LLM scaling laws tutorial pairs intuition with hands-on learning
Its creator describes coverage of classic papers and recent developments, plus a toy associative-memory model meant to make scaling predictions more intuitive.
TLDR
Scaling laws are central to thinking about scaling large language models, but understanding why they work can be difficult, a tutorial creator says. They shared material they say they presented at MLSS Tübingen, describing a mix of research, a toy associative-memory model where scaling can be predicted from simple intuitions, and a hands-on environment for learning to use scaling laws.
Combined views
—
1 Source, first seen ago
