paper-with-me

Papers

Communication-Efficient Adaptive Batch Size Strategies for Distributed Local Gradient Methods

2024-06-20 · Tim Tsz-Kit Lau, Weijian Li, Chenwei Xu, Han Liu, Mladen Kolar

Modern deep neural networks often require distributed training with many workers due to their large size. As the number of workers increases, communication overheads become the main bottleneck in data-parallel minibatch stochastic gradient methods with per-iteration gradient synchronization. Local gradient methods like Local SGD reduce communication by only synchronizing model parameters and/or gradients after several local steps. Despite an understanding of their convergence and the importance of batch sizes for training efficiency and generalization, optimal batch sizes for local gradient methods are difficult to determine. We introduce adaptive batch size strategies for local gradient methods that increase batch sizes adaptively to reduce minibatch gradient variance. We provide convergence guarantees under homogeneous data conditions and support our claims with image classification and language modeling experiments, demonstrating the effectiveness of our strategies for both training efficiency and generalization.

📄 PDF Abstract BibTeX arXiv:2406.13936

Code (0)

등록된 구현이 없습니다.

Tasks

image-classificationImage ClassificationLanguage ModelingLanguage Modelling

Methods 이 논문이 사용한 방법론

Local SGD Local SGD is a distributed training technique that runs SGD independently in parallel on different workers and averages the sequences…
SGD Stochastic Gradient Descent is an iterative optimization technique that uses minibatches of data to form an expectation of the gradient, rather than the full gradient using…

Similar Papers 제목 키워드 기반

1-bit LAMB: Communication Efficient Large-Scale Large-Batch Training with LAMB's Convergence Speed

2021-04-13 · Conglong Li, Ammar Ahmad Awan, Hanlin Tang, Samyam Rajbhandari 외

To train large models (like BERT and GPT-3) on hundreds of GPUs, communication has become a major bottleneck, especially on commodity systems with limited-bandwidth TCP network. On one side large batch-size optimization …

8k

AdLoCo: adaptive batching significantly improves communications efficiency and convergence for Large Language Models

2025-08-25 · Nikolay Kutuzov, Makar Baderko, Stepan Kulibaba, Artem Dzhalilov 외 arxiv

Scaling distributed training of Large Language Models (LLMs) requires not only algorithmic advances but also efficient utilization of heterogeneous hardware resources. While existing methods such as DiLoCo have demonstra…

Communication trade-offs for synchronized distributed SGD with large step size

2019-04-25 · Kumar Kshitij Patel, Aymeric Dieuleveut

Synchronous mini-batch SGD is state-of-the-art for large-scale distributed machine learning. However, in practice, its convergence is bottlenecked by slow communication rounds between worker nodes. A natural solution to …

Adaptive Learning of the Optimal Batch Size of SGD

2020-05-03 · Motasem Alfarra, Slavomir Hanzely, Alyazeed Albasyoni, Bernard Ghanem 외

Recent advances in the theoretical understanding of SGD led to a formula for the optimal batch size minimizing the number of effective data passes, i.e., the number of iterations times the batch size. However, this formu…

Adaptive Batch Size Schedules for Distributed Training of Language Models with Data and Model Parallelism

2024-12-30 · Tim Tsz-Kit Lau, Weijian Li, Chenwei Xu, Han Liu 외

An appropriate choice of batch sizes in large-scale model training is crucial, yet it involves an intrinsic yet inevitable dilemma: large-batch training improves training efficiency in terms of memory utilization, while …