How to slice a tensor without copying it?

[InAnYan]

Why .slice() takes a self argument?

Is there any variant that would make a new tensor that is a slice of an existing tensor but without copying the original tensor (e.g.: a method that accepts &self)?

[InAnYan]

How would I properly split into train and test tensors?

[laggui]

If you still want to use the tensor after the slice, simply do tensor.clone().slice(...). Cloning does not copy the data it just increases a reference count. And then you can proceed to use tensor in following operations.

[InAnYan]

Oh, it's not cloning literally. I see, thanks!