fix(optimizers-losses-activations): from_config

Author: marcpinet

neuralnetlib/activations.py

@@ -17,26 +17,13 @@ def get_config(self) -> dict:
     @staticmethod
     def from_config(config: dict):
         name = config['name']
-        if name == 'Sigmoid':
-            return Sigmoid()
-        elif name == 'ReLU':
-            return ReLU()
-        elif name == 'Tanh':
-            return Tanh()
-        elif name == 'Softmax':
-            return Softmax()
-        elif name == 'Linear':
-            return Linear()
-        elif name == 'LeakyReLU':
-            return LeakyReLU(alpha=config['alpha'])
-        elif name == 'ELU':
-            return ELU()
-        elif name == 'SELU':
-            return SELU(alpha=config['alpha'], scale=config['scale'])
-        elif name == 'GELU':
-            return GELU()
-        else:
-            raise ValueError(f'Unknown activation function: {name}')
+        
+        for activation_class in ActivationFunction.__subclasses__():
+            if activation_class.__name__ == name:
+                constructor_params = {k: v for k, v in config.items() if k != 'name'}
+                return activation_class(**constructor_params)
+                
+        raise ValueError(f'Unknown activation function: {name}')
 
 
 class Sigmoid(ActivationFunction):

neuralnetlib/losses.py

@@ -16,57 +16,53 @@ def get_config(self) -> dict:
 
     @staticmethod
     def from_config(config: dict) -> 'LossFunction':
-        if config['name'] == 'MeanSquaredError':
-            return MeanSquaredError()
-        elif config['name'] == 'BinaryCrossentropy':
-            return BinaryCrossentropy()
-        elif config['name'] == 'CategoricalCrossentropy':
-            return CategoricalCrossentropy()
-        elif config['name'] == 'MeanAbsoluteError':
-            return MeanAbsoluteError()
-        elif config['name'] == 'HuberLoss':
-            return HuberLoss(config['delta'])
-        elif config['name'] == 'KullbackLeiblerDivergence':
-            return KullbackLeiblerDivergence()
-        elif config['name'] == 'CrossEntropyWithLabelSmoothing':
-            return CrossEntropyWithLabelSmoothing(config['label_smoothing'])
-        elif config['name'] == 'Wasserstein':
-            return Wasserstein()
-        elif config['name'] == 'FocalLoss':
-            return FocalLoss(config['gamma'], config['alpha'])
-        else:
-            raise ValueError(f'Unknown loss function: {config["name"]}')
+        loss_name = config['name']
+        
+        for loss_class in LossFunction.__subclasses__():
+            if loss_class.__name__ == loss_name:
+                constructor_params = {k: v for k, v in config.items() if k != 'name'}
+                return loss_class(**constructor_params)
 
     @staticmethod
     def from_name(name: str) -> "LossFunction":
+        aliases = {
+            "mse": "MeanSquaredError",
+            "bce": "BinaryCrossentropy",
+            "cce": "CategoricalCrossentropy",
+            "scce": "SparseCategoricalCrossentropy",
+            "mae": "MeanAbsoluteError",
+            "kld": "KullbackLeiblerDivergence",
+            "cels": "CrossEntropyWithLabelSmoothing",
+            "wass": "Wasserstein",
+            "focal": "FocalLoss",
+            "fl": "FocalLoss"
+        }
+        
+        original_name = name
         name = name.lower().replace("_", "")
-        if name == "mse" or name == "meansquarederror":
-            return MeanSquaredError()
-        elif name == "bce" or name == "binarycrossentropy":
-            return BinaryCrossentropy()
-        elif name == "cce" or name == "categorycrossentropy":
-            return CategoricalCrossentropy()
-        elif name == "scce" or name == "sparsecategoricalcrossentropy":
-            return SparseCategoricalCrossentropy()
-        elif name == "mae" or name == "meanabsoluteerror":
-            return MeanAbsoluteError()
-        elif name == "kld" or name == "kullbackleiblerdivergence":
-            return KullbackLeiblerDivergence()
-        elif name == "crossentropywithlabelsmoothing" or name == "cels":
-            return CrossEntropyWithLabelSmoothing()
-        elif name == "Wasserstein" or name == "wasserstein" or name == "wass":
-            return Wasserstein()
-        elif name == "focalloss" or name == "focal" or name == "fl":
-            return FocalLoss()
-        elif name.startswith("huber") and len(name.split("_")) == 2:
-            delta = float(name.split("_")[-1])
-            return HuberLoss(delta)
-        else:
-            for subclass in LossFunction.__subclasses__():
-                if subclass.__name__.lower() == name:
-                    return subclass()
-
-        raise ValueError(f"No loss function found for the name: {name}")
+        
+        if name.startswith("huber") and len(original_name.split("_")) == 2:
+            try:
+                delta = float(original_name.split("_")[-1])
+                return Huber(delta=delta)
+            except ValueError:
+                pass
+        
+        if name in aliases:
+            name = aliases[name]
+        
+        for loss_class in LossFunction.__subclasses__():
+            if loss_class.__name__.lower() == name or loss_class.__name__ == name:
+                if loss_class.__name__ == "Huber":
+                    return loss_class(delta=1.0)
+                elif loss_class.__name__ == "CrossEntropyWithLabelSmoothing":
+                    return loss_class(label_smoothing=0.1)
+                elif loss_class.__name__ == "FocalLoss":
+                    return loss_class(gamma=2.0, alpha=0.25)
+                else:
+                    return loss_class()
+                
+        raise ValueError(f"No loss function found for the name: {original_name}")
 
 
 class MeanSquaredError(LossFunction):
@@ -142,7 +138,7 @@ def __str__(self):
         return "MeanAbsoluteError"
 
 
-class HuberLoss(LossFunction):
+class Huber(LossFunction):
     def __init__(self, delta: float = 1.0):
         super().__init__()
         self.delta = delta

neuralnetlib/optimizers.py

@@ -16,16 +16,14 @@ def get_config(self) -> dict:
 
     @staticmethod
     def from_config(config: dict):
-        if config['name'] == 'SGD':
-            return SGD.from_config(config)
-        elif config['name'] == 'Momentum':
-            return Momentum.from_config(config)
-        elif config['name'] == 'RMSprop':
-            return RMSprop.from_config(config)
-        elif config['name'] == 'Adam':
-            return Adam.from_config(config)
-        else:
-            raise ValueError(f"Unknown optimizer name: {config['name']}")
+        optimizer_name = config['name']
+        
+        for optimizer_class in Optimizer.__subclasses__():
+            if optimizer_class.__name__ == optimizer_name:
+                constructor_params = {k: v for k, v in config.items() if k != 'name'}
+                return optimizer_class(**constructor_params)
+                
+        raise ValueError(f"No optimizer found for the name: {optimizer_name}")
 
     @staticmethod
     def from_name(name: str) -> "Optimizer":

tests/test_losses.py

@@ -3,7 +3,7 @@
 import numpy as np
 
 from neuralnetlib.losses import MeanSquaredError, BinaryCrossentropy, CategoricalCrossentropy, MeanAbsoluteError, \
-    HuberLoss
+    Huber
 
 
 class TestLossFunctions(unittest.TestCase):
@@ -41,7 +41,7 @@ def test_mean_absolute_error(self):
         self.assertAlmostEqual(calculated_loss, expected_loss)
 
     def test_huber_loss(self):
-        huber = HuberLoss(delta=1.0)
+        huber = Huber(delta=1.0)
         y_true = np.array([1, 2, 3])
         y_pred = np.array([1, 2, 4])
         error = y_true - y_pred