Instructions to use kontextdev/agent-gemma with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use kontextdev/agent-gemma with Transformers:

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

pipe = pipeline("image-text-to-text", model="kontextdev/agent-gemma")
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, AutoModelForImageTextToText

processor = AutoProcessor.from_pretrained("kontextdev/agent-gemma")
model = AutoModelForImageTextToText.from_pretrained("kontextdev/agent-gemma")
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

vLLM

How to use kontextdev/agent-gemma with vLLM:

Install from pip and serve model

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

SGLang

How to use kontextdev/agent-gemma 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 "kontextdev/agent-gemma" \
    --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": "kontextdev/agent-gemma",
		"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 "kontextdev/agent-gemma" \
        --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": "kontextdev/agent-gemma",
		"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 kontextdev/agent-gemma with Docker Model Runner:
```
docker model run hf.co/kontextdev/agent-gemma
```

Problems with LiteRT-LM v0.9.0-beta

by dvillevalois - opened Mar 22

Discussion

dvillevalois

Mar 22

Hello!

I fail to use your model with v0.9.0-beta. I call it from the Kotlin official bindings and I get the following error from the native runtime:

Failed to apply template: unknown function: format_function_declaration is unknown (in template:21)

I am no expert so I naively believe that the jinja template embedded in the litertml file is invalid.
Am I right?

If that's indeed the problem, unfortunatelly I don't have the possibility to override the template.

Thanks in advance for your help.

macmacmacmac

Kontext.dev org Mar 23

👋 dvillevalois ,i just fixxed the format_function_declaration template error you were hitting.

the stock Gemma 3n chat template calls format_function_declaration(), a custom Jinja function that exists in Google's Python tokenizer, but isn't supported by LiteRT-LM's on device template engine. So when your app tries to register tools, the template render fails.

i fine-tuned the model with a custom chat template that replaces format_function_declaration() with standard Jinja2 (tojson filter + / markers). Same function-calling behavior, fully compatible with LiteRT-LM's native engine.

tested fix in android (SDK 36, JDK 21, LiteRT-LM v0.9.0-alpha01) that loads the model, registers tools, and makes a function call. pushed the updated model.

dvillevalois

Mar 23

Hey @macmacmacmac ! Thanks a lot! That works. Tested on Linux/JVM with LiteRT-LM 0.9.0-beta. Thanks again.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment