Best use of jax.jit

Excited to share Penzai, a JAX research toolkit from Google DeepMind for building, editing, and visualizing neural networks! Penzai makes it easy to see model internals and lets you inject custom logic anywhere.

Check it out on GitHub: github.com/google-deepmin…

Core pytorch NNs (forward and backward) written in clean, readable Triton (from @bobmcdear)
github.com/BobMcDear/atto…

I'm thrilled to join Princeton's faculty as an assistant professor in the ECE department starting Fall 2025 🐯

Stay tuned for the launch of my lab. We will develop generally helpful robots that learn and plan 🤖

my version would have stuff like: 'empathy', 'the first law of robotics', 'libertarian ideals' in the outer bubble - really show the reader what's at stake

Sasha Rush Somehow the disentangled arch also makes the gradients cleaner and 'interpretable'

today was an intense day of research. running commands, writing code, checking numbers, being confused, looking at notes, running more commands, taking a nap, literally going 🤔

then at the end of the day i figured out one (1) Very Important Thing. so today counts as a good day

✨Excited to finally drop our new paper: SSMs “look like” RNNs, but we show their statefulness is an illusion🪄🐇

Current SSMs cannot express basic state tracking, but a minimal change fixes this! 👀

w/ jackson petty, Ashish Sabharwal
arxiv.org/abs/2404.08819

wow this must feel good when training

(from Megalodon arxiv.org/abs/2404.08801 )