pytorch深入学习(六)

• Resnet需要注意的是： 因为是x和处理后的特征相加， 所以经过处理后的特征的维度和输入是一致的，假如说有变化的话，那么x可以加一个1*1卷积，变换通道数
• 用个例程说明
• Bn层会减去均值， 所以cnn的bias可以设置为False加快运算

def conv3_3 (in_planes, out_planes, stride=1, groups=1, dilation=1):
	return nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride, 
	padding=dilation, groups=groups,bias=False, dilation=dilation)

class BasicBlock(nn.Module):
	expansion=1
	__constants__ = ['downsample']
	def __init__(self, inplanes, planes, stride=1, downsample=None, groups=1, 
	base_width=64, dialation=1, norm_layer=None):
		super(BasicBlock, self).__init__()
		if norm_layer is None:
			norm_layer = nn.BatchNorm2d
		
		self.conv1 = conv3_3(inplanes, planes, stride)
		self.bn1 = norm_layer(planes) # Bn层的输入是通道的数量
		self.relu = nn.Relu(inplace=True)
		self.conv2 = nn.conv3_3(planes, planes)
		self.bn2 = norm_layer(planes)
		self.downsample = downsample
		self.stride = stride
	
	def forward(self, x):
		identity = x
		out = self.conv1(x)
		out = self.bn1(out)
		out = self.relu(out)
		out = self.conv2(out)
		out = self.bn2(out)
		
		if self.downsample is Not None:
			identity = self.downsample(x)
			out += identity
			out = self.relu(out)
			return out

• 一般cnn构架是bn后加Relu， Resnet最后相加后也需要Relu， 所以cnn的后面没有Relu