Added logging print levels and verbosity

executables/bestest_hydronic_heat_pump/P_hp.py

@@ -12,7 +12,7 @@
 """
 
 # Setup up logger for saving
-Logger.setup_logger(folder_name='P_hp', override=True)
+Logger.setup_logger(folder_name='P_hp', override=True, print_level='warning')
 
 # File path to data
 file_path = r"data/bestest_hydronic_heat_pump/pid_data.csv"

physXAI/evaluation/metrics.py

@@ -2,6 +2,7 @@
 import numpy as np
 from sklearn.metrics import mean_squared_error, r2_score
 from physXAI.preprocessing.training_data import TrainingData, TrainingDataMultiStep, TrainingDataGeneric
+from physXAI.utils.logging import Logger
 
 
 class Metrics:
@@ -10,8 +11,6 @@ class Metrics:
     for training, validation, and test datasets.
     """
 
-    print_evaluate = True
-
     def __init__(self, td: TrainingDataGeneric):
         """
                 Initializes the Metrics object by calculating metrics for train, validation (if available),
@@ -52,10 +51,9 @@ def evaluate(y_true: np.ndarray, y_pred: np.ndarray, label: str = '') -> dict[st
         kpis['RMSE' + ' ' + label] = rmse
         kpis['R2' + ' ' + label] = r2
 
-        if Metrics.print_evaluate:
-            # print(f"{label} MSE: {mse:.2f}")
-            print(f"{label} RMSE: {rmse:.2f}")
-            print(f"{label} R2: {r2:.2f}")
+        Logger.print(f"{label} MSE: {mse:.2f}", 'debug')
+        Logger.print(f"{label} RMSE: {rmse:.2f}", 'info')
+        Logger.print(f"{label} R2: {r2:.2f}", 'info')
 
         return kpis
 
@@ -130,7 +128,7 @@ def evaluate(y_true: np.ndarray, y_pred: np.ndarray, label: str = '', **kwargs)
         for loss in kwargs['pinn_losses']:
             val = float(loss(y_true, y_pred))
             kpis[loss.__name__ + ' ' + label] = val
-            print(f"{loss.__name__ + ' ' + label}: {val:.2f}")
+            Logger.print(f"{loss.__name__ + ' ' + label}: {val:.2f}", 'info')
 
         return kpis
 

physXAI/feature_selection/recursive_feature_elimination.py

@@ -35,12 +35,12 @@ def search_best_features(runs: dict, multi_step: bool, use_multi_step_error: boo
     except ValueError:
         max_features = np.inf
 
-    print('Selected features:')
+    Logger.print('Selected features:', 'info')
     if max_features == np.inf:
         inputs = sorted_kpis[min_index]['inputs']
     else:
         inputs = sorted_kpis[max_features]['inputs']
-    print(inputs)
+    Logger.print(inputs, 'info')
     return inputs
 
 
@@ -53,12 +53,11 @@ def recursive_feature_elimination(file_path: str, preprocessing: PreprocessingDa
     if fixed_inputs is None:
         fixed_inputs = list()
 
-    print('Feature Selection')
-    Metrics.print_evaluate = False
-
     if Logger._logger is None:
         Logger.setup_logger()
 
+    Logger.print('Feature Selection', 'info')
+
     org_inputs = preprocessing.inputs
     inputs = preprocessing.inputs
     input_length = len(inputs)
@@ -83,8 +82,8 @@ def recursive_feature_elimination(file_path: str, preprocessing: PreprocessingDa
 
     # Recursive feature elimination
     for j in range(input_length - 1, 0, -1):
-        print(f'Features {j + 1}')
-        print(inputs)
+        Logger.print(f'Features {j + 1}', 'info')
+        Logger.print(inputs, 'info')
 
         # Reduced input features
         new_inputs = list()
@@ -137,8 +136,8 @@ def recursive_feature_elimination(file_path: str, preprocessing: PreprocessingDa
             key_filter = int(min(kpis, key=kpis.get))
         inputs = new_inputs[key_filter]
         runs[j] = run
-    print(f'Features {1}')
-    print(inputs)
+    Logger.print(f'Features {1}', 'info')
+    Logger.print(inputs, 'info')
 
     preprocessing.inputs = org_inputs
 

physXAI/models/ann/ann_design.py

@@ -80,7 +80,7 @@ def fit_model(self, model, td: TrainingDataGeneric):
         callbacks = list()
         if self.early_stopping_epochs is not None:
             es = keras.callbacks.EarlyStopping(monitor='val_loss', mode='min', patience=self.early_stopping_epochs,
-                                               restore_best_weights=True, verbose=1)
+                                               restore_best_weights=True, verbose=Logger.verbosity_int())
             callbacks.append(es)
 
         # Fit model, track training time
@@ -111,14 +111,16 @@ def fit_model(self, model, td: TrainingDataGeneric):
         training_history = model.fit(train_ds,
                                      validation_data=val_ds,
                                      epochs=self.epochs,
-                                     callbacks=callbacks)
+                                     callbacks=callbacks,
+                                     verbose=Logger.verbosity())
         stop_time = time.perf_counter()
 
         # Add metrics to training data
         td.add_training_time(stop_time - start_time)
         td.add_training_record(training_history)
 
-        model.summary()
+        if Logger.check_print_level('info'):
+            model.summary()
 
     def plot(self, td: TrainingDataGeneric):
         """
@@ -587,10 +589,10 @@ def evaluate(self, model, td: TrainingDataGeneric):
             td (TrainingData): The training data
         """
 
-        y_pred_train = model.predict(td.X_train_single)
-        y_pred_test = model.predict(td.X_test_single)
+        y_pred_train = model.predict(td.X_train_single, verbose=Logger.verbosity())
+        y_pred_test = model.predict(td.X_test_single, verbose=Logger.verbosity())
         if td.X_val is not None:
-            y_pred_val = model.predict(td.X_val_single)
+            y_pred_val = model.predict(td.X_val_single, verbose=Logger.verbosity())
         else:
             y_pred_val = None
         td.add_predictions(y_pred_train, y_pred_val, y_pred_test)
@@ -703,7 +705,7 @@ def fit_model(self, model, td: TrainingDataMultiStep):
         callbacks = list()
         if self.early_stopping_epochs is not None:
             es = keras.callbacks.EarlyStopping(monitor='val_loss', mode='min', patience=self.early_stopping_epochs,
-                                               restore_best_weights=True, verbose=1)
+                                               restore_best_weights=True, verbose=Logger.verbosity_int())
             callbacks.append(es)
 
         # Fit model, track training time

physXAI/models/ann/model_construction/ann_models.py

@@ -67,8 +67,6 @@ def ClassicalANNConstruction(config: dict, td: TrainingDataGeneric):
     if config['rescale_output']:
         model.add(keras.layers.Rescaling(scale=rescale_sigma, offset=rescale_mean))
 
-    model.summary()
-
     return model
 
 
@@ -182,6 +180,4 @@ def CMNNModelConstruction(config: dict, td: TrainingDataGeneric):
 
     model = keras.models.Model(inputs=input_layer, outputs=x)
 
-    model.summary()
-
     return model

physXAI/models/ann/model_construction/rbf_models.py

@@ -5,6 +5,7 @@
 from physXAI.preprocessing.training_data import TrainingDataGeneric
 from physXAI.models.ann.configs.ann_model_configs import RBFConstruction_config
 from physXAI.models.ann.keras_models.keras_models import RBFLayer
+from physXAI.utils.logging import Logger
 
 
 def gamma_init(centers, overlap=0.5) -> float:
@@ -26,7 +27,7 @@ def gamma_init(centers, overlap=0.5) -> float:
         return 1.0  # Fallback
 
     gamma = -np.log(overlap) / avg_dist_sq
-    # print(f"Calculated Gamma: {gamma}")
+    Logger.print(f"Calculated Gamma: {gamma}", 'info')
     return gamma
 
 
@@ -111,6 +112,4 @@ def RBFModelConstruction(config: dict, td: TrainingDataGeneric):
 
     model = keras.Model(inputs=input_layer, outputs=x)
 
-    model.summary()
-
     return model

physXAI/models/ann/model_construction/residual_models.py

@@ -57,6 +57,4 @@ def LinResidualANNConstruction(config: dict, td: TrainingDataGeneric, lin_model:
     lin.set_weights([lin_model.coef_.reshape(-1, 1), np.array(lin_model.intercept_)])
     lin.trainable = False
 
-    model.summary()
-
     return model

physXAI/models/models.py

@@ -1,3 +1,4 @@
+import os
 import time
 from abc import ABC, abstractmethod
 from typing import Type
@@ -10,6 +11,9 @@
 from physXAI.evaluation.metrics import Metrics, MetricsMultiStep
 from physXAI.plotting.plotting import (plot_prediction_correlation, plot_metrics_table, subplots,
                                               plot_predictions, plot_multi_rmse)
+os.environ['TF_CPP_MIN_LOG_LEVEL'] = '1'
+import keras
+os.environ['TF_CPP_MIN_LOG_LEVEL'] = '0'
 
 
 MODEL_CLASS_REGISTRY: dict[str, Type['AbstractModel']] = dict()
@@ -199,7 +203,7 @@ def register_model(cls):  # pragma: no cover
     The class is registered using its __name__.
     """
     if cls.__name__ in MODEL_CLASS_REGISTRY:
-        print(f"Warning: Class '{cls.__name__}' is already registered. Overwriting.")
+        Logger.print(f"Warning: Class '{cls.__name__}' is already registered. Overwriting.", 'warning')
     MODEL_CLASS_REGISTRY[cls.__name__] = cls
     return cls  # Decorators must return the class (or a replacement)
 
@@ -224,11 +228,14 @@ def evaluate(model, td: TrainingDataGeneric):
             model: The trained model instance.
             td (TrainingDataGeneric): The TrainingData object containing datasets and for storing results.
         """
+        kwargs = {}
+        if isinstance(model, keras.Model):
+            kwargs['verbose'] = Logger.verbosity()
 
-        y_pred_train = model.predict(td.X_train_single)
-        y_pred_test = model.predict(td.X_test_single)
+        y_pred_train = model.predict(td.X_train_single, **kwargs)
+        y_pred_test = model.predict(td.X_test_single, **kwargs)
         if td.X_val_single is not None:
-            y_pred_val = model.predict(td.X_val_single)
+            y_pred_val = model.predict(td.X_val_single, **kwargs)
         else:
             y_pred_val = None
         td.add_predictions(y_pred_train, y_pred_val, y_pred_test)
@@ -307,7 +314,7 @@ def _evaluate_multi_inner_loop(model, X: np.ndarray, y: np.ndarray, X_columns: l
             current_val = X[:, 0, index].reshape(-1, 1)
             current_true_val = current_val.copy()
             for t in range(X.shape[1]):
-                pred = model.predict(X[:, t, :], verbose=0)
+                pred = model.predict(X[:, t, :], verbose=Logger.verbosity())
                 if delta_prediction:
                     current_val += pred
                     current_true_val += y[:, t, 0].reshape(-1, 1)
@@ -463,12 +470,12 @@ def evaluate(model, td: TrainingDataMultiStep):
             td (TrainingDataMultistep): The TrainingDataMultiStep object containing datasets and for storing results.
         """
 
-        y_pred_train = model.predict(td.X_train)
+        y_pred_train = model.predict(td.X_train, verbose=Logger.verbosity())
         if td.X_val is not None:
-            y_pred_val = model.predict(td.X_val)
+            y_pred_val = model.predict(td.X_val, verbose=Logger.verbosity())
         else:
             y_pred_val = None
-        y_pred_test = model.predict(td.X_test)
+        y_pred_test = model.predict(td.X_test, verbose=Logger.verbosity())
         td.add_predictions(y_pred_train, y_pred_val, y_pred_test)
 
         metrics = MetricsMultiStep(td)

physXAI/preprocessing/constructed.py

@@ -1,4 +1,5 @@
 from abc import ABC, abstractmethod
+from physXAI.utils.logging import Logger
 from typing import Type, Union
 import numpy as np
 from pandas import DataFrame, Series
@@ -135,7 +136,7 @@ def register_feature(cls):
     The class is registered using its __name__.
     """
     if cls.__name__ in CONSTRUCTED_CLASS_REGISTRY:  # pragma: no cover
-        print(f"Warning: Class '{cls.__name__}' is already registered. Overwriting.")  # pragma: no cover
+        Logger.print(f"Warning: Class '{cls.__name__}' is already registered. Overwriting.", 'warning')  # pragma: no cover
     CONSTRUCTED_CLASS_REGISTRY[cls.__name__] = cls
     return cls  # Decorators must return the class (or a replacement)
 

physXAI/utils/logging.py

@@ -3,12 +3,9 @@
 import os
 import shutil
 from datetime import datetime
+from typing import Union
 import git
-from physXAI.preprocessing.constructed import FeatureConstruction
 import pickle
-from pathlib import Path
-
-from physXAI.preprocessing.training_data import TrainingDataMultiStep
 
 
 def get_parent_working_directory() -> str:
@@ -28,10 +25,10 @@ def get_parent_working_directory() -> str:
         git_root = repo.working_tree_dir
         return git_root
     except git.InvalidGitRepositoryError:  # pragma: no cover
-        print(f"Error: Cannot find git root directory.")  # pragma: no cover
+        Logger.print(f"Error: Cannot find git root directory.", 'error')  # pragma: no cover
         return ''  # pragma: no cover
     except Exception as e:  # pragma: no cover
-        print(f"Error: An unexpected error occurred when searching for parent directory: {e}")  # pragma: no cover
+        Logger.print(f"Error: An unexpected error occurred when searching for parent directory: {e}", 'error')  # pragma: no cover
         return ''  # pragma: no cover
 
 
@@ -110,9 +107,40 @@ class Logger:
     save_name_model: str = 'model'
     save_name_model_online_learning: str = 'model_ol'
 
+    print_level: str = 'info'  # options: 'debug', 'info', 'warning', 'error'
+    _print_levels = ['debug', 'info', 'warning', 'error']
+
     _logger = None
     _override = False
 
+    @staticmethod
+    def print(message: str, print_level: str = 'info'):
+        if Logger.check_print_level(print_level):
+            print(message)
+
+    @staticmethod
+    def check_print_level(print_level: str) -> bool:
+        if str(print_level).lower() not in Logger._print_levels:
+            raise ValueError(f"Invalid print level: {str(print_level).lower()}. Valid options are: {Logger._print_levels}")
+        if Logger._print_levels.index(str(print_level).lower()) >= Logger._print_levels.index(Logger.print_level):
+            return True
+        else:
+            return False
+
+    @staticmethod
+    def verbosity() -> Union[int, str]:
+        if Logger._print_levels.index(Logger.print_level) >= Logger._print_levels.index('warning'):
+            return 0
+        else:
+            return "auto"
+
+    @staticmethod
+    def verbosity_int() -> int:
+        if Logger._print_levels.index(Logger.print_level) >= Logger._print_levels.index('warning'):
+            return 0
+        else:
+            return 1
+
     @staticmethod
     def override_question(path: str):  # pragma: no cover
         if os.path.exists(path) and not Logger._override:
@@ -138,7 +166,7 @@ def already_exists_question(path: str):  # pragma: no cover
                 raise e
 
     @staticmethod
-    def setup_logger(folder_name: str = None, override: bool = False, base_path: str = None):
+    def setup_logger(folder_name: str = None, override: bool = False, base_path: str = None, print_level: str = None):
         if base_path is None:
             base_path = Logger.base_path
         if folder_name is None:
@@ -153,6 +181,11 @@ def setup_logger(folder_name: str = None, override: bool = False, base_path: str
 
         Logger._logger = path
         Logger._override = override
+
+        if print_level is not None:
+            if str(print_level).lower() not in Logger._print_levels:
+                raise ValueError(f"Invalid print level: {str(print_level).lower()}. Valid options are: {Logger._print_levels}")
+            Logger.print_level = str(print_level).lower()
 
     @staticmethod
     def log_setup(preprocessing=None, model=None, save_name_preprocessing=None, save_name_model=None,
@@ -173,6 +206,7 @@ def log_setup(preprocessing=None, model=None, save_name_preprocessing=None, save
             with open(path, "w") as f:
                 json.dump(preprocessing_dict, f, indent=4)
 
+            from physXAI.preprocessing.constructed import FeatureConstruction
             constructed_config = FeatureConstruction.get_config()
             if len(constructed_config) > 0:
                 if save_name_constructed is None:
@@ -221,6 +255,7 @@ def save_training_data(training_data, path: str = None):
         with open(p, "w") as f:
             json.dump(td_dict, f, indent=4)
 
+        from physXAI.preprocessing.training_data import TrainingDataMultiStep
         if isinstance(training_data, TrainingDataMultiStep):
             training_data = copy.copy(training_data)
             training_data.train_ds = None