ActTune: Action-Aware Precision and GPU Operating-Point Adaptation for Energy-Efficient Vision-Language-Action Inference

Key points

• Our approach builds on two observations: quantization sensitivity varies across action classes, model layers, and weights versus activations; and numerical precision changes the workload, shifting favorable GPU operating points.1
• We introduce ActTune, an action-aware framework that connects layer-wise precision allocation with workload-dependent GPU operating-point selection over requested frequency--power-cap pairs.1
• A lightweight decision tree learns its splits and leaf precision configurations directly from configuration action errors, then selects precision before each policy call.1
• On LIBERO, a benchmark for lifelong robot learning, ActTune improves mean task success by up to 2.3% relative to state of the art.1

Sources (1)

• [1]ActTune: Action-Aware Precision and GPU Operating-Point Adaptation for Energy-Efficient Vision-Language-Action Inference
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 6, 02:35 PM

  Vision-language-action (VLA) policies repeatedly invoke inference to control robots, making graphics processing unit (GPU) energy a recurring cost of task execution.

  Our approach builds on two observations: quantization sensitivity varies across action classes, model layers, and weights versus activations; and numerical precision changes the workload, shifting favorable GPU operating points.

Extractive summary: sentences quoted from the sources.