I want to replace the model layer

1

I want to replace the model layer.

But since resnet has a block I have to hardcode it.

Recommend a good way?

model.layer1[0].conv1 = nn.Conv()
model.layer1[0].conv2 = nn.Conv()

model stucture

ResNet(
   (conv1): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
   (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
   (layer1): Sequential(
     (0): BasicBlock(
       (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
       (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
       (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
       (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
       (shortcut): Sequential()
     )
     (1): BasicBlock(
       (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
       (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
       (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
       (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
       (shortcut): Sequential()
     )
   )
2

How would you like to replace the layers?
You could use some condition, but we would need to see your use case first.

3

I want all layer replace

example:

pre:

ResNet(
   (conv1): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
   (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
   (layer1): Sequential(
     (0): BasicBlock(
       (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
       (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
       (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
       (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
       (shortcut): Sequential()
     )
     (1): BasicBlock(
       (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
       (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
       (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
       (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
       (shortcut): Sequential()
     )
   )

after:

ResNet(
   (conv1): Conv2d(3, 48, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
   (bn1): BatchNorm2d(48, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
   (layer1): Sequential(
     (0): BasicBlock(
       (conv1): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
       (bn1): BatchNorm2d(48, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
       (conv2): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
       (bn2): BatchNorm2d(48, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
       (shortcut): Sequential()
     )
     (1): BasicBlock(
       (conv1): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
       (bn1): BatchNorm2d(48, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
       (conv2): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
       (bn2): BatchNorm2d(48, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
       (shortcut): Sequential()
     )
   )

I can put them one by one, but is there an easier way?

4

If you want to change a lot of layers, it might be easier to use the Resnet base class and adapt the feature calculation to your use case.