# SkipPipe: Partial and Reordered Pipelining Framework for Training LLMs in Heterogeneous Networks

arXiv: https://arxiv.org/abs/2502.19913 · HTML: https://mlsanity.com/paper/2502.19913/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Builds on 19 of its 30 references that are in the map.
- First posted on arXiv 27 Feb 2025.

## Builds on (most cited first, 19 of 19)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [LLaMA: Open and Efficient Foundation Language Models](https://mlsanity.com/paper/2302.13971/) (2023)
- [SentencePiece: A simple and language independent subword tokenizer and detokenizer for Neural Text Processing](https://mlsanity.com/paper/1808.06226/) (2018)
- [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://mlsanity.com/paper/1909.08053/) (2019)
- [Deep Networks with Stochastic Depth](https://mlsanity.com/paper/1603.09382/) (2016)
- [Reducing Transformer Depth on Demand with Structured Dropout](https://mlsanity.com/paper/1909.11556/) (2019)
- [LayerSkip: Enabling Early Exit Inference and Self-Speculative Decoding](https://mlsanity.com/paper/2404.16710/) (2024)
- [Understanding Robustness of Transformers for Image Classification](https://mlsanity.com/paper/2103.14586/) (2021)
- [DiLoCo: Distributed Low-Communication Training of Language Models](https://mlsanity.com/paper/2311.08105/) (2023)
- [PipeDream: Fast and Efficient Pipeline Parallel DNN Training](https://mlsanity.com/paper/1806.03377/) (2018)
- [On the Use of ArXiv as a Dataset](https://mlsanity.com/paper/1905.00075/) (2019)
- [Decentralized Training of Foundation Models in Heterogeneous Environments](https://mlsanity.com/paper/2206.01288/) (2022)
- [Low-Resource Domain Adaptation for Compositional Task-Oriented Semantic Parsing](https://mlsanity.com/paper/2010.03546/) (2020)
- [SWARM Parallelism: Training Large Models Can Be Surprisingly Communication-Efficient](https://mlsanity.com/paper/2301.11913/) (2023)
- [HetPipe: Enabling Large DNN Training on (Whimpy) Heterogeneous GPU Clusters through Integration of Pipelined Model Parallelism and Data Parallelism](https://mlsanity.com/paper/2005.14038/) (2020)
- [OpenDiLoCo: An Open-Source Framework for Globally Distributed Low-Communication Training](https://mlsanity.com/paper/2407.07852/) (2024)
- [Improving Continuous-time Conflict Based Search](https://mlsanity.com/paper/2101.09723/) (2021)
- [HexiScale: Facilitating Large Language Model Training over Heterogeneous Hardware](https://mlsanity.com/paper/2409.01143/) (2024)

## Built on by (most cited first, 4 of 4)

- [TAH-QUANT: Effective Activation Quantization in Pipeline Parallelism over Slow Network](https://mlsanity.com/paper/2506.01352/) (2025)
- [All is Not Lost: LLM Recovery without Checkpoints](https://mlsanity.com/paper/2506.15461/) (2025)
- [Pipeline Parallelism is All You Need for Optimized Early-Exit Based Self-Speculative Decoding](https://mlsanity.com/paper/2509.19368/) (2025)
- [Backdoor Attacks on Decentralised Post-Training](https://mlsanity.com/paper/2604.02372/) (2026)

## Abstract (arXiv)

Data and pipeline parallelism are ubiquitous for training of Large Language Models (LLM) on distributed nodes. Driven by the need for cost-effective training, recent work explores efficient communication arrangement for end to end training. Motivated by LLM's resistance to layer skipping and layer reordering, in this paper, we explore stage (several consecutive layers) skipping in pipeline training, and challenge the conventional practice of sequential pipeline execution. We derive convergence and throughput constraints (guidelines) for pipelining with skipping and swapping pipeline stages. Based on these constraints, we propose SkipPipe, the first partial pipeline framework to reduce the end-to-end training time for LLMs while preserving the convergence. The core of SkipPipe is a path scheduling algorithm that optimizes the paths for individual microbatches and reduces idle time (due to microbatch collisions) on the distributed nodes, complying with the given stage skipping ratio. We extensively evaluate SkipPipe on LLaMa models from 500M to 8B parameters on up to 20 nodes. Our results show that SkipPipe reduces training iteration time by up to 55% compared to full pipeline. Our partial pipeline training also improves resistance to layer omission during inference, experiencing a drop in perplexity of only 7% when running only half the model. Our code is available at https://github.com/gensyn-ai/skippipe.