Q

Qwen2.5 VL 7B Instruct FP8 Dynamic

Developed by RedHatAI
The FP8 quantized version of Qwen2.5-VL-7B-Instruct, supporting efficient vision-text inference through vLLM
Downloads 25.18k
Release Time : 2/6/2025

Model Overview

An FP8 dynamic quantized model based on Qwen2.5-VL-7B-Instruct, supporting vision-text input and text output, suitable for multimodal understanding and generation tasks

Model Features

FP8 Dynamic Quantization
Both weights and activations use FP8 quantization technology to improve inference efficiency while maintaining model accuracy
vLLM Optimization Support
Optimized for the vLLM inference engine, supporting efficient deployment and inference acceleration
Multimodal Understanding
Supports joint input of vision and text, capable of understanding and analyzing image content

Model Capabilities

Visual Question Answering
Image Content Understanding
Document Parsing
Chart Analysis
Mathematical Visual Inference
Multimodal Text Generation

Use Cases

Document Processing
Document Visual Question Answering
Parse and understand the content in document images and answer questions
Achieved a 94.27 ANLS score on the DocVQA dataset
Visual Inference
Chart Analysis
Understand and interpret chart data
Achieved an 86.80% accuracy rate on the ChartQA test set
Mathematical Visual Problem Solving
Solve mathematical problems containing visual elements
Achieved a 71.07% accuracy rate on the Mathvista test set
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Š 2025AIbase