Saved in:
Bibliographic Details
Main Authors: Zou, Haosheng, Lv, Xiaowei, Jia, Shousheng, Li, Lin, Gong, Xiaochun, Zhang, Xiangzheng
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2505.22296
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911198093508608
author Zou, Haosheng
Lv, Xiaowei
Jia, Shousheng
Li, Lin
Gong, Xiaochun
Zhang, Xiangzheng
author_facet Zou, Haosheng
Lv, Xiaowei
Jia, Shousheng
Li, Lin
Gong, Xiaochun
Zhang, Xiangzheng
contents Adding sequence parallelism into LLaMA-Factory, we open-sourced 360-LLaMA-Factory at https://github.com/Qihoo360/360-LLaMA-Factory. 360-LLaMA-Factory has received wide recognition and used in models such as Light-R1 arXiv:2503.10460, TinyR1 arXiv:2503.04872, Kaggle AIMO math models and also in large companies' training frameworks. This technical report delves deeper into the different sequence parallel modes behind 360-LLaMA-Factory and discusses our implementation insights.
format Preprint
id arxiv_https___arxiv_org_abs_2505_22296
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle 360-LLaMA-Factory: Plug & Play Sequence Parallelism for Long Post-Training
Zou, Haosheng
Lv, Xiaowei
Jia, Shousheng
Li, Lin
Gong, Xiaochun
Zhang, Xiangzheng
Computation and Language
Machine Learning
Adding sequence parallelism into LLaMA-Factory, we open-sourced 360-LLaMA-Factory at https://github.com/Qihoo360/360-LLaMA-Factory. 360-LLaMA-Factory has received wide recognition and used in models such as Light-R1 arXiv:2503.10460, TinyR1 arXiv:2503.04872, Kaggle AIMO math models and also in large companies' training frameworks. This technical report delves deeper into the different sequence parallel modes behind 360-LLaMA-Factory and discusses our implementation insights.
title 360-LLaMA-Factory: Plug & Play Sequence Parallelism for Long Post-Training
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2505.22296