Error: Input type (torch.cuda.FloatTensor) and weight type (torch.FloatTensor) should be the same

pytorch使用中出现以上问题解决办法

以上错误是因为网络部分模块用列表list放置，没有被nn.ModuleList()处理；导致pytorch无法识别，也就没放到cuda上；

利用nn.ModuleList()可解决：

class OneModule(nn.Module):
def __init__(self):
    super(OneModule, self).__init__()
   	self.layers = []#RuntimeError: Input type (torch.cuda.FloatTensor) and weight type (torch.FloatTensor) should be the same
    self.layers.append(block(self.res_inplanes, planes, stride, downsample))
    self.res_inplanes = planes * block.expansion
    for i in range(1, blocks):
        self.layers.append(block(self.res_inplanes, planes))
    self.layers = nn.ModuleList(self.layers)#缺少这行代码就会报错
     

def forward(self, x):
    pass