# PipeOffload: Improving Scalability of Pipeline Parallelism with Memory Optimization

arXiv: https://arxiv.org/abs/2503.01328 · HTML: https://mlsanity.com/paper/2503.01328/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Builds on 17 of its 24 references that are in the map.
- First posted on arXiv 3 Mar 2025.

## Builds on (most cited first, 17 of 17)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://mlsanity.com/paper/1909.08053/) (2019)
- [Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour](https://mlsanity.com/paper/1706.02677/) (2017)
- [ZeRO: Memory Optimizations Toward Training Trillion Parameter Models](https://mlsanity.com/paper/1910.02054/) (2019)
- [Training Deep Nets with Sublinear Memory Cost](https://mlsanity.com/paper/1604.06174/) (2016)
- [Efficient Large-Scale Language Model Training on GPU Clusters Using Megatron-LM](https://mlsanity.com/paper/2104.04473/) (2021)
- [Reducing Activation Recomputation in Large Transformer Models](https://mlsanity.com/paper/2205.05198/) (2022)
- [ZeRO-Infinity: Breaking the GPU Memory Wall for Extreme Scale Deep Learning](https://mlsanity.com/paper/2104.07857/) (2021)
- [PipeDream: Fast and Efficient Pipeline Parallel DNN Training](https://mlsanity.com/paper/1806.03377/) (2018)
- [DAPPLE: A Pipelined Data Parallel Approach for Training Large Models](https://mlsanity.com/paper/2007.01045/) (2020)
- [TeraPipe: Token-Level Pipeline Parallelism for Training Large-Scale Language Models](https://mlsanity.com/paper/2102.07988/) (2021)
- [Zero Bubble Pipeline Parallelism](https://mlsanity.com/paper/2401.10241/) (2023)
- [Hanayo: Harnessing Wave-like Pipeline Parallelism for Enhanced Large Model Training Efficiency](https://mlsanity.com/paper/2308.15762/) (2023)
- [Pipeline Parallelism with Controllable Memory](https://mlsanity.com/paper/2405.15362/) (2024)
- [Seq1F1B: Efficient Sequence-Level Pipeline Parallelism for Large Language Model Training](https://mlsanity.com/paper/2406.03488/) (2024)
- [Balancing Pipeline Parallelism with Vocabulary Parallelism](https://mlsanity.com/paper/2411.05288/) (2024)

## Built on by (most cited first, 5 of 5)

- [TokenSmith: Streamlining Data Editing, Search, and Inspection for Large-Scale Language Model Training and Interpretability](https://mlsanity.com/paper/2507.19419/) (2025)
- [OptPipe: Memory- and Scheduling-Optimized Pipeline Parallelism for LLM Training](https://mlsanity.com/paper/2510.05186/) (2025)
- [PipeMax: Enhancing Offline LLM Inference on Commodity GPU Servers](https://mlsanity.com/paper/2605.02189/) (2026)
- [GNMR: Runtime Stability Control for Low-Precision Large Language Model Training](https://mlsanity.com/paper/2606.00539/) (2026)
- [CoCoScale: Leveraging Layer-wise Scaling to Unlock the Potential of Online LLM Serving](https://mlsanity.com/paper/2607.04181/) (2026)

## Abstract (arXiv)

Pipeline parallelism (PP) is widely used for training large language models (LLMs), yet its scalability is often constrained by high activation memory consumption as the number of in-flight microbatches grows with the degree of PP. In this paper, we focus on addressing this challenge by leveraging the under-explored memory offload strategy in PP. With empirical study, we discover that in the majority of standard configurations, at least half, and potentially all, of the activations can be offloaded with negligible overhead. In the cases where full overload is not possible, we introduce a novel selective offload strategy that decreases peak activation memory in a better-than-linear manner. Furthermore, we integrate memory offload with other techniques to jointly consider overall throughput and memory limitation. Our experiments proves that the per-device activation memory effectively reduces with the total number of stages, making PP a stronger alternative than TP, offering up to a 19% acceleration with even lower memory consumption. The implementation is open-sourced at hrefhttps://github.com/sail-sg/zero-bubble-pipeline-parallelismthis url.