Accelerating Multimodal Large Language Models via Dynamic Visual-Token Exit and the Empirical Findings
- URL: http://arxiv.org/abs/2411.19628v1
- Date: Fri, 29 Nov 2024 11:24:23 GMT
- Title: Accelerating Multimodal Large Language Models via Dynamic Visual-Token Exit and the Empirical Findings
- Authors: Qiong Wu, Wenhao Lin, Weihao Ye, Yiyi Zhou, Xiaoshuai Sun, Rongrong Ji,
- Abstract summary: Excessive use of visual tokens in existing Multimoal Large Language Models (MLLMs) often exhibits obvious redundancy and brings in prohibitively expensive computation.
We propose a simple yet effective method to improve the efficiency of MLLMs, termed dynamic visual-token exit (DyVTE)
DyVTE uses lightweight hyper-networks to perceive the text token status and decide the removal of all visual tokens after a certain layer.
- Score: 69.35226485836641
- License:
- Abstract: The excessive use of visual tokens in existing Multimoal Large Language Models (MLLMs) often exhibits obvious redundancy and brings in prohibitively expensive computation. To gain insights into this problem, we first conduct extensive empirical studies on the attention behaviors of MLLMs, and summarize three main inference stages in MLLMs: (i) Early fusion between tokens is first accomplished quickly. (ii) Intra-modality modeling then comes to play. (iii) Multimodal reasoning} resumes and lasts until the end of inference. In particular, we reveal that visual tokens will stop contributing to reasoning when the text tokens receive enough image information, yielding obvious visual redundancy. Based on these generalized observations, we propose a simple yet effective method to improve the efficiency of MLLMs, termed dynamic visual-token exit (DyVTE). DyVTE uses lightweight hyper-networks to perceive the text token status and decide the removal of all visual tokens after a certain layer, thereby addressing the observed visual redundancy. To validate VTE, we apply it to a set of MLLMs, including LLaVA, VILA, Eagle and InternVL, and conduct extensive experiments on a bunch of benchmarks. The experiment results not only show the effectiveness of our VTE in improving MLLMs' efficiency, but also yield the general modeling patterns of MLLMs, well facilitating the in-depth understanding of MLLMs. Our code is anonymously released at https://github.com/DoubtedSteam/DyVTE.
Related papers
- Towards Text-Image Interleaved Retrieval [49.96332254241075]
We introduce the text-image interleaved retrieval (TIIR) task, where the query and document are interleaved text-image sequences.
We construct a TIIR benchmark based on naturally interleaved wikiHow tutorials, where a specific pipeline is designed to generate interleaved queries.
We propose a novel Matryoshka Multimodal Embedder (MME), which compresses the number of visual tokens at different granularity.
arXiv Detail & Related papers (2025-02-18T12:00:47Z) - FoPru: Focal Pruning for Efficient Large Vision-Language Models [11.36025001578531]
We propose Focal Pruning (FoPru), a training-free method that prunes visual tokens based on the attention-based token significance derived from the vision encoder.
Our method can prune a large number of redundant tokens while maintaining high accuracy, leading to significant improvements in inference efficiency.
arXiv Detail & Related papers (2024-11-21T14:22:38Z) - Treat Visual Tokens as Text? But Your MLLM Only Needs Fewer Efforts to See [37.7015406019386]
Multimodal Large Language Models (MLLMs) treat visual tokens from visual encoders as text tokens.
As token counts grow, the quadratic scaling of computation in LLMs introduces an efficiency bottleneck.
In this study, we investigate the redundancy in visual computation at both the parameter and computational pattern levels within LLaVA.
arXiv Detail & Related papers (2024-10-08T16:13:24Z) - ControlMLLM: Training-Free Visual Prompt Learning for Multimodal Large Language Models [73.34709921061928]
We propose a training-free method to inject visual prompts into Multimodal Large Language Models (MLLMs)
We optimize a learnable latent variable based on an energy function, enhancing the strength of referring regions in the attention map.
Our method offers a promising direction for integrating referring abilities into MLLMs, and supports referring with box, mask, scribble and point.
arXiv Detail & Related papers (2024-07-31T11:40:29Z) - TokenPacker: Efficient Visual Projector for Multimodal LLM [37.1071749188282]
The visual projector serves as an essential bridge between the visual encoder and the Large Language Model (LLM)
We propose a novel visual projector, which adopts a coarse-to-fine scheme to inject the enriched characteristics to generate the condensed visual tokens.
Our approach compresses the visual tokens by 75%89%, while achieves comparable or even better performance across diverse benchmarks.
arXiv Detail & Related papers (2024-07-02T16:10:55Z) - AIM: Let Any Multi-modal Large Language Models Embrace Efficient In-Context Learning [15.770849688170477]
In-context learning (ICL) facilitates Large Language Models exhibiting emergent ability on downstream tasks without updating billions of parameters.
Most primary MLLMs are only trained on single-image datasets, making them unable to read multi-modal demonstrations.
We propose a general and light-weighted framework textbfAIM to tackle the mentioned problems through textbfAggregating textbfImage information of textbfMultimodal demonstrations to the dense latent space of the corresponding linguistic part.
arXiv Detail & Related papers (2024-06-11T08:12:43Z) - Towards Semantic Equivalence of Tokenization in Multimodal LLM [149.11720372278273]
Vision tokenization is essential for semantic alignment between vision and language.
This paper proposes a novel dynamic Semantic-Equivalent Vision Tokenizer (SeTok)
SeTok groups visual features into semantic units via a dynamic clustering algorithm.
The resulting vision tokens effectively preserve semantic integrity and capture both low-frequency and high-frequency visual features.
arXiv Detail & Related papers (2024-06-07T17:55:43Z) - NoteLLM-2: Multimodal Large Representation Models for Recommendation [71.87790090964734]
Large Language Models (LLMs) have demonstrated exceptional proficiency in text understanding and embedding tasks.
Their potential in multimodal representation, particularly for item-to-item (I2I) recommendations, remains underexplored.
We propose an end-to-end fine-tuning method that customizes the integration of any existing LLMs and vision encoders for efficient multimodal representation.
arXiv Detail & Related papers (2024-05-27T03:24:01Z) - Boosting Multimodal Large Language Models with Visual Tokens Withdrawal for Rapid Inference [59.91176945361035]
We introduce Visual Tokens Withdrawal (VTW), a plug-and-play module to boost MLLMs for rapid inference.
VTW strategically withdraws vision tokens at a certain layer, enabling only text tokens to engage in subsequent layers.
Our approach can cut computational overhead by over 40% across diverse multimodal tasks while maintaining performance.
arXiv Detail & Related papers (2024-05-09T14:38:53Z)
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.