Papers

LoongForge-Embodied 发布:统一框架加速异构具身模型训练

LoongForge-Embodied: A Unified Framework for Training Heterogeneous Embodied Models

Why we picked it

训练 VLA 模型太贵、各家训练栈互不通用,这篇论文给了一套 PyTorch 原生解决方案,训练吞吐最高提 4.38 倍,评测快 23 倍,还带机器人数据合成流水线。做具身智能或机器人方向的同学可以直接看它的抽象层设计和 FP8 通信优化怎么落地。

论文提出 LoongForge-Embodied,一个原生 PyTorch 框架,用于统一训练 Vision-Language-Action(VLA)模型和 World-Action Models(WAM)这两类异构具身模型。框架核心是抽象层设计,把模型特定实现与共享训练基础设施解耦,使系统优化可以跨模型复用,而不约束各模型自身的架构和执行需求。

Body

在统一基础设施之上,框架在计算、执行、内存、数据 I/O 和分布式通信五个层面提供可复用的加速机制与硬件感知优化,其中包括面向 DDP 和 FSDP 的 FP8 通信优化。此外还集成了统一数据基础设施,支持从第一视角数据合成机器人数据的流水线,以及支持并发和模拟器计算优化的评测框架。

在代表性 VLA 和 WAM 模型上的实验显示,端到端训练吞吐量较各自的开源基线提升 1.47x 到 4.38x,相同协议下评测速度快 4.1x 到 23.2x。对从事具身智能训练的团队来说,这套框架可能减少在碎片化训练栈上的重复工程投入。

Original · arXiv cs.AI

LoongForge-Embodied: A Unified Framework for Training Heterogeneous Embodied Models

Training Vision-Language-Action (VLA) models and World-Action Models (WAMs) is computationally expensive, while their heterogeneous architectures and fragmented training stacks make system optimizations difficult to reuse across models. To address these challenges, we present LoongForge-Embodied, a native-PyTorch framework that unifies and accelerates the training of heterogeneous embodied models. At its core, LoongForge-Embodied introduces a common abstraction layer that decouples model-specific implementations from shared training infrastructure, enabling cross-model reuse without constraining individual architectures or execution requirements. Buildingon this unified infrastructure, the framework combines reusable acceleration mechanisms with model-specific and hardware-aware optimization strategies across computation, execution, memory, data I/O, and distributed communication, including FP8-based communication optimizations for DDP and FSDP. Beyond training acceleration, LoongForge-Embodied integrates a unified data infrastructure with a pipeline for synthesizing robot data from egocentric data, and an efficient evaluation framework that supports concurrency and optimized computation in simulator. Experiments across representative VLA and WAM models demonstrate 1.47x-4.38x end-to-end training throughput improvements over their corresponding open-source baselines, as well as 4.1x-23.2x faster evaluation under the same protocols.

Industry Coordinates & Source Verification

This report is categorized under AITOP's "Papers" desk, focusing on China's AI industry and domestic foundation model ecosystem. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows and US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include VLA, LoongForge-Embodied, embodied AI, PyTorch.

Source attribution: curated from arXiv cs.AI (@ LoongForge Team).