arXiv AI

Accelerating Sharded Data Parallelism at Scale with Federated Learning

The paper proposes two hybrid algorithms, FL+FSDP and FL+HSDP, that combine sharded data parallelism with federated learning-style aggregations to reduce communication overhead in large-scale AI training. By partitioning GPUs into loosely‑coupled federation groups, the methods keep inter‑group traffic minimal while maintaining a bounded global batch size. Experiments on a Llama3.1 8B model trained on 512 A100 GPUs show up to 8.04× faster data processing and 4.48 lower evaluation perplexity compared to conventional sharded DP.

Hugging Face Trending Papers
5d ago

Accelerating Sharded Data Parallelism at Scale with Federated Learning

The paper proposes two hybrid algorithms, FL+FSDP and FL+HSDP, that combine sharded data parallelism with federated learning-style aggregations to reduce communication overhead in large-scale AI training. By forming loosely‑coupled federation groups, the methods keep inter‑group traffic minimal while maintaining a bounded global batch size. Experiments on a Llama3.1 8B model trained on 512 A100 GPUs show up to 8.04× faster data processing and 4.48 lower evaluation perplexity compared to traditional sharded DP approaches.

arXiv AI
5d ago

Efficiently Distributed Federated Learning

The paper introduces FastFederatedLearning (FFL), an open‑source federated learning framework written in C/C++ that prioritizes performance and customizability. FFL allows users to define arbitrary communication graphs between clients and servers, and it consistently outperforms Intel OpenFL with speedups of 2.5x to 3.69x across x86‑64, ARM‑v8, and RISC‑V platforms. Future plans include a Python wrapper and middleware for diverse communication backends, aiming to support dynamic, evolving federations as long‑term services.

By Gianluca Mittone, Robert Birke, Marco Aldinucci
arXiv Machine Learning
Jun 16

Photon: Federated LLM Pre-Training

arXiv:2411. 02908v2 Announce Type: replace Abstract: Scaling large language models (LLMs) demands extensive data and computing resources, which are traditionally constrained to data centers by the high-bandwidth requirements of distributed training.

By Lorenzo Sani, Alex Iacob, Zeyu Cao, Royson Lee, Bill Marino, Yan Gao, Dongqi Cai, Zexi Li, Wanru Zhao, Xinchi Qiu, Nicholas D. Lane
arXiv AI
Jun 10

Piper: A Programmable Distributed Training System

arXiv:2606. 11169v1 Announce Type: cross Abstract: Large-scale model training increasingly relies on composing multiple parallelism strategies, such as data, pipeline, and expert parallelism, together with memory-saving optimizations like ZeRO.

By Megan Frisella, Shubham Tiwari, Andy Ruan, Yi Pan, Parker Gustafson, Mat Jacob, Gilbert Bernstein, Stephanie Wang
arXiv Machine Learning
Sep 2

FlexP-SFT: A Flexible Aggregation-Free Framework for On-Device Personalized Split Federated Fine-Tuning of LLMs

FlexP-SFT introduces an aggregation-free framework for personalized split federated fine-tuning of large language models, eliminating the client-side aggregation step that traditionally causes communication bottlenecks and straggler issues. The method employs a layer‑flexible alignment strategy to balance personalization and generalization without global synchronization, and formulates split‑ratio selection as a resource‑aware discrete optimization problem. Experiments demonstrate that FlexP-SFT improves both accuracy and latency compared to baselines, achieving a superior resource‑accuracy trade‑off.

By Jiaxiang Geng, Tianjun Yuan, Pengchao Han, Ying Gao, Xianhao Chen, Bing Luo
arXiv AI
Aug 26

ShardMeter: Sharded and Geo-Distributed Training Without the Guesswork

ShardMeter is a lightweight analytical performance model that predicts end-to-end runtime for transformer-based workloads across sharded, distributed, and decentralized training setups. By taking a model’s characteristics and a target hardware topology as input, it estimates per-GPU and per-island throughput, training cost, total wall-clock time, and pinpoints performance bottlenecks. The model reveals diminishing-return regimes with increasing island size, quantifies compute- versus communication-bound scaling, evaluates hyperparameter trade-offs, and models cost-throughput for large-scale decentralized training, enabling rapid exploration of configuration space and near-optimal deployment plans.

By Tim Beringer (Technical University of Darmstadt), Patrick Diem (Technical University of Darmstadt), Felix Wolf (Technical University of Darmstadt), Arya Mazaheri (Technical University of Darmstadt, PanocularAI)