analiza aiqbench
vla.cpp establishes a critical infrastructure layer for deploying Vision-Language-Action models by eliminating PyTorch dependencies and enabling efficient inference across heterogeneous hardware like NVIDIA, Apple, and Intel chips. The runtime's ability to share model loading and serving logic while supporting specific action heads significantly lowers the barrier for integrating large multimodal models into resource-constrained robotic systems. This work addresses a key bottleneck in edge robotics, providing a standardized path to deploy complex AI agents on devices with limited memory and compute power.
vla.cpp: A Unified Inference Runtime for Vision-Language-Action Models
arXiv:2606.08094v2 Announce Type: replace-cross Abstract: Deploying vision--language--action (VLA) models on robots requires adapting model-specific inference pipelines to heterogeneous processors and limited onboard memory. We present vla.cpp, a unified C++ inference runtime for eleven VLA models, with no PyTorch dependency for model execution. The runtime shares model loading, tensor execution, and serving while retaining architecture-specific attention, conditioning, and action heads. Iterative policies reuse observation-dependent computation across solver steps, while regression policies predict actions directly. We evaluate task success on LIBERO-Object and profile supported configurations on NVIDIA, Apple, and Intel hardware. BitVLA completes 200/200 LIBERO-Object episodes on an 8,GB Jetson Orin Nano. A ternary tensor-core kernel accelerates its client inference by $4.0$--$4.6\times$ over the CUDA-core baseline on RTX 3060 and AGX Orin. A SmolVLA case study links positional-index precision to gripper commands and task success, showing why fixed-input numerical checks should accompany rollout evaluation. Deployments on UR10e and ALOHA demonstrate physical robot integration; delay and execution-horizon studies characterize synchronous chunked control. The results demonstrate a common deployment path across VLA architectures and hardware, with numerical validation and control settings guiding deployment alongside inference efficiency.
Title:vla.cpp: A Unified Inference Runtime for Vision-Language-Action Models
Abstract:Deploying vision--language--action (VLA) models on robots requires adapting model-specific inference pipelines to heterogeneous processors and limited onboard memory. We present this http URL, a unified C++ inference runtime for eleven VLA models, with no PyTorch dependency for model execution. The runtime shares model loading, tensor execution, and serving while retaining architecture-specific attention, conditioning, and action heads. Iterative policies reuse observation-dependent computation across solver steps, while regression policies predict actions directly. We evaluate task success on LIBERO-Object and profile supported configurations on NVIDIA, Apple, and Intel hardware. BitVLA completes 200/200 LIBERO-Object episodes on an 8,GB Jetson Orin Nano. A ternary tensor-core kernel accelerates its client inference by $4.0$--$4.6\times$ over the CUDA-core baseline on RTX 3060 and AGX Orin. A SmolVLA case study links positional-index precision to gripper commands and task success, showing why fixed-input numerical checks should accompany rollout evaluation. Deployments on UR10e and ALOHA demonstrate physical robot integration; delay and execution-horizon studies characterize synchronous chunked control. The results demonstrate a common deployment path across VLA architectures and hardware, with numerical validation and control settings guiding deployment alongside inference efficiency.
8 pages, 5 figures. Code available at this https URL
Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Systems and Control (eess.SY)
arXiv:2606.08094 [cs.RO]
(or arXiv:2606.08094v2 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2606.08094 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 298 cuvinte
Subiecte conexe
Vertical Advantage: Transforming Industries with Lakebase and Agentic AI
Databricks · 27 aug. 2026
NVIDIA GTC Taipei at COMPUTEX: Live Updates on What’s Next in AI
NVIDIA · 21 mai 2026
Vision-Language-Action Safety: Threats, Challenges, Evaluations, and Mechanisms
arXiv · 26 aug. 2026
NVIDIA Nemotron 3 Nano Omni Powers Multimodal Agent Reasoning in a Single Efficient Open Model
NVIDIA · 28 apr. 2026