adapt init for PyTorch >= 0.4

Author: fitsumreda

models.py

@@ -81,13 +81,13 @@ def __init__(self, args, batchNorm=False, div_flow = 20.):
         for m in self.modules():
             if isinstance(m, nn.Conv2d):
                 if m.bias is not None:
-                    init.uniform(m.bias)
-                init.xavier_uniform(m.weight)
+                    init.uniform_(m.bias)
+                init.xavier_uniform_(m.weight)
 
             if isinstance(m, nn.ConvTranspose2d):
                 if m.bias is not None:
-                    init.uniform(m.bias)
-                init.xavier_uniform(m.weight)
+                    init.uniform_(m.bias)
+                init.xavier_uniform_(m.weight)
                 # init_deconv_bilinear(m.weight)
 
     def init_deconv_bilinear(self, weight):

networks/FlowNetC.py

@@ -58,13 +58,13 @@ def __init__(self,args, batchNorm=True, div_flow = 20):
         for m in self.modules():
             if isinstance(m, nn.Conv2d):
                 if m.bias is not None:
-                    init.uniform(m.bias)
-                init.xavier_uniform(m.weight)
+                    init.uniform_(m.bias)
+                init.xavier_uniform_(m.weight)
 
             if isinstance(m, nn.ConvTranspose2d):
                 if m.bias is not None:
-                    init.uniform(m.bias)
-                init.xavier_uniform(m.weight)
+                    init.uniform_(m.bias)
+                init.xavier_uniform_(m.weight)
                 # init_deconv_bilinear(m.weight)
         self.upsample1 = nn.Upsample(scale_factor=4, mode='bilinear')
 

networks/FlowNetFusion.py

@@ -35,13 +35,13 @@ def __init__(self,args, batchNorm=True):
         for m in self.modules():
             if isinstance(m, nn.Conv2d):
                 if m.bias is not None:
-                    init.uniform(m.bias)
-                init.xavier_uniform(m.weight)
+                    init.uniform_(m.bias)
+                init.xavier_uniform_(m.weight)
 
             if isinstance(m, nn.ConvTranspose2d):
                 if m.bias is not None:
-                    init.uniform(m.bias)
-                init.xavier_uniform(m.weight)
+                    init.uniform_(m.bias)
+                init.xavier_uniform_(m.weight)
                 # init_deconv_bilinear(m.weight)
 
     def forward(self, x):

networks/FlowNetS.py

@@ -47,13 +47,13 @@ def __init__(self, args, input_channels = 12, batchNorm=True):
         for m in self.modules():
             if isinstance(m, nn.Conv2d):
                 if m.bias is not None:
-                    init.uniform(m.bias)
-                init.xavier_uniform(m.weight)
+                    init.uniform_(m.bias)
+                init.xavier_uniform_(m.weight)
 
             if isinstance(m, nn.ConvTranspose2d):
                 if m.bias is not None:
-                    init.uniform(m.bias)
-                init.xavier_uniform(m.weight)
+                    init.uniform_(m.bias)
+                init.xavier_uniform_(m.weight)
                 # init_deconv_bilinear(m.weight)
         self.upsample1 = nn.Upsample(scale_factor=4, mode='bilinear')
 

networks/FlowNetSD.py

@@ -51,13 +51,13 @@ def __init__(self, args, batchNorm=True):
         for m in self.modules():
             if isinstance(m, nn.Conv2d):
                 if m.bias is not None:
-                    init.uniform(m.bias)
-                init.xavier_uniform(m.weight)
+                    init.uniform_(m.bias)
+                init.xavier_uniform_(m.weight)
 
             if isinstance(m, nn.ConvTranspose2d):
                 if m.bias is not None:
-                    init.uniform(m.bias)
-                init.xavier_uniform(m.weight)
+                    init.uniform_(m.bias)
+                init.xavier_uniform_(m.weight)
                 # init_deconv_bilinear(m.weight)
         self.upsample1 = nn.Upsample(scale_factor=4, mode='bilinear')