline_config["result_logger_dir"])
            id2config = res.get_id2config_mapping()
            incumbent_trajectory = res.get_incumbent_trajectory(bigger_is_better=False, non_decreasing_budget=False)
        except Exception as e:
            raise RuntimeError("Error parsing results. Check results.json and output for more details. An empty results.json is usually caused by a misconfiguration of AutoNet.")

        if (len(incumbent_trajectory['config_ids']) == 0):
            return dict()
        
        final_config_id = incumbent_trajectory['config_ids'][-1]
        final_budget = incumbent_trajectory['budgets'][-1]
        best_run = [r for r in res.get_runs_by_id(final_config_id) if r.budget == final_budget][0]
        return {'optimized_hyperparameter_config': id2config[final_config_id]['config'],
                'budget': final_budget,
                'loss': best_run.loss,
                'info': best_run.info} 

# 需要導入模塊: from hpbandster.core import result [as 別名]
# 或者: from hpbandster.core.result import logged_results_to_HBS_result [as 別名]
def convert(self, folders, ta_exec_dirs=None, output_dir=None, converted_dest='converted_input_data'):
        try:
            from hpbandster.core.result import Result as HPBResult
            from hpbandster.core.result import logged_results_to_HBS_result
        except ImportError as e:
            raise ImportError("To analyze BOHB-data, please install hpbandster (e.g. `pip install hpbandster`)")

        self.logger.debug("Converting BOHB-data to SMAC3-data. Called with: folders=%s, ta_exec_dirs=%s, output_dir=%s,"
                          " converted_dest=%s", str(folders), str(ta_exec_dirs), str(output_dir), str(converted_dest))

        # Using temporary files for the intermediate smac-result-like format if no output_dir specified
        if not output_dir:
            output_dir = tempfile.mkdtemp()
            self.logger.debug("Temporary directory for intermediate SMAC3-results: %s", output_dir)
        if ta_exec_dirs is None or len(ta_exec_dirs) == 0:
            ta_exec_dirs = ['.']
        if len(ta_exec_dirs) != len(folders):
            ta_exec_dirs = [ta_exec_dirs[0] for _ in folders]

        # Get a list with alternative interpretations of the configspace-file
        # (if it's a .pcs-file, for .json-files it's a length-one-list)
        cs_interpretations = self.load_configspace(folders[0])
        self.logger.debug("Loading with %d configspace alternative options...", len(cs_interpretations))
        self.logger.info("Assuming BOHB treats target algorithms as deterministic (and does not re-evaluate)")

        #####################
        # Actual conversion #
        #####################
        folder_basenames = get_folder_basenames(folders)
        result = OrderedDict()
        for f, f_base, ta_exec_dir in zip(folders, folder_basenames, ta_exec_dirs):  # Those are the parallel runs
            converted_folder_path = os.path.join(output_dir, converted_dest, f_base)
            self.logger.debug("Processing folder=%s, f_base=%s, ta_exec_dir=%s. Saving to %s.",
                              f, f_base, ta_exec_dir, converted_folder_path)
            if not os.path.exists(converted_folder_path):
                self.logger.debug("%s doesn't exist. Creating...", converted_folder_path)
                os.makedirs(converted_folder_path)

            # Original hpbandster-formatted result-object
            hp_result = logged_results_to_HBS_result(f)
            result[f] = self.hpbandster2smac(f, hp_result, cs_interpretations, converted_folder_path)

        return result 

# 需要導入模塊: from hpbandster.core import result [as 別名]
# 或者: from hpbandster.core.result import logged_results_to_HBS_result [as 別名]
def build_run_trajectories(results_folder, autonet_config, metrics, log_functions):
    # parse results
    try:
        res = logged_results_to_HBS_result(results_folder)
        incumbent_trajectory = res.get_incumbent_trajectory(bigger_is_better=False, non_decreasing_budget=False)
    except:
        print("No incumbent trajectory found")
        return dict()

    # prepare
    metric_name = autonet_config["optimize_metric"]
    all_metrics = autonet_config["additional_metrics"] + [metric_name]
    additional_metric_names = [("val_" + m, metrics[m]) for m in all_metrics]
    additional_metric_names += [("train_" + m, metrics[m]) for m in all_metrics]
    additional_metric_names += [(l, log_functions[l]) for l in autonet_config["additional_logs"]]

    # initialize incumbent trajectories
    incumbent_trajectories = dict()
    
    # save incumbent trajectories
    for name, obj in additional_metric_names:
        tj = copy(incumbent_trajectory)
        log_available = [name in run["info"] for config_id, budget in zip(tj["config_ids"], tj["budgets"])
                                             for run in res.get_runs_by_id(config_id)
                                             if run["budget"] == budget]
        tj["values"] = [run["info"][name] for config_id, budget in zip(tj["config_ids"], tj["budgets"])
                                          for run in res.get_runs_by_id(config_id)
                                          if run["budget"] == budget and name in run["info"]]
        tj["losses"] = [obj.loss_transform(x) for x in tj["values"]]

        for key, value_list in tj.items():
            if key in ["losses"]:
                continue
            tj[key] = [value for i, value in enumerate(value_list) if log_available[i]]
        if tj["losses"]:
            incumbent_trajectories[name] = tj
    
    # assume first random config has been evaluated already at time 0
    for name, trajectory in incumbent_trajectories.items():
        for key, value_list in trajectory.items():
            if not isinstance(value_list, (list, tuple)):
                continue
            trajectory[key] = [value_list[0] if key != "times_finished" else 0] + value_list

    return incumbent_trajectories