TL;DR
This paper introduces BAC, a caching method that accelerates diffusion policy inference in robotics by adaptively reusing features, achieving up to 3x speedup without retraining.
Contribution
BAC is a novel, training-free caching technique that adaptively updates features at the block level to speed up diffusion policy inference in robotic models.
Findings
BAC achieves up to 3x inference speedup.
It is compatible with existing transformer-based models.
Extensive experiments validate its effectiveness across benchmarks.
Abstract
Diffusion Policy has demonstrated strong visuomotor modeling capabilities, but its high computational cost renders it impractical for real-time robotic control. Despite huge redundancy across repetitive denoising steps, existing diffusion acceleration techniques fail to generalize to Diffusion Policy due to fundamental architectural and data divergences. In this paper, we propose lock-wise daptive aching (), a method to accelerate Diffusion Policy by caching intermediate action features. BAC achieves lossless action generation acceleration by adaptively updating and reusing cached features at the block level, based on a key observation that feature similarities exhibit non-uniform temporal dynamics and distinct block-specific patterns. To operationalize this insight, we first design an Adaptive Caching Scheduler to identify optimal…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Code & Models
Videos
