Deep learning library written in Futhark

I've only used Futhark a handful of times, but found it to be really easy, and intuitive (coming from extensively using SML).

For anyone looking to get started with it, writing a Monte Carlo simulation is a good first exercise to get familiar. Monte Carlo simulations also happen to perfectly showcase its power to exploit parallelism for massive speedups.

Wonderful to see this getting attention! I met with Minh since we were both using Futhark for our thesis. It's blazing fast and quite fun to work with.

If anyone's interested, my thesis contains some more examples on deep learning with Futhark: https://github.com/Jegp/thesis/

Every time the Futhark language comes up, I think it should look like this:

ᛈᚱᛖᚾᛏᚨ("ᚺᛖᛚᛟ ᚹᛟᚱᛚᛞ!")

Great to see a futhark lib in the wild, the code seems really concise and readable. I feel the main reason for Futhark's ergonomics is FP combined with its array prelude (e.g. fold/transpose/flatten/rotate/indices).

There's also the thesis linked from https://futhark-lang.org/publications.html - https://futhark-lang.org/student-projects/svend-msc-thesis.p... - and this paper: https://futhark-lang.org/publications/fhpnc19.pdf

Great to see Futhark improving, we need more options on GPGPU ecosystem.

Not much documentation but the example program looks nice https://github.com/HnimNart/deeplearning/blob/master/program...

I've seen some impressive results from Futhark in other areas. Would be very interested to see how it performs when applied to DL. Couldn't find any benchmarks here?