RuntimeError: Given groups=1, weight of size [128, 512, 1, 1], expected input[4, 2048, 1, 1] to have 512 channels, but got 2048 channels instead

[AngLancer]

大神你好，我是个初学者，在用voc训练psp的时候默认设置可以跑通，但是我想把Ci弄小一点，于是改成了：
class _PSPHead(nn.Module):
def init(self, nclass, norm_layer=nn.BatchNorm2d, norm_kwargs=None, **kwargs):
super(_PSPHead, self).init()
self.psp = _PyramidPooling(512, norm_layer=norm_layer, norm_kwargs=norm_kwargs)
self.block = nn.Sequential(
nn.Conv2d(1024, 128, 3, padding=1, bias=False),#Ci，Co，kernelsize
norm_layer(128, **({} if norm_kwargs is None else norm_kwargs)),
nn.ReLU(True),
nn.Dropout(0.1),
nn.Conv2d(128, nclass, 1)
)
运行就报错了：
Traceback (most recent call last):
File "./train.py", line 335, in
trainer.train()
File "./train.py", line 224, in train
outputs = self.model(images)
File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 727, in _call_impl
result = self.forward(*input, **kwargs)
File "/home/cse/lza/awesome-semantic-segmentation-pytorch/core/models/pspnet.py", line 51, in forward
x = self.head(c4)
File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 727, in _call_impl
result = self.forward(*input, **kwargs)
File "/home/cse/lza/awesome-semantic-segmentation-pytorch/core/models/pspnet.py", line 113, in forward
x = self.psp(x)
File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 727, in _call_impl
result = self.forward(*input, **kwargs)
File "/home/cse/lza/awesome-semantic-segmentation-pytorch/core/models/pspnet.py", line 87, in forward
feat1 = F.interpolate(self.conv1(self.avgpool1(x)), size, mode='bilinear', align_corners=True)
File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 727, in _call_impl
result = self.forward(*input, **kwargs)
File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/container.py", line 117, in forward
input = module(input)
File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 727, in _call_impl
result = self.forward(*input, **kwargs)
File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/conv.py", line 423, in forward
return self._conv_forward(input, self.weight)
File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/conv.py", line 420, in _conv_forward
self.padding, self.dilation, self.groups)
RuntimeError: Given groups=1, weight of size [128, 512, 1, 1], expected input[4, 2048, 1, 1] to have 512 channels, but got 2048 channels instead
调了好久都是同类型的错，请问我该如何让Ci变小呢？

[luojunyi520]

我也是初学者，试着回答一下，不对的话请大神指正，nn.Conv2d第一个参数是输入通道数，这个输入通道数是不可改变的，是由你的图片通道数决定的

[AngLancer]

我也是初学者，试着回答一下，不对的话请大神指正，nn.Conv2d第一个参数是输入通道数，这个输入通道数是不可改变的，是由你的图片通道数决定的

但是代码里本来是这么写的：
class _PSPHead(nn.Module):
def init(self, nclass, norm_layer=nn.BatchNorm2d, norm_kwargs=None, **kwargs):
super(_PSPHead, self).init()
self.psp = _PyramidPooling(2048, norm_layer=norm_layer, norm_kwargs=norm_kwargs)
self.block = nn.Sequential(
nn.Conv2d(4096, 512, 3, padding=1, bias=False),
norm_layer(512, **({} if norm_kwargs is None else norm_kwargs)),
nn.ReLU(True),
nn.Dropout(0.1),
nn.Conv2d(512, nclass, 1)
)
第一个参数是4096，我是不想要4096这么大的，voc图片通道数本身有这么大吗老哥

[luojunyi520]

我也是初学者，试着回答一下，不对的话请大神指正，nn.Conv2d第一个参数是输入通道数，这个输入通道数是不可改变的，是由你的图片通道数决定的

但是代码里本来是这么写的： class _PSPHead(nn.Module): def init(self, nclass, norm_layer=nn.BatchNorm2d, norm_kwargs=None, **kwargs): super(_PSPHead, self).init() self.psp = _PyramidPooling(2048, norm_layer=norm_layer, norm_kwargs=norm_kwargs) self.block = nn.Sequential( nn.Conv2d(4096, 512, 3, padding=1, bias=False), norm_layer(512, **({} if norm_kwargs is None else norm_kwargs)), nn.ReLU(True), nn.Dropout(0.1), nn.Conv2d(512, nclass, 1) ) 第一个参数是4096，我是不想要4096这么大的，voc图片通道数本身有这么大吗老哥

   VOC使用的是RGB 图片，本身通道数只有3，随着图片不断进行下采样，图片尺寸缩小，图片通道数会不断增大。下一层卷积的输入通道数是上一层卷积的输出通道数，你改动其中一层卷积的输入输出通道，肯定会造成某一个环节的通道不匹配。
  
   你要想改通道数，当然也是可以的，原始self.psp = _PyramidPooling(2048, norm_layer=norm_layer, norm_kwargs=norm_kwargs)输入2048，输出4096，你设定输入是512，输出是1024，那么你就得一步步修改class _PyramidPooling(nn.Module):这个类里边self.avgpool1，self.avgpool2等池化层的参数，使其满足_PyramidPooling输入是512，输出是1024，计算每一步的通道数。当然不建议你那么做，因为很麻烦

[AngLancer]

我也是初学者，试着回答一下，不对的话请大神指正，nn.Conv2d第一个参数是输入通道数，这个输入通道数是不可改变的，是由你的图片通道数决定的

但是代码里本来是这么写的： class _PSPHead(nn.Module): def init(self, nclass, norm_layer=nn.BatchNorm2d, norm_kwargs=None, **kwargs): super(_PSPHead, self).init() self.psp = _PyramidPooling(2048, norm_layer=norm_layer, norm_kwargs=norm_kwargs) self.block = nn.Sequential( nn.Conv2d(4096, 512, 3, padding=1, bias=False), norm_layer(512, **({} if norm_kwargs is None else norm_kwargs)), nn.ReLU(True), nn.Dropout(0.1), nn.Conv2d(512, nclass, 1) ) 第一个参数是4096，我是不想要4096这么大的，voc图片通道数本身有这么大吗老哥

   VOC使用的是RGB 图片，本身通道数只有3，随着图片不断进行下采样，图片尺寸缩小，图片通道数会不断增大。下一层卷积的输入通道数是上一层卷积的输出通道数，你改动其中一层卷积的输入输出通道，肯定会造成某一个环节的通道不匹配。
  
   你要想改通道数，当然也是可以的，原始self.psp = _PyramidPooling(2048, norm_layer=norm_layer, norm_kwargs=norm_kwargs)输入2048，输出4096，你设定输入是512，输出是1024，那么你就得一步步修改class _PyramidPooling(nn.Module):这个类里边self.avgpool1，self.avgpool2等池化层的参数，使其满足_PyramidPooling输入是512，输出是1024，计算每一步的通道数。当然不建议你那么做，因为很麻烦

好的，太感谢了老哥！我回头试试