Pytorch C++ - Block Access?

In Python I can do tensor[0:3, :] etc. How do I do the same thing with the Pytorch C++ API

tensor.slice(0, 0, 3) (API ref)

You can also use Tensor::narrow (API ref, Python docs) just note that the third argument to narrow is “length” not “end” (where start + length = end)

How can I use the slice operation to “set” a particular region? All this does is give me a new memory of that slice. I want to say, write to only a particular column

Any update? This is not what I want