X · @teortaxesTex
· X / Twitter
RT Lazarz: You’re asking why they’re underscoring on benchmarks I’m telling you they spent 98% of the development time on getting the muP right and…
RT LazarzYou’re asking why they’re underscoring on benchmarksI’m telling you they spent 98% of the development time on getting the muP right and they scored my heartstochasm: the mup multiplier also implies they used mup which is quite cool to see. also of course this is the largest anyone has scaled up short convs Can