2/11/2024
•
EN
How to Generate and Use Synthetic Data for Finetuning
Explores methods for generating synthetic data (distillation & self-improvement) to fine-tune LLMs for pretraining, instruction-tuning, and preference-tuning.