Perfect!
#2
by Harley-ml - opened
This is really cool. I think I'll use it as a benchmark from now on for every model that I create.
It'd be nice to see if a model trained on general web stuff with no mathematical content (like a filter) would be able to get above like 25% or so. Not asking you to do that, though.
Anyway, keep up the good work like always!! 😉
Ty!
Will chuck some relevant model scores on the readme a bit later, just gotta get round to it.