Skip to content
 
 

Latest commit

 

History

2 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

LowDiff: Efficient Frequent Checkpointing via Low-Cost Differential for High-Performance Distributed Training Systems

Abstract

This paper proposes LowDiff, an efficient frequent checkpointing framework that reuses compressed gradients, serving as differential checkpoints to reduce cost. Furthermore, LowDiff incorporates a batched gradient write optimization to efficiently persist these differentials to storage. It also dynamically tunes both the checkpoint frequency and the batching size to maximize the performance. Experiments on various workloads show that LowDiff can achieve checkpointing frequency up to once per iteration with less than 3.1% runtime overhead.

Architecture

Setup

Software library

  • Ubuntu 22.04
  • CUDA-12.4
  • NCCL-2.23.4
  • OpenMPI-4.0.5
  • Python-3.10.15
  • PyTorch-2.6.0
  • Deepspeed-0.16.4

Conda environment

conda create -n LowDiff python=3.10
conda activate LowDiff
pip install -r requirements.txt 

Quick start

To run CV jobs:

bash ./scripts/cv_lowdiff.sh

To run NLP jobs:

bash ./scripts/gpt_lowdiff.sh

Datasets

License

See LICENSE.

About

SC25-LowDiff Reproduction

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages