RecIS: Sparse to Dense, A Unified Training Framework for Recommendation Models
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866911175821754368 |
|---|---|
| author | Zong, Hua Zeng, Qingtao Zhou, Zhengxiong Han, Zhihua Yan, Zhensong Liu, Mingjie Sun, Hechen Liu, Jiawei Hu, Yiwen Wang, Qi Xian, YiHan Guo, Wenjie Xiang, Houyuan Zeng, Zhiyuan Sheng, Xiangrong Yan, Bencheng Hu, Nan Huang, Yuheng Lian, Jinqing Xu, Ziru Zhang, Yan Huang, Ju Yang, Siran Yi, Huimin Wang, Jiamang Wang, Pengjie Zhu, Han Wu, Jian Ou, Dan Xu, Jian Tang, Haihong Jiang, Yuning Zheng, Bo Qu, Lin |
| author_facet | Zong, Hua Zeng, Qingtao Zhou, Zhengxiong Han, Zhihua Yan, Zhensong Liu, Mingjie Sun, Hechen Liu, Jiawei Hu, Yiwen Wang, Qi Xian, YiHan Guo, Wenjie Xiang, Houyuan Zeng, Zhiyuan Sheng, Xiangrong Yan, Bencheng Hu, Nan Huang, Yuheng Lian, Jinqing Xu, Ziru Zhang, Yan Huang, Ju Yang, Siran Yi, Huimin Wang, Jiamang Wang, Pengjie Zhu, Han Wu, Jian Ou, Dan Xu, Jian Tang, Haihong Jiang, Yuning Zheng, Bo Qu, Lin |
| contents | In this paper, we propose RecIS, a unified Sparse-Dense training framework designed to achieve two primary goals: 1. Unified Framework To create a Unified sparse-dense training framework based on the PyTorch ecosystem that meets the training needs of industrial-grade recommendation models that integrated with large models. 2.System Optimization To optimize the sparse component, offering superior efficiency over the TensorFlow-based recommendation models. The dense component, meanwhile, leverages existing optimization technologies within the PyTorch ecosystem. Currently, RecIS is being used in Alibaba for numerous large-model enhanced recommendation training tasks, and some traditional sparse models have also begun training in it. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_20883 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | RecIS: Sparse to Dense, A Unified Training Framework for Recommendation Models Zong, Hua Zeng, Qingtao Zhou, Zhengxiong Han, Zhihua Yan, Zhensong Liu, Mingjie Sun, Hechen Liu, Jiawei Hu, Yiwen Wang, Qi Xian, YiHan Guo, Wenjie Xiang, Houyuan Zeng, Zhiyuan Sheng, Xiangrong Yan, Bencheng Hu, Nan Huang, Yuheng Lian, Jinqing Xu, Ziru Zhang, Yan Huang, Ju Yang, Siran Yi, Huimin Wang, Jiamang Wang, Pengjie Zhu, Han Wu, Jian Ou, Dan Xu, Jian Tang, Haihong Jiang, Yuning Zheng, Bo Qu, Lin Information Retrieval Distributed, Parallel, and Cluster Computing Machine Learning In this paper, we propose RecIS, a unified Sparse-Dense training framework designed to achieve two primary goals: 1. Unified Framework To create a Unified sparse-dense training framework based on the PyTorch ecosystem that meets the training needs of industrial-grade recommendation models that integrated with large models. 2.System Optimization To optimize the sparse component, offering superior efficiency over the TensorFlow-based recommendation models. The dense component, meanwhile, leverages existing optimization technologies within the PyTorch ecosystem. Currently, RecIS is being used in Alibaba for numerous large-model enhanced recommendation training tasks, and some traditional sparse models have also begun training in it. |
| title | RecIS: Sparse to Dense, A Unified Training Framework for Recommendation Models |
| topic | Information Retrieval Distributed, Parallel, and Cluster Computing Machine Learning |
| url | https://arxiv.org/abs/2509.20883 |