---
library_name: pytorch
---

![llava_onevison_logo](resource/LLaVA_onevision.png)

LLaVA-OneVision is a multimodal vision-language model that integrates a pretrained Qwen-2 language model with a visual encoder, enabling instruction-tuned understanding and reasoning across text and images.

Original paper: [LLaVA-OneVision: Easy Visual Task Transfer](https://arxiv.org/abs/2408.03326)

# LLaVA-OneVision-Qwen2-7B

This model uses LLaVA-OneVision with Qwen-2 as the language backbone, allowing rich multimodal reasoning and generation capabilities. It is well suited for applications such as image-grounded question answering, multimodal dialogue, and tasks requiring aligned understanding of visual and textual information.

Model Configuration:
- Reference implementation: [LLaVA_OneVision](https://github.com/LLaVA-VL/LLaVA-NeXT)
- Original Weight: [llava-onevision-qwen2-7b-ov-chat](https://huggingface.co/lmms-lab/llava-onevision-qwen2-7b-ov-chat)
- Vision Encoder: SO400M
- Language Model: Qwen-2.0
- Resolution: 3x384x384
- Support Cooper version:
    - Cooper SDK: [2.5.2]
    - Cooper Foundry: [2.2]

| Model | Device | Model Link |
| :-----: | :-----: | :-----: |
| LLaVA-OneVision | N1-655 | [Model_Link](https://huggingface.co/Ambarella/LLaVA-OneVision/blob/main/n1-655_llava_onevision_7B_1NVP.tar) |