HN Explorer

Mougatine

↗ HN profile [ 76.5 ms ] full profile
Karma
296
Created
January 2, 2018 (8y ago)
Submissions
0
RS @ DeepMind.

Doing distributed stuff, such as DiLoCo and DiPaCo

  1. Fault Tolerant Llama training (pytorch.org)
    65 points 15
  2. MuLoCo: Muon is a practical inner optimizer for DiLoCo (arxiv.org)
    2 points 0
  3. OpenDiLoCo: Open-Source Framework for Distributed Low-Communication Training (arxiv.org)
    4 points 0
  4. Show HN: Deep Learning for Computer Vision course with colabs and Anki cards (arthurdouillard.com)
    12 points 2
  5. Continual Learning at CVPR 2020 (medium.com)
    1 points 0
  6. Operation Red Falcon (2015) (magazine.atavist.com)
    1 points 0
  7. Lifelong Learning for Deep Neural Networks (2019) (medium.com)
    2 points 0
  8. Seeing Is Not Necessarily Believing: Limitations of GANs for Data Augmentation (openreview.net)
    2 points 0
  9. Cars detection from satellite imagery with RetinaNet (medium.com)
    2 points 0
  10. Human or Company (twitter.com)
    1 points 0
  11. 3 Small but Powerful Convolutional Networks (towardsdatascience.com)
    4 points 0
  12. An Explanation of Densely Connected Convolutional Networks (arthurdouillard.com)
    1 points 0
  13. Amazon launches an Android app in India called “Internet” (techcrunch.com)
    1 points 0
  14. Summary of “Deep Learning Scaling Is Predictable, Empirically” (arthurdouillard.com)
    1 points 0
  15. How RoI Pooling and RPN Work in Faster-RCNN (arthurdouillard.com)
    2 points 0
  16. Selective Search Explained (arthurdouillard.com)
    1 points 0
  17. Why Some Clocks Have Been Running Slow in Europe (twitter.com)
    2 points 0
  18. Efficient Graph-Based Segmentation (arthurdouillard.com)
    2 points 0
  19. A Few Useful Things to Know About Machine Learning (arthurdouillard.com)
    1 points 0
  20. Job One for Quantum Computers: Boost Artificial Intelligence (quantamagazine.org)
    1 points 0
  21. The case for learned index structures (blog.acolyer.org)
    3 points 0
  22. Data-Intensive Systems for the Next 1000x (2016) (futuredata.stanford.edu)
    2 points 0
  23. The Morning Paper: CS Papers Explained Every Weekday (blog.acolyer.org)
    4 points 0
  24. Useful Mental Models (2016) (defmacro.org)
    586 points 74
  25. Shut Up and Calculate (arxiv.org)
    1 points 0