Instructions to use kjunh/v1-7B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use kjunh/v1-7B with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="kjunh/v1-7B", trust_remote_code=True)
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoModelForConditionalGeneration
model = AutoModelForConditionalGeneration.from_pretrained("kjunh/v1-7B", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use kjunh/v1-7B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "kjunh/v1-7B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "kjunh/v1-7B",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/kjunh/v1-7B

SGLang

How to use kjunh/v1-7B with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "kjunh/v1-7B" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "kjunh/v1-7B",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "kjunh/v1-7B" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "kjunh/v1-7B",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use kjunh/v1-7B with Docker Model Runner:
```
docker model run hf.co/kjunh/v1-7B
```

v1-7B

File size: 2,208 Bytes

57536f4

---
license: apache-2.0
pipeline_tag: image-text-to-text
library_name: transformers
---

# Don't Look Only Once: Towards Multimodal Interactive Reasoning with Selective Visual Revisitation  

<p align="left">
    <a href='https://jiwanchung.github.io/' target='_blank'>Jiwan Chung<sup>*</sup></a>&emsp;
    <a href='https://junhyeok.kim/' target='_blank'>Junhyeok Kim<sup>*</sup></a>&emsp;
    <a href='https://scholar.google.com/citations?user=w3hOuRoAAAAJ' target='_blank'>Siyeol Kim</a>&emsp;
    <a href='https://jaeyoung-l.github.io/' target='_blank'>Jaeyoung Lee</a>&emsp;
    <a href="https://scholar.google.com/citations?user=Og3gN_AAAAAJ" target='_blank'>Minsoo Kim</a>&emsp;
    <a href='https://mirlab.yonsei.ac.kr/' target='_blank'>Youngjae Yu</a>
</p>

[![arXiv](https://img.shields.io/badge/arXiv-2505.18842-b31b1b.svg)](https://arxiv.org/abs/2505.18842) [![HuggingFace](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-kjunh/v1--7B-FFD21E)](https://huggingface.co/kjunh/v1-7B)

<p align="center">
  <img src="assets/figure.png">
</p>

## Installation
```bash
conda create -n v1 python=3.10 -y
conda activate v1
pip install -r requirements.txt
pip install flash-attn --no-build-isolation
```

## Demo

### Gradio Web UI
Highly Recommended as the copy tokens are displayed on image.

<p align="center">
  <img src="assets/demo.png">
</p>

```bash
python run_gradio.py
```

### Inference
```bash
python inference.py
```
The script uses a default image URL and text prompt. To use your own inputs, you can modify the `image` variable within the `messages` list and the `text` field for the user prompt.

## Coming Soon
- [x] Inference code
- [ ] Training data
- [ ] Evaluation code
- [ ] Training code


## Citation
If you find our work valuable, please cite:
```bibtex
@misc{chung2025dontlookoncemultimodal,
      title={Don't Look Only Once: Towards Multimodal Interactive Reasoning with Selective Visual Revisitation}, 
      author={Jiwan Chung and Junhyeok Kim and Siyeol Kim and Jaeyoung Lee and Min Soo Kim and Youngjae Yu},
      year={2025},
      eprint={2505.18842},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2505.18842}, 
}
```