Draft-and-Target Sampling for Video Generation Policy

TL;DR

This paper introduces Draft-and-Target Sampling, a training-free diffusion inference method that significantly speeds up video generation policies by using a self-play denoising approach, token chunking, and progressive acceptance strategies.

Contribution

It proposes a novel, training-free diffusion inference paradigm for efficient video generation, combining draft and target sampling with new speedup techniques.

Findings

Achieves up to 2.1x speedup in video generation

Improves efficiency of state-of-the-art methods with minimal success rate loss

Demonstrates effectiveness on three benchmark datasets

Abstract

Video generation models have been used as a robot policy to predict the future states of executing a task conditioned on task description and observation. Previous works ignore their high computational cost and long inference time. To address this challenge, we propose Draft-and-Target Sampling, a novel diffusion inference paradigm for video generation policy that is training-free and can improve inference efficiency. We introduce a self-play denoising approach by utilizing two complementary denoising trajectories in a single model, draft sampling takes large steps to generate a global trajectory in a fast manner and target sampling takes small steps to verify it. To further speedup generation, we introduce token chunking and progressive acceptance strategy to reduce redundant computation. Experiments on three benchmarks show that our method can achieve up to 2.1x speedup and improve the efficiency of current state-of-the-art methods with minimal compromise to the success rate. Our code is available.