SlimPipe: Memory-Thrifty and Efficient Pipeline Parallelism for Long-Context LLM Training

• URL: http://arxiv.org/abs/2504.14519v1
• Date: Sun, 20 Apr 2025 07:33:33 GMT
• Title: SlimPipe: Memory-Thrifty and Efficient Pipeline Parallelism for Long-Context LLM Training
• Authors: Zhouyang Li, Yuliang Liu, Wei Zhang, Tailing Yuan, Bin Chen, Chengru Song, Di Zhang,
• Abstract summary: SlimPipe is a novel approach to fine-grained pipeline parallelism.<n>It reduces the accumulated activations from several microbatches to just one, which is split into several slices.<n>It achieves near-zero memory overhead and (2) minimal pipeline bubbles simultaneously.
• Score: 21.93724007255793
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Pipeline Parallelism (PP) serves as a crucial technique for training Large Language Models (LLMs), owing to its capability to alleviate memory pressure from model states with relatively low communication overhead. However, in long-context scenarios, existing pipeline parallelism methods fail to address the substantial activation memory pressure, primarily due to the peak memory consumption resulting from the accumulation of activations across multiple microbatches. Moreover, these approaches inevitably introduce considerable pipeline bubbles, further hindering efficiency. To tackle these challenges, we propose SlimPipe, a novel approach to fine-grained pipeline parallelism that employs uniform sequence slicing coupled with one-forward-one-backward (1F1B) schedule. It reduces the accumulated activations from several microbatches to just one, which is split into several slices. Although the slices are evenly partitioned, the computation cost is not equal across slices due to causal attention. We develop a sophisticated workload redistribution technique to address this load imbalance. SlimPipe achieves (1) near-zero memory overhead and (2) minimal pipeline bubbles simultaneously. The effectiveness of SlimPipe has been proven by thorough testing with diverse model architectures, context window sizes, and SlimPipe-specific configurations. For example, on the Llama 70B model, compared to state-of-the-art methods, SlimPipe significantly boosts the Model FLOPs Utilization (MFU) to up to $1.57\times$ for a context length of 512K. More notably, for a context length of 2048K, it maintains over 45% utilization on 256 NVIDIA Hopper 80GB GPUs, while other approaches either suffer significant performance drops or fail entirely due to memory constraints.

Related papers

• PipeOffload: Improving Scalability of Pipeline Parallelism with Memory Optimization [6.583624095434974]
  Pipeline parallelism (PP) is widely used for training large language models (LLMs) PP is often constrained by high activation memory consumption as the number of in-flight microbatches grows with the degree of PP. We focus on addressing this challenge by leveraging the under-explored memory offload strategy in PP.
  arXiv  Detail & Related papers  (2025-03-03T09:11:06Z)
• SkipPipe: Partial and Reordered Pipelining Framework for Training LLMs in Heterogeneous Networks [0.1638581561083717]
  SkipPipe is a partial pipeline framework to reduce the end-to-end training time for Large Language Models.<n>Our results show that SkipPipe reduces training time by up to $55%$ compared to full pipeline.
  arXiv  Detail & Related papers  (2025-02-27T09:34:23Z)
• 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)
• Breaking the Memory Barrier: Near Infinite Batch Size Scaling for Contrastive Loss [59.835032408496545]
  We propose a tile-based strategy that partitions the contrastive loss calculation into arbitrary small blocks. We also introduce a multi-level tiling strategy to leverage the hierarchical structure of distributed systems. Compared to SOTA memory-efficient solutions, it achieves a two-order-of-magnitude reduction in memory while maintaining comparable speed.
  arXiv  Detail & Related papers  (2024-10-22T17:59:30Z)
• EPS-MoE: Expert Pipeline Scheduler for Cost-Efficient MoE Inference [49.94169109038806]
  This paper introduces EPS-MoE, a novel expert pipeline scheduler for MoE that surpasses the existing parallelism schemes. Our results demonstrate at most 52.4% improvement in prefill throughput compared to existing parallel inference methods.
  arXiv  Detail & Related papers  (2024-10-16T05:17:49Z)
• LoRC: Low-Rank Compression for LLMs KV Cache with a Progressive Compression Strategy [59.1298692559785]
  Key-Value ( KV) cache is crucial component in serving transformer-based autoregressive large language models (LLMs) Existing approaches to mitigate this issue include: (1) efficient attention variants integrated in upcycling stages; (2) KV cache compression at test time; and (3) KV cache compression at test time. We propose a low-rank approximation of KV weight matrices, allowing plug-in integration with existing transformer-based LLMs without model retraining. Our method is designed to function without model tuning in upcycling stages or task-specific profiling in test stages.
  arXiv  Detail & Related papers  (2024-10-04T03:10:53Z)
• SHERL: Synthesizing High Accuracy and Efficient Memory for Resource-Limited Transfer Learning [63.93193829913252]
  We propose an innovative METL strategy called SHERL for resource-limited scenarios. In the early route, intermediate outputs are consolidated via an anti-redundancy operation. In the late route, utilizing minimal late pre-trained layers could alleviate the peak demand on memory overhead.
  arXiv  Detail & Related papers  (2024-07-10T10:22:35Z)
• Bypass Back-propagation: Optimization-based Structural Pruning for Large Language Models via Policy Gradient [57.9629676017527]
  We propose an optimization-based structural pruning on Large-Language Models. We learn the pruning masks in a probabilistic space directly by optimizing the loss of the pruned model. Our method operates for 2.7 hours with around 35GB memory for the 13B models on a single A100 GPU.
  arXiv  Detail & Related papers  (2024-06-15T09:31:03Z)
• Pipeline Parallelism with Controllable Memory [6.135123843073223]
  We show that almost all existing pipeline schedules are memory inefficient. We introduce a family of memory efficient building blocks with controllable activation memory. We can achieve almost zero pipeline bubbles while maintaining the same activation memory as 1F1B.
  arXiv  Detail & Related papers  (2024-05-24T08:54:36Z)
• Zero Bubble Pipeline Parallelism [6.7021820542657045]
  Experimental evaluations show that our method outperforms the 1F1B schedule up to 23% in throughput under a similar memory limit. We believe our results mark a major step forward in harnessing the true potential of pipeline parallelism.
  arXiv  Detail & Related papers  (2023-11-30T10:40:34Z)
• UniPT: Universal Parallel Tuning for Transfer Learning with Efficient Parameter and Memory [69.33445217944029]
  PETL is an effective strategy for adapting pre-trained models to downstream domains. Recent PETL works focus on the more valuable memory-efficient characteristic. We propose a new memory-efficient PETL strategy, Universal Parallel Tuning (UniPT)
  arXiv  Detail & Related papers  (2023-08-28T05:38:43Z)
• Pipeline MoE: A Flexible MoE Implementation with Pipeline Parallelism [91.9372563527801]
  Existing MoE models suffer from tremendous inner-node and inter-node communication overhead. We propose a novel MoE architecture called Pipeline MoE (PPMoE) to tackle them. PPMoE builds expert parallel incorporating with tensor parallel and replaces communication-intensive all-to-all dispatching and gathering.
  arXiv  Detail & Related papers  (2023-04-22T14:09:14Z)

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.