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

有numpy和torch两种类型的数据

z = mean.clone().detach() + eps * torch.exp(logstd)

　　源代码这边报错了，修改如下

eps = eps.cuda()    
z = mean.cuda()+ eps * torch.exp(logstd).cuda()

　　就是把和numpy有关的，比如reshape什么的后面加上cuda()