How to specify which dimension to reduce along when using sum ?

[mspronesti]

Basically, I want to convert torch.sum(tensor, 1). tch::Tensor has a sum(dtype) method which doesn't allow to specify which dimension to reduce along: it reduces along all of them, like torch.sum does when no dim is passed.

I guess the correct method to call is sum_dim_intlist, however dim must implement Into<Option<&'a [i64]>>, which forces to use a slice. Isn't there anything simpler than this call ?

some_tensor.sum_dim_intlist([1].as_slice(), false, Kind::Int64)

[LaurentMazare]

I think you're right and that's the appropriate way to do it, I wouldn't know about a simpler way now but I've just prototype in #621 a way to simplify this and allow one to write the following:

some_tensor.sum_dim_intlist(1, false, Kind::Int64)

However this is a bit of a breaking change so we probably want to think a bit more about it (besides only making it a major revision change which is already the case).

[LaurentMazare]

I've merged these changes via #682 so the following should work now.

some_tensor.sum_dim_intlist(1, false, Kind::Int64)

Hopefully this makes it simpler to write code that involve reduction across a dimension. Feel free to re-open if you run into further issues.