RuntimeError: Input type (torch.cuda.FloatTensor) and weight type (torch.DoubleTensor) should be the

出错位置： 

train_label_batch = torch.from_numpy(train_label_batch)

解决办法：将数据类型转换成FloatTensor即可，如下，加一行代码Tensor.type(torch.FloatTensor)

train_label_batch = torch.from_numpy(train_label_batch)
train_label_batch = train_label_batch.type(torch.FloatTensor)  # 转Float
train_label_batch = train_label_batch.cuda()  # 转cuda