ddi-fw 0.0.179__tar.gz → 0.0.181__tar.gz

{ddi_fw-0.0.179 → ddi_fw-0.0.181}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ddi_fw
-Version: 0.0.179
+Version: 0.0.181
 Summary: Do not use :)
 Author-email: Kıvanç Bayraktar <bayraktarkivanc@gmail.com>
 Maintainer-email: Kıvanç Bayraktar <bayraktarkivanc@gmail.com>

{ddi_fw-0.0.179 → ddi_fw-0.0.181}/pyproject.toml

@@ -6,7 +6,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "ddi_fw"
-version = "0.0.179"
+version = "0.0.181"
 description = "Do not use :)"
 readme = "README.md"
 authors = [

{ddi_fw-0.0.179 → ddi_fw-0.0.181}/src/ddi_fw/ml/tensorflow_wrapper.py

@@ -28,9 +28,10 @@ class TFModelWrapper(ModelWrapper):
         self.epochs = kwargs.get('epochs', 100)
         self.use_mlflow = use_mlflow
 
+    # TODO think different settings for num_classes 
     def fit_model(self, X_train, y_train, X_valid, y_valid):
         self.kwargs['input_shape'] = self.train_data.shape
-        self.num_classes = len(np.unique(y_train))
+        self.num_classes = len(np.unique(y_train, axis=0))
         model = self.model_func(**self.kwargs)
         checkpoint = ModelCheckpoint(
             filepath=f'{self.descriptor}_validation.weights.h5',
@@ -136,8 +137,11 @@ class TFModelWrapper(ModelWrapper):
                 print(best_model_key)
                 self.best_model: Model = best_model
                 pred = self.predict()
-                pred = tf.keras.utils.to_categorical(np.argmax(pred,axis=1), num_classes=self.num_classes)
-                actual = tf.keras.utils.to_categorical(self.test_label, num_classes=self.num_classes)
+                actual = self.test_label
+                if not utils.is_binary_encoded(pred):
+                    pred = tf.keras.utils.to_categorical(np.argmax(pred,axis=1), num_classes=self.num_classes)
+                if not utils.is_binary_encoded(actual):
+                    actual = tf.keras.utils.to_categorical(actual, num_classes=self.num_classes)
 
                 logs, metrics = evaluate(
                     actual=actual, pred=pred, info=self.descriptor, print_detail=print_detail)
@@ -155,8 +159,11 @@ class TFModelWrapper(ModelWrapper):
             print(best_model_key)
             self.best_model = best_model
             pred = self.predict()
-            pred = tf.keras.utils.to_categorical(np.argmax(pred,axis=1), num_classes=self.num_classes)
-            actual = tf.keras.utils.to_categorical(self.test_label, num_classes=self.num_classes)
+            actual = self.test_label
+            if not utils.is_binary_encoded(pred):
+                pred = tf.keras.utils.to_categorical(np.argmax(pred,axis=1), num_classes=self.num_classes)
+            if not utils.is_binary_encoded(actual):
+                actual = tf.keras.utils.to_categorical(actual, num_classes=self.num_classes)
 
             logs, metrics = evaluate(
                 actual=actual, pred=pred, info=self.descriptor)

{ddi_fw-0.0.179 → ddi_fw-0.0.181}/src/ddi_fw/utils/__init__.py

@@ -3,4 +3,5 @@ from .zip_helper import ZipHelper
 from .py7zr_helper import Py7ZipHelper
 from .enums import UMLSCodeTypes, DrugBankTextDataTypes
 from .package_helper import get_import
-from .kaggle import create_kaggle_dataset
+from .kaggle import create_kaggle_dataset
+from .categorical_data_encoding_checker import is_one_hot_encoded, is_binary_encoded, is_binary_vector

ddi_fw-0.0.181/src/ddi_fw/utils/categorical_data_encoding_checker.py

@@ -0,0 +1,32 @@
+import numpy as np
+
+
+def is_one_hot_encoded(arr):
+    # Check if the array is one-hot encoded
+    # Ensure the input is a numpy ndarray and is 2D
+    if not isinstance(arr, np.ndarray):
+        raise ValueError("Input must be a NumPy ndarray.")
+    if not np.all(np.isin(arr, [0, 1])):
+        return False
+    # Check if each row (or column) has exactly one "1"
+    return np.all(np.sum(arr, axis=1) == 1)  # For row-wise checking
+
+
+def is_binary_encoded(arr):
+    # Ensure the input is a numpy ndarray and is 2D
+    if not isinstance(arr, np.ndarray):
+        raise ValueError("Input must be a NumPy ndarray.")
+    if arr.ndim != 2:
+        raise ValueError("Input must be a 2D array.")
+
+    # Check if all elements are either 0 or 1
+    return np.all(np.isin(arr, [0, 1]))
+
+
+def is_binary_vector(arr):
+    # Ensure the input is a numpy ndarray and is 1D
+    if not isinstance(arr, np.ndarray):
+        raise ValueError("Input must be a NumPy ndarray.")
+    if arr.ndim != 1:
+        raise ValueError("Input must be a 1D array.")
+    return arr.ndim == 1 and np.all(np.isin(arr, [0, 1]))

{ddi_fw-0.0.179 → ddi_fw-0.0.181}/src/ddi_fw.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ddi_fw
-Version: 0.0.179
+Version: 0.0.181
 Summary: Do not use :)
 Author-email: Kıvanç Bayraktar <bayraktarkivanc@gmail.com>
 Maintainer-email: Kıvanç Bayraktar <bayraktarkivanc@gmail.com>

{ddi_fw-0.0.179 → ddi_fw-0.0.181}/src/ddi_fw.egg-info/SOURCES.txt

@@ -94,6 +94,7 @@ src/ddi_fw/pipeline/multi_pipeline.py
 src/ddi_fw/pipeline/ner_pipeline.py
 src/ddi_fw/pipeline/pipeline.py
 src/ddi_fw/utils/__init__.py
+src/ddi_fw/utils/categorical_data_encoding_checker.py
 src/ddi_fw/utils/enums.py
 src/ddi_fw/utils/json_helper.py
 src/ddi_fw/utils/kaggle.py