Max-Heinrich Laves

I already created a pull request for that: https://github.com/avibrazil/RDM/pull/51 You should be able to check out the `arch_arm64` branch of my fork and compile it with `make`. For those of...

@rennsport Maybe it's better to keep this issue open. The original RDM repository seems inactive and others will find this issue more easily when it's still directly visible.

@lockieluke @gmcmullengrc Sorry, I accidentally deleted the file when I was cleaning up my drive. I have re-uploaded the file, the link should work again.

I also noticed the slowness of `conv` on the cpu when I implemented `conv3d` in #993. I looked into it and the operation on the cpu could be substantially faster...

> @mlaves you are planning to come back to this? Would be great to land it. Sorry for the delay, I'll finish that in the next couple of days.

@awni I added a `transpose` flag to the existing `Convolution` primitive and removed the initially introduced primitive that is now obsolete. I also rebased onto main.

> I remove the tranpose flag and attempted to resolve the incorrect gradient w.r.t. the `weight` when the `flip` argument is `True`. I don't think all the gradients are correct...

> Looks good to me. Do you mind sharing the results of a benchmark to see how well the groups parameter works for 3D convs? Ideally we should get a...