AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Benchmarks for Hardware Back Ends
M per has become kind of the industry wide bench mark for hardware and which models are performing better ware. But i get the sense that it is not a very good indicator of which hardware back ends are really good across variety of applications. And so one thing i wonder about is how a bench mark like m l purf could look better. One piece i read recently actually was poking a different as, and i think that it suggested a sort of normalization. I'm uncurious if you've given that any thought, and how how something like na perf could be more indicative of the kinds of things we wantyes, a hundred %. So what you are talking about is really, if you