LLM Training Pipeline

The chapters in this module, in reading order.

#	Chapter
00	LLM Training Lifecycle — The Five-Year-Old Version
01	Base Model Product Contract — fluency is not usefulness
02	Curriculum Data Mix — the model learns what it repeatedly reads
03	Next-Token Training Loop — the tiny contract repeated billions of times
04	Memory and Parallelism — training breaks where bytes multiply
05	PyTorch and Hugging Face Tooling — abstractions with escape hatches
06	SFT Behavior Copying — same loss, different scenes
07	Chat Protocol and Data Quality — behavior lives in tiny boundaries
08	Preferences, Reward Models, PPO, and DPO — choosing better answers without breaking the model
09	Lifecycle Decisions and Evals — choosing the next knob without fooling yourself
10	Honest Admission — the lifecycle works, but the knobs are not fully understood