Gert-Jan Both

@ethanluoyc This is very useful, thanks! I had some issues building it myself - any chance you could do a quick python 3.12 wheel?

There's two problems: 1) Tensorflow 2.8 doesnt have wheels for apple silicon - so it won't work there. 2) I'm surprised your test passes as it seems ray doesn't have...

If I remember correctly, I originally designed this with @awennersteen. Indeed we didn't consider fixing some parameters, and having different grad backends. There's two options, both without changing the API,...

In the new version it's indeed fixed, thanks. It hasn't been in the pip version, though. Additional question: what settings do I need to reproduce scikit-learn Lasso results? I'm struggling...

Thanks, this was very helpful - I was unaware this was optimistix default behaviour. Additionally, do you know if there's a difference between the jaxopt matrix-free solvers and the optimistix...

Allright, I'll get to it then! My money's on a divide by zero somewhere (I do some padding), but it doesn't seem to be an issue for jaxopt, so that's...

A warning in the code could be a good idea, although it would require us to decide on a size limit - which would be fairly system dependent... Looking at...

Hi @hartikainen, I've actually written something like this a few days ago and so far it's working quite well, (but I'm still testing it, so no guarantees :-)). The trick...