AI Practices AI实践 6h ago Updated 3h ago 更新于 3小时前 49

Build a Physical AI model factory with NVIDIA Cosmos 3 on SageMaker HyperPod 在 SageMaker HyperPod 上使用 NVIDIA Cosmos 3 构建物理 AI 模型工厂

NVIDIA Cosmos 3 is an omnimodal world foundation model using a Mixture-of-Transformers (MoT) design with per-layer joint attention between a reasoner and generator, treating video, image, action, and sound as a single token stream The architecture enables three distinct operational modes—forward-dynamics world model for synthetic video generation, inverse-dynamics action labeler, and deployable action policy—using the same transformer trunk A deliberate train-versus-inference asymmetry allows fu NVIDIA Cosmos 3 采用 Mixture-of-Transformers (MoT) 架构,通过单层联合注意力机制实现推理器与生成器的深度融合,支持视频、图像、动作和声音的统一多模态处理 模型采用训练-推理非对称设计:训练时运行完整去噪流程并解码视频,推理时仅执行少量去噪步骤并跳过视频解码,直接输出动作令牌 单一模型家族可运行三种模式:前向动力学世界模型(合成视频生成)、逆动力学动作标注器、可部署的动作策略,实现生成、后训练和评估的统一 在 Amazon SageMaker HyperPod 上构建 Physical AI 模型工厂,通过时间共享容量替代分阶段独立 GPU 池,以

70
Hot 热度
72
Quality 质量
68
Impact 影响力

Analysis 深度分析

TL;DR

  • NVIDIA Cosmos 3 is an omnimodal world foundation model using a Mixture-of-Transformers (MoT) design with per-layer joint attention between a reasoner and generator, treating video, image, action, and sound as a single token stream
  • The architecture enables three distinct operational modes—forward-dynamics world model for synthetic video generation, inverse-dynamics action labeler, and deployable action policy—using the same transformer trunk
  • A deliberate train-versus-inference asymmetry allows full denoising during training while skipping video decoding at inference, enabling efficient real-time robot control
  • The model integrates cleanly with Amazon SageMaker HyperPod and EKS, replacing the need for separate GPU pools per pipeline stage with a single persistent, time-shared node pool
  • GPU cost efficiency is reframed around "GPU goodput"—useful pipeline progress per reserved GPU-hour across the entire loop—rather than peak throughput of individual jobs

Why It Matters

This article demonstrates a practical blueprint for building a continuous Physical AI model factory, addressing a critical bottleneck in deploying robots and autonomous vehicles that require iterative synthetic data generation, post-training, and closed-loop evaluation. By unifying generation, perception, and policy into one model architecture, it eliminates the operational complexity and cost overhead of managing separate GPU infrastructure per pipeline stage—a significant shift for teams building production Physical AI systems.

Technical Details

  • Mixture-of-Transformers (MoT) architecture: Each layer contains a reasoner (autoregressive token prediction) and a generator (diffusion-based denoising for video, audio, and actions), joined by dual-stream attention at every layer—unlike common approaches that bolt a DiT onto a VLM and cross-attend only at the final output
  • Single shared token stream: All modalities feed into one sequence with an autoregressive (AR) zone (text and vision tokens the model reads) preceding a diffusion-model (DM) zone (video, audio, and action tokens it generates); images use a ViT encoder, pixels use a frozen Wan2.2 video VAE, and actions use a compact per-embodiment vector of pose deltas and grasp state
  • Train-inference asymmetry: During training, the full denoising schedule runs and video is decoded back to pixels as part of the loss; at inference on a robot, only a few denoise steps are executed and video decoding is skipped entirely—video latents remain internal to ground actions, but only action tokens are decoded into joint positions
  • Three action modes from one checkpoint: The mid-trained base model runs three jobs by changing which tokens start as noise; post-training then specializes a checkpoint to a single mode and control frequency
  • Deployment on SageMaker HyperPod with EKS: The article provides a complete end-to-end walkthrough including cluster setup, shared multi-terabyte storage layer, distributed post-training for three workloads, and a robot-policy stage trained on the public DROID dataset, with all manifests and configurations available in the awsome-distributed-ai GitHub repository

Industry Insight

  • The shift from per-stage GPU provisioning to time-shared capacity across a persistent node pool represents a fundamental operational model change for Physical AI teams—organizations should evaluate capacity commitment strategies (flexible training plans or reservations) across the entire pipeline loop rather than optimizing individual job throughput
  • The unified MoT architecture that natively supports generation, perception, and policy in one model family suggests a converging direction for Physical AI: future systems will likely favor integrated omni-models over modular stacks of specialized components, reducing integration complexity and enabling tighter closed-loop improvement cycles
  • The concept of "GPU goodput" as the primary cost metric—useful pipeline progress per reserved GPU-hour across the whole loop—should replace traditional throughput benchmarks when planning and budgeting Physical AI infrastructure, as idle capacity between pipeline iterations is an inherent and unavoidable cost of continuous operation

TL;DR

  • NVIDIA Cosmos 3 采用 Mixture-of-Transformers (MoT) 架构,通过单层联合注意力机制实现推理器与生成器的深度融合,支持视频、图像、动作和声音的统一多模态处理
  • 模型采用训练-推理非对称设计:训练时运行完整去噪流程并解码视频,推理时仅执行少量去噪步骤并跳过视频解码,直接输出动作令牌
  • 单一模型家族可运行三种模式:前向动力学世界模型(合成视频生成)、逆动力学动作标注器、可部署的动作策略,实现生成、后训练和评估的统一
  • 在 Amazon SageMaker HyperPod 上构建 Physical AI 模型工厂,通过时间共享容量替代分阶段独立 GPU 池,以 GPU goodput 作为核心成本指标
  • 提供完整的端到端实现,包括集群配置、多太字节共享存储层设置,以及在 DROID 数据集上的机器人策略后训练演示

为什么值得看

本文展示了如何将 NVIDIA Cosmos 3 的多模态世界模型与 AWS SageMaker HyperPod 基础设施深度集成,为 Physical AI 系统(机器人、自动驾驶)提供可扩展的持续训练流水线架构。对 AI 从业者而言,这提供了从模型设计到生产部署的完整参考实现,揭示了多模态统一架构在 Physical AI 领域的技术优势与工程实践路径。

技术解析

  • MoT 架构设计:Cosmos 3 采用 Mixture-of-Transformers 设计,每层包含推理器(预测下一个令牌)和生成器(去噪视频、音频和动作)两个专家模块,通过双流注意力机制实现逐层联合,而非仅在输出层进行交叉注意力。自回归(AR)子序列保持因果注意力,扩散模型(DM)子序列使用全注意力。
  • 统一令牌流与多模态处理:所有模态(视频、图像、动作、声音)汇入单一共享序列。图像理解由 ViT 处理,像素生成由冻结的 Wan2.2 视频 VAE 处理,动作由紧凑的每具身向量(姿态增量和抓取状态)表示,使同一模型可驱动自动驾驶车辆和机械臂。
  • 训练-推理非对称性:训练阶段运行完整去噪调度并将预测视频解码回像素以计算损失;部署阶段仅执行少量去噪步骤,跳过视频解码,内部产生的视频潜变量仅用于 grounding 动作,最终仅解码动作令牌至关节位置。
  • SageMaker HyperPod 集成:基于 Amazon EKS 构建持久化 GPU 节点池,通过时间共享容量替代分阶段独立资源池。关键成本指标为 GPU goodput(每预留 GPU 小时的流水线有效进展),而非单任务峰值吞吐量。
  • 三种动作模式与后训练:中间训练基础检查点通过改变噪声起始令牌运行三种任务;后训练阶段将检查点专门化至单一模式和控制频率。完整代码和配置托管在 awsome-distributed-ai GitHub 仓库。

行业启示

  • Physical AI 基础设施范式转变:从分阶段独立 GPU 池转向统一持久化容量池,通过时间共享提升资源利用率,以 goodput 而非峰值吞吐量作为成本优化核心指标,为机器人/自动驾驶系统的持续迭代提供经济可行的工程路径。
  • 多模态统一架构成为 Physical AI 新标准:Cosmos 3 的单一令牌流设计证明视频、图像、动作和声音可在同一 Transformer 主干中统一处理,减少系统复杂度并提升跨模态一致性,预示世界模型将向全模态融合方向演进。
  • 训练-推理非对称设计优化部署效率:通过牺牲推理阶段的计算完整性换取实时性,同时保持训练阶段的性能保障,为 Physical AI 系统在实际部署中的延迟-精度权衡提供可复用的架构模式。

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

Robotics 机器人 Autonomous Driving 自动驾驶 Training 训练 GPU GPU Deployment 部署