从全连接层反推输入尺寸大小
程序员文章站
2022-03-03 14:37:12
...
全连接层:输出 512维
batch_size = 128
nn.Linear(128*8*8, 512),
输入尺寸(3通道图片):32x32=1024
batch_size = 128
channel = 3
torch.Size([128, 3, 32, 32])
完整的网络结构:
import torch.nn as nn
class _Encoder(nn.Module):
def __init__(self, layers):
super(_Encoder, self).__init__()
self.layers = nn.Sequential(*layers)
def forward(self, x):
x = self.layers(x)
x = x.view(x.size(0), -1)
return x
class _Decoder(nn.Module):
def __init__(self, output_size):
super(_Decoder, self).__init__()
self.layers = nn.Sequential(
nn.Linear(128*8*8, 512),
nn.BatchNorm1d(512),
nn.ReLU(),
nn.Linear(512, output_size)
)
def forward(self, x):
x = self.layers(x)
return x
def Model(num_classes, num_channels):
layers = [
nn.Conv2d(num_channels, 32, kernel_size=3, padding=1),
nn.BatchNorm2d(32),
nn.ReLU(),
nn.Conv2d(32, 64, kernel_size=3, padding=1, stride=2),
nn.BatchNorm2d(64),
nn.ReLU(),
nn.Conv2d(64, 64, kernel_size=3, padding=1),
nn.BatchNorm2d(64),
nn.ReLU(),
nn.Conv2d(64, 128, kernel_size=3, padding=1, stride=2),
nn.BatchNorm2d(128),
nn.ReLU(),
nn.Conv2d(128, 128, kernel_size=3, padding=1),
nn.BatchNorm2d(128),
nn.ReLU(),
]
encoders = [_Encoder(layers=layers) for _ in num_classes]
print(num_classes, encoders) # num_classes: [2, 2, 2, 2, 2, 2, 2, 2, 2, 2]
return [_Model(output_size=cls, encoder=encoder) for cls, encoder in zip(num_classes, encoders)]
推荐阅读