pytorch forward两个参数实例

以channel Attention Block为例子

class CAB(nn.Module):
 
 def __init__(self, in_channels, out_channels):
 super(CAB, self).__init__()
 self.global_pooling = nn.AdaptiveAvgPool2d(output_size=1)
 self.conv1 = nn.Conv2d(in_channels, out_channels, kernel_size=1, stride=1, padding=0)
 self.relu = nn.ReLU()
 self.conv2 = nn.Conv2d(out_channels, out_channels, kernel_size=1, stride=1, padding=0)
 self.sigmod = nn.Sigmoid()
 
 def forward(self, x):
 x1, x2 = x # high, low
 x = torch.cat([x1,x2],dim=1)
 x = self.global_pooling(x)
 x = self.conv1(x)
 x = self.relu(x)
 x = self.conv2(x)
 x = self.sigmod(x)
 x2 = x * x2
 res = x2 + x1
 return res

以上这篇pytorch forward两个参数实例就是小编分享给大家的全部内容了，希望能给大家一个参考，也希望大家多多支持脚本之家。