Diffusion Model Meetup & Paper Reading — Diffusion Transformers & Scalable Diffusion Models with Transformers, with Paper and Code
TL;DR
In this session, we’ll walk through one of the landmark papers connecting transformers and diffusion models — “Scalable Diffusion Models with Transformers” (Diffusion Transformer / DiT).
We’ll unpack why replacing U‑Nets with transformers in diffusion models is a big deal, how the DiT architecture is designed, and what this means for scaling high‑quality image generation.
This session is part of our ongoing Diffusion Model Paper Reading Group, a friendly, online community across NY, SF, Toronto, and Boston — open to anyone curious about modern generative models.
👌 Learning requirements
You’ll be fine as long as you’re:
Curious about how diffusion models and transformers can be combined
Comfortable skimming a technical paper and engaging in discussion
Open to learning visually and conceptually (light math, optional code intuition, no coding required during the session)
🗓 Schedule
First 60 min: High‑level walkthrough of the “Scalable Diffusion Models with Transformers” paper and core DiT architecture, with Code.
Second 30 min: Open group discussion and Q&A to clarify concepts and connect DiTs to earlier diffusion and transformer sessions.
📚 Pre‑class learning
Paper: Scalable Diffusion Models with Transformers (DiT) — arXiv:2212.09748
https://arxiv.org/pdf/2212.09748Recommended video (concise, < 30 min):
Scalable Diffusion Models with Transformers — DiT Explanation and Implementation by hu‑po
https://www.youtube.com/watch?v=aSLDXdc2hkk
Clear, focused walkthrough of the DiT paper and architecture, ideal if you want a streamlined explanation.Additional video (~1 hour+ deep dive):
Stanford CS25: Transformers in Diffusion Models for Image Generation and Beyond by Sayak Paul (Hugging Face)
https://www.youtube.com/watch?v=vXtapCFctTI
Comprehensive lecture from Stanford’s CS25 series that connects U‑Nets, diffusion models, and Diffusion Transformers in a broader research context.
👥 Speakers
Led by master’s and PhD students in AI, IBM‑style applied AI consultants, and CTO‑level builders from award‑winning AI startups — all experienced in explaining diffusion models and transformer architectures clearly to practitioners.
The highlight of this session is clarity — by the end, you’ll understand how Diffusion Transformers are structured, why they scale so well, and how they fit into the broader generative model landscape.
🧠 About the Diffusion Model Reading Group
A peer‑led, 5‑month learning journey for engineers, students, researchers, and builders exploring diffusion architectures and modern AI.
No ML background required — just curiosity and some comfort with technical ideas
2–4 hours/week with paper readings, discussions, and final projects
Supportive community of people already working in or entering the AI industry