OSP-Next: Efficient High-Quality Video Generation with Sparse Sequence Parallelism, HiF8 Quantization, and Reinforcement Learning

2026-05-27

AI Summary

This work proposes an efficient text-to-video generation framework addressing the quadratic computational complexity of full attention mechanisms in diffusion Transformers. The approach introduces several key innovations: Skiparse-2D sparse attention combined with Sparse Sequence Parallelism (SSP), reducing communication overhead by 75%; a hybrid full-sparse attention architecture compatible with FlashAttention; HiF8 8-bit quantization; and Mix-GRPO, a post-training reinforcement learning strategy. Evaluated on VBench, the method achieves a total score of 83.73%, delivers 1.64× faster single-GPU inference, and accelerates eight-GPU training by over 1.52×. On Ascend 950PR hardware, it attains up to 2.27× speedup with only a 0.4% performance drop, significantly enhancing cross-platform training and inference efficiency.

Abstract

Diffusion Transformers achieve strong video generation quality, but the quadratic cost of full attention limits efficiency. We introduce OSP-Next, an efficient text-to-video generation model that integrates sparse attention, parallelism, quantization, and reinforcement learning. OSP-Next uses a hybrid full-sparse attention architecture, where the sparse component is implemented with Skiparse-2D Attention. This fixed-pattern mechanism applies token-wise and group-wise sparse attention along spatial dimensions, leveraging locality while maintaining native compatibility with FlashAttention kernels. Based on the local equivalence of rearrangement in Skiparse-2D Attention, we further propose Sparse Sequence Parallelism (SSP), which partitions subsequences across ranks and switches sparse patterns through a single All-to-All communication. Compared with Ulysses Sequence Parallelism (SP), SSP provides a native parallel strategy for sparse attention and reduces communication volume by 75%. OSP-Next also incorporates HiF8 quantization to enable stable joint training with 8-bit quantization and sparse fine-tuning, and applies Mix-GRPO post-training to improve the performance of the sparse model. Experiments show that OSP-Next achieves a VBench total score of 83.73%, surpassing the Wan2.1 baseline. Under the 5-second 720P and 5-second 768P settings, OSP-Next achieves up to 1.64$\times$ single-GPU speedup and over 1.52$\times$ eight-GPU speedup on NVIDIA H200 GPUs. In addition, with only a 0.4% drop in VBench total score, OSP-Next-HiF8 achieves 1.69$\times$ and 2.27$\times$ speedups under the two settings on a single Ascend 950PR, demonstrating the efficiency and performance of OSP-Next across hardware platforms.

Problem

• video generation
• diffusion transformers
• attention efficiency
• computational cost
• scalability

Innovation

• Sparse Sequence Parallelism
• Skiparse-2D Attention
• HiF8 Quantization
• Mix-GRPO
• Diffusion Transformer