Merge pull request #248 from basf/loss_func_fix

fix num_classes argument for binary classification

Author: AnFreTh

mambular/__version__.py

@@ -17,5 +17,4 @@
 
 # The following line *must* be the last in the module, exactly as formatted:
 
-__version__ = "1.3.1"
-
+__version__ = "1.3.2"

mambular/models/utils/sklearn_base_classifier.py

@@ -4,6 +4,7 @@
 import torch
 from sklearn.metrics import accuracy_score, log_loss
 from .sklearn_parent import SklearnBase
+import numpy as np
 
 
 class SklearnBaseClassifier(SklearnBase):
@@ -85,6 +86,8 @@ def build_model(
             The built classifier.
         """
 
+        num_classes = len(np.unique(y))
+
         return super()._build_model(
             X,
             y,
@@ -94,6 +97,7 @@ def build_model(
             y_val=y_val,
             embeddings=embeddings,
             embeddings_val=embeddings_val,
+            num_classes=num_classes,
             random_state=random_state,
             batch_size=batch_size,
             shuffle=shuffle,
@@ -190,6 +194,7 @@ def fit(
             The fitted classifier.
         """
 
+        num_classes = len(np.unique(y))
         return super().fit(
             X=X,
             y=y,
@@ -215,6 +220,7 @@ def fit(
             train_metrics=train_metrics,
             val_metrics=val_metrics,
             rebuild=rebuild,
+            num_classes=num_classes,
             **trainer_kwargs,
         )
 

mambular/models/utils/sklearn_base_regressor.py

@@ -93,6 +93,7 @@ def build_model(
             y_val=y_val,
             embeddings=embeddings,
             embeddings_val=embeddings_val,
+            num_classes=1,
             random_state=random_state,
             batch_size=batch_size,
             shuffle=shuffle,
@@ -198,6 +199,7 @@ def fit(
             y_val=y_val,
             embeddings=embeddings,
             embeddings_val=embeddings_val,
+            num_classes=1,
             max_epochs=max_epochs,
             random_state=random_state,
             batch_size=batch_size,

mambular/models/utils/sklearn_parent.py

@@ -120,6 +120,7 @@ def _build_model(
         y_val=None,
         embeddings=None,
         embeddings_val=None,
+        num_classes: int = None,
         random_state: int = 101,
         batch_size: int = 128,
         shuffle: bool = True,
@@ -223,6 +224,7 @@ def _build_model(
             weight_decay=(
                 weight_decay if weight_decay is not None else self.config.weight_decay
             ),
+            num_classes=num_classes,
             train_metrics=train_metrics,
             val_metrics=val_metrics,
             optimizer_type=self.optimizer_type,
@@ -273,6 +275,7 @@ def fit(
         y_val=None,
         embeddings=None,
         embeddings_val=None,
+        num_classes: int = None,
         max_epochs: int = 100,
         random_state: int = 101,
         batch_size: int = 128,
@@ -357,6 +360,7 @@ def fit(
                 y_val=y_val,
                 embeddings=embeddings,
                 embeddings_val=embeddings_val,
+                num_classes=num_classes,
                 random_state=random_state,
                 batch_size=batch_size,
                 shuffle=shuffle,

pyproject.toml

@@ -1,7 +1,7 @@
 [tool.poetry]
 name = "mambular"
 
-version = "1.3.1"
+version = "1.3.2"
 
 description = "A python package for tabular deep learning with mamba blocks."
 authors = ["Anton Thielmann", "Manish Kumar", "Christoph Weisser"]