# Lightweight Vision-Language Model

This is a 256M-parameter preview version of a document understanding model, specifically designed for document structure parsing and content extraction tasks, supporting the conversion of image documents into structured data.

Qwen2.5-VL-7B-Captioner-Relaxed is a multimodal vision-language model based on the Qwen2.5 architecture, focusing on image-to-text generation tasks.

MLX format model converted from SmolVLM2-500M-Video-Instruct, supporting video-to-text tasks

This is a video-text-to-text model based on the MLX format, developed by HuggingFaceTB, supporting English language processing.

LlavaGuard is a safety evaluation guardian based on vision-language models, primarily used for safety classification and violation detection of image content.

Doubutsu is a lightweight vision-language model series, specifically designed for customized scenario fine-tuning.

Cerule is a lightweight yet powerful vision-language model built on Google's Gemma-2b and SigLIP, focusing on image-text processing.

UForm-Gen2-dpo is a small generative vision-language model, aligned for image caption generation and visual question answering tasks through Direct Preference Optimization (DPO) on VLFeedback and LLaVA-Human-Preference-10K preference datasets.

A fine-tuned version of Moondream2, optimized for image prompt generation. It is a lightweight vision-language model suitable for efficient operation on edge devices.

LLaVa-Phi-2-3B is an open-source multimodal chatbot model, fine-tuned based on the Phi-2 architecture, capable of processing image and text inputs to generate natural language responses.