Small modifications in MCCNN and FASNet trainer

bob/learn/pytorch/architectures/MCCNN.py

@@ -43,7 +43,7 @@ class MCCNN(nn.Module):
   	The path to download the pretrained LightCNN model from.
   
   """
-  def __init__(self, block=resblock, layers=[1, 2, 3, 4], num_channels=4, verbosity_level=2):
+  def __init__(self, block=resblock, layers=[1, 2, 3, 4], num_channels=4, verbosity_level=2, use_sigmoid=True):
     """ Init function
 
     Parameters
@@ -51,6 +51,9 @@ class MCCNN(nn.Module):
 
     num_channels: int
       The number of channels present in the input
+    use_sigmoid: bool
+      Whether to use sigmoid in eval phase. If set to `False` do not use
+      sigmoid in eval phase. Training phase is not affected.
     verbosity_level: int
       Verbosity level.
     
@@ -58,6 +61,7 @@ class MCCNN(nn.Module):
     super(MCCNN, self).__init__()
 
     self.num_channels=num_channels
+    self.use_sigmoid=use_sigmoid
 
     self.lcnn_layers=['conv1','block1','group1','block2', 'group2','block3','group3','block4','group4','fc']
 
@@ -211,9 +215,9 @@ class MCCNN(nn.Module):
 
     output = self.linear2fc(output)
 
-    #if self.training:
+    if  self.training or self.use_sigmoid:
 
-    output=nn.Sigmoid()(output)
+      output=nn.Sigmoid()(output)
 
     return output
 

bob/learn/pytorch/trainers/FASNetTrainer.py

@@ -178,7 +178,7 @@ class FASNetTrainer(object):
 
 		# setup optimizer
 
-		optimizer = optim.Adam(filter(lambda p: p.requires_grad, self.network.parameters()),lr = learning_rate )
+		optimizer = optim.Adam(filter(lambda p: p.requires_grad, self.network.parameters()),lr = learning_rate, weight_decay=0.000001)
 
 		self.network.train(True)