Orignal Relu

WideResNet_28-10.txt

WideResNet(
  (init_conv): Conv2d(3, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (conv2): Sequential(
    (0): WideBasic(
      (residual): Sequential(
        (0): BatchNorm2d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (1): ReLU(inplace=True)
        (2): Conv2d(16, 160, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        (3): BatchNorm2d(160, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (4): ReLU(inplace=True)
        (5): Dropout(p=0.5, inplace=False)
        (6): Conv2d(160, 160, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      )
      (shortcut): Sequential(
        (0): Conv2d(16, 160, kernel_size=(1, 1), stride=(1, 1))
      )
    )
    (1): WideBasic(
      (residual): Sequential(
        (0): BatchNorm2d(160, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (1): ReLU(inplace=True)
        (2): Conv2d(160, 160, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        (3): BatchNorm2d(160, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (4): ReLU(inplace=True)
        (5): Dropout(p=0.5, inplace=False)
        (6): Conv2d(160, 160, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      )
      (shortcut): Sequential()
    )
    (2): WideBasic(
      (residual): Sequential(
        (0): BatchNorm2d(160, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (1): ReLU(inplace=True)
        (2): Conv2d(160, 160, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        (3): BatchNorm2d(160, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (4): ReLU(inplace=True)
        (5): Dropout(p=0.5, inplace=False)
        (6): Conv2d(160, 160, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      )
      (shortcut): Sequential()
    )
    (3): WideBasic(
      (residual): Sequential(
        (0): BatchNorm2d(160, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (1): ReLU(inplace=True)
        (2): Conv2d(160, 160, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        (3): BatchNorm2d(160, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (4): ReLU(inplace=True)
        (5): Dropout(p=0.5, inplace=False)
        (6): Conv2d(160, 160, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      )
      (shortcut): Sequential()
    )
  )
  (conv3): Sequential(
    (0): WideBasic(
      (residual): Sequential(
        (0): BatchNorm2d(160, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (1): ReLU(inplace=True)
        (2): Conv2d(160, 320, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))
        (3): BatchNorm2d(320, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (4): ReLU(inplace=True)
        (5): Dropout(p=0.5, inplace=False)
        (6): Conv2d(320, 320, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      )
      (shortcut): Sequential(
        (0): Conv2d(160, 320, kernel_size=(1, 1), stride=(2, 2))
      )
    )
    (1): WideBasic(
      (residual): Sequential(
        (0): BatchNorm2d(320, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (1): ReLU(inplace=True)
        (2): Conv2d(320, 320, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        (3): BatchNorm2d(320, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (4): ReLU(inplace=True)
        (5): Dropout(p=0.5, inplace=False)
        (6): Conv2d(320, 320, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      )
      (shortcut): Sequential()
    )
    (2): WideBasic(
      (residual): Sequential(
        (0): BatchNorm2d(320, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (1): ReLU(inplace=True)
        (2): Conv2d(320, 320, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        (3): BatchNorm2d(320, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (4): ReLU(inplace=True)
        (5): Dropout(p=0.5, inplace=False)
        (6): Conv2d(320, 320, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      )
      (shortcut): Sequential()
    )
    (3): WideBasic(
      (residual): Sequential(
        (0): BatchNorm2d(320, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (1): ReLU(inplace=True)
        (2): Conv2d(320, 320, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        (3): BatchNorm2d(320, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (4): ReLU(inplace=True)
        (5): Dropout(p=0.5, inplace=False)
        (6): Conv2d(320, 320, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      )
      (shortcut): Sequential()
    )
  )
  (conv4): Sequential(
    (0): WideBasic(
      (residual): Sequential(
        (0): BatchNorm2d(320, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (1): ReLU(inplace=True)
        (2): Conv2d(320, 640, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))
        (3): BatchNorm2d(640, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (4): ReLU(inplace=True)
        (5): Dropout(p=0.5, inplace=False)
        (6): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      )
      (shortcut): Sequential(
        (0): Conv2d(320, 640, kernel_size=(1, 1), stride=(2, 2))
      )
    )
    (1): WideBasic(
      (residual): Sequential(
        (0): BatchNorm2d(640, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (1): ReLU(inplace=True)
        (2): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        (3): BatchNorm2d(640, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (4): ReLU(inplace=True)
        (5): Dropout(p=0.5, inplace=False)
        (6): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      )
      (shortcut): Sequential()
    )
    (2): WideBasic(
      (residual): Sequential(
        (0): BatchNorm2d(640, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (1): ReLU(inplace=True)
        (2): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        (3): BatchNorm2d(640, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (4): ReLU(inplace=True)
        (5): Dropout(p=0.5, inplace=False)
        (6): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      )
      (shortcut): Sequential()
    )
    (3): WideBasic(
      (residual): Sequential(
        (0): BatchNorm2d(640, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (1): ReLU(inplace=True)
        (2): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
        (3): BatchNorm2d(640, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (4): ReLU(inplace=True)
        (5): Dropout(p=0.5, inplace=False)
        (6): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      )
      (shortcut): Sequential()
    )
  )
  (bn): BatchNorm2d(640, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (relu): ReLU(inplace=True)
  (avg_pool): AdaptiveAvgPool2d(output_size=(1, 1))
  (linear): Linear(in_features=640, out_features=10, bias=True)
)



extract_characteristics.py

    ###################
    model.conv2[0].residual[1].register_forward_hook( get_activation('conv2_0_relu_1') )
    model.conv2[0].residual[4].register_forward_hook( get_activation('conv2_0_relu_4') )

    model.conv2[1].residual[1].register_forward_hook( get_activation('conv2_1_relu_1') )
    model.conv2[1].residual[4].register_forward_hook( get_activation('conv2_1_relu_4') )

    model.conv2[2].residual[1].register_forward_hook( get_activation('conv2_2_relu_1') )
    model.conv2[2].residual[4].register_forward_hook( get_activation('conv2_2_relu_4') )

    model.conv2[3].residual[1].register_forward_hook( get_activation('conv2_3_relu_1') )
    model.conv2[3].residual[4].register_forward_hook( get_activation('conv2_3_relu_4') )
    ######################

    model.conv3[0].residual[1].register_forward_hook(get_activation('conv3_0_relu_1'))
    model.conv3[0].residual[4].register_forward_hook(get_activation('conv3_0_relu_4'))

    model.conv3[1].residual[1].register_forward_hook(get_activation('conv3_1_relu_1'))
    model.conv3[1].residual[4].register_forward_hook(get_activation('conv3_1_relu_4'))

    model.conv3[2].residual[1].register_forward_hook(get_activation('conv3_2_relu_1'))
    model.conv3[2].residual[4].register_forward_hook(get_activation('conv3_2_relu_4'))

    model.conv3[3].residual[1].register_forward_hook(get_activation('conv3_3_relu_1'))
    model.conv3[3].residual[4].register_forward_hook(get_activation('conv3_3_relu_4'))


    model.conv4[0].residual[1].register_forward_hook(get_activation('conv4_0_relu_1'))
    model.conv4[0].residual[4].register_forward_hook(get_activation('conv4_0_relu_4'))

    model.conv4[1].residual[1].register_forward_hook(get_activation('conv4_1_relu_1'))
    model.conv4[1].residual[4].register_forward_hook(get_activation('conv4_1_relu_4'))

    model.conv4[2].residual[1].register_forward_hook(get_activation('conv4_2_relu_1'))
    model.conv4[2].residual[4].register_forward_hook(get_activation('conv4_2_relu_4'))

    model.conv4[3].residual[1].register_forward_hook(get_activation('conv4_3_relu_1'))
    model.conv4[3].residual[4].register_forward_hook(get_activation('conv4_3_relu_4'))

    model.relu.register_forward_hook(get_activation('relu'))