MMInference: Accelerating Pre-filling for Long-Context VLMs via Modality-Aware Permutation Sparse Attention

• URL: http://arxiv.org/abs/2504.16083v1
• Date: Tue, 22 Apr 2025 17:59:51 GMT
• Title: MMInference: Accelerating Pre-filling for Long-Context VLMs via Modality-Aware Permutation Sparse Attention
• Authors: Yucheng Li, Huiqiang Jiang, Chengruidong Zhang, Qianhui Wu, Xufang Luo, Surin Ahn, Amir H. Abdi, Dongsheng Li, Jianfeng Gao, Yuqing Yang, Lili Qiu,
• Abstract summary: MMInference is a dynamic sparse attention method that accelerates the prefilling stage for long-context multi-modal inputs.<n>We show that MMInference accelerates the pre-filling stage by up to 8.3x at 1M tokens while maintaining accuracy.
• Score: 61.025422435235456
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: The integration of long-context capabilities with visual understanding unlocks unprecedented potential for Vision Language Models (VLMs). However, the quadratic attention complexity during the pre-filling phase remains a significant obstacle to real-world deployment. To overcome this limitation, we introduce MMInference (Multimodality Million tokens Inference), a dynamic sparse attention method that accelerates the prefilling stage for long-context multi-modal inputs. First, our analysis reveals that the temporal and spatial locality of video input leads to a unique sparse pattern, the Grid pattern. Simultaneously, VLMs exhibit markedly different sparse distributions across different modalities. We introduce a permutation-based method to leverage the unique Grid pattern and handle modality boundary issues. By offline search the optimal sparse patterns for each head, MMInference constructs the sparse distribution dynamically based on the input. We also provide optimized GPU kernels for efficient sparse computations. Notably, MMInference integrates seamlessly into existing VLM pipelines without any model modifications or fine-tuning. Experiments on multi-modal benchmarks-including Video QA, Captioning, VisionNIAH, and Mixed-Modality NIAH-with state-of-the-art long-context VLMs (LongVila, LlavaVideo, VideoChat-Flash, Qwen2.5-VL) show that MMInference accelerates the pre-filling stage by up to 8.3x at 1M tokens while maintaining accuracy. Our code is available at https://aka.ms/MMInference.

Related papers

• Token-Efficient Long Video Understanding for Multimodal LLMs [101.70681093383365]
  STORM is a novel architecture incorporating a dedicated temporal encoder between the image encoder and the Video-LLMs.<n>We show that STORM achieves state-of-the-art results across various long video understanding benchmarks.
  arXiv  Detail & Related papers  (2025-03-06T06:17:38Z)
• MInference 1.0: Accelerating Pre-filling for Long-Context LLMs via Dynamic Sparse Attention [36.49445805074941]
  MInference (Milliontokens Inference) is a sparse calculation method designed to accelerate pre-filling of long-sequence processing. We demonstrate that MInference effectively reduces inference latency by up to 10x for pre-filling on an A100, while maintaining accuracy.
  arXiv  Detail & Related papers  (2024-07-02T17:59:56Z)
• MULTIFLOW: Shifting Towards Task-Agnostic Vision-Language Pruning [28.254318215697527]
  Vision-Language models (VLMs) come with high computational costs due to their large number of parameters. Existing techniques for VLMs are task-specific, and thus require pruning the network from scratch for each new task of interest. We explore a new direction: Task-Agnostic Vision-Language Pruning (TA-language) We propose Multimodal FlowPruning (MULTIFLOW), a first, gradient-free, pruning framework for TA-language.
  arXiv  Detail & Related papers  (2024-04-08T15:51:21Z)
• CALF: Aligning LLMs for Time Series Forecasting via Cross-modal Fine-Tuning [59.88924847995279]
  We propose a novel Cross-Modal LLM Fine-Tuning (CALF) framework for MTSF.<n>To reduce the distribution discrepancy, we develop the cross-modal match module.<n>CALF establishes state-of-the-art performance for both long-term and short-term forecasting tasks.
  arXiv  Detail & Related papers  (2024-03-12T04:04:38Z)
• ProbVLM: Probabilistic Adapter for Frozen Vision-Language Models [69.50316788263433]
  We propose ProbVLM, a probabilistic adapter that estimates probability distributions for the embeddings of pre-trained vision-language models. We quantify the calibration of embedding uncertainties in retrieval tasks and show that ProbVLM outperforms other methods. We present a novel technique for visualizing the embedding distributions using a large-scale pre-trained latent diffusion model.
  arXiv  Detail & Related papers  (2023-07-01T18:16:06Z)
• Unmasked Teacher: Towards Training-Efficient Video Foundation Models [50.19560876891811]
  Video Foundation Models (VFMs) have received limited exploration due to high computational costs and data scarcity. This paper proposes a training-efficient method for temporal-sensitive VFMs that integrates the benefits of existing methods. Our model can handle various tasks including scene-related, temporal-related, and complex video-language understanding.
  arXiv  Detail & Related papers  (2023-03-28T15:39:28Z)
• MIST: Multi-modal Iterative Spatial-Temporal Transformer for Long-form Video Question Answering [73.61182342844639]
  We introduce a new model named Multi-modal Iterative Spatial-temporal Transformer (MIST) to better adapt pre-trained models for long-form VideoQA. MIST decomposes traditional dense spatial-temporal self-attention into cascaded segment and region selection modules. Visual concepts at different granularities are then processed efficiently through an attention module.
  arXiv  Detail & Related papers  (2022-12-19T15:05:40Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.