Sublime

The chunked sigmoid loss was quite fun to implement (see "Sigmoid Loss for Language Image Pre-Training" paper). By changing cross-entropy loss to sigmoid, we can now calculate total loss independently on each device without all-gather (mem + compute). https://t.co/FW92RSIM8c

What's one machine learning topic that you wish was taught better in courses/books?

I'm excited to introduce my AI Machine Learning Agent that built 32 ML models in 30 seconds. Today, I'll share with you how to automate building 100s of ML models with the AI ML Agent, which is available on GitHub. We'll create an ML Agent focusing on a Customer Churn Problem. I'll guide you... See more

Here's what @browsercompany's AI eng & ML teams are working on for @diabrowser right now: (This is a pitch to come work for us; info at end) 🤖 COMPUTER USE – we've built our own bespoke APIs on top of Chromium to optimize latency, accuracy, and cost of computer-using agents. Demo attached.... See more

Folks have asked me about my favorite tutorials on diffusion models. 1> https://t.co/lufwvZiDR8, Miika Aittala on EDM 2> https://t.co/e4g3H3AE5P, Jia-Bin Huang on Flow-Matching 3> https://t.co/NvXfhXougc, Yang Song on Score-based Diffusion... See more