Ionel Gog (@ICGog) profile picture

Ionel Gog

@ICGog

Machine learning infrastructure @GoogleDeepMind. Previously Postdoc @ucberkeley and PhD @Cambridge_CL.

 San Francisco, CA
ionelgog.org
Joined January 2010
  • Tweets215
  • Following155
  • Followers379
6Media
Photo by Ionel Gog (@ICGog)Photo by Ionel Gog (@ICGog)Photo by Ionel Gog (@ICGog)Photo by Ionel Gog (@ICGog)Photo by Ionel Gog (@ICGog)
The DiLoCo team at Google DeepMind and Google Research is proud to release Decoupled DiLoCo, the next frontier for resilient AI pre-training. Decoupled DiLoCo enables training with datacenters across the world, using heterogeneous hardware, and never halting the system despite hardware failures.
35
87
33
615
2,729,664
I'm super excited to release DiPaCo, a new kind of mixture of experts, that can scale engineering-wise to data centers across the entire world! A few words about it in this thread 🧵
AK (@_akhaliq) profile pictureAK
@_akhaliq19 Mar 2024
Google presents DiPaCo Distributed Path Composition Progress in machine learning (ML) has been fueled by scaling neural network models. This scaling has been enabled by ever more heroic feats of engineering, necessary for accommodating ML approaches that require high
Image from AK (@_akhaliq)
12
49
14
288
108,920
Ionel Gog retweeted
Google AI (@GoogleAI) profile picture
Google AI
@GoogleAI16 Dec 2022
Presenting Confident Adaptive Language Modeling (CALM), a novel method that allows language models to dynamically modify computational effort when generating text. Learn how CALM can accelerate text generation while preserving output quality → goo.gle/3HJKzbM
9
108
17
520
90,444
Excited to announce that @sukritkalra and @pschafhalter will be presenting at @EuroSys_conf on Thursday our work on developing a new execution model (D3), and an execution engine (ERDOS) that enables autonomous vehicles to adapt and to avoid collisions.
1
4
21
Attend their talk to learn about how we validated our execution engine by building an open-source self-driving car stack, which we deployed both on simulated and real-world cars. Joint work with the above, @mejoeyg and Ion Stoica.
1
1
2
Paper: dl.acm.org/doi/abs/10.1145/3… Execution Engine Code: github.com/erdos-project/erd… Self-driving Car Stack Code: github.com/erdos-project/pyl…
D3 | Proceedings of the Seventeenth European Conference on Computer Systems

D3 | Proceedings of the Seventeenth European Conference on Computer Systems

dl.acm.org
1
At SoCC today, I will be presenting our work (with @m_isard and Martin Abadi) on offering efficient fault-tolerance for complex dataflow applications that combine batch, stream, and iterative processing. #socc21 Paper: dl.acm.org/doi/abs/10.1145/3…
1
1
9
Our solution enables an application to pick the most appropriate fault-tolerance mechanism for each part of the application, and to combine the chosen fault-tolerance mechanisms such that the application recovers in a globally consistent state in the event of a failure.
Today in the Unmanned Vehicle award session of #ICRA2021, @sukritkalra will present our work on Pylot, an adaptable platform for developing autonomous vehicles. Our platform provides plug and play modules to deploy AVs in simulation and real-world.
1
2
7
The paper is available here: ionelgog.org/data/papers/202…, and the code is github.com/erdos-project/pyl… in case you want to take Pylot for a spin.
1
Working on self-driving cars? Check out our paper on Pylot, a modular pipeline for exploring latency-accuracy tradeoffs in self-driving cars (accepted to ICRA 2021). arxiv.org/abs/2104.07830
Peter Schafhalter (@pschafhalter) profile picturePeter Schafhalter@pschafhalter28 Feb 2021
First ever accepted paper 🥳🎉
1
5
3
27
Bought a mechanical keyboard with more resistance so my code will be strongly typed.
78
2,237
153
9,608
Slides for our #osdi18 talk on Noria, a new data-flow system that makes web applications fast, are now available: pdos.csail.mit.edu/papers/no… (joint work with @Jonhoo, @xexd and others). Will take no more than 17 minutes to view, promised!
Image from Malte Schwarzkopf (@ms705)
4
14
41
Interested in AI and ML? Consider applying to the first Transylvanian Machine Learning Summer School, which has a great line-up of lecturers and instructions: tmlss.ro
1
3
"Firmament: Fast, centralized cluster scheduling at scale" - Gog et al. 2016 blog.acolyer.org/2016/11/30/… #themorningpaper
30
3
41
. @mrry gave a great talk about TensorFlow and the challenges encountered while building it. #osdi16
Image from Ionel Gog (@ICGog)
1
3
Today, @ICGog presents our Firmament paper at @usenix #osdi16: usenix.org/system/files/conf… – at Savannah, GA
Image from Cambridge Sys@Scale (@CamSysAtScale)
2
8
Ionel Gog retweeted
Cristian Strat (@cgst) profile picture
Cristian Strat
@cgst29 Aug 2016
Bayesians against discrimination! Ban genetic algorithms! Free variables! 😂 #MLProtestSigns flickr.com/photos/30686429@N…
CMU G20 protests Pittsburgh

CMU G20 protests Pittsburgh

CMU machine learning dept members protest at G20. Update: we're on the Oct. 1 Daily Show! For about 3 seconds!

flickr.com
1
Q: Why did the deep net cross the road? A: We don't know. But look, it did it really well... #deeplearning – at Cambridge, MA
3
464
8
616
We'll soon write a blog post about it! teapaw.org/CamSysAtScale/status/7…
Cambridge Sys@Scale (@CamSysAtScale) profile pictureCambridge Sys@Scale@CamSysAtScale31 Jul 2016
Our paper on scaling Firmament's scheduling logic to very large clusters has been accepted at @usenix #OSDI16!
5
Load more