This website uses cookies to anonymously analyze website traffic using Google Analytics.
Research

CocktailSGD: Fine-tuning foundation models over 500Mbps networks

April 24, 2023

By 

Jue Wang, Binhang Yuan, Luka Rimanic, Yongjun He, Tri Dao, Beidi Chen, Christopher Re, Ce Zhang

Distributed training of foundation models, especially large language models (LLMs), is communication-intensive and so has heavily relied on centralized data centers with fast interconnects. Can we train on slow networks and unlock the potential of decentralized infrastructure for foundation models? In this paper, we propose CocktailSGD, a novel communication-efficient training framework that combines three distinct compression techniques -- random sparsification, top-K sparsification, and quantization -- to achieve much greater compression than each individual technique alone. We justify the benefit of such a hybrid approach through a theoretical analysis of convergence. Empirically, we show that CocktailSGD achieves up to 117x compression in fine-tuning LLMs up to 20 billion parameters without hurting convergence. On a 500Mbps network, CocktailSGD only incurs ∼1.2x slowdown compared with data center networks.

  • Lower
    Cost
    20%
  • faster
    training
    4x
  • network
    compression
    117x

Start
building
yours
here →