今天在看代码时候遇到了for循环错误,看的百思不得其解,最后又重新翻了翻书才理解。下面看错误分析:
in_channels = [ 768, 192, 192,192,192,]
out_channels = [192,192,192,192,37,]
# zip函数生成[(768,192),(192,192),(192,192),(192,192),(192,37)]
# enumerate函数生成(0,(768,192)), (1,(192,192)) , (2,(192,192)), (3,(192,192)), (4,(192,37))
modules = []
for i, (in_channel, out_channel) in enumerate(
zip(in_channels, out_channels)):
modules.append(
nn.Conv2d(
in_channels=in_channel,
out_channels=out_channel,
kernel_size=1,
stride=1))
if i != 4:
modules.append(nn.Upsample(scale_factor=2, mode='bilinear'))
print(modules)
in_channels = [ 768, 192, 192,192,192,]
out_channels = [192,192,192,192,37,]
# zip函数生成[(768,192),(192,192),(192,192),(192,192),(192,37)]
# enumerate函数生成(0,(768,192)), (1,(192,192)) , (2,(192,192)), (3,(192,192)), (4,(192,37))
modules = []
for i, (in_channel, out_channel) in enumerate(
zip(in_channels, out_channels)):
modules.append(
nn.Conv2d(
in_channels=in_channel,
out_channels=out_channel,
kernel_size=1,
stride=1))
if i != 4:
modules.append(nn.Upsample(scale_factor=2, mode='bilinear'))
print(modules)
找不同,这两个代码段只有最后一行不同,且缩进不同。造成的结果千差万别:
分别对应第一个和第二个:
[Conv2d(768, 192, kernel_size=(1, 1), stride=(1, 1)), Upsample(scale_factor=2.0, mode=bilinear)]
[Conv2d(768, 192, kernel_size=(1, 1), stride=(1, 1)), Upsample(scale_factor=2.0, mode=bilinear), Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1)), Upsample(scale_factor=2.0, mode=bilinear)]
[Conv2d(768, 192, kernel_size=(1, 1), stride=(1, 1)), Upsample(scale_factor=2.0, mode=bilinear), Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1)), Upsample(scale_factor=2.0, mode=bilinear), Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1)), Upsample(scale_factor=2.0, mode=bilinear)]
[Conv2d(768, 192, kernel_size=(1, 1), stride=(1, 1)), Upsample(scale_factor=2.0, mode=bilinear), Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1)), Upsample(scale_factor=2.0, mode=bilinear), Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1)), Upsample(scale_factor=2.0, mode=bilinear), Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1)), Upsample(scale_factor=2.0, mode=bilinear)]
[Conv2d(768, 192, kernel_size=(1, 1), stride=(1, 1)), Upsample(scale_factor=2.0, mode=bilinear), Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1)), Upsample(scale_factor=2.0, mode=bilinear), Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1)), Upsample(scale_factor=2.0, mode=bilinear), Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1)), Upsample(scale_factor=2.0, mode=bilinear), Conv2d(192, 37, kernel_size=(1, 1), stride=(1, 1))
首先这是for循环缩进错误,正确的是第二个,第一个也没报错,就是不符合逻辑,
print缩进,表示一直循环,直到跳出循环,print不缩进,表示只执行一次
开始modules=[ ],执行第一次循环往里面添加第一个卷积块Conv2d(768, 192),这时正确的应该是继续执行第二次循环,再添加第二个卷积块Conv2d(192, 192),一直添加直到遍历结束。
对于错误的,执行第一次遍历,往module添加第一个卷积块,然后就直接的打印了
,同时module里面还放有第一个卷积块,执行第二次遍历,添加第二个卷积块,打印module,此时module里面有两个卷积块,同样,每一次打印都有之前的卷积块,即每次打印都比上一次多一个模块。
所以注意print的缩进,注意打印次数。