NestPipe: Large-Scale Recommendation Training on 1,500+ Accelerators via Nested Pipelining

• URL: http://arxiv.org/abs/2604.06956v1
• Date: Wed, 08 Apr 2026 11:19:17 GMT
• Title: NestPipe: Large-Scale Recommendation Training on 1,500+ Accelerators via Nested Pipelining
• Abstract summary: NestPipe is a large-scale decentralized embedding training framework.<n>It exploits two hierarchical sparse parallelism opportunities through nested pipelining.<n> Experiments on production GPU and NPU clusters with 1,536 workers demonstrate that NestPipe achieves up to 3.06x speedup and 94.07% scaling efficiency.
• Score: 9.69998937447445
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Modern recommendation models have increased to trillions of parameters. As cluster scales expand to O(1k), distributed training bottlenecks shift from computation and memory to data movement, especially lookup and communication latency associated with embeddings. Existing solutions either optimize only one bottleneck or improve throughput by sacrificing training consistency. This paper presents NestPipe, a large-scale decentralized embedding training framework that tackles both bottlenecks while preserving synchronous training semantics. NestPipe exploits two hierarchical sparse parallelism opportunities through nested pipelining. At the inter-batch level, Dual-Buffer Pipelining (DBP) constructs a staleness-free five-stage pipeline through dual-buffer synchronization, mitigating lookup bottlenecks without embedding staleness. At the intra-batch level, we identify the embedding freezing phenomenon, which inspires Frozen-Window Pipelining (FWP) to overlap All2All communication with dense computation via coordinated stream scheduling and key-centric sample clustering. Experiments on production GPU and NPU clusters with 1,536 workers demonstrate that NestPipe achieves up to 3.06x speedup and 94.07% scaling efficiency.

Related papers

• Mixtures of Subspaces for Bandwidth Efficient Context Parallel Training [54.43685601795611]
  We propose a compression method for communication-efficient context parallelism in decentralized settings.<n>Our key insight is to exploit the intrinsic low-rank structure of activation outputs by dynamically constraining them to learned mixtures of subspaces.
  arXiv  Detail & Related papers  (2026-06-15T08:17:13Z)
• Efficient Learned Data Compression via Dual-Stream Feature Decoupling [11.752785837648199]
  We propose a Dual-Stream Multi-Scale Decoupler that disentangles local and global contexts to replace deep serial processing with shallow parallel streams.<n>Our method achieves state-of-the-art performance in both compression ratio and throughput, while maintaining the lowest latency and memory usage.
  arXiv  Detail & Related papers  (2026-04-08T16:05:24Z)
• AsyncMesh: Fully Asynchronous Optimization for Data and Pipeline Parallelism [54.8494905524997]
  We introduce asynchronous updates across both parallelism axes, relaxing the co-location requirement.<n>We provide convergence guarantees for both sparse averaging and asynchronous updates.<n>Experiments on large-scale language models demonstrate that our approach matches the performance of the fully synchronous baseline.
  arXiv  Detail & Related papers  (2026-01-30T01:24:47Z)
• Eliminating Multi-GPU Performance Taxes: A Systems Approach to Efficient Distributed LLMs [61.953548065938385]
  We introduce the ''Three Taxes'' (Bulk Synchronous, Inter- Kernel Data Locality, and Kernel Launch Overhead) as an analytical framework.<n>We propose moving beyond the rigid BSP model to address key inefficiencies in distributed GPU execution.<n>We observe a 10-20% speedup in end-to-end latency over BSP-based approaches.
  arXiv  Detail & Related papers  (2025-11-04T01:15:44Z)
• Nesterov Method for Asynchronous Pipeline Parallel Optimization [59.79227116582264]
  We introduce a variant of Nesterov Accelerated Gradient (NAG) for asynchronous optimization in Pipeline Parallelism.<n>Specifically, we modify the look-ahead step in NAG to effectively address the staleness in gradients.<n>We theoretically prove that our approach converges at a sublinear rate in the presence of fixed delay in gradients.
  arXiv  Detail & Related papers  (2025-05-02T08:23:29Z)
• BitPipe: Bidirectional Interleaved Pipeline Parallelism for Accelerating Large Models Training [5.7294516069851475]
  BitPipe is a bidirectional interleaved pipeline parallelism for accelerating large models training. We show that BitPipe improves the training throughput of GPT-style and BERT-style models by 1.05x-1.28x compared to the state-of-the-art synchronous approaches.
  arXiv  Detail & Related papers  (2024-10-25T08:08:51Z)
• FusionLLM: A Decentralized LLM Training System on Geo-distributed GPUs with Adaptive Compression [55.992528247880685]
  Decentralized training faces significant challenges regarding system design and efficiency. We present FusionLLM, a decentralized training system designed and implemented for training large deep neural networks (DNNs) We show that our system and method can achieve 1.45 - 9.39x speedup compared to baseline methods while ensuring convergence.
  arXiv  Detail & Related papers  (2024-10-16T16:13:19Z)
• ReCycle: Resilient Training of Large DNNs using Pipeline Adaptation [2.0181279529015925]
  ReCycle is a system designed for efficient training in the presence of failures. It exploits the inherent functional redundancy in distributed training systems. We show it achieves high training throughput under multiple failures.
  arXiv  Detail & Related papers  (2024-05-22T21:35:56Z)
• Accelerating Training and Inference of Graph Neural Networks with Fast Sampling and Pipelining [58.10436813430554]
  Mini-batch training of graph neural networks (GNNs) requires a lot of computation and data movement. We argue in favor of performing mini-batch training with neighborhood sampling in a distributed multi-GPU environment. We present a sequence of improvements to mitigate these bottlenecks, including a performance-engineered neighborhood sampler. We also conduct an empirical analysis that supports the use of sampling for inference, showing that test accuracies are not materially compromised.
  arXiv  Detail & Related papers  (2021-10-16T02:41:35Z)
• Accelerating Distributed K-FAC with Smart Parallelism of Computing and Communication Tasks [13.552262050816616]
  Kronecker-Factored Approximate Curvature (KFAC) is one of the most efficient approximation algorithms for training deep models. Yet, when leveraging GPU clusters to train models with KFAC, it incurs extensive computation as well as introduces extra communications during each iteration. We propose D-KFAC with smart parallelism of computing and communication tasks to reduce the iteration time.
  arXiv  Detail & Related papers  (2021-07-14T08:01:07Z)
• Stochastic Optimization with Laggard Data Pipelines [65.20044914532221]
  We show that "dataechoed" extensions of common optimization methods exhibit provable improvements over their synchronous counterparts. Specifically, we show that in convex optimization with minibatches, data echoing affords speedups on the curvature-dominated part of the convergence rate, while maintaining the optimal statistical rate.
  arXiv  Detail & Related papers  (2020-10-26T14:55:31Z)
• Pipelined Training with Stale Weights of Deep Convolutional Neural Networks [0.1921787217122713]
  We explore the impact of stale weights on the statistical efficiency and performance in a pipelined backpropagation scheme. We show that when pipelining is limited to early layers in a network, training with stale weights converges and results in models with comparable inference accuracies. We propose combining pipelined and non-pipelined training in a hybrid scheme to address this drop.
  arXiv  Detail & Related papers  (2019-12-29T15:28:13Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.