Thanks for Ryan's valuable advice!

I saw in the chat channel that there's another contributor who wants to take up
this CUDA project, and he/she wants to start with CNN. To avoid duplication,
I'd love to start with multi-head attention. If this is plausible, I'll draft
a proposal based on this.

And, if I may ask, would Ryan or Marcus like to be my GSoC mentor?

Thanks
Liu, Zhuojin
On Apr 12, 2022, 23:13 +0800, Ryan Curtin <r...@ratml.org>, wrote:
> On Mon, Apr 11, 2022 at 11:55:00PM +0800, Zhuojin Liu wrote:
> > The worry I have is whether mlpack is ready for adding GPU supports now.
> > Though it's on the mlpack vision list, but it's not on the GSoC idea list, 
> > plus
> > that the bandicoot library is not release-ready now, so I'm not sure it's 
> > time
> > to add GPU support to mlpack this year.
>
> Indeed, bandicoot is still in an early stage. At least personally I
> think it would be reasonable to propose a project to complete some
> functionality there, if you can find an interested mentor. In essence
> bandicoot becomes ready when we have 'enough' functionality implemented
> (where 'enough' is defined as the set of functions we need to, e.g.,
> implement some simple neural networks in mlpack using coot::mat as the
> matrix type).
>
> --
> Ryan Curtin | "Like, with jetpacks?"
> r...@ratml.org | - Scott
_______________________________________________
mlpack mailing list
mlpack@lists.mlpack.org
http://knife.lugatgt.org/cgi-bin/mailman/listinfo/mlpack

Reply via email to