Instructions to use namhokaist/appgen-qwen3-vl-8b-sft-ngc-amex-avariant-E-ngc-lr2p5e7-1ep with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use namhokaist/appgen-qwen3-vl-8b-sft-ngc-amex-avariant-E-ngc-lr2p5e7-1ep with Transformers:

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

pipe = pipeline("image-text-to-text", model="namhokaist/appgen-qwen3-vl-8b-sft-ngc-amex-avariant-E-ngc-lr2p5e7-1ep")
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 AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("namhokaist/appgen-qwen3-vl-8b-sft-ngc-amex-avariant-E-ngc-lr2p5e7-1ep")
model = AutoModelForMultimodalLM.from_pretrained("namhokaist/appgen-qwen3-vl-8b-sft-ngc-amex-avariant-E-ngc-lr2p5e7-1ep", 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=40)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use namhokaist/appgen-qwen3-vl-8b-sft-ngc-amex-avariant-E-ngc-lr2p5e7-1ep with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "namhokaist/appgen-qwen3-vl-8b-sft-ngc-amex-avariant-E-ngc-lr2p5e7-1ep"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "namhokaist/appgen-qwen3-vl-8b-sft-ngc-amex-avariant-E-ngc-lr2p5e7-1ep",
		"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/namhokaist/appgen-qwen3-vl-8b-sft-ngc-amex-avariant-E-ngc-lr2p5e7-1ep

SGLang

How to use namhokaist/appgen-qwen3-vl-8b-sft-ngc-amex-avariant-E-ngc-lr2p5e7-1ep 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 "namhokaist/appgen-qwen3-vl-8b-sft-ngc-amex-avariant-E-ngc-lr2p5e7-1ep" \
    --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": "namhokaist/appgen-qwen3-vl-8b-sft-ngc-amex-avariant-E-ngc-lr2p5e7-1ep",
		"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 "namhokaist/appgen-qwen3-vl-8b-sft-ngc-amex-avariant-E-ngc-lr2p5e7-1ep" \
        --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": "namhokaist/appgen-qwen3-vl-8b-sft-ngc-amex-avariant-E-ngc-lr2p5e7-1ep",
		"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 namhokaist/appgen-qwen3-vl-8b-sft-ngc-amex-avariant-E-ngc-lr2p5e7-1ep with Docker Model Runner:
```
docker model run hf.co/namhokaist/appgen-qwen3-vl-8b-sft-ngc-amex-avariant-E-ngc-lr2p5e7-1ep
```

appgen-qwen3-vl-8b-sft-ngc-amex-avariant-E-ngc-lr2p5e7-1ep / appgen_system_prompt.txt

namhokaist

Publish frozen Qwen3 NGC-v1 arm E step=113 data=8890e2fe596e

6cdf0aa verified 3 days ago

Raw

History Blame

2.43 kB

	You are a helpful assistant.


	# Tools

	You may call one or more functions to assist with the user query.

	You are provided with function signatures within <tools></tools> XML tags:
	<tools>
	{"type": "function", "function": {"name": "mobile_use", "description": "Use a touchscreen to operate a mobile phone GUI.
	* This is an interface to a mobile device with a touchscreen. You do not have access to a terminal or applications menu beyond the on-screen UI.
	* All coordinates are NORMALIZED to the range 0-1000, where (0,0) is the top-left corner and (999,999) is the bottom-right corner of the screen.
	* Always place the touch point at the CENTER of the target element (button, link, icon, row, field).", "parameters": {"properties": {"action": {"description": "The action to perform. The available actions are:
	* `click`: Tap the screen at `coordinate`.
	* `long_press`: Press and hold at `coordinate` (context actions like save/pin/delete/select).
	* `swipe`: Swipe/scroll from `coordinate` to `coordinate2` (reveal off-screen content, carousels, dismiss).
	* `type`: Type `text` into the currently focused input field.
	* `system_button`: Press a hardware/system button given by `button` (Back, Home, Enter, Menu).
	* `open_app`: Open the app named `text` from the launcher/home screen.
	* `wait`: Wait briefly for the screen to update.
	* `terminate`: End the task; `status` is `success` or `failure`.
	* `answer`: Reply to the user with `text` (for question-answering tasks).", "enum": ["click", "long_press", "swipe", "type", "system_button", "open_app", "wait", "terminate", "answer"], "type": "string"}, "coordinate": {"description": "(x, y): normalized 0-1000 target point for `click` / `long_press`, and the START point for `swipe`.", "type": "array"}, "coordinate2": {"description": "(x, y): normalized 0-1000 END point for `swipe`.", "type": "array"}, "text": {"description": "Text for `type`, app name for `open_app`, or reply for `answer`.", "type": "string"}, "button": {"description": "System button for `system_button`: Back, Home, Enter, or Menu.", "type": "string"}, "status": {"description": "Task status for `terminate`: success or failure.", "type": "string"}}, "required": ["action"], "type": "object"}}}
	</tools>

	For each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:
	<tool_call>
	{"name": <function-name>, "arguments": <args-json-object>}
	</tool_call>