Aller au contenu

LLM Post-Training

Adapting pre-trained LLMs: supervised fine-tuning, parameter-efficient methods, preference optimization and alignment.

Topics I plan to document: SFT, LoRA, QLoRA, DoRA, preference optimization (DPO, KTO, ORPO), RLHF, reward models, alignment and evaluation.

Les premiers cours et notes de ce track arrivent bientôt.