.fuse() gradient introduction bug fix

utils/torch_utils.py

@@ -104,8 +104,8 @@ def prune(model, amount=0.3):
 
 
 def fuse_conv_and_bn(conv, bn):
-    # https://tehnokv.com/posts/fusing-batchnorm-and-conv/
-    with torch.no_grad():
+    # Fuse convolution and batchnorm layers https://tehnokv.com/posts/fusing-batchnorm-and-conv/
+
     # init
     fusedconv = nn.Conv2d(conv.in_channels,
                           conv.out_channels,
@@ -113,7 +113,7 @@ def fuse_conv_and_bn(conv, bn):
                           stride=conv.stride,
                           padding=conv.padding,
                           groups=conv.groups,
-                              bias=True).to(conv.weight.device)
+                          bias=True).requires_grad_(False).to(conv.weight.device)
 
     # prepare filters
     w_conv = conv.weight.clone().view(conv.out_channels, -1)