返回全部动态
NVIDIA FLARE 构建联邦多模态 AI 工作流
原标题：Building Federated Multimodal AI Workflows with NVIDIA FLARE
AI 摘要
NVIDIA 技术博客介绍了如何使用 NVIDIA FLARE 构建联邦多模态 AI 工作流，重点讨论了跨网络传输的模型状态以及高效传输和聚合的方法。文章通过 FedUMM 示例展示了在冻结的多模态骨干上联邦轻量级适配器，并利用外部化、张量流和磁盘支持聚合来处理大型模型更新。该工作流支持参数高效和全模型通信模式，解决了数据分布在不同机构时的训练挑战。
以上摘要由 AI 生成，可能存在误差。事实请以原文为准。
正文节选
Modern vision-language models (VLMs) can support tasks such as visual question answering, captioning, and image-text reasoning. In practice, however, the data needed to adapt these models may be distributed across institutions or organizations that cannot centralize their raw records. Federated learning provides a way to coordinate training across these data-local sites. For VLMs, the challenge is not only orchestration. Sites may contribute different task or modality mixes, and model updates ca
发布时间：2026-08-20 01:50
抓取时间：2026-08-20 01:56
来源机构：NVIDIA