JoyNexus: Service-Oriented Multi-Tenant Post-Training for VLA Models

📄 arXiv: 2607.16074 📥 PDF

作者: Haoran Sun, Wentao Zhang, Junyang Hua, Hedan Yang, Yongjian Guo, Yifei Zhang, Xiaolong Xiang, Mingxi Luo, Jing Long, Chen Zhao, Chen Zhou, Wanting Xu, Qiming Yang, Hui Zhang, Song Wang, Xiaodong Bai, Shuai Di, Xu Chu, Xiaotie Deng, Yicheng Gong, Junwu Xiong

分类: cs.DC, cs.AI, cs.SE

发布日期: 2026-07-20


💡 一句话要点

提出JoyNexus以解决多租户VLA模型后训练效率问题

🎯 匹配领域: 支柱二:RL算法与架构 (RL & Architecture) 支柱九:具身大模型 (Embodied Foundation Models)

关键词: 多租户服务 后训练 VLA模型 资源调度 强化学习 API设计 组批处理

📋 核心要点

  1. 现有方法通常将计算资源独占分配给单一租户,导致基础设施适配负担和低效的资源利用。
  2. JoyNexus通过解耦不同服务并引入高层和低层API,支持多租户并发工作负载,提高了训练和评估的灵活性。
  3. 实验结果显示,JoyNexus在共享资源上通过跨租户调度减少了GPU时间,提高了服务利用率。

📝 摘要(中文)

Vision-Language-Action (VLA) 模型的后训练因模拟器多样性、机器人形态和任务目标的不同而显得尤为重要。现有计算服务通常将GPU和CPU资源独占分配给单一租户,虽然这种模式最大化了客户灵活性,但也给用户带来了基础设施适配的负担。此外,固定的卡时计费模式使得短期或突发工作负载对租户而言成本高昂,对服务提供者而言效率低下。为了解决这些挑战,本文提出了JoyNexus,一个统一的多租户VLA监督微调、强化学习和评估服务。JoyNexus通过API解耦训练模型服务、推理模型服务和环境服务,支持租户直接调用高层语义API进行训练、回滚和评估,或使用低层API和分配的端点组合自定义算法。多个租户可以并发提交工作负载,服务通过全局训练队列和推理队列进行调度,从而提高多租户训练效率。

🔬 方法详解

问题定义:本文旨在解决多租户VLA模型后训练中的资源独占和效率低下的问题。现有方法在处理短期或突发工作负载时,往往导致资源浪费和高昂的成本。

核心思路:JoyNexus通过解耦训练、推理和环境服务,允许多个租户共享资源,并通过API调用实现灵活的训练和评估。该设计旨在提高资源利用率和降低租户的基础设施适配负担。

技术框架:JoyNexus的整体架构包括训练模型服务、推理模型服务和环境服务,所有服务通过API进行访问,支持租户特定的插槽。系统通过全局训练队列和推理队列调度多个租户的工作负载,确保各租户的模块和记录相互隔离。

关键创新:JoyNexus的主要创新在于引入了多租户的共享训练机制和组批处理策略,允许不同数据模式的租户共享模型前缀,从而实现高效的前向传播。与现有方法相比,JoyNexus显著提高了多租户训练的效率。

关键设计:在设计中,JoyNexus采用了高层和低层API,允许租户根据需求选择合适的调用方式。此外,系统支持异构VLA数据模式的组批处理,优化了训练过程中的资源利用。具体的参数设置和损失函数设计尚未详细说明。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

实验结果表明,与传统的单租户执行相比,JoyNexus在共享资源上减少了整体GPU时间,并通过跨租户调度提高了服务利用率,具体性能提升幅度未在摘要中详细说明。

🎯 应用场景

JoyNexus的研究成果具有广泛的应用潜力,尤其在机器人学习、自动化控制和多模态交互等领域。通过提高多租户模型训练的效率,能够降低企业在AI模型开发和部署过程中的成本,促进更快速的技术迭代和应用落地。

📄 摘要(原文)

The post-training of Vision-Language-Action (VLA) models is essential due to the diversity of simulators, robot embodiments, and task objectives. Existing compute services, whether offered as direct accelerator rental or batch-workload submission, typically allocate an exclusive set of GPU and CPU resources to a single tenant. While this paradigm maximizes client flexibility, it burdens users with infrastructure adaptation, and the fixed card-hour accounting model renders short or bursty workloads both expensive for tenants and inefficient for the service provider. To address these challenges, we present JoyNexus, a unified service for multi-tenant VLA supervised fine-tuning, reinforcement learning, and evaluation. JoyNexus decouples the Training Model Service, Inference Model Service, and Environment Service, each accessed through APIs and backed by resident shared base models with tenant-specific slots. Tenants can directly invoke high-level semantic APIs for training, rollout, and evaluation, or compose custom algorithms using lower-level APIs and their assigned endpoints. Multiple tenants submit workloads concurrently; their action modules, optimizers, rollout records, and policy versions remain isolated, and the service is scheduled by the global Training Queue and Inference Queue. To further improve multi-tenant training efficiency, JoyNexus introduces group batching for heterogeneous VLA data schemas that share a compatible model-facing prefix, enabling a single shared backbone forward pass over grouped samples. Finally, we evaluate JoyNexus through workload simulation and a group-batching pipeline in a realistic embodied scenario. Results show that, compared with isolated single-tenant execution, JoyNexus reduces aggregate GPU time and improves service utilization via cross-tenant scheduling on shared resources.