George E. Dahl's Avatar

George E. Dahl

@georgeedahl.bsky.social

Machine learning researcher @Google DeepMind. My opinions do not necessarily represent my employer. Prefer email over DMs. https://scholar.google.com/citations?hl=e&user=ghbWy-0AAAAJ https://www.cs.toronto.edu/~gdahl/

78 Followers  |  63 Following  |  1 Posts  |  Joined: 10.03.2025  |  1.3359

Latest posts by georgeedahl.bsky.social on Bluesky

Preview
GitHub - mlcommons/algorithmic-efficiency: MLCommons Algorithmic Efficiency is a benchmark and competition measuring neural network training speedups due to algorithmic improvements in both training a... MLCommons Algorithmic Efficiency is a benchmark and competition measuring neural network training speedups due to algorithmic improvements in both training algorithms and models. - mlcommons/algori...

We just released AlgoPerf v0.6! ๐ŸŽ‰
โœ… Rolling leaderboard
โœ… Lower compute costs
โœ… JAX jit migration
โœ… Bug fixes & flexible API
Coming soon: More contemporary baselines + an LM workloadโ€ฆ
github.com/mlcommons/al...

08.09.2025 13:29 โ€” ๐Ÿ‘ 3    ๐Ÿ” 2    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0
Research Engineer/Scientist, Training Algorithms Mountain View, California, US

My team is hiring a research scientist or engineer to work on methodological deep learning research! We study how to improve the deep learning "workflow" (developers.google.com/machine-lear...) with a special emphasis on training algorithms and recipes job-boards.greenhouse.io/deepmind/job...

18.07.2025 20:45 โ€” ๐Ÿ‘ 7    ๐Ÿ” 3    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0
ICLR 2025: Accelerating Neural Network Training (AlgoPerf)
YouTube video by Tรผbingen Machine Learning ICLR 2025: Accelerating Neural Network Training (AlgoPerf)

The explainer video: www.youtube.com/watch?v=_yX1...

03.04.2025 11:15 โ€” ๐Ÿ‘ 7    ๐Ÿ” 2    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

We're all about acceleration! ๐Ÿ˜‰
Watch @priya-kasimbeg.bsky.social & @fsschneider.bsky.social speedrun an explanation of the AlgoPerf benchmark, rules, and results all within a tight 5 minutes for our #ICLR2025 paper video on "Accelerating Neural Network Training". See you in Singapore!

03.04.2025 11:15 โ€” ๐Ÿ‘ 5    ๐Ÿ” 4    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0

Hi there! This account will post about the AlgoPerf benchmark and leaderboard updates for faster neural network training via better training algorithms. But let's start with what AlgoPerf is, what we have done so far, and how you can train neural nets ~30% faster.

14.03.2025 20:56 โ€” ๐Ÿ‘ 6    ๐Ÿ” 3    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 1
Post image

Making LLMs run efficiently can feel scary, but scaling isnโ€™t magic, itโ€™s math! We wanted to demystify the โ€œsystems viewโ€ of LLMs and wrote a little textbook called โ€œHow To Scale Your Modelโ€ which weโ€™re releasing today. 1/n

04.02.2025 18:54 โ€” ๐Ÿ‘ 95    ๐Ÿ” 28    ๐Ÿ’ฌ 3    ๐Ÿ“Œ 8

@georgeedahl is following 20 prominent accounts