Stanford CME296 Diffusion & Large Vision Models | Spring 2026 | Lecture 6 - Model Training

Learn more details about this course: https://online.stanford.edu/courses/cme296-diffusion-and-large-vision-models To follow along with the course schedule and syllabus, visit: https://cme296.stanford.edu/syllabus/ Chapters: 00:00:00 Introduction 00:07:45 Training lifecycle overview 00:10:39 Loss parameterization 00:16:08 Timestep sampling 00:22:27 Logit normal distribution 00:25:44 Sampling shift for different resolutions 00:43:56 Representation alignment (REPA) 00:49:21 Pre-training 00:52:46 Continued training (CT) 00:54:02 Supervised finetuning (SFT) 00:54:39 Preference tuning 00:56:48 Reward feedback learning (ReFL) 01:00:16 Flow-GRPO 01:03:23 Diffusion-DPO 01:05:06 Prompt enhancement (PE) 01:10:00 DreamBooth, low-rank adaptation (LoRA) 01:16:35 Distillation 01:21:42 Progressive distillation (PD) 01:25:14 InstaFlow 01:30:18 Consistency models (CM) 01:34:23 Distribution matching distillation (DMD) 01:39:16 Adversarial diffusion distillation (ADD) For more information about Stanford’s graduate programs, visit: https://online.stanford.edu/graduate-education Afshine Amidi is an Adjunct Lecturer at Stanford University. Shervine Amidi is an Adjunct Lecturer at Stanford University. View the course playlist: https://www.youtube.com/playlist?list=PLoROMvodv4rNdy8rt2rZ4T2xM0OjADnfu