VGGBnDrop(
  (features): Sequential(
    (0): ConvBNReLU(
      (conv): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (1): Dropout(p=0.3, inplace=False)
    (2): ConvBNReLU(
      (conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
    (4): ConvBNReLU(
      (conv): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (5): Dropout(p=0.4, inplace=False)
    (6): ConvBNReLU(
      (conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (7): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
    (8): ConvBNReLU(
      (conv): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (9): Dropout(p=0.4, inplace=False)
    (10): ConvBNReLU(
      (conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (11): Dropout(p=0.4, inplace=False)
    (12): ConvBNReLU(
      (conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (13): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
    (14): ConvBNReLU(
      (conv): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (15): Dropout(p=0.4, inplace=False)
    (16): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (17): Dropout(p=0.4, inplace=False)
    (18): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (19): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
    (20): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (21): Dropout(p=0.4, inplace=False)
    (22): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (23): Dropout(p=0.4, inplace=False)
    (24): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (25): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
  )
  (classifier): Sequential(
    (0): Dropout(p=0.5, inplace=False)
    (1): Linear(in_features=512, out_features=512, bias=True)
    (2): BatchNorm1d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (3): ReLU(inplace=True)
    (4): Dropout(p=0.5, inplace=False)
    (5): Linear(in_features=512, out_features=10, bias=True)
  )
)
Executing method neural_path_kmeans on model CIFAR-VGG and dataset CIFAR10
Ratio, Test accuracy: 1.000 90.32
Ratio, Val accuracy: 0.900 98.05
Ratio, Test accuracy: 0.900 89.95
Ratio, Val accuracy: 0.700 98.15
Ratio, Test accuracy: 0.700 89.80
epoch 1/5 | acc: train=0.00% val=0.00%:   0%|          | 0/5 [1:41:20<?, ?it/s]
Ratio, Val accuracy: 0.500 98.06
Ratio, Test accuracy: 0.500 89.80
Ratio, Val accuracy: 0.300 97.70
Ratio, Test accuracy: 0.300 89.56
Ratio, Val accuracy: 0.200 97.99
Ratio, Test accuracy: 0.200 89.79
Ratio, Val accuracy: 0.150 97.89
Ratio, Test accuracy: 0.150 89.91
Ratio, Val accuracy: 0.100 97.77
Ratio, Test accuracy: 0.100 89.40
Ratio, Val accuracy: 0.050 97.30
Ratio, Test accuracy: 0.050 89.19
Ratio, Val accuracy: 0.020 84.10
Ratio, Test accuracy: 0.020 77.54
Ratio, Val accuracy: 0.010 50.98
Ratio, Test accuracy: 0.010 48.31
Ratio, Val accuracy: 0.005 20.12
Ratio, Test accuracy: 0.005 18.99
Executing method random_structured on model CIFAR-VGG and dataset CIFAR10
Ratio, Test accuracy: 1.000 90.32
Ratio, Val accuracy: 0.900 98.14
Ratio, Test accuracy: 0.900 89.97
Ratio, Val accuracy: 0.700 98.30
Ratio, Test accuracy: 0.700 90.06
Ratio, Val accuracy: 0.500 98.24
Ratio, Test accuracy: 0.500 90.22
Ratio, Val accuracy: 0.300 98.38
Ratio, Test accuracy: 0.300 90.09
Ratio, Val accuracy: 0.200 98.05
Ratio, Test accuracy: 0.200 89.71
Ratio, Val accuracy: 0.150 97.26
Ratio, Test accuracy: 0.150 88.98
Ratio, Val accuracy: 0.100 93.94
Ratio, Test accuracy: 0.100 85.68
Ratio, Val accuracy: 0.050 83.77
Ratio, Test accuracy: 0.050 77.90
Ratio, Val accuracy: 0.020 40.58
Ratio, Test accuracy: 0.020 36.77
Ratio, Val accuracy: 0.010 36.94
Ratio, Test accuracy: 0.010 34.66
Ratio, Val accuracy: 0.005 25.36
Ratio, Test accuracy: 0.005 22.86
Executing method l1_structured on model CIFAR-VGG and dataset CIFAR10
Ratio, Test accuracy: 1.000 90.32
Ratio, Val accuracy: 0.900 98.14
Ratio, Test accuracy: 0.900 90.04
Ratio, Val accuracy: 0.700 98.42
Ratio, Test accuracy: 0.700 90.07
Ratio, Val accuracy: 0.500 97.69
Ratio, Test accuracy: 0.500 89.55
Ratio, Val accuracy: 0.300 98.31
Ratio, Test accuracy: 0.300 89.82
Ratio, Val accuracy: 0.200 97.05
Ratio, Test accuracy: 0.200 88.91
Ratio, Val accuracy: 0.150 97.72
Ratio, Test accuracy: 0.150 89.38
Ratio, Val accuracy: 0.100 97.10
Ratio, Test accuracy: 0.100 88.86
Ratio, Val accuracy: 0.050 96.39
Ratio, Test accuracy: 0.050 88.36
Ratio, Val accuracy: 0.020 87.18
Ratio, Test accuracy: 0.020 79.34
Ratio, Val accuracy: 0.010 32.85
Ratio, Test accuracy: 0.010 31.32
Ratio, Val accuracy: 0.005 24.30
Ratio, Test accuracy: 0.005 23.06
VGGBnDrop(
  (features): Sequential(
    (0): ConvBNReLU(
      (conv): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (1): Dropout(p=0.3, inplace=False)
    (2): ConvBNReLU(
      (conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
    (4): ConvBNReLU(
      (conv): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (5): Dropout(p=0.4, inplace=False)
    (6): ConvBNReLU(
      (conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (7): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
    (8): ConvBNReLU(
      (conv): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (9): Dropout(p=0.4, inplace=False)
    (10): ConvBNReLU(
      (conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (11): Dropout(p=0.4, inplace=False)
    (12): ConvBNReLU(
      (conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (13): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
    (14): ConvBNReLU(
      (conv): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (15): Dropout(p=0.4, inplace=False)
    (16): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (17): Dropout(p=0.4, inplace=False)
    (18): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (19): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
    (20): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (21): Dropout(p=0.4, inplace=False)
    (22): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (23): Dropout(p=0.4, inplace=False)
    (24): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (25): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
  )
  (classifier): Sequential(
    (0): Dropout(p=0.5, inplace=False)
    (1): Linear(in_features=512, out_features=512, bias=True)
    (2): BatchNorm1d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (3): ReLU(inplace=True)
    (4): Dropout(p=0.5, inplace=False)
    (5): Linear(in_features=512, out_features=10, bias=True)
  )
)
Executing method neural_path_kmeans on model CIFAR-VGG and dataset CIFAR10
Ratio, Test accuracy: 1.000 90.01
Ratio, Val accuracy: 0.900 97.99
Ratio, Test accuracy: 0.900 89.56
Ratio, Val accuracy: 0.700 97.93
Ratio, Test accuracy: 0.700 89.27
Ratio, Val accuracy: 0.500 97.64
Ratio, Test accuracy: 0.500 88.96
Ratio, Val accuracy: 0.300 97.17
Ratio, Test accuracy: 0.300 88.78
Ratio, Val accuracy: 0.200 98.08
Ratio, Test accuracy: 0.200 89.54
Ratio, Val accuracy: 0.150 98.02
Ratio, Test accuracy: 0.150 89.40
Ratio, Val accuracy: 0.100 96.81
Ratio, Test accuracy: 0.100 88.44
Ratio, Val accuracy: 0.050 97.48
Ratio, Test accuracy: 0.050 89.12
Ratio, Val accuracy: 0.020 82.24
Ratio, Test accuracy: 0.020 75.46
Ratio, Val accuracy: 0.010 55.81
Ratio, Test accuracy: 0.010 52.59
Ratio, Val accuracy: 0.005 18.86
Ratio, Test accuracy: 0.005 17.70
Executing method random_structured on model CIFAR-VGG and dataset CIFAR10
Ratio, Test accuracy: 1.000 90.01
Ratio, Val accuracy: 0.900 98.07
Ratio, Test accuracy: 0.900 89.64
Ratio, Val accuracy: 0.700 98.00
Ratio, Test accuracy: 0.700 89.64
Ratio, Val accuracy: 0.500 97.60
Ratio, Test accuracy: 0.500 89.01
Ratio, Val accuracy: 0.300 97.61
Ratio, Test accuracy: 0.300 89.44
Ratio, Val accuracy: 0.200 97.86
Ratio, Test accuracy: 0.200 88.66
Ratio, Val accuracy: 0.150 97.72
Ratio, Test accuracy: 0.150 88.81
Ratio, Val accuracy: 0.100 96.79
Ratio, Test accuracy: 0.100 88.18
Ratio, Val accuracy: 0.050 96.77
Ratio, Test accuracy: 0.050 88.75
Ratio, Val accuracy: 0.020 44.99
Ratio, Test accuracy: 0.020 41.42
Ratio, Val accuracy: 0.010 56.75
Ratio, Test accuracy: 0.010 52.28
Ratio, Val accuracy: 0.005 17.04
Ratio, Test accuracy: 0.005 16.19
Executing method l1_structured on model CIFAR-VGG and dataset CIFAR10
Ratio, Test accuracy: 1.000 90.01
Ratio, Val accuracy: 0.900 98.03
Ratio, Test accuracy: 0.900 89.62
Ratio, Val accuracy: 0.700 98.23
Ratio, Test accuracy: 0.700 89.72
Ratio, Val accuracy: 0.500 98.12
Ratio, Test accuracy: 0.500 89.54
Ratio, Val accuracy: 0.300 97.77
Ratio, Test accuracy: 0.300 89.31
Ratio, Val accuracy: 0.200 97.90
Ratio, Test accuracy: 0.200 89.32
Ratio, Val accuracy: 0.150 97.25
Ratio, Test accuracy: 0.150 88.95
Ratio, Val accuracy: 0.100 97.61
Ratio, Test accuracy: 0.100 89.29
Ratio, Val accuracy: 0.050 81.35
Ratio, Test accuracy: 0.050 73.88
Ratio, Val accuracy: 0.020 38.10
Ratio, Test accuracy: 0.020 36.91
Ratio, Val accuracy: 0.010 36.83
Ratio, Test accuracy: 0.010 34.48
Ratio, Val accuracy: 0.005 27.77
Ratio, Test accuracy: 0.005 26.78
VGGBnDrop(
  (features): Sequential(
    (0): ConvBNReLU(
      (conv): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (1): Dropout(p=0.3, inplace=False)
    (2): ConvBNReLU(
      (conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
    (4): ConvBNReLU(
      (conv): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (5): Dropout(p=0.4, inplace=False)
    (6): ConvBNReLU(
      (conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (7): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
    (8): ConvBNReLU(
      (conv): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (9): Dropout(p=0.4, inplace=False)
    (10): ConvBNReLU(
      (conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (11): Dropout(p=0.4, inplace=False)
    (12): ConvBNReLU(
      (conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (13): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
    (14): ConvBNReLU(
      (conv): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (15): Dropout(p=0.4, inplace=False)
    (16): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (17): Dropout(p=0.4, inplace=False)
    (18): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (19): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
    (20): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (21): Dropout(p=0.4, inplace=False)
    (22): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (23): Dropout(p=0.4, inplace=False)
    (24): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (25): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
  )
  (classifier): Sequential(
    (0): Dropout(p=0.5, inplace=False)
    (1): Linear(in_features=512, out_features=512, bias=True)
    (2): BatchNorm1d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (3): ReLU(inplace=True)
    (4): Dropout(p=0.5, inplace=False)
    (5): Linear(in_features=512, out_features=10, bias=True)
  )
)
Executing method neural_path_kmeans on model CIFAR-VGG and dataset CIFAR10
Ratio, Test accuracy: 1.000 90.19
Ratio, Val accuracy: 0.900 97.80
Ratio, Test accuracy: 0.900 89.60
Ratio, Val accuracy: 0.700 97.49
Ratio, Test accuracy: 0.700 89.22
Ratio, Val accuracy: 0.500 97.47
Ratio, Test accuracy: 0.500 89.25
Ratio, Val accuracy: 0.300 97.62
Ratio, Test accuracy: 0.300 89.64
Ratio, Val accuracy: 0.200 97.76
Ratio, Test accuracy: 0.200 89.57
Ratio, Val accuracy: 0.150 97.64
Ratio, Test accuracy: 0.150 89.35
Ratio, Val accuracy: 0.100 97.17
Ratio, Test accuracy: 0.100 89.16
Ratio, Val accuracy: 0.050 96.42
Ratio, Test accuracy: 0.050 88.36
Ratio, Val accuracy: 0.020 82.22
Ratio, Test accuracy: 0.020 76.10
Ratio, Val accuracy: 0.010 54.86
Ratio, Test accuracy: 0.010 51.81
Ratio, Val accuracy: 0.005 19.82
Ratio, Test accuracy: 0.005 19.28
Executing method random_structured on model CIFAR-VGG and dataset CIFAR10
Ratio, Test accuracy: 1.000 90.19
Ratio, Val accuracy: 0.900 97.74
Ratio, Test accuracy: 0.900 89.63
Ratio, Val accuracy: 0.700 97.74
Ratio, Test accuracy: 0.700 89.39
Ratio, Val accuracy: 0.500 97.45
Ratio, Test accuracy: 0.500 89.24
Ratio, Val accuracy: 0.300 97.32
Ratio, Test accuracy: 0.300 88.86
Ratio, Val accuracy: 0.200 96.78
Ratio, Test accuracy: 0.200 88.21
Ratio, Val accuracy: 0.150 97.39
Ratio, Test accuracy: 0.150 88.93
Ratio, Val accuracy: 0.100 97.55
Ratio, Test accuracy: 0.100 88.85
Ratio, Val accuracy: 0.050 76.77
Ratio, Test accuracy: 0.050 70.52
Ratio, Val accuracy: 0.020 67.42
Ratio, Test accuracy: 0.020 63.12
Ratio, Val accuracy: 0.010 44.03
Ratio, Test accuracy: 0.010 41.41
Ratio, Val accuracy: 0.005 30.38
Ratio, Test accuracy: 0.005 28.38
Executing method l1_structured on model CIFAR-VGG and dataset CIFAR10
Ratio, Test accuracy: 1.000 90.19
Ratio, Val accuracy: 0.900 97.87
Ratio, Test accuracy: 0.900 89.50
Ratio, Val accuracy: 0.700 97.79
Ratio, Test accuracy: 0.700 89.46
Ratio, Val accuracy: 0.500 97.70
Ratio, Test accuracy: 0.500 89.46
Ratio, Val accuracy: 0.300 97.26
Ratio, Test accuracy: 0.300 89.18
Ratio, Val accuracy: 0.200 96.66
Ratio, Test accuracy: 0.200 88.68
Ratio, Val accuracy: 0.150 96.54
Ratio, Test accuracy: 0.150 87.88
Ratio, Val accuracy: 0.100 96.70
Ratio, Test accuracy: 0.100 88.30
Ratio, Val accuracy: 0.050 88.85
Ratio, Test accuracy: 0.050 82.62
Ratio, Val accuracy: 0.020 64.70
Ratio, Test accuracy: 0.020 60.43
Ratio, Val accuracy: 0.010 20.86
Ratio, Test accuracy: 0.010 20.52
Ratio, Val accuracy: 0.005 22.93
Ratio, Test accuracy: 0.005 21.52
VGGBnDrop(
  (features): Sequential(
    (0): ConvBNReLU(
      (conv): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (1): Dropout(p=0.3, inplace=False)
    (2): ConvBNReLU(
      (conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
    (4): ConvBNReLU(
      (conv): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (5): Dropout(p=0.4, inplace=False)
    (6): ConvBNReLU(
      (conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (7): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
    (8): ConvBNReLU(
      (conv): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (9): Dropout(p=0.4, inplace=False)
    (10): ConvBNReLU(
      (conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (11): Dropout(p=0.4, inplace=False)
    (12): ConvBNReLU(
      (conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (13): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
    (14): ConvBNReLU(
      (conv): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (15): Dropout(p=0.4, inplace=False)
    (16): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (17): Dropout(p=0.4, inplace=False)
    (18): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (19): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
    (20): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (21): Dropout(p=0.4, inplace=False)
    (22): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (23): Dropout(p=0.4, inplace=False)
    (24): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (25): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
  )
  (classifier): Sequential(
    (0): Dropout(p=0.5, inplace=False)
    (1): Linear(in_features=512, out_features=512, bias=True)
    (2): BatchNorm1d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (3): ReLU(inplace=True)
    (4): Dropout(p=0.5, inplace=False)
    (5): Linear(in_features=512, out_features=10, bias=True)
  )
)
Executing method neural_path_kmeans on model CIFAR-VGG and dataset CIFAR10
Ratio, Test accuracy: 1.000 89.75
Ratio, Val accuracy: 0.900 97.82
Ratio, Test accuracy: 0.900 89.40
Ratio, Val accuracy: 0.700 97.80
Ratio, Test accuracy: 0.700 89.25
Ratio, Val accuracy: 0.500 97.70
Ratio, Test accuracy: 0.500 89.09
Ratio, Val accuracy: 0.300 97.50
Ratio, Test accuracy: 0.300 89.02
Ratio, Val accuracy: 0.200 98.01
Ratio, Test accuracy: 0.200 89.49
Ratio, Val accuracy: 0.150 97.96
Ratio, Test accuracy: 0.150 89.52
Ratio, Val accuracy: 0.100 97.67
Ratio, Test accuracy: 0.100 89.06
Ratio, Val accuracy: 0.050 96.97
Ratio, Test accuracy: 0.050 88.52
Ratio, Val accuracy: 0.020 85.38
Ratio, Test accuracy: 0.020 77.76
Ratio, Val accuracy: 0.010 49.79
Ratio, Test accuracy: 0.010 46.37
Ratio, Val accuracy: 0.005 20.07
Ratio, Test accuracy: 0.005 18.80
Executing method random_structured on model CIFAR-VGG and dataset CIFAR10
Ratio, Test accuracy: 1.000 89.75
Ratio, Val accuracy: 0.900 97.88
Ratio, Test accuracy: 0.900 89.39
Ratio, Val accuracy: 0.700 97.81
Ratio, Test accuracy: 0.700 89.34
Ratio, Val accuracy: 0.500 97.74
Ratio, Test accuracy: 0.500 89.19
Ratio, Val accuracy: 0.300 97.60
Ratio, Test accuracy: 0.300 89.06
Ratio, Val accuracy: 0.200 97.84
Ratio, Test accuracy: 0.200 89.60
Ratio, Val accuracy: 0.150 97.29
Ratio, Test accuracy: 0.150 88.58
Ratio, Val accuracy: 0.100 93.53
Ratio, Test accuracy: 0.100 84.96
Ratio, Val accuracy: 0.050 89.66
Ratio, Test accuracy: 0.050 81.24
Ratio, Val accuracy: 0.020 38.63
Ratio, Test accuracy: 0.020 36.32
Ratio, Val accuracy: 0.010 34.26
Ratio, Test accuracy: 0.010 31.95
Ratio, Val accuracy: 0.005 17.37
Ratio, Test accuracy: 0.005 16.55
Executing method l1_structured on model CIFAR-VGG and dataset CIFAR10
Ratio, Test accuracy: 1.000 89.75
Ratio, Val accuracy: 0.900 98.00
Ratio, Test accuracy: 0.900 89.40
Ratio, Val accuracy: 0.700 97.91
Ratio, Test accuracy: 0.700 89.40
Ratio, Val accuracy: 0.500 98.05
Ratio, Test accuracy: 0.500 89.44
Ratio, Val accuracy: 0.300 97.61
Ratio, Test accuracy: 0.300 89.03
Ratio, Val accuracy: 0.200 97.42
Ratio, Test accuracy: 0.200 89.01
Ratio, Val accuracy: 0.150 96.95
Ratio, Test accuracy: 0.150 88.51
Ratio, Val accuracy: 0.100 83.98
Ratio, Test accuracy: 0.100 76.87
Ratio, Val accuracy: 0.050 74.49
Ratio, Test accuracy: 0.050 69.02
Ratio, Val accuracy: 0.020 60.25
Ratio, Test accuracy: 0.020 55.66
Ratio, Val accuracy: 0.010 46.59
Ratio, Test accuracy: 0.010 42.83
Ratio, Val accuracy: 0.005 19.94
Ratio, Test accuracy: 0.005 19.12
VGGBnDrop(
  (features): Sequential(
    (0): ConvBNReLU(
      (conv): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (1): Dropout(p=0.3, inplace=False)
    (2): ConvBNReLU(
      (conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
    (4): ConvBNReLU(
      (conv): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (5): Dropout(p=0.4, inplace=False)
    (6): ConvBNReLU(
      (conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (7): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
    (8): ConvBNReLU(
      (conv): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (9): Dropout(p=0.4, inplace=False)
    (10): ConvBNReLU(
      (conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (11): Dropout(p=0.4, inplace=False)
    (12): ConvBNReLU(
      (conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (13): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
    (14): ConvBNReLU(
      (conv): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (15): Dropout(p=0.4, inplace=False)
    (16): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (17): Dropout(p=0.4, inplace=False)
    (18): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (19): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
    (20): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (21): Dropout(p=0.4, inplace=False)
    (22): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (23): Dropout(p=0.4, inplace=False)
    (24): ConvBNReLU(
      (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (25): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=True)
  )
  (classifier): Sequential(
    (0): Dropout(p=0.5, inplace=False)
    (1): Linear(in_features=512, out_features=512, bias=True)
    (2): BatchNorm1d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (3): ReLU(inplace=True)
    (4): Dropout(p=0.5, inplace=False)
    (5): Linear(in_features=512, out_features=10, bias=True)
  )
)
Executing method neural_path_kmeans on model CIFAR-VGG and dataset CIFAR10
Ratio, Test accuracy: 1.000 90.62
Ratio, Val accuracy: 0.900 98.26
Ratio, Test accuracy: 0.900 90.28
Ratio, Val accuracy: 0.700 98.31
Ratio, Test accuracy: 0.700 90.19
Ratio, Val accuracy: 0.500 98.13
Ratio, Test accuracy: 0.500 90.02
Ratio, Val accuracy: 0.300 98.23
Ratio, Test accuracy: 0.300 90.02
Ratio, Val accuracy: 0.200 97.76
Ratio, Test accuracy: 0.200 89.72
Ratio, Val accuracy: 0.150 97.93
Ratio, Test accuracy: 0.150 89.81
Ratio, Val accuracy: 0.100 98.12
Ratio, Test accuracy: 0.100 90.12
Ratio, Val accuracy: 0.050 97.25
Ratio, Test accuracy: 0.050 89.00
Ratio, Val accuracy: 0.020 86.88
Ratio, Test accuracy: 0.020 79.62
Ratio, Val accuracy: 0.010 46.22
Ratio, Test accuracy: 0.010 43.82
Ratio, Val accuracy: 0.005 18.09
Ratio, Test accuracy: 0.005 17.30
Executing method random_structured on model CIFAR-VGG and dataset CIFAR10
Ratio, Test accuracy: 1.000 90.62
Ratio, Val accuracy: 0.900 98.33
Ratio, Test accuracy: 0.900 90.20
Ratio, Val accuracy: 0.700 98.26
Ratio, Test accuracy: 0.700 90.34
Ratio, Val accuracy: 0.500 98.34
Ratio, Test accuracy: 0.500 89.91
Ratio, Val accuracy: 0.300 98.10
Ratio, Test accuracy: 0.300 89.92
Ratio, Val accuracy: 0.200 97.39
Ratio, Test accuracy: 0.200 89.47
Ratio, Val accuracy: 0.150 97.74
Ratio, Test accuracy: 0.150 89.79
Ratio, Val accuracy: 0.100 92.63
Ratio, Test accuracy: 0.100 85.24
Ratio, Val accuracy: 0.050 93.13
Ratio, Test accuracy: 0.050 85.79
Ratio, Val accuracy: 0.020 59.31
Ratio, Test accuracy: 0.020 53.66
Ratio, Val accuracy: 0.010 36.22
Ratio, Test accuracy: 0.010 34.45
Ratio, Val accuracy: 0.005 17.85
Ratio, Test accuracy: 0.005 17.87
Executing method l1_structured on model CIFAR-VGG and dataset CIFAR10
Ratio, Test accuracy: 1.000 90.62
Ratio, Val accuracy: 0.900 98.27
Ratio, Test accuracy: 0.900 90.30
Ratio, Val accuracy: 0.700 98.44
Ratio, Test accuracy: 0.700 90.28
Ratio, Val accuracy: 0.500 98.30
Ratio, Test accuracy: 0.500 90.23
Ratio, Val accuracy: 0.300 98.33
Ratio, Test accuracy: 0.300 90.43
Ratio, Val accuracy: 0.200 97.97
Ratio, Test accuracy: 0.200 89.99
Ratio, Val accuracy: 0.150 97.58
Ratio, Test accuracy: 0.150 89.41
Ratio, Val accuracy: 0.100 97.53
Ratio, Test accuracy: 0.100 89.57
Ratio, Val accuracy: 0.050 71.94
Ratio, Test accuracy: 0.050 66.73
Ratio, Val accuracy: 0.020 57.45
Ratio, Test accuracy: 0.020 53.55
Ratio, Val accuracy: 0.010 20.90
Ratio, Test accuracy: 0.010 20.38
Ratio, Val accuracy: 0.005 22.54
Ratio, Test accuracy: 0.005 21.39
