chore: update sumpool tensor operation docs

[field-worker]

The flag was missing from arg description. Also added a doctest that shows the usage of the flag.
Just another small PR as I was going through and understanding the codebase.

[field-worker]

As a sidenote, I noticed that some tensor ops are generic over T (e.g. pack) while others are tied to specific types, in this case i128. Do you think there'd be any value in making sumpool and others generic on T? (though I'm not mistaken, the trait bounds would have to be a little more restrictive than T: TensorType + std::marker::Send + std::marker::Sync to allow for multiplication, for example.

[alexander-camuto]

As a sidenote, I noticed that some tensor ops are generic over T (e.g. pack) while others are tied to specific types, in this case i128. Do you think there'd be any value in making sumpool and others generic on T? (though I'm not mistaken, the trait bounds would have to be a little more restrictive than T: TensorType + std::marker::Send + std::marker::Sync to allow for multiplication, for example.

Hey thanks for the PR. We could make it more generic -- but it would require Div and Mul if I'm not mistaken. and then it stops applying to field elements -- hence why some ops are over integers -- so as to let the end user know exactly what type works here (else a bit of a pain to debug unfortunately).

[field-worker]

That's right, also pub fn const_div would need to become generic on T. I might tackle this in a follow-up

[alexander-camuto]

That's right, also pub fn const_div would need to become generic on T. I might tackle this in a follow-up

just as a note I edited the above with

then it stops applying to field elements -- hence why some ops are over integers -- so as to let the end user know exactly what type works here (else a bit of a pain to debug unfortunately).

Down to make it more generic though :)

[field-worker]

I'm not sure I follow - you want to be able to have a Tensor over field elements, or just over primitive types?
If the former, AFAIK Field already implements Mul. While Div trait is not directly specified in the bounds, in theory it's possible to divide two prime field elements.

But now that I look at it, I think Div won't be needed after all: https://github.com/field-worker/ezkl/blob/b1582bde89df7b87f52d374d8930a5dc1a6abec5/src/tensor/ops.rs#L3731

[alexander-camuto]

DM me on twitter so we can discuss in full (@CamutoDante)

[field-worker]

Happy to, but you'd have to msg me or follow back: https://twitter.com/field_worker99