0.17.3

architecture.py

Author: bartzbeielstein

pyproject.toml

@@ -7,7 +7,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "spotpython"
-version = "0.17.2"
+version = "0.17.3"
 authors = [
   { name="T. Bartz-Beielstein", email="tbb@bartzundbartz.de" }
 ]

src/spotpython/hyperparameters/architecture.py

@@ -0,0 +1,46 @@
+def generate_div2_list(n, n_min) -> list:
+    result = []
+    current = n
+    repeats = 1
+    max_repeats = 4
+    while current >= n_min:
+        result.extend([current] * min(repeats, max_repeats))
+        current = current // 2
+        repeats = repeats + 1
+    return result
+
+
+def get_hidden_sizes(_L_in, l1, n=10) -> list:
+    """
+    Generates a list of hidden sizes for a neural network with a given input size and l1 regularization.
+    The list is generated by dividing the input size by 2 until the minimum size is reached.
+
+    Args:
+        _L_in (int):
+            input size.
+        l1 (int):
+            l1 regularization.
+        n (int):
+            number of hidden sizes to generate.
+
+    Returns:
+        (list):
+            list of hidden sizes.
+
+    Examples:
+        >>> from spotpython.hyperparameters.architecture import get_hidden_sizes
+            _L_in = 10
+            l1 = 10
+            n = 10
+            get_hidden_sizes(_L_in, l1, n)
+            [10, 5, 2, 1, 1, 1, 1, 1, 1, 1]
+    """
+    if l1 < 4:
+        raise ValueError("l1 must be at least 4")
+    n_low = _L_in // 4
+    n_high = max(l1, 2 * n_low)
+    hidden_sizes = generate_div2_list(n_high, n_low)
+    # keep only the first n values of hidden_sizes list
+    if len(hidden_sizes) > n:
+        hidden_sizes = hidden_sizes[:n]
+    return hidden_sizes

src/spotpython/light/regression/nn_linear_regressor.py

@@ -4,6 +4,7 @@
 from spotpython.hyperparameters.optimizer import optimizer_handler
 import torchmetrics.functional.regression
 import torch.optim as optim
+from spotpython.hyperparameters.architecture import get_hidden_sizes
 
 
 class NNLinearRegressor(L.LightningModule):
@@ -186,9 +187,7 @@ def __init__(
         # set dummy input array for Tensorboard Graphs
         # set log_graph=True in Trainer to see the graph (in traintest.py)
         self.example_input_array = torch.zeros((batch_size, self._L_in))
-        if self.hparams.l1 < 4:
-            raise ValueError("l1 must be at least 4")
-        hidden_sizes = self._get_hidden_sizes()
+        hidden_sizes = get_hidden_sizes(_L_in=self._L_in, l1=l1, n=10)
 
         if batch_norm:
             # Add batch normalization layers
@@ -238,26 +237,6 @@ def _init_weights(self, module):
             if module.bias is not None:
                 nn.init.zeros_(module.bias)
 
-    def _generate_div2_list(self, n, n_min) -> list:
-        result = []
-        current = n
-        repeats = 1
-        max_repeats = 4
-        while current >= n_min:
-            result.extend([current] * min(repeats, max_repeats))
-            current = current // 2
-            repeats = repeats + 1
-        return result
-
-    def _get_hidden_sizes(self, n=10) -> list:
-        n_low = self._L_in // 4
-        n_high = max(self.hparams.l1, 2 * n_low)
-        hidden_sizes = self._generate_div2_list(n_high, n_low)
-        # keep only the first 10 values of hidden_sizes list
-        if len(hidden_sizes) > n:
-            hidden_sizes = hidden_sizes[:n]
-        return hidden_sizes
-
     def forward(self, x: torch.Tensor) -> torch.Tensor:
         """
         Performs a forward pass through the model.