Sebastian Bodenstein

150 posts

Sebastian Bodenstein

Sebastian Bodenstein

@sebbodenstein

Research Engineer @DeepMind, working on proteins. Ex-physicist.

London Katılım Ekim 2014
345 Takip Edilen542 Takipçiler
Sebastian Bodenstein
Sebastian Bodenstein@sebbodenstein·
@AlexLaterre @sarahookr Note that the the JAX graph does not allow for any conclusions to be drawn: JIT is not used and these are mostly tiny single ops completely dominated by Python/CPU overhead (different CPUs are used on the GPU/TPU machines). I've reported this issue to the authors.
English
0
0
1
126
Alex Laterre
Alex Laterre@AlexLaterre·
@sarahookr I'm amazed by this plot! JAX/XLA was designed to be compatible with both GPUs and TPUs, ensuring reliable performance across various accelerators. In contrast, PyTorch appears to adopt a different approach...
Alex Laterre tweet media
English
1
0
7
521
Alex Laterre
Alex Laterre@AlexLaterre·
Enjoyed this paper from @sarahookr team 👏 Key point -- JAX significantly outperforms PyTorch in hardware portability, which suffers 44% GPU to TPU failure rate. This echoes my first-hand experience and confirms our choice to adopt JAX early on, and its ongoing benefits 💪
Cohere Labs@Cohere_Labs

How portable are popular ML software frameworks? 🚚 Our recent cross-institutional collaboration reveals how costly straying from a narrow set of hardware-software combinations can be. 📜 arxiv.org/pdf/2309.07181…

English
4
8
70
23.3K
Sebastian Bodenstein retweetledi
Tim Green
Tim Green@tfgg2·
We have updated the AlphaFold-Multimer paper and code with a new set of protein complex models trained with an improved loss. These models reduce the number of steric clashes and improve accuracy. Code: dpmd.ai/af-multimer-os Paper: dpmd.ai/alphafold-mult…
Tim Green tweet media
English
4
141
536
0
Sebastian Bodenstein retweetledi
Google DeepMind
Google DeepMind@GoogleDeepMind·
The #AlphaFold source code has been updated and now accounts for multi-chain protein complexes - providing a significant improvement in accuracy for predicting protein interactions: dpmd.ai/af-multimer-os Generate predictions from your browser via: dpmd.ai/alphafold-colab
Google DeepMind tweet media
English
16
677
2K
0
Sebastian Bodenstein
Sebastian Bodenstein@sebbodenstein·
An amazing contribution to human knowledge, 350,000 protein structure predictions from AlphaFold 2 including for every protein in a human body.
Google DeepMind@GoogleDeepMind

Today with @emblebi, we're launching the #AlphaFold Protein Structure Database, which offers the most complete and accurate picture of the human proteome, doubling humanity’s accumulated knowledge of high-accuracy human protein structures - for free: dpmd.ai/alphafolddb 1/

English
0
0
8
0
Sebastian Bodenstein retweetledi
Demis Hassabis
Demis Hassabis@demishassabis·
Brief update on some exciting progress on #AlphaFold! We’ve been heads down working flat out on our full methods paper (currently under review) with accompanying open source code and on providing broad free access to AlphaFold for the scientific community. More very soon!
Demis Hassabis tweet media
English
68
660
3.5K
0
Sebastian Bodenstein retweetledi
Wessel
Wessel@ikwess·
@CambridgeMLG is launching a blog, featuring a first two-part post about what keeps a Bayesian awake at night by Richard E. Turner and me. 🧵 mlg-blog.com
English
2
36
171
0
Sebastian Bodenstein retweetledi
David Pfau
David Pfau@pfau·
If you follow me on Twitter, you probably know that I am pretty allergic to hype, especially around deep learning. So believe me when I say - this is a big f---ing deal. deepmind.com/blog/article/a…
English
13
169
1.1K
0
Sebastian Bodenstein retweetledi
Sundar Pichai
Sundar Pichai@sundarpichai·
.@DeepMind's incredible AI-powered protein folding breakthrough will help us better understand one of life’s fundamental building blocks + enable researchers to tackle new and hard problems, from fighting diseases to environmental sustainability. deepmind.com/blog/article/a…
English
93
802
5.1K
0
Thomas Wolf
Thomas Wolf@Thom_Wolf·
They are loosely connected to what I’m working on these days but these three books are still very clearly the most enjoyable read I’ve had since I joined the field. What a pleasure it was to read them!
Thomas Wolf tweet media
English
21
104
983
0
Sebastian Bodenstein
Sebastian Bodenstein@sebbodenstein·
@sharky6000 Even the thought of calling a Python-implemented non-trivial RL environment ~millions of times whilst I'm fighting with some obstreperous learning algorithm causes me deep psychic pain 😜
English
0
0
0
0