Merge pull request #18 from RMeli/weightsandbiases

Remove weight initialisation from model __init__

Author: RMeli

gnina/models.py

@@ -21,6 +21,28 @@
 from torch import nn
 
 
+def weights_and_biases_init(m: nn.Module) -> None:
+    """
+    Initialize the weights and biases of the model.
+
+    Parameters
+    ----------
+    m : nn.Module
+        Module (layer) to initialize
+
+    Notes
+    -----
+    This function is used to initialize the weights of the model for both convolutional
+    and linear layers. Weights are initialized using uniform Xavier initialization
+    while biases are set to zero.
+
+    https://github.com/gnina/libmolgrid/blob/e6d5f36f1ae03f643ca69cdec1625ac52e653f88/test/test_torch_cnn.py#L45-L48
+    """
+    if isinstance(m, nn.Conv3d) or isinstance(m, nn.Linear):
+        nn.init.xavier_uniform_(m.weight.data)
+        nn.init.constant_(m.bias.data, 0.0)
+
+
 class Default2017(nn.Module):
     """
     GNINA default2017 model architecture.
@@ -134,13 +156,6 @@ def __init__(self, input_dims: Tuple):
             )
         )
 
-        # Xavier initialization for convolutional and linear layers
-        for m in self.modules():
-            if isinstance(m, nn.Conv3d) or isinstance(m, nn.Linear):
-                nn.init.xavier_uniform_(m.weight.data)
-                # TODO: Initialize bias to zero?
-                # TODO: See https://github.com/gnina/libmolgrid/blob/e6d5f36f1ae03f643ca69cdec1625ac52e653f88/test/test_torch_cnn.py#L48
-
     def forward(self, x: torch.Tensor):
         """
         Parameters
@@ -197,13 +212,6 @@ def __init__(self, input_dims: Tuple):
             )
         )
 
-        # Xavier initialization for convolutional and linear layers
-        for m in self.modules():
-            if isinstance(m, nn.Conv3d) or isinstance(m, nn.Linear):
-                nn.init.xavier_uniform_(m.weight.data)
-                # TODO: Initialize bias to zero?
-                # TODO: See https://github.com/gnina/libmolgrid/blob/e6d5f36f1ae03f643ca69cdec1625ac52e653f88/test/test_torch_cnn.py#L48
-
     def forward(self, x: torch.Tensor):
         """
         Parameters
@@ -368,13 +376,6 @@ def __init__(self, input_dims: Tuple):
             )
         )
 
-        # Xavier initialization for convolutional and linear layers
-        for m in self.modules():
-            if isinstance(m, nn.Conv3d) or isinstance(m, nn.Linear):
-                nn.init.xavier_uniform_(m.weight.data)
-                # TODO: Initialize bias to zero?
-                # TODO: See https://github.com/gnina/libmolgrid/blob/e6d5f36f1ae03f643ca69cdec1625ac52e653f88/test/test_torch_cnn.py#L48
-
     def forward(self, x: torch.Tensor):
         """
         Parameters
@@ -430,13 +431,6 @@ def __init__(self, input_dims: Tuple):
             )
         )
 
-        # Xavier initialization for convolutional and linear layers
-        for m in self.modules():
-            if isinstance(m, nn.Conv3d) or isinstance(m, nn.Linear):
-                nn.init.xavier_uniform_(m.weight.data)
-                # TODO: Initialize bias to zero?
-                # TODO: See https://github.com/gnina/libmolgrid/blob/e6d5f36f1ae03f643ca69cdec1625ac52e653f88/test/test_torch_cnn.py#L48
-
     def forward(self, x: torch.Tensor):
         """
         Parameters
@@ -656,11 +650,6 @@ def __init__(
 
         self.features = nn.Sequential(features)
 
-        # Xavier initialization for convolutional and linear layers
-        for m in self.modules():
-            if isinstance(m, nn.Conv3d) or isinstance(m, nn.Linear):
-                nn.init.xavier_uniform_(m.weight.data)
-
     def forward(self, x):
         """
         Parameters
@@ -719,11 +708,6 @@ def __init__(
             )
         )
 
-        # Xavier initialization for convolutional and linear layers
-        for m in self.modules():
-            if isinstance(m, nn.Conv3d) or isinstance(m, nn.Linear):
-                nn.init.xavier_uniform_(m.weight.data)
-
     def forward(self, x):
         """
         Parameters
@@ -795,11 +779,6 @@ def __init__(
             )
         )
 
-        # Xavier initialization for convolutional and linear layers
-        for m in self.modules():
-            if isinstance(m, nn.Conv3d) or isinstance(m, nn.Linear):
-                nn.init.xavier_uniform_(m.weight.data)
-
     def forward(self, x):
         """
         Parameters
@@ -927,11 +906,6 @@ def __init__(self, input_dims: Tuple):
             )
         )
 
-        # Xavier initialization for convolutional and linear layers
-        for m in self.modules():
-            if isinstance(m, nn.Conv3d) or isinstance(m, nn.Linear):
-                nn.init.xavier_uniform_(m.weight.data)
-
     def forward(self, x: torch.Tensor):
         """
         Parameters
@@ -1066,11 +1040,6 @@ def __init__(self, input_dims: Tuple):
             )
         )
 
-        # Xavier initialization for convolutional and linear layers
-        for m in self.modules():
-            if isinstance(m, nn.Conv3d) or isinstance(m, nn.Linear):
-                nn.init.xavier_uniform_(m.weight.data)
-
     def forward(self, x: torch.Tensor):
         """
         Parameters

gnina/training.py

@@ -18,7 +18,7 @@
 from gnina import metrics, setup, utils
 from gnina.dataloaders import GriddedExamplesLoader
 from gnina.losses import AffinityLoss
-from gnina.models import models_dict
+from gnina.models import models_dict, weights_and_biases_init
 
 
 def options(args: Optional[List[str]] = None):
@@ -587,6 +587,7 @@ def training(args):
     # Create model
     # Select model based on architecture and affinity flag (pose vs affinity)
     model = models_dict[(args.model, affinity)](train_loader.dims).to(device)
+    model.apply(weights_and_biases_init)
 
     # TODO: Compile model into TorchScript
     # Requires model refactoring to avoid branching based on affinity