DUL β€” Deep Unsupervised Learning (UC-Berkeley)

Lecture notes from UC Berkeley’s Deep Unsupervised Learning course, covering language modeling, scaling laws, and modern architectures.

Child Notes

  • Lecture 8: Large Language Models β€” Language model abstraction, pretraining stages, tokenization, scaling laws (Chinchilla), compute cost estimation, and architectural comparisons between LSTMs and Transformers.