two device

[thestars-maker]

源代码
conv_weight = self.conv.weight
conv_shape = conv_weight.shape
conv_weight_hd = torch.cuda.FloatTensor(conv_shape[0], conv_shape[1], 3 * 3).fill_(0)
conv_weight_hd[:, :, [0, 3, 6]] = conv_weight[:, :, :]

RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!

这儿赋值时显示conv_weight是cpu,conv_weight_hd是cuda，即使此处改了device问题但其他地方依然存在该问题，请问应该如何有效处理呢

[thestars-maker]

def forward(self, x):

    w1, b1 = self.conv1_1.get_weight()
    w2, b2 = self.conv1_2.get_weight()
    w3, b3 = self.conv1_3.get_weight()
    w4, b4 = self.conv1_4.get_weight()
    w5, b5 = self.conv1_5.weight, self.conv1_5.bias

    w = w1 + w2 + w3 + w4 + w5
    b = b1 + b2 + b3 + b4 + b5
    res = nn.functional.conv2d(input=x, weight=w, bias=b, stride=1, padding=1, groups=1)

    return res

此处的w1和后面的w也存在该问题