"The critical IT power required for a 100k H100 cluster is ~150MW. " This paper on the computing scales and associated engineering challenges for training large AI systems is absolutely vertiginous semianalysis.com/p/100000-h100-…

Maybe I'm out of the loop here, but did Stability.ai already release SD3's weights ? Did they even announce a release ?

arxiv.org/pdf/2406.13661 A very nice survey on generative learning from a statistical physics perspective, credits to Davide Carbone !

arXiv -> alphaXiv Students at Stanford have built alphaXiv, an open discussion forum for arXiv papers. alphaXiv You can post questions and comments directly on top of any arXiv paper by changing arXiv to alphaXiv in any URL!

New blog post: sander.ai/2024/09/02/spe…

The continuity equation: the very fibre of generative modelling video: youtu.be/uK-apwLuEk8?si… I'm explaining the intuitive way to derive this equation. This is important not only for understanding generative modelling but also for using it in different contexts. Also, you can

Today is the day! 🎉 After 1.5 years of hard work, "Build A Large Language Model (From Scratch)" is finally out! Print or ebook copies are available on Manning’s site: mng.bz/M96o 📚. And it's also available on Amazon soon: amazon.com/Build-Large-La… 📦.

KL guarantees for Flow-Matching, by Silveri, Conforti and Durmus 👍 arxiv.org/pdf/2409.08311

It's always a pleasure to see these integrable systems papers with so many beautiful figures - here , the (2+1) Solid-on-Solid model, by wonderful Benoît Laslier and Eyal Lubetzki : arxiv.org/pdf/2409.08745

Book: "What is Entropy?" (by John C. Baez): arxiv.org/abs/2409.09232

😍😍

Dijkstra himself, on why numbering in array programming should always start at zero. Checkmate, The Julia Language 🫵 cs.utexas.edu/~EWD/transcrip…