$\nabla$NABLA: Neighborhood Adaptive Block-Level Attention

TL;DR

NABLA introduces a neighborhood adaptive block-level attention mechanism that reduces computational complexity in video diffusion transformers, enabling faster training and inference without significant loss in quality.

Contribution

The paper presents NABLA, a novel attention mechanism that adapts to sparsity patterns, improving efficiency in video transformers without requiring custom operators.

Findings

Achieves up to 2.7x faster training and inference

Maintains comparable quantitative metrics and visual quality

Seamlessly integrates with PyTorch's Flex Attention

Abstract

Recent progress in transformer-based architectures has demonstrated remarkable success in video generation tasks. However, the quadratic complexity of full attention mechanisms remains a critical bottleneck, particularly for high-resolution and long-duration video sequences. In this paper, we propose NABLA, a novel Neighborhood Adaptive Block-Level Attention mechanism that dynamically adapts to sparsity patterns in video diffusion transformers (DiTs). By leveraging block-wise attention with adaptive sparsity-driven threshold, NABLA reduces computational overhead while preserving generative quality. Our method does not require custom low-level operator design and can be seamlessly integrated with PyTorch's Flex Attention operator. Experiments demonstrate that NABLA achieves up to 2.7x faster training and inference compared to baseline almost without compromising quantitative metrics (CLIP score, VBench score, human evaluation score) and visual quality drop. The code and model weights are available here: https://github.com/gen-ai-team/Wan2.1-NABLA