Skip to content
All tags

#mup

1 posts

CS336 Lecture 11: Scaling Laws in Practice Must Scale Learning Rate and Batch Too

Lecture 11 reads public recipes from MiniCPM, DeepSeek, Qwen, and Llama 3: hold most architectural ratios fixed, sweep learning rate and batch at small scale, then choose model/data allocation with IsoFLOPs. μP helps, but normalization, optimizers, and weight decay can break transfer.