TensorX
返回文献探索

Paper · arXiv 2412.13303

FastVLM: Efficient Vision Encoding for Vision Language Models

Pavan Kumar Anasosalu Vasu, Fartash Faghri, Chun-Liang Li, Cem Koc, Nate True, Albert Antony, Gokul Santhanam, James Gabriel, Peter Grasch, Oncel Tuzel, Hadi Pouransari

77 upvotesDecember 17, 2024arXiv 预印本
AI 摘要

FastVLM, using FastViTHD, optimizes Vision Language Models by scaling image resolution to reduce latency and token count, improving time-to-first-token and maintaining performance.

ViTsself-attention layersFastVLMFastViTHDtime-to-first-tokenTTFTSeedBenchMMMU

Abstract

Scaling the input image resolution is essential for enhancing the performance of Vision Language Models (VLMs), particularly in text-rich image understanding tasks. However, popular visual encoders such as ViTs become inefficient at high resolutions due to the large number of tokens and high encoding latency caused by stacked self-attention layers. At different operational resolutions, the vision encoder of a VLM can be optimized along two axes: reducing encoding latency and minimizing the number of visual tokens passed to the LLM, thereby lowering overall latency. Based on a comprehensive efficiency analysis of the interplay between image resolution, vision latency, token count, and LLM size, we introduce FastVLM, a model that achieves an optimized trade-off between latency, model size and accuracy. FastVLM incorporates FastViTHD, a novel hybrid vision encoder designed to output fewer tokens and significantly reduce encoding time for high-resolution images. Unlike previous methods, FastVLM achieves the optimal balance between visual token count and image resolution solely by scaling the input image, eliminating the need for additional token pruning and simplifying the model design. In the LLaVA-1.5 setup, FastVLM achieves 3.2times improvement in time-to-first-token (TTFT) while maintaining similar performance on VLM benchmarks compared to prior works. Compared to LLaVa-OneVision at the highest resolution (1152times1152), FastVLM achieves comparable performance on key benchmarks like SeedBench and MMMU, using the same 0.5B LLM, but with 85times faster TTFT and a vision encoder that is 3.4times smaller.

北京市昌平区探索星信息技术及软件开发工作室

京ICP备2026059466号
FastVLM: Efficient Vision Encoding for Vision Language Models | TensorX