Scale ML

We are exploring a new domain in Scale ML this week! @jeremyWohlwend, will be presenting his exciting work on Boltz-1! Time: April 9th 4PM EST, sign up at scale-ml.org to join our mailing list for the zoom link.

This week we will be having a @YifeiZhou02 present his exciting work on: Self-improvement of LLM agents through Reinforcement Learning At Scale ⚡️ Time: Mar 26 12PM EST, sign up at scale-ml.org to join our mailing list for the zoom link.

We are excited to have Heyi Tang from @Kimi_Moonshot present: Design and Optimization of Large-Scale Inference Systems of kimi.ai🚀 Time: Mar 12 7:15PM EST, sign up at scale-ml.org to join our mailing list for the zoom link.

We are excited to have @SonglinYang4 present: Linear Attention and Beyond 🚀🚀🚀 Time: Mar 5, 4pm EST, sign up at scale-ml.org to join our mailing list for the zoom link.

This week we will be having @Yikang_Shen present his exciting work on efficient training of LLMs! Date: Wednesday Feb 26, 4pm EST To attend the talk virtually, sign up via scale-ml.org @MITIBMLab

This week we will be having @BoyuanChen0 present his exciting work on diffusion models! Diffusion Forcing: Next-token Prediction Meets Full-Sequence Diffusion Time: Feb 5, 3pm EST, sign up at scale-ml.org to join our mailing list for the zoom link.

We are back with @SimonXinDong from NVIDIA Research presenting Hymba, a state-of-the-art hybrid architecture that’s leading the way in the small language model revolution! Time: Jan 22, 4pm EST Sign up at scale-ml.org to join our mailing list for the zoom link

For our last seminar of the year we will end with Lucas Wilkinson from @neuralmagic presenting! Machete: a cutting-edge mixed-input GEMM GPU kernel targeting NVIDIA Hopper GPUs Time: Dec 4, 3pm EST Sign up via scale-ml.org to join our mailing list for the zoom link

Hello everyone, this week at 3pm EST Nov 20 (Wed) we will be having @Guangxuan_Xiao present his work about efficient/effective long sequence modeling! Sign up via scale-ml.org to join our mailing list and zoom access.

This week we have Yuka Ikarashi (@c20) present: Exocompilation for Specialized Hardware! Date/Time: Nov 13 3pm Sign up via scale-ml.org to join our mailing list for the zoom link

Excited to host @thecharlieblake and Constantin tomorrow to talk about their work u-μP! Sign up via scale-ml.org to join remotely via Zoom!

Tomorrow, we will have Moshik Hershcovitch present: ZipNN - A Lossless Compression Library tailored for AI models. Join via zoom by signing up on our website! scale-ml.org

After a mini break, we are back! Tomorrow we will have @EranMalach from Harvard giving a talk titled: Computational Benefits and Limitations of Transformers and State-Space Models For those interested in listening virtually, signup on our website!

Today we welcome @vyasnikhil96 to speak about his work on understanding the Shampoo Optimizer at 3pm! Sign up via our website scale-ml.org to join via zoom! Paper: arxiv.org/pdf/2406.17748…

Yesterday in @scaleml we implemented autograd, so today I'm implementing modular normalization. I'll try to code up the Newton-Schulz iteration for Linear.normalize if there's time. Remember you can sign up for the zoom on atlas-mlc.github.io if you have an academic email

Excited to continue building Modula in NumPy with @jxbz ! Hope you can join us in-person or through zoom