Anthony Costa

@anthonybcosta.bsky.social

Digital Biology @NVIDIA

Just released BioNeMo Framework v2.6! Check it out here: github.com/NVIDIA/bione... Particularly notable include support for AMPLIFY, with a 70% speed over the existing xformers-based attention backend (+ validation), alpha LoRA ESM, and a ton of fixes and improvements.

Release NVIDIA BioNeMo Framework v2.6 · NVIDIA/bionemo-framework

New Features Adds support for AMPLIFY doi:10.1101/2024.09.23.614603 pre-training and inference, offering a 70% speedup over the xformers-based attention backend with similar final perplexity value...

github.com

We at LilaSciences are excited to come out of stealth! Join us at #NeurIPS2024 for a happy hour co-hosted with Nvidia, featuring a discussion on AI in science with Max Welling and Chris Bishop. If you’ll be at NeurIPS this year, we’d love to see you! RSVP here: lu.ma/nt1c4vf4

Toward Scientific Superintelligence with Lila Sciences and NVIDIA · Luma

Join Lila Sciences at NeurIPS at an exclusive event cohosted with NVIDIA to explore the transformative potential of AI in scientific discovery. Featuring a…

lu.ma

So many great collaborators, friends, colleagues, and #techbio leaders will be here. Unfortunately I’m going to miss the conference this year, but come find all my awesome colleagues!

Recursion@recursionpharma.bsky.social · 2y ago

Planning to attend the Superbowl of ML conferences? #NeurIPS (@neuripsconf.bsky.social) is coming to Vancouver Dec. 10-15 & @valenceai.bsky.social will be on site presenting 10+ accepted papers & cohosting a “Night of Science” #TechBio social w/ us & NVIDIA on Thurs. 12/12 👉 RSVP: lu.ma/biikt7ox

I feel like I’ve put together a pretty informative network of folks in AI/ML for chem, bio, genomics, materials, physics, and science generally, at least on par with what I had on Twitter. Thanks to all of you for doing the hard work up front. Don’t understand the vibe yet, but it feels exciting!

GPUs are *fast* at aligning protein sequences (and profiles!). 178x faster than JackHMMER! In ColabFold, 23x faster end-to-end compared to AlphaFold2 reaching the same accuracy! You get immediate speedups for all methods leveraging MSAs, from DCA, to PoET, and even AlphaFold3!

Bild
Kieran Didi@kdidi.bsky.social · 2y ago

MSAs go brrr with MMseqs2-GPU! Super fun project, happy to work with and learn from a stellar team of engineers and scientists. Try it out and stay tuned! 📄 Preprint: www.biorxiv.org/content/10.1... 💾 Code: mmseqs.com 🗞️ Blog: developer.nvidia.com/blog/boost-a...

You all have done some wonderful things curating a community while I’ve been lurking. There may be some phantom me accounts lying around in ways I can’t figure out to access anymore, but this is me for real, the one from NVIDIA interested in all things computing and digital bio!