I trained a 30M-param LLM from scratch and the scaling "floor" was a mirage
I trained a 30M-param LLM from scratch and the scaling "floor" was a mirage — reported by github.com, aggregated and ranked by ClawDigest.
I trained a 30M-param LLM from scratch and the scaling "floor" was a mirage — reported by github.com, aggregated and ranked by ClawDigest.