nkululeko 0.86.8__py3-none-any.whl → 0.88.0__py3-none-any.whl

nkululeko/models/model_mlp.py

@@ -1,25 +1,33 @@
 # model_mlp.py
+import ast
+from collections import OrderedDict
+
+import numpy as np
 import pandas as pd
+from sklearn.metrics import recall_score
+import torch
 
-from nkululeko.utils.util import Util
 import nkululeko.glob_conf as glob_conf
+from nkululeko.losses.loss_softf1loss import SoftF1Loss
 from nkululeko.models.model import Model
 from nkululeko.reporting.reporter import Reporter
-import torch
-import ast
-import numpy as np
-from sklearn.metrics import recall_score
-from collections import OrderedDict
-from nkululeko.losses.loss_softf1loss import SoftF1Loss
+from nkululeko.utils.util import Util
 
 
-class MLP_model(Model):
+class MLPModel(Model):
     """MLP = multi layer perceptron."""
 
     is_classifier = True
 
     def __init__(self, df_train, df_test, feats_train, feats_test):
-        """Constructor taking the configuration and all dataframes."""
+        """Constructor, taking all dataframes.
+
+        Args:
+            df_train (pd.DataFrame): The train labels.
+            df_test (pd.DataFrame): The test labels.
+            feats_train (pd.DataFrame): The train features.
+            feats_test (pd.DataFrame): The test features.
+        """
         super().__init__(df_train, df_test, feats_train, feats_test)
         super().set_model_type("ann")
         self.name = "mlp"
@@ -97,7 +105,7 @@ class MLP_model(Model):
             self.optimizer.step()
         self.loss = (np.asarray(losses)).mean()
 
-    def evaluate_model(self, model, loader, device):
+    def evaluate(self, model, loader, device):
         logits = torch.zeros(len(loader.dataset), self.class_num)
         targets = torch.zeros(len(loader.dataset))
         model.eval()
@@ -119,14 +127,28 @@ class MLP_model(Model):
         self.loss_eval = (np.asarray(losses)).mean()
         predictions = logits.argmax(dim=1)
         uar = recall_score(targets.numpy(), predictions.numpy(), average="macro")
-        return uar, targets, predictions
+        return uar, targets, predictions, logits
+
+    def get_probas(self, logits):
+        # make a dataframe for probabilites (logits)
+        proba_d = {}
+        classes = self.df_test[self.target].unique()
+        classes.sort()
+        for c in classes:
+            proba_d[c] = []
+        for i, c in enumerate(classes):
+            proba_d[c] = list(logits.numpy().T[i])
+        probas = pd.DataFrame(proba_d)
+        probas = probas.set_index(self.df_test.index)
+        return probas
 
     def predict(self):
-        _, truths, predictions = self.evaluate_model(
+        _, truths, predictions, logits = self.evaluate(
             self.model, self.testloader, self.device
         )
-        uar, _, _ = self.evaluate_model(self.model, self.trainloader, self.device)
-        report = Reporter(truths, predictions, self.run, self.epoch)
+        uar, _, _, _ = self.evaluate(self.model, self.trainloader, self.device)
+        probas = self.get_probas(logits)
+        report = Reporter(truths, predictions, self.run, self.epoch, probas=probas)
         try:
             report.result.loss = self.loss
         except AttributeError:  # if the model was loaded from disk the loss is unknown
@@ -139,9 +161,7 @@ class MLP_model(Model):
         return report
 
     def get_predictions(self):
-        _, truths, predictions = self.evaluate_model(
-            self.model, self.testloader, self.device
-        )
+        _, _, predictions, _ = self.evaluate(self.model, self.testloader, self.device)
         return predictions.numpy()
 
     def get_loader(self, df_x, df_y, shuffle):

nkululeko/models/model_mlp_regression.py

@@ -97,7 +97,9 @@ class MLP_Reg_model(Model):
             self.model, self.testloader, self.device
         )
         result, _, _ = self.evaluate_model(self.model, self.trainloader, self.device)
-        report = Reporter(truths.numpy(), predictions.numpy(), self.run, self.epoch)
+        report = Reporter(
+            truths.numpy(), predictions.numpy(), None, self.run, self.epoch
+        )
         try:
             report.result.loss = self.loss
         except AttributeError:  # if the model was loaded from disk the loss is unknown

nkululeko/plots.py

@@ -48,7 +48,7 @@ class Plots:
             )
             ax.set_ylabel(f"number of speakers")
             ax.set_xlabel("number of samples")
-            self._save_plot(
+            self.save_plot(
                 ax,
                 "Samples per speaker",
                 f"Samples per speaker ({df_speakers.shape[0]})",
@@ -70,9 +70,9 @@ class Plots:
                     rot=0,
                 )
             )
-            ax.set_ylabel(f"number of speakers")
+            ax.set_ylabel("number of speakers")
             ax.set_xlabel("number of samples")
-            self._save_plot(
+            self.save_plot(
                 ax,
                 "Sample value counts",
                 f"Samples per speaker ({df_speakers.shape[0]})",
@@ -96,7 +96,7 @@ class Plots:
             binned_data = self.util.continuous_to_categorical(df[class_label])
             ax = binned_data.value_counts().plot(kind="bar")
             filename_binned = f"{class_label}_discreet"
-            self._save_plot(
+            self.save_plot(
                 ax,
                 "Sample value counts",
                 filename_binned,
@@ -106,7 +106,7 @@ class Plots:
             dist_type = self.util.config_val("EXPL", "dist_type", "hist")
             ax = df[class_label].plot(kind=dist_type)
 
-        self._save_plot(
+        self.save_plot(
             ax,
             "Sample value counts",
             filename,
@@ -131,17 +131,17 @@ class Plots:
                             df, class_label, att1, self.target, type_s
                         )
                     else:
-                        ax, caption = self._plotcatcont(
+                        ax, caption = self.plotcatcont(
                             df, class_label, att1, att1, type_s
                         )
                 else:
                     if self.util.is_categorical(df[att1]):
-                        ax, caption = self._plotcatcont(
+                        ax, caption = self.plotcatcont(
                             df, att1, class_label, att1, type_s
                         )
                     else:
                         ax, caption = self._plot2cont(df, class_label, att1, type_s)
-                self._save_plot(
+                self.save_plot(
                     ax,
                     caption,
                     f"Correlation of {self.target} and {att[0]}",
@@ -171,15 +171,11 @@ class Plots:
                             ax, caption = self._plot2cat(df, att1, att2, att1, type_s)
                         else:
                             # class_label = cat, att1 = cat, att2 = cont
-                            ax, caption = self._plotcatcont(
-                                df, att1, att2, att1, type_s
-                            )
+                            ax, caption = self.plotcatcont(df, att1, att2, att1, type_s)
                     else:
                         if self.util.is_categorical(df[att2]):
                             # class_label = cat, att1 = cont, att2 = cat
-                            ax, caption = self._plotcatcont(
-                                df, att2, att1, att2, type_s
-                            )
+                            ax, caption = self.plotcatcont(df, att2, att1, att2, type_s)
                         else:
                             # class_label = cat, att1 = cont, att2 = cont
                             ax, caption = self._plot2cont_cat(
@@ -205,7 +201,7 @@ class Plots:
                             # class_label = cont, att1 = cont, att2 = cont
                             ax, caption = self._plot2cont(df, att1, att2, type_s)
 
-                self._save_plot(
+                self.save_plot(
                     ax, caption, f"Correlation of {att1} and {att2}", filename, type_s
                 )
 
@@ -215,16 +211,16 @@ class Plots:
                     f" {att} has more than 2 values. Perhaps you forgot to state a list of lists?"
                 )
 
-    def _save_plot(self, ax, caption, header, filename, type_s):
+    def save_plot(self, ax, caption, header, filename, type_s):
         # one up because of the runs
         fig_dir = self.util.get_path("fig_dir") + "../"
-        fig = ax.figure
+        fig_plots = ax.figure
         # avoid warning
         # plt.tight_layout()
         img_path = f"{fig_dir}{filename}_{type_s}.{self.format}"
         plt.savefig(img_path)
-        plt.close(fig)
-        # fig.clear()   # avoid error
+        plt.close(fig_plots)
+        self.util.debug(f"Saved plot to {img_path}")
         glob_conf.report.add_item(
             ReportItem(
                 Header.HEADER_EXPLORE,
@@ -244,35 +240,29 @@ class Plots:
         return att, df
 
     def _plot2cont_cat(self, df, cont1, cont2, cat, ylab):
-        """
-        plot relation of two continuous distributions with one categorical
-        """
+        """Plot relation of two continuous distributions with one categorical."""
         pearson = stats.pearsonr(df[cont1], df[cont2])
         # trunc to three digits
         pearson = int(pearson[0] * 1000) / 1000
         pearson_string = f"PCC: {pearson}"
         ax = sns.lmplot(data=df, x=cont1, y=cont2, hue=cat)
         caption = f"{ylab} {df.shape[0]}. {pearson_string}"
-        ax.fig.suptitle(caption)
+        ax.figure.suptitle(caption)
         return ax, caption
 
     def _plot2cont(self, df, col1, col2, ylab):
-        """
-        plot relation of two continuous distributions
-        """
+        """Plot relation of two continuous distributions."""
         pearson = stats.pearsonr(df[col1], df[col2])
         # trunc to three digits
         pearson = int(pearson[0] * 1000) / 1000
         pearson_string = f"PCC: {pearson}"
         ax = sns.lmplot(data=df, x=col1, y=col2)
         caption = f"{ylab} {df.shape[0]}. {pearson_string}"
-        ax.fig.suptitle(caption)
+        ax.figure.suptitle(caption)
         return ax, caption
 
-    def _plotcatcont(self, df, cat_col, cont_col, xlab, ylab):
-        """
-        plot relation of categorical distribution with continuous
-        """
+    def plotcatcont(self, df, cat_col, cont_col, xlab, ylab):
+        """Plot relation of categorical distribution with continuous."""
         dist_type = self.util.config_val("EXPL", "dist_type", "hist")
         cats, cat_str, es = su.get_effect_size(df, cat_col, cont_col)
         if dist_type == "hist":
@@ -287,13 +277,11 @@ class Plots:
             )
             ax.set(xlabel=f"{cont_col}")
             caption = f"{ylab} {df.shape[0]}. {cat_str} ({cats}):" f" {es}"
-            ax.fig.suptitle(caption)
+            ax.figure.suptitle(caption)
         return ax, caption
 
     def _plot2cat(self, df, col1, col2, xlab, ylab):
-        """
-        plot relation of 2 categorical distributions
-        """
+        """Plot relation of 2 categorical distributions."""
         crosstab = pd.crosstab(index=df[col1], columns=df[col2])
         res_pval = stats.chi2_contingency(crosstab)
         res_pval = int(res_pval[1] * 1000) / 1000
@@ -320,8 +308,8 @@ class Plots:
         max = self.util.to_3_digits(df.duration.max())
         title = f"Duration distr. for {sample_selection} {df.shape[0]}. min={min}, max={max}"
         ax.set_title(title)
-        ax.set_xlabel(f"duration")
-        ax.set_ylabel(f"number of samples")
+        ax.set_xlabel("duration")
+        ax.set_ylabel("number of samples")
         fig = ax.figure
         # plt.tight_layout()
         img_path = f"{fig_dir}{filename}_{sample_selection}.{self.format}"

nkululeko/reporting/reporter.py

@@ -2,16 +2,21 @@ import ast
 import glob
 import json
 import math
+import os
 
 from confidence_intervals import evaluate_with_conf_int
 import matplotlib.pyplot as plt
 import numpy as np
+from scipy.special import softmax
+from scipy.stats import entropy
 from scipy.stats import pearsonr
-from sklearn.metrics import ConfusionMatrixDisplay, roc_curve
+from sklearn.metrics import ConfusionMatrixDisplay
+from sklearn.metrics import auc
 from sklearn.metrics import classification_report
 from sklearn.metrics import confusion_matrix
 from sklearn.metrics import r2_score
-from sklearn.metrics import roc_curve, auc, roc_auc_score
+from sklearn.metrics import roc_auc_score
+from sklearn.metrics import roc_curve
 from torch import is_tensor
 
 from audmetric import accuracy
@@ -21,6 +26,7 @@ from audmetric import mean_squared_error
 from audmetric import unweighted_average_recall
 
 import nkululeko.glob_conf as glob_conf
+from nkululeko.plots import Plots
 from nkululeko.reporting.defines import Header
 from nkululeko.reporting.report_item import ReportItem
 from nkululeko.reporting.result import Result
@@ -46,9 +52,18 @@ class Reporter:
                 self.MEASURE = "CCC"
                 self.result.measure = self.MEASURE
 
-    def __init__(self, truths, preds, run, epoch):
-        """Initialization with ground truth und predictions vector."""
+    def __init__(self, truths, preds, run, epoch, probas=None):
+        """Initialization with ground truth und predictions vector.
+
+        Args:
+            truths (list): the ground truth
+            preds (list): the predictions
+            run (int): number of run
+            epoch (int): number of epoch
+            probas (pd.Dataframe, optional): probabilities per class. Defaults to None.
+        """
         self.util = Util("reporter")
+        self.probas = probas
         self.format = self.util.config_val("PLOT", "format", "png")
         self.truths = np.asarray(truths)
         self.preds = np.asarray(preds)
@@ -108,6 +123,47 @@ class Reporter:
                 self.result.test = test_result
                 self.result.set_upper_lower(upper, lower)
                 # train and loss are being set by the model
+        # print out the class  probilities
+
+    def print_probabilities(self):
+        """Print the probabilities per class to a file in the store."""
+        if (
+            self.util.exp_is_classification()
+            and self.probas is not None
+            and "uncertainty" not in self.probas
+        ):
+            probas = self.probas
+            probas["predicted"] = self.preds
+            probas["truth"] = self.truths
+            # softmax the probabilities or logits
+            uncertainty = probas.apply(softmax, axis=1)
+            try:
+                le = glob_conf.label_encoder
+                mapping = dict(zip(le.classes_, range(len(le.classes_))))
+                mapping_reverse = {value: key for key, value in mapping.items()}
+                probas = probas.rename(columns=mapping_reverse)
+                probas["predicted"] = probas["predicted"].map(mapping_reverse)
+                probas["truth"] = probas["truth"].map(mapping_reverse)
+            except AttributeError as ae:
+                self.util.debug(f"Can't label categories: {ae}")
+            # compute entropy per sample
+            uncertainty = uncertainty.apply(entropy)
+            # scale it to 0-1
+            max_ent = math.log(len(glob_conf.labels))
+            uncertainty = (uncertainty - uncertainty.min()) / (
+                max_ent - uncertainty.min()
+            )
+            probas["uncertainty"] = uncertainty
+            probas["correct"] = probas.predicted == probas.truth
+            sp = os.path.join(self.util.get_path("store"), "pred_df.csv")
+            self.probas = probas
+            probas.to_csv(sp)
+            self.util.debug(f"Saved probabilities to {sp}")
+            plots = Plots()
+            ax, caption = plots.plotcatcont(
+                probas, "correct", "uncertainty", "uncertainty", "correct"
+            )
+            plots.save_plot(ax, caption, "Uncertainty", "uncertainty", "samples")
 
     def set_id(self, run, epoch):
         """Make the report identifiable with run and epoch index."""
@@ -123,6 +179,12 @@ class Reporter:
         self.preds = np.digitize(self.preds, bins) - 1
 
     def plot_confmatrix(self, plot_name, epoch=None):
+        """Plot a confusionmatrix to the store.
+
+        Args:
+            plot_name (str): name for the image file.
+            epoch (int, optional): Number of epoch. Defaults to None.
+        """
         if not self.util.exp_is_classification():
             self.continuous_to_categorical()
         self._plot_confmat(self.truths, self.preds, plot_name, epoch)
@@ -212,10 +274,11 @@ class Reporter:
             )
         img_path = f"{fig_dir}{plot_name}{self.filenameadd}.{self.format}"
         plt.savefig(img_path)
+        self.util.debug(f"Saved confusion plot to {img_path}")
         fig.clear()
         plt.close(fig)
-        plt.savefig(img_path)
-        plt.close(fig)
+        plt.close()
+        plt.clf()
         glob_conf.report.add_item(
             ReportItem(
                 Header.HEADER_RESULTS,

nkululeko/runmanager.py

@@ -11,7 +11,7 @@ from nkululeko.utils.util import Util
 
 
 class Runmanager:
-    """Class to manage the runs of the experiment (e.g. when results differ caused by random initialization)"""
+    """Class to manage the runs of the experiment (e.g. when results differ caused by random initialization)."""
 
     model = None  # The underlying model
     df_train, df_test, feats_train, feats_test = (
@@ -23,15 +23,14 @@ class Runmanager:
     reports = []
 
     def __init__(self, df_train, df_test, feats_train, feats_test):
-        """Constructor setting up the dataframes
+        """Constructor setting up the dataframes.
+
         Args:
             df_train: train dataframe
             df_test: test dataframe
             feats_train: train features
             feats_train: test features
 
-        Returns:
-
         """
         self.df_train, self.df_test, self.feats_train, self.feats_test = (
             df_train,
@@ -46,7 +45,7 @@ class Runmanager:
         # self._select_model(model_type)
 
     def do_runs(self):
-        """Start the runs"""
+        """Start the runs."""
         self.best_results = []  # keep the best result per run
         self.last_epochs = []  # keep the epoch of best result per run
         # for all runs
@@ -105,15 +104,13 @@ class Runmanager:
                 )
                 self.print_model(best_report, plot_name)
             # finally, print out the numbers for this run
-            # self.reports[-1].print_results(
-            #     int(self.util.config_val("EXP", "epochs", 1))
-            # )
             best_report.print_results(best_report.epoch)
+            best_report.print_probabilities()
             self.best_results.append(best_report)
             self.last_epochs.append(last_epoch)
 
     def print_best_result_runs(self):
-        """Print the best result for all runs"""
+        """Print the best result for all runs."""
         best_report = self.get_best_result(self.best_results)
         self.util.debug(
             f"best result all runs with run {best_report.run}             and"
@@ -177,7 +174,7 @@ class Runmanager:
         return self.load_model(best_report)
 
     def get_best_result(self, reports):
-        best_r = Reporter([], [], 0, 0)
+        best_r = Reporter([], [], None, 0, 0)
         if self.util.high_is_good():
             best_r = self.search_best_result(reports, "ascending")
         else:
@@ -185,7 +182,7 @@ class Runmanager:
         return best_r
 
     def search_best_result(self, reports, order):
-        best_r = Reporter([], [], 0, 0)
+        best_r = Reporter([], [], None, 0, 0)
         if order == "ascending":
             best_result = 0
             for r in reports:

nkululeko/test_predictor.py

@@ -6,13 +6,12 @@
 
 import ast
 
-import numpy as np
 import pandas as pd
 from sklearn.preprocessing import LabelEncoder
 
+import nkululeko.glob_conf as glob_conf
 from nkululeko.data.dataset import Dataset
 from nkululeko.feature_extractor import FeatureExtractor
-import nkululeko.glob_conf as glob_conf
 from nkululeko.scaler import Scaler
 from nkululeko.utils.util import Util
 
@@ -42,7 +41,6 @@ class TestPredictor:
             scale = self.util.config_val("FEATS", "scale", False)
             labelenc = LabelEncoder()
             data_df[self.target] = labelenc.fit_transform(data_df[self.target])
-            #            data_df[self.target] = self.label_encoder.fit_transform(data_df[self.target])
             if scale:
                 self.scaler = Scaler(data_df, None, feats_df, None, scale)
                 feats_df, _ = self.scaler.scale()
@@ -56,18 +54,13 @@ class TestPredictor:
         else:
             test_dbs = ast.literal_eval(glob_conf.config["DATA"]["tests"])
             test_dbs_string = "_".join(test_dbs)
-            predictions = self.model.get_predictions()
+            predictions, _ = self.model.get_predictions()
             report = self.model.predict()
             result = report.result.get_result()
             report.set_filename_add(f"test-{test_dbs_string}")
             self.util.print_best_results([report])
             report.plot_confmatrix(self.util.get_plot_name(), 0)
             report.print_results(0)
-            # print(predictions)
-            # df = pd.DataFrame(index=self.orig_df.index)
-            # df["speaker"] = self.orig_df["speaker"]
-            # df["gender"] = self.orig_df["gender"]
-            # df[self.target] = self.orig_df[self.target]
             df = self.orig_df.copy()
             df["predictions"] = self.label_encoder.inverse_transform(predictions)
             target = self.util.config_val("DATA", "target", "emotion")

nkululeko/utils/stats.py

@@ -70,12 +70,16 @@ def get_effect_size(df, target, variable):
         cats[c] = df[df[target] == c][variable].values
     combos = all_combinations(categories)
     results = {}
-    for combo in combos:
-        one = combo[0]
-        other = combo[1]
-        results[f"{one}-{other}"] = cohen_d(cats[one], cats[other])
-    max_cat = max(results, key=results.get)
-    cat_s = cohens_D_to_string(float(results[max_cat]))
+    if len(categories) == 1:
+        cat_s = cohens_D_to_string(0)
+        return categories[0], cat_s, 0
+    else:
+        for combo in combos:
+            one = combo[0]
+            other = combo[1]
+            results[f"{one}-{other}"] = cohen_d(cats[one], cats[other])
+        max_cat = max(results, key=results.get)
+        cat_s = cohens_D_to_string(float(results[max_cat]))
     return max_cat, cat_s, results[max_cat]
 
 
@@ -92,7 +96,7 @@ def cohens_D_to_string(val):
 
 
 def normalize(values):
-    """Do a z-transformation of a distribution. 
+    """Do a z-transformation of a distribution.
 
     So that mean = 0 and variance = 1
     """

nkululeko/utils/util.py

@@ -37,8 +37,7 @@ class Util:
                 import nkululeko.glob_conf as glob_conf
 
                 self.config = glob_conf.config
-                self.got_data_roots = self.config_val(
-                    "DATA", "root_folders", False)
+                self.got_data_roots = self.config_val("DATA", "root_folders", False)
                 if self.got_data_roots:
                     # if there is a global data rootfolder file, read from
                     # there
@@ -108,19 +107,17 @@ class Util:
             if self.got_data_roots:
                 try:
                     if len(key) > 0:
-                        return self.data_roots["DATA"][dataset +
-                                                       "." + key].strip("'\"")
+                        return self.data_roots["DATA"][dataset + "." + key].strip("'\"")
                     else:
                         return self.data_roots["DATA"][dataset].strip("'\"")
                 except KeyError:
                     if default not in self.stopvals:
                         self.debug(
-                            f"value for {key} not found, using default:"
-                            f" {default}")
+                            f"value for {key} not found, using default:" f" {default}"
+                        )
                     return default
             if default not in self.stopvals:
-                self.debug(
-                    f"value for {key} not found, using default: {default}")
+                self.debug(f"value for {key} not found, using default: {default}")
             return default
 
     def set_config(self, config):
@@ -131,6 +128,10 @@ class Util:
         store = self.get_path("store")
         return f"{store}/{self.get_exp_name()}.pkl"
 
+    def get_pred_name(self):
+        store = self.get_path("store")
+        return f"{store}/pred_df.csv"
+
     def is_categorical(self, pd_series):
         """Check if a dataframe column is categorical"""
         return pd_series.dtype.name == "object" or isinstance(
@@ -163,10 +164,8 @@ class Util:
         if len(df) == 0:
             return df
         if not isinstance(df.index, pd.MultiIndex):
-            self.debug(
-                "converting to segmented index, this might take a while...")
-            df.index = audformat.utils.to_segmented_index(
-                df.index, allow_nat=False)
+            self.debug("converting to segmented index, this might take a while...")
+            df.index = audformat.utils.to_segmented_index(df.index, allow_nat=False)
         return df
 
     def _get_value_descript(self, section, name):
@@ -209,7 +208,11 @@ class Util:
         mt = f'{self.config["MODEL"]["type"]}'
         # ft = "_".join(ast.literal_eval(self.config["FEATS"]["type"]))
         ft_value = self.config["FEATS"]["type"]
-        if isinstance(ft_value, str) and ft_value.startswith("[") and ft_value.endswith("]"):
+        if (
+            isinstance(ft_value, str)
+            and ft_value.startswith("[")
+            and ft_value.endswith("]")
+        ):
             ft = "_".join(ast.literal_eval(ft_value))
         else:
             ft = ft_value
@@ -237,8 +240,9 @@ class Util:
             ["FEATS", "wav2vec2.layer"],
         ]
         for option in options:
-            return_string += self._get_value_descript(
-                option[0], option[1]).replace(".", "-")
+            return_string += self._get_value_descript(option[0], option[1]).replace(
+                ".", "-"
+            )
         return return_string
 
     def get_plot_name(self):
@@ -284,8 +288,7 @@ class Util:
             return self.config[section][key]
         except KeyError:
             if default not in self.stopvals:
-                self.debug(
-                    f"value for {key} not found, using default: {default}")
+                self.debug(f"value for {key} not found, using default: {default}")
             return default
 
     def config_val_list(self, section, key, default):
@@ -293,10 +296,12 @@ class Util:
             return ast.literal_eval(self.config[section][key])
         except KeyError:
             if default not in self.stopvals:
-                self.debug(
-                    f"value for {key} not found, using default: {default}")
+                self.debug(f"value for {key} not found, using default: {default}")
             return default
 
+    def get_labels(self):
+        return ast.literal_eval(self.config["DATA"]["labels"])
+
     def continuous_to_categorical(self, series):
         """
         discretize a categorical variable.