据京东云消息,近日,京东云联合顶尖学术机构,发表了题为《RL-VLA³: Reinforcement Learning VLA Accelerating via Full Asynchronism》的研究成果论文,首次提出并支持面向视觉-语言-动作(VLA)模型的全异步强化学习训练框架。在LIBERO基准测试中,该框架相比现有同步训练策略吞吐量提升高达59.25%,深度优化后更可提升126.67%,为大规模通用机器人智能的训练提供了全新的AI基础设施新范式。
特别声明:以上内容(如有图片或视频亦包括在内)为自媒体平台“网易号”用户上传并发布,本平台仅提供信息存储服务。
Notice: The content above (including the pictures and videos if any) is uploaded and posted by a user of NetEase Hao, which is a social media platform and only provides information storage services.