Exploring PyTorch and Open-Source Communities with Soumith Chintala, VP/Fellow of Meta, Co-Creator of PyTorch
Gradient Dissent: Conversations on AI
00:00
The Philosophical Direction of Jaxie and Tiny Grad
Jaxie is trying to be a purely functional graph acquisition framework that uses its purely functional nature to write elegant functional transforms. Ggml and TinyGrad are taking a full vertical integration approach like, hey, if all we had to do was build a framework for Lava, then how can we take that idea to its extreme and see what comes out of it? I pretty much care about competition, especially friendly competition where I think it is strictly always helpful,. Not just to user land, not just to consumers but to yourself because it lets us push ourselves.
Transcript
Play full episode
