It seems that Pytorch tensor does not support unsigned int32 and int64 currently. While Numpy and TensorFlow both support them. I am wondering whether Pytorch has any plan about it? If not, any short cut i could do to make extension from Numpy? Or write C++ code is necessary?