juliaup VS tiny-cuda-nn

There is juliaup, available in the Microsoft store, for managing julia versions: https://github.com/JuliaLang/juliaup

Better use the official binaries from https://julialang.org/downloads/. For Windows, best use the Julia app in the MS Store (https://github.com/JuliaLang/juliaup ).

Thanks - fwiw, I currently use juliaup, which basically does the same thing. Unfortunately, the problem isn't that Julia isn't on the path (it is), it's that Neovim's exepath can't find it, for some reason. :)

But it’s considered prerelease for Mac and Linux according to its git repository. So presumably it won’t become the default until it’s not experimental any more.

GitHub - JuliaLang/juliaup: Julia installer and version multiplexer GitHub - JuliaPluto/PlutoUI.jl GitHub - fonsp/Pluto.jl: 🎈 Simple reactive notebooks for Julia

Use the right tool for the job! I would never write a system utility or OS or user application or compiler in Julia, for example. juliaup is a Julia utility being written in Rust for this exact reason!

And yes, I know about jill.py and have considered using it as well. It seems fairly straightforwad to use. A similar project that I found interesting is juliaup. Still, I would definitely prefer just using zypper instead of relying on third party tools. That's why I figured I might as well ask on here.

In the opposite direction from your question is a very interesting project, TinyNN all implemented as close to the metal as possible and very fast: https://github.com/NVlabs/tiny-cuda-nn

This repo only implemented the rendering part of the NGP but is more simple and has a lesser amount of code compared to the original (Instant-NGP and tiny-cuda-nn).

...a C++ library with a CUDA backend. But these high-performance building blocks might only be saturating the GPU fully if the data is large enough.

I haven't looked at implementing these things, but I imagine uf you have smaller networks and thus less data, the large building blocks may not be optimal. You may for example want to fuse some operations to reduce memory latency from repeated memory access.

In PyTorch world, there are approaches for small networks as well, there is https://github.com/NVlabs/tiny-cuda-nn - as far as I understand from the first link in the README, it makes clever use of the CUDA shared memory, which can hold all the weights of a tiny network (but not larger ones).