Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2507.01663 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908430371913728 |
|---|---|
| author | Han, Zhenyu You, Ansheng Wang, Haibo Luo, Kui Yang, Guang Shi, Wenqi Chen, Menglong Zhang, Sicheng Lan, Zeshun Deng, Chunshi Ji, Huazhong Liu, Wenjie Huang, Yu Zhang, Yixiang Pan, Chenyi Wang, Jing Huang, Xin Li, Chunsheng Wu, Jianping |
| author_facet | Han, Zhenyu You, Ansheng Wang, Haibo Luo, Kui Yang, Guang Shi, Wenqi Chen, Menglong Zhang, Sicheng Lan, Zeshun Deng, Chunshi Ji, Huazhong Liu, Wenjie Huang, Yu Zhang, Yixiang Pan, Chenyi Wang, Jing Huang, Xin Li, Chunsheng Wu, Jianping |
| contents | Reinforcement learning (RL) has become a pivotal technology in the post-training phase of large language models (LLMs). Traditional task-colocated RL frameworks suffer from significant scalability bottlenecks, while task-separated RL frameworks face challenges in complex dataflows and the corresponding resource idling and workload imbalance. Moreover, most existing frameworks are tightly coupled with LLM training or inference engines, making it difficult to support custom-designed engines. To address these challenges, we propose AsyncFlow, an asynchronous streaming RL framework for efficient post-training. Specifically, we introduce a distributed data storage and transfer module that provides a unified data management and fine-grained scheduling capability in a fully streamed manner. This architecture inherently facilitates automated pipeline overlapping among RL tasks and dynamic load balancing. Moreover, we propose a producer-consumer-based asynchronous workflow engineered to minimize computational idleness by strategically deferring parameter update process within staleness thresholds. Finally, the core capability of AsynFlow is architecturally decoupled from underlying training and inference engines and encapsulated by service-oriented user interfaces, offering a modular and customizable user experience. Extensive experiments demonstrate an average of 1.59 throughput improvement compared with state-of-the-art baseline. The presented architecture in this work provides actionable insights for next-generation RL training system designs. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2507_01663 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | AsyncFlow: An Asynchronous Streaming RL Framework for Efficient LLM Post-Training Han, Zhenyu You, Ansheng Wang, Haibo Luo, Kui Yang, Guang Shi, Wenqi Chen, Menglong Zhang, Sicheng Lan, Zeshun Deng, Chunshi Ji, Huazhong Liu, Wenjie Huang, Yu Zhang, Yixiang Pan, Chenyi Wang, Jing Huang, Xin Li, Chunsheng Wu, Jianping Machine Learning Artificial Intelligence Reinforcement learning (RL) has become a pivotal technology in the post-training phase of large language models (LLMs). Traditional task-colocated RL frameworks suffer from significant scalability bottlenecks, while task-separated RL frameworks face challenges in complex dataflows and the corresponding resource idling and workload imbalance. Moreover, most existing frameworks are tightly coupled with LLM training or inference engines, making it difficult to support custom-designed engines. To address these challenges, we propose AsyncFlow, an asynchronous streaming RL framework for efficient post-training. Specifically, we introduce a distributed data storage and transfer module that provides a unified data management and fine-grained scheduling capability in a fully streamed manner. This architecture inherently facilitates automated pipeline overlapping among RL tasks and dynamic load balancing. Moreover, we propose a producer-consumer-based asynchronous workflow engineered to minimize computational idleness by strategically deferring parameter update process within staleness thresholds. Finally, the core capability of AsynFlow is architecturally decoupled from underlying training and inference engines and encapsulated by service-oriented user interfaces, offering a modular and customizable user experience. Extensive experiments demonstrate an average of 1.59 throughput improvement compared with state-of-the-art baseline. The presented architecture in this work provides actionable insights for next-generation RL training system designs. |
| title | AsyncFlow: An Asynchronous Streaming RL Framework for Efficient LLM Post-Training |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2507.01663 |