Silero[0] seems to have decent performance (although you will have to some minimal coding). I believe there are better ones if you're willing to tinker a bit more.
This can argument can be made for all essential products/services, what you are arguing seems to be effectively UBI causes additional inflation, which may or may not be true.
The goal of the paper was not to get the best pattern fitter, but to see if it's even possible to do symbolic regression with a simple token-to-token prediction method.
They used a fixed size transformer, where the vocab determines the functions and input/output range. So unless the model needs more 'memory' for your class of expression there wouldn't necessarily be a big change in performance. They have experiments in the paper with bigger/smaller vocabs.
UI looks nice but is janky. Documentation is lacking, but sufficient. Longhorn works great (if you ignore CPU usage due to syncing when there's a lot of IOPS). Overal though self-hosting it has been pretty great for us from a usability perspective. Not sure if it's as enterprise-ready as they say though. Also haven't had to deal with their sales/support (yet).
IANAL, but I think "to the extent allowed by the law" is a way to allow an overreaching (doing more than the law allows) contract to not be fully voided. It effectively allows you to specify ideals, but also means that the license my not work fully as intended.
> Unsigned does not mean "not negative", it means "modular arithmetic".
Unsigned does mean non-negative (or rather no indication of negatives) as the sign of a number is just the positive or negative factor of a number: 1 or -1.
