Instructions to use hanchaow/QTuneVL1_5-2B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use hanchaow/QTuneVL1_5-2B with Transformers:

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

pipe = pipeline("image-text-to-text", model="hanchaow/QTuneVL1_5-2B", 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 AutoModel
model = AutoModel.from_pretrained("hanchaow/QTuneVL1_5-2B", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use hanchaow/QTuneVL1_5-2B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "hanchaow/QTuneVL1_5-2B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "hanchaow/QTuneVL1_5-2B",
		"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/hanchaow/QTuneVL1_5-2B

SGLang

How to use hanchaow/QTuneVL1_5-2B 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 "hanchaow/QTuneVL1_5-2B" \
    --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": "hanchaow/QTuneVL1_5-2B",
		"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 "hanchaow/QTuneVL1_5-2B" \
        --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": "hanchaow/QTuneVL1_5-2B",
		"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 hanchaow/QTuneVL1_5-2B with Docker Model Runner:
```
docker model run hf.co/hanchaow/QTuneVL1_5-2B
```

hanchaow commited on Aug 6, 2025

Commit

7a08d91

verified ·

1 Parent(s): 6789701

Update README.md

Browse files

Files changed (1) hide show

README.md +1 -1

README.md CHANGED Viewed

@@ -20,7 +20,7 @@ Like QTuneVL1-2B, QTuneVL1.5-2B is a lightweight MLLM that incorporates cropping
 # Evaluation
-By evaluating our model on eight benchmarks in the [OpenCompass](https://rank.opencompass.org.cn/leaderboard-multimodal) leaderboard using [VLMEvalKit](https://github.com/open-compass/VLMEvalKit), we found that it outperformed its predecessor(QTuneVL1-2B) in terms of average scores, particularly on MMStar and MMMU_DEV_VAL benchmarks. The eight benchmarks  and specific experimental results are as follows:
 **Eight benchmark:** ` 'MMBench_DEV_EN_V11', 'MMStar', 'MMMU_DEV_VAL', 'MathVista_MINI', 'HallusionBench', 'AI2D_TEST', 'OCRBench', 'MMVet' `.

 # Evaluation
+By evaluating our model on eight benchmarks in the [OpenCompass](https://rank.opencompass.org.cn/leaderboard-multimodal) leaderboard using [VLMEvalKit](https://github.com/open-compass/VLMEvalKit), we found that it outperformed its predecessor(QTuneVL1-2B) in terms of average scores, particularly on MMStar MMMU_DEV_VAL and OCRBench benchmarks. The eight benchmarks  and specific experimental results are as follows:
 **Eight benchmark:** ` 'MMBench_DEV_EN_V11', 'MMStar', 'MMMU_DEV_VAL', 'MathVista_MINI', 'HallusionBench', 'AI2D_TEST', 'OCRBench', 'MMVet' `.