An official website of the United States government
We propose Vision Token Turing Machines (ViTTM), an efficient, low-latency, memory-augmented Vision Transformer (ViT). Our approach builds on Neural Turing Machines and Token Turing Machines, which were applied to NLP and sequential visual understanding tasks. ViTTMs are designed for non-sequential computer vision tasks such as image classification and segmentation. Our model creates two sets of tokens: process tokens and memory tokens; process tokens pass through encoder blocks and read-write from memory tokens at each encoder block in the network, allowing them to store and retrieve information from memory. By ensuring that there are fewer process tokens than memory tokens, we are able to reduce the inference time of the network while maintaining its accuracy. On ImageNet-1K, the state-of-the-art ViT-B has median latency of 529.5ms and 81.0% accuracy, while our ViTTM-B is 56% faster (234.1ms), with 2.4 times fewer FLOPs, with an accuracy of 82.9%. On ADE20K semantic segmentation, ViT-B achieves 45.65mIoU at 13.8 frame-per-second (FPS) whereas our ViTTM-B model acheives a 45.17 mIoU with 26.8 FPS (+94%).
more »
« less
This content will become publicly available on April 11, 2026
HiRED: Attention-Guided Token Dropping for Efficient Inference of High-Resolution Vision-Language Models
High-resolution Vision-Language Models (VLMs) are widely used in multimodal tasks to enhance accuracy by preserving detailed image information. However, these models often generate an excessive number of visual tokens due to the need to encode multiple partitions of a high-resolution image input. Processing such a large number of visual tokens poses significant computational challenges, particularly for resource-constrained commodity GPUs. To address this challenge, we propose High-Resolution Early Dropping (HiRED), a plug-and-play token-dropping method designed to operate within a fixed token budget. HiRED leverages the attention of CLS token in the vision transformer (ViT) to assess the visual content of the image partitions and allocate an optimal token budget for each partition accordingly. The most informative visual tokens from each partition within the allocated budget are then selected and passed to the subsequent Large Language Model (LLM). We showed that HiRED achieves superior accuracy and performance, compared to existing token-dropping methods. Empirically, HiRED-20% (i.e., a 20% token budget) on LLaVA-Next-7B achieves a 4.7x increase in token generation throughput, reduces response latency by 78%, and saves 14% of GPU memory for single inference on an NVIDIA TESLA P40 (24 GB). For larger batch sizes (e.g., 4), HiRED-20% prevents out-of-memory errors by cutting memory usage by 30%, while preserving throughput and latency benefits.
more »
« less
- Award ID(s):
- 2315851
- PAR ID:
- 10621667
- Publisher / Repository:
- Association for the Advancement of Artificial Intelligence!
- Date Published:
- Journal Name:
- Proceedings of the AAAI Conference on Artificial Intelligence
- Volume:
- 39
- Issue:
- 2
- ISSN:
- 2159-5399
- Page Range / eLocation ID:
- 1773 to 1781
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
This paper investigates how to efficiently deploy vision transformers on edge devices for small workloads. Recent methods reduce the latency of transformer neural networks by removing or merging tokens, with small accuracy degradation. However, these methods are not designed with edge device deployment in mind: they do not leverage information about the latency-workload trends to improve efficiency. We address this shortcoming in our work. First, we identify factors that affect ViT latency-workload relationships. Second, we determine token pruning schedule by leveraging non-linear latency-workload relationships. Third, we demonstrate a training-free, token pruning method utilizing this schedule. We show other methods may increase latency by 2-30%, while we reduce latency by 9-26%. For similar latency (within 5.2% or 7ms) across devices we achieve 78.6%-84.5% ImageNet1K accuracy, while the state-of-the-art, Token Merging, achieves 45.8%-85.4%.more » « less
-
In our research paper, we introduce a revolutionary approach to designing energy-aware dynamically prunable Vision Trans- formers for use in edge applications. Our solution denoted as Incremental Resolution Enhancing Transformer (IRET), works by the sequential sampling of the input image. However, in our case, the embedding size of input tokens is considerably smaller than prior-art solutions. This embedding is used in the first few layers of the IRET vision transformer until a reliable attention matrix is formed. Then the attention matrix is used to sample additional information using a learnable 2D lifting scheme only for important tokens and IRET drops the tokens receiving low attention scores. Hence, as the model pays more attention to a subset of tokens for its task, its focus and resolu- tion also increase. This incremental attention-guided sampling of input and dropping of unattended tokens allow IRET to sig- nificantly prune its computation tree on demand. By controlling the threshold for dropping unattended tokens and increasing the focus of attended ones, we can train a model that dynami- cally trades off complexity for accuracy. This is especially useful for edge devices, where accuracy and complexity could be dy- namically traded based on factors such as battery life, reliability, etc.more » « less
-
null (Ed.)High spatiotemporal resolution can offer high precision for vision applications, which is particularly useful to capture the nuances of visual features, such as for augmented reality. Unfortunately, capturing and processing high spatiotemporal visual frames generates energy-expensive memory traffic. On the other hand, low resolution frames can reduce pixel memory throughput, but reduce also the opportunities of high-precision visual sensing. However, our intuition is that not all parts of the scene need to be captured at a uniform resolution. Selectively and opportunistically reducing resolution for different regions of image frames can yield high-precision visual computing at energy-efficient memory data rates. To this end, we develop a visual sensing pipeline architecture that flexibly allows application developers to dynamically adapt the spatial resolution and update rate of different “rhythmic pixel regions” in the scene. We develop a system that ingests pixel streams from commercial image sensors with their standard raster-scan pixel read-out patterns, but only encodes relevant pixels prior to storing them in the memory. We also present streaming hardware to decode the stored rhythmic pixel region stream into traditional frame-based representations to feed into standard computer vision algorithms. We integrate our encoding and decoding hardware modules into existing video pipelines. On top of this, we develop runtime support allowing developers to flexibly specify the region labels. Evaluating our system on a Xilinx FPGA platform over three vision workloads shows 43 − 64% reduction in interface traffic and memory footprint, while providing controllable task accuracy.more » « less
-
The ability to accurately interpret complex visual information is a crucial topic of multimodal large language models (MLLMs). Recent work indicates that enhanced visual perception significantly reduces hallucinations and improves performance on resolution-sensitive tasks, such as optical character recognition and document analysis. A number of recent MLLMs achieve this goal using a mixture of vision encoders. Despite their success, there is a lack of systematic comparisons and detailed ablation studies addressing critical aspects, such as expert selection and the integration of multiple vision experts. This study provides an extensive exploration of the design space for MLLMs using a mixture of vision encoders and resolutions. Our findings reveal several underlying principles common to various existing strategies, leading to a streamlined yet effective design approach. We discover that simply concatenating visual tokens from a set of complementary vision encoders is as effective as more complex mixing architectures or strategies. We additionally introduce Pre-Alignment to bridge the gap between vision-focused encoders and language tokens, enhancing model coherence. The resulting family of MLLMs, Eagle, surpasses other leading open-source models on major MLLM benchmarks.more » « less
