Tag: 多模态
8 articles
AI Research 42 - Multimodal Large Model Quantization: Fro...
Survey outline for multimodal large model quantization schemes: from FP32 to INT4. Core goal is model capability retention, compression efficiency 50-75%, inference speedup 2-4x. Analyze comparison...
AI Research 41 - Multimodal Large Model Quantization: Qwe...
Qwen2.5-VL is the new generation multimodal large model launched by Alibaba, significantly leading in visual understanding, video analysis, and cross-modal reasoning. Provides multiple versions fro...
AI Research 40 - Multimodal Large Model Quantization: Pat...
Multimodal large models are developing rapidly, with representative models like BLIP-2, MiniGPT-4, Flamingo, LLaVA, and Qwen2.5-VL emerging. Analyze each model's architectural innovations, performa...
AI Research 39 - Multimodal Large Model Quantization: How...
In multimodal large model optimization, the order choice of fine-tuning and quantization directly affects the final model's performance and efficiency. There are three main strategies: fine-tune fi...
AI Research 38 - Multimodal Large Model Quantization: Ana...
To systematically evaluate the impact of model quantization on performance, need to combine multiple vision-language datasets and metrics. Commonly used datasets include Flickr30k and MS COCO, usin...
AI Research 37 - Multimodal Large Model Quantization: Imp...
Model quantization compresses FP32 weights into low-precision representations, significantly reducing inference resource consumption. Experiments show quantized models have 60% lower latency and 70...
AI Research 36 - Comprehensive Analysis of Multimodal Lar...
This comprehensive overview systematically introduces mainstream quantization techniques in multimodal models, including principles and practices of...
AI Research 13 - LLM and Agent Research: The Rise and Dev...
2024 is called the 'Year of Agents'. LLM trends show parallel development of 'bigger and stronger' and 'smaller and more specialized'. OpenAI o1 series, Claude, and other multimodal models continue...