RuntimeError: Expected tensor for argument #1 ‘indices’ to have scalar type Long; but got torch.IntTensor instead (while checking arguments for embedding)

RuntimeError: Expected tensor for argument #1 'indices' to have scalar type Long; but got torch.IntTensor instead (while checking arguments for embedding)

题目限制，错误信息不完整。
定位到了这一行：

        out = model.forward(batch.src, batch.trg, 
                            batch.src_mask, batch.trg_mask)

打印出来看看类型：

        print(batch.src.type())
        print(batch.src_mask.type())
        print(batch.trg.type())
        print(batch.trg_mask.type())

把torch.IntTensor 对应的变量转化为Long,理论上batch.src.long()应该可以解决，但是又提示让转化为int64,总之就是打印变量类型，根据提示去转化吧。
print大法好啊。

        out = model.forward(batch.src.to(torch.int64), batch.trg.to(torch.int64), 
                            batch.src_mask, batch.trg_mask)