paper-with-me

Papers

AsyncMesh: Fully Asynchronous Optimization for Data and Pipeline Parallelism

2026-01-30 · Thalaiyasingam Ajanthan, Sameera Ramasinghe, Gil Avraham, Hadi Mohaghegh Dolatabadi, Chamin P Hewa Koneputugodage, Violetta Shevchenko, Yan Zuo, Alexander Long arxiv

Data and pipeline parallelism are key strategies for scaling neural network training across distributed devices, but their high communication cost necessitates co-located computing clusters with fast interconnects, limiting their scalability. We address this communication bottleneck by introducing asynchronous updates across both parallelism axes, relaxing the co-location requirement at the expense of introducing staleness between pipeline stages and data parallel replicas. To mitigate staleness, for pipeline parallelism, we adopt a weight look-ahead approach, and for data parallelism, we introduce an asynchronous sparse averaging method equipped with an exponential moving average based correction mechanism. We provide convergence guarantees for both sparse averaging and asynchronous updates. Experiments on large-scale language models (up to \em 1B parameters) demonstrate that our approach matches the performance of the fully synchronous baseline, while significantly reducing communication overhead.

📄 PDF Abstract BibTeX arXiv:2601.22442

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Breaking the Bubble: Asynchronous Pipeline Parallel Training with Bounded Weight Inconsistency

2026-06-05 · Itay Elam, Eliron Rahimi, Avi Mendelson, Chaim Baskin arxiv

Pipeline parallelism is essential for training large neural networks, but existing schedules trade off throughput, memory, and optimization consistency. Synchronous pipelines preserve forward/backward weight consistency …

Single-Rollout Asynchronous Optimization for Agentic Reinforcement Learning

2026-07-08 · Zhenyu Hou, Yujiang Li, Jie Tang, Yuxiao Dong arxiv

Reinforcement learning (RL) is becoming increasingly important for post-training large language models (LLMs). Previous RL pipelines for LLMs were mostly synchronous and batch-interleaved, which is inefficient for long-h…

Reinforcement Learning

Nesterov Method for Asynchronous Pipeline Parallel Optimization

2025-05-02 · Thalaiyasingam Ajanthan, Sameera Ramasinghe, Yan Zuo, Gil Avraham 외

Pipeline Parallelism (PP) enables large neural network training on small, interconnected devices by splitting the model into multiple stages. To maximize pipeline utilization, asynchronous optimization is appealing as it…

DecoderLanguage Modelling

Online AutoML: An adaptive AutoML framework for online learning

2022-01-24 · Bilge Celik, Prabhant Singh, Joaquin Vanschoren

Automated Machine Learning (AutoML) has been used successfully in settings where the learning task is assumed to be static. In many real-world scenarios, however, the data distribution will evolve over time, and it is ye…

AutoMLBIG-bench Machine Learning

AMDP: Asynchronous Multi-Directional Pipeline Parallelism for Large-Scale Models Training

2026-05-28 · Ling Chen, Houming Wu, Wenjie Yu arxiv

Pipeline parallelism is essential for large-scale model training, but existing asynchronous approaches often degrade convergence due to parameter mismatch between forward and backward passes. We propose Asynchronous Mult…