Keith Rush's picture

1 2

Keith Rush

jkr26

·

https://www.jkrush.com

jkr26

AI & ML interests

Distributed ML, autodiff, scale and speed, and private ML

Recent Activity

authored a paper about 8 hours ago

Streaming DiLoCo with overlapping communication: Towards a Distributed Free Lunch

authored a paper about 8 hours ago

Communication-Efficient Language Model Training Scales Reliably and Robustly: Scaling Laws for DiLoCo

upvoted a paper 3 days ago

Communication-Efficient Language Model Training Scales Reliably and Robustly: Scaling Laws for DiLoCo

View all activity

Organizations

jkr26's activity

authored 2 papers about 8 hours ago

Streaming DiLoCo with overlapping communication: Towards a Distributed Free Lunch

Paper • 2501.18512 • Published Jan 30 • 28

Communication-Efficient Language Model Training Scales Reliably and Robustly: Scaling Laws for DiLoCo

Paper • 2503.09799 • Published 4 days ago • 11

upvoted a paper 3 days ago

Communication-Efficient Language Model Training Scales Reliably and Robustly: Scaling Laws for DiLoCo

Paper • 2503.09799 • Published 4 days ago • 11

commented a paper about 1 year ago

FAX: Scalable and Differentiable Federated Primitives in JAX

Paper • 2403.07128 • Published Mar 11, 2024 • 13 •

upvoted a paper about 1 year ago

FAX: Scalable and Differentiable Federated Primitives in JAX

Paper • 2403.07128 • Published Mar 11, 2024 • 13

authored 2 papers about 1 year ago

Multi-Epoch Matrix Factorization Mechanisms for Private Machine Learning

Paper • 2211.06530 • Published Nov 12, 2022

FAX: Scalable and Differentiable Federated Primitives in JAX

Paper • 2403.07128 • Published Mar 11, 2024 • 13