Merge pull request #31 from marcpinet/refactor-better-compat-handling

Refactor better compat handling

neuralnetlib/layers.py

@@ -963,3 +963,23 @@ def _pool_backward(output_error: np.ndarray, input_data: np.ndarray, pool_size:
             d_input = d_input[:, pad_steps:-pad_steps, :]
 
         return d_input
+
+
+# --------------------------------------------------------------------------------------------------------------
+
+
+compatibility_dict = {
+    Input: [Dense, Conv2D, Conv1D, Embedding],
+    Dense: [Dense, Activation, Dropout, BatchNormalization],
+    Activation: [Dense, Conv2D, Conv1D, MaxPooling2D, AveragePooling2D, MaxPooling1D, AveragePooling1D, Flatten, Dropout],
+    Conv2D: [Conv2D, MaxPooling2D, AveragePooling2D, Activation, Dropout, Flatten, BatchNormalization],
+    MaxPooling2D: [Conv2D, MaxPooling2D, AveragePooling2D, Flatten],
+    AveragePooling2D: [Conv2D, MaxPooling2D, AveragePooling2D, Flatten],
+    Conv1D: [Conv1D, MaxPooling1D, AveragePooling1D, Activation, Dropout, Flatten, BatchNormalization],
+    MaxPooling1D: [Conv1D, MaxPooling1D, AveragePooling1D, Flatten],
+    AveragePooling1D: [Conv1D, MaxPooling1D, AveragePooling1D, Flatten],
+    Flatten: [Dense, Dropout],
+    Dropout: [Dense, Conv2D, Conv1D, Activation],
+    Embedding: [Conv1D, Flatten, Dense],
+    BatchNormalization: [Dense, Conv2D, Conv1D, Activation]
+}

neuralnetlib/model.py

@@ -3,7 +3,7 @@
 
 import numpy as np
 
-from neuralnetlib.layers import Layer, Input, Dense, Activation, Conv2D, MaxPooling2D, Conv1D, MaxPooling1D, AveragePooling2D, AveragePooling1D, Flatten, Dropout, Embedding, BatchNormalization
+from neuralnetlib.layers import Layer, Input, Activation, Dropout, compatibility_dict
 from neuralnetlib.losses import LossFunction, CategoricalCrossentropy
 from neuralnetlib.optimizers import Optimizer
 from neuralnetlib.utils import shuffle, progress_bar
@@ -38,39 +38,8 @@ def add(self, layer: Layer):
                 raise ValueError("The first layer must be an Input layer.")
         else:
             previous_layer = self.layers[-1]
-            if isinstance(previous_layer, Input):
-                if not isinstance(layer, (Dense, Conv2D, Conv1D, Embedding)):
-                    raise ValueError("Input layer can only be followed by Dense, Conv2D, Conv1D, or Embedding.")
-            elif isinstance(previous_layer, Dense):
-                if not isinstance(layer, (Dense, Activation, Dropout, BatchNormalization)):
-                    raise ValueError("Dense layer can only be followed by Dense, Activation, BatchNormalization, or Dropout.")
-            elif isinstance(previous_layer, Activation):
-                if not isinstance(layer, (Dense, Conv2D, Conv1D, MaxPooling2D, AveragePooling2D, MaxPooling1D, AveragePooling1D, Flatten, Dropout)):
-                    raise ValueError("Activation layer can only be followed by Dense, Conv2D, Conv1D, MaxPooling2D, AveragePooling2D, MaxPooling1D, AveragePooling1D, Flatten, or Dropout.")
-            elif isinstance(previous_layer, Conv2D):
-                if not isinstance(layer, (Conv2D, MaxPooling2D, AveragePooling2D, Activation, Dropout, Flatten, BatchNormalization)):
-                    raise ValueError("Conv2D layer can only be followed by Conv2D, MaxPooling2D, AveragePooling2D, Activation, Dropout, BatchNormalization, or Flatten.")
-            elif isinstance(previous_layer, MaxPooling2D) or isinstance(previous_layer, AveragePooling2D):
-                if not isinstance(layer, (Conv2D, MaxPooling2D, AveragePooling2D, Flatten)):
-                    raise ValueError("MaxPooling2D or AveragePooling2D layer can only be followed by Conv2D, MaxPooling2D, AveragePooling2D, or Flatten.")
-            elif isinstance(previous_layer, Conv1D):
-                if not isinstance(layer, (Conv1D, MaxPooling1D, AveragePooling1D, Activation, Dropout, Flatten, BatchNormalization)):
-                    raise ValueError("Conv1D layer can only be followed by Conv1D, MaxPooling1D, AveragePooling1D, Activation, Dropout, BatchNormalization, or Flatten.")
-            elif isinstance(previous_layer, MaxPooling1D) or isinstance(previous_layer, AveragePooling1D):
-                if not isinstance(layer, (Conv1D, MaxPooling1D, AveragePooling1D, Flatten)):
-                    raise ValueError("MaxPooling1D or AveragePooling1D layer can only be followed by Conv1D, MaxPooling1D, AveragePooling1D, or Flatten.")
-            elif isinstance(previous_layer, Flatten):
-                if not isinstance(layer, (Dense, Dropout)):
-                    raise ValueError("Flatten layer can only be followed by Dense or Dropout.")
-            elif isinstance(previous_layer, Dropout):
-                if not isinstance(layer, (Dense, Conv2D, Conv1D, Activation)):
-                    raise ValueError("Dropout layer can only be followed by Dense, Conv2D, Conv1D, or Activation.")
-            elif isinstance(previous_layer, Embedding):
-                if not isinstance(layer, (Conv1D, Flatten, Dense)):
-                    raise ValueError("Embedding layer can only be followed by Conv1D, Flatten, or Dense.")
-            elif isinstance(previous_layer, BatchNormalization):
-                if not isinstance(layer, (Dense, Conv2D, Conv1D, Activation)):
-                    raise ValueError("BatchNormalization layer can only be followed by Dense, Conv2D, Conv1D, or Activation.")
+            if type(layer) not in compatibility_dict[type(previous_layer)]:
+                raise ValueError(f"{type(layer).__name__} layer cannot follow {type(previous_layer).__name__} layer.")
 
         self.layers.append(layer)
 

setup.py

@@ -2,7 +2,7 @@
 
 setup(
     name='neuralnetlib',
-    version='2.3.1',
+    version='2.3.2',
     author='Marc Pinet',
     description='A simple convolutional neural network library with only numpy as dependency',
     long_description=open('README.md', encoding="utf-8").read(),