Model Sizes and Deployment
Joe discusses the importance of making models usable at scale, considering factors like compute envelope and memory capacity. Lukas questions the rationale behind model sizes, leading to insights on memory bandwidth limitations in mobile devices. Joe highlights the iterative learning process in determining optimal model sizes for different environments.In this clip
From this podcast
Gradient Dissent - A Machine Learning Podcast
Bridging AI & Science: The Impact of Machine Learning on Material Innovation with Joe Spisak of Meta
Related Questions
How are billion parameter models different from smaller ones?
How will large language models (LLMs) and AI change software engineering and the software development lifecycle (SDLC) as discussed in the episode Meta’s Joe Spisak on Llama 3.1 405B and the Democratization of Frontier Models | Training Data, and in the clip Scaling Innovations?