PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
June 26, 20241 citationsOpen Access

LoongTrain: Efficient Training of Long-Sequence LLMs with Head-Context Parallelism

View Full Paper
DGDiandian GuPSPeng SunQHQinghao Hu

Key Points

Key points are not available for this paper at this time.

Abstract

Efficiently training LLMs with long sequences is important yet challenged by the massive computation and memory requirements. Sequence parallelism has been proposed to tackle these problems, but existing methods suffer from scalability or efficiency issues. We propose LoongTrain, a novel system to efficiently train LLMs with long sequences at scale. The core of LoongTrain is the 2D-Attention mechanism, which combines both head-parallel and context-parallel techniques to break the scalability constraints while maintaining efficiency. We introduce Double-Ring-Attention and analyze the performance of device placement strategies to further speed up training. We implement LoongTrain with the hybrid ZeRO and Selective Checkpoint++ techniques. Experiment results show that LoongTrain outperforms state-of-the-art baselines, i.e., DeepSpeed-Ulysses and Megatron Context Parallelism, in both end-to-end training speed and scalability, and improves Model FLOPs Utilization (MFU) by up to 2.88x.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Gu et al. (2024) studied this question.

synapsesocial.com/papers/68e634cdb6db6435875c63b9https://doi.org/10.48550/arxiv.2406.18485
Ask AI
Helpful
Bookmark
Share
View Full Paper