Expected tensor for argument #1 'indices' to have scalar type Long;but got torch.IntTensor instead

日萌社

人工智能AI：Keras PyTorch MXNet TensorFlow PaddlePaddle 深度学习实战（不定时更新）

---

 

        报错：RuntimeError: Expected tensor for argument #1 'indices' to have scalar type Long; 
              but got torch.IntTensor instead (while checking arguments for embedding)
        分析：训练的批量样本数据输入值需要是long值的Tensor数据，而不是int值的Tensor数据.
        解决：
            1.把输入数据data和目标数据data的类型值都从int转换为long
                1.可以加long()
                    source = Variable(data, requires_grad=False).long()
                    target = Variable(data, requires_grad=False).long()
                2.也可以加torch.LongTensor
                    source = Variable(torch.LongTensor(data), requires_grad=False)
                    target = Variable(torch.LongTensor(data), requires_grad=False)

            2.如果有使用以下这个第三方pyitcast包下的transformer_utils.py的话，可以如下修改
                    class SimpleLossCompute
                        def __call__(self, x, y, norm):
                            #return loss.data[0] * norm
                            #loss.data[0] 改成 loss.data.item()
                            return loss.data.item() * norm
            3.tensor变量.item() 的用法
                x = torch.randn(1)
                print(x)        #tensor([-0.4464])
                print(x.item()) #-0.44643348455429077