Instructions to use Fatha/vev-9b-v2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Fatha/vev-9b-v2 with Transformers:

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

pipe = pipeline("image-text-to-text", model="Fatha/vev-9b-v2")
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)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("Fatha/vev-9b-v2")
model = AutoModelForMultimodalLM.from_pretrained("Fatha/vev-9b-v2", device_map="auto")
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?"}
        ]
    },
]
inputs = processor.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Fatha/vev-9b-v2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Fatha/vev-9b-v2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Fatha/vev-9b-v2",
		"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/Fatha/vev-9b-v2

SGLang

How to use Fatha/vev-9b-v2 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 "Fatha/vev-9b-v2" \
    --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": "Fatha/vev-9b-v2",
		"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 "Fatha/vev-9b-v2" \
        --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": "Fatha/vev-9b-v2",
		"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 Fatha/vev-9b-v2 with Docker Model Runner:
```
docker model run hf.co/Fatha/vev-9b-v2
```

vev-9b-v2

Small, calibrated vision decision model. Given one image, it answers yes/no or multiple-choice questions and returns calibrated probabilities over the options. Built mainly for evaluating GenAI-generated images (prompt match, artifacts, rendered text, etc.).

Code, serving, CLI, benchmark: https://github.com/Fathaah/vev
Base model: Qwen/Qwen3.5-9B-Base, rank-32 LoRA, merged into the weights
Training: 846 steps, 59,378 rows, public human-labeled single-image datasets (no rating or aesthetic data)
Calibration: one temperature (T = 1.163) fitted on a held-out split of 1,999 examples and stored in vev_config.json

Held-out calibration split	Accuracy	ECE
after temperature scaling	0.906	0.0159

Usage

This repo is the merged checkpoint. Serving, the question/answer API, and vev eval-dir are in the GitHub repo; see its README for how to point vev.serve --run at a model directory.

Licensing

The code is Apache-2.0, but the weights were trained on datasets with their own terms, some research-only. Those restrictions carry over to the weights. See DATA_LICENSES.md before any commercial use.

Downloads last month: -

Safetensors

Model size

9B params

Tensor type

BF16

Model tree for Fatha/vev-9b-v2

Base model

Qwen/Qwen3.5-9B-Base

Finetuned

(613)

this model