When does OO syntax pass the Tensor as the resulting Tensor?

[xzhangcs]

maths.md states:

"The Torch package adopts the same concept, so that calling a function directly on the Tensor itself using an object-oriented syntax is equivalent to passing the Tensor as the optional resulting Tensor. The following two calls are equivalent.

torch.log(x, x)
x:log()
"

However, this doesn't always seem to be the case. For example, if we call

x:cat(y), x.reshape(2,3), x:t()

then x does NOT change after the call. In particular, I can't see why x:add(y) changes x, while x:cat(y) doesn't.

So is there any rule on when a tensor will be changed by calling on itself in OO syntax? Thanks.

[soumith]

the three functions you listed, i.e. cat, reshape and t do not have an in-place mode, and return a new tensor (their documentation should cover that detail).

There is unfortunately no rule.