1 paper · 1 filter
Ruotong Sun, Ermin Wei
Optimizers that exploit the matrix structure of gradients are central to modern LLM pre-training, with two distinct frontiers: explicit Kronecker-factored preconditioning -- most r…