Newly published today in ACS Central Science — Meta and Georgia Tech researchers released a dataset + SOTA AI models to help accelerate research on Direct Air Capture — a key technology to combat climate change ➡️ go.fb.me/eojy3k

Proud of this work coming out of a collaboration between FAIR researchers & Georgia Tech that was just published in ACS Central Science! The dataset & AI models that we released together could help meaningfully accelerate research combating climate change. open-dac.github.io

Super proud of our work at #ICLR2024! 🎉 If you're interested in large generalizable models for chemistry, come check out our posters and chat with the team. "From Molecules to Materials: Pre-training Large Generalizable Models for Atomic Property Prediction" - We show that

(May the Force be with you!) Here’s a look at the vector field learned by a normalizing flow for water's electron density (orbital-free DFT framework). author: Alexande collab: Ricky T. Q. Chen 📜:arxiv.org/abs/2404.08764 #StarWarsDay #MayThe4th #CompChem

And a third FAIR Chemistry paper at #ICLR2024 with Nate Gruver ! "Fine-Tuned Language Models Generate Stable Inorganic Materials as Text" -We fine-tune LLaMA2 models to predict 3d point clouds for crystals! Poster: Wed May 8. Thread here: x.com/zackulissi/sta…

🎉Super excited to have presented our poster for "From Molecules to Materials: Pre-training Large Generalizable Models for Atomic Property Prediction" at ICLR 2024! Check out our webpage for our code, interactive visualization, and more! nima.sh/jmp/

Happy to share that this work (w/ Yaron Lipman) was awarded an Outstanding Paper Honorable Mention at ICLR 2024. Come see me talk about "Flow Matching on General Geometries" tomorrow in the afternoon orals session!

We are excited to share CatTSunami: arxiv.org/abs/2405.02078! We show that a zero-shot application of machine learning models trained only on OC20 can calculate a transition state with 91% accuracy in less than a minute.

Check out the updated repo! 🚀

Two years ago we released 'No Language Left Behind', a first-of-its-kind project for neural machine translation directly across 200 languages. This has fuelled countless new innovations. Today the work was published in nature — congratulations Marta R. Costa-jussa + the NLLB team!

🚀We're excited to emerge from stealth and announce our $30M Seed financing round. A huge thank you to our incredible investor group including Hoxton Ventures, Basis Set, Lightspeed, Northzone, LocalGlobe, Touring Capital, Giant Ventures, FJ Labs, Zero Prime VC & Tiferes

Announcing our new model for materials! FlowMM... - Generates stable & novel materials efficiently - Predicts crystal structure accurately - Generalizes Riemannian Flow Matching to point clouds w/ periodic boundaries arxiv.org/abs/2406.04713 Ricky T. Q. Chen Anuroop Sriram Brandon Wood

We use Flow Matching for generating materials (infinitely repeating structures). Combining manifolds with invariance was harder than we thought! e.g., can't simply remove the "mean" (undefined) for translation-invariance. To be presented at ICML24. Reach out to Benjamin Kurt Miller for Qs.

So proud of Nikhila Ravi for being named a 2024 Young Engineer of the Year! She has done amazing work at FAIR, including our Segment Anything Model, and there's so much more to come! Congratulations!! raeng.org.uk/news/royal-aca…

🌊 GNNs for particle-based fluid simulations have been struggling to generate stable predictions over long rollouts. Not anymore with Neural SPH! Accepted at ICML '24 Joint work with Jonas A. Erbesdobler under the supervision of N. Adams and Johannes Brandstetter <3 1/8

⚡️ Excited to share that I am starting an AI+Education company called Eureka Labs. The announcement: --- We are Eureka Labs and we are building a new kind of school that is AI native. How can we approach an ideal experience for learning something new? For example, in the case

Excited to present FlowMM at ICML Conference today! 🤖 If you're in Vienna, stop by and explore our latest research in generative models for materials. 📍 Poster Location: Hall C 4-9 #113 ⏰ Time: 1:30-3:00 PM CEST

Why do 16k GPU jobs fail? The Llama3 paper has many cool details -- but notably, has a huge infrastructure section that covers how we parallelize, keep things reliable, etc. We hit an overall 90% effective-training-time. ai.meta.com/research/publi…

🌍 Interested in AI and its role in combating climate change? Come to the #ICML2024 Meta booth tomorrow from 10:30 AM to 12:30 PM CEST to discuss our Open Direct Air Capture project (OpenDAC). I’ll be filling in for Anuroop Sriram. Learn more: open-dac.github.io

Many have asked about tighter convergence settings for OC20. Here we showcase how a cancellation of errors results in similar adsorption energies and ML performance between the original OC20 and a more tightly converged variant.