Instructions to use rbrmafort/functiongemma-270m-event-commands-gguf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use rbrmafort/functiongemma-270m-event-commands-gguf with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf rbrmafort/functiongemma-270m-event-commands-gguf:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf rbrmafort/functiongemma-270m-event-commands-gguf:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf rbrmafort/functiongemma-270m-event-commands-gguf:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf rbrmafort/functiongemma-270m-event-commands-gguf:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf rbrmafort/functiongemma-270m-event-commands-gguf:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf rbrmafort/functiongemma-270m-event-commands-gguf:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf rbrmafort/functiongemma-270m-event-commands-gguf:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf rbrmafort/functiongemma-270m-event-commands-gguf:Q4_K_M

Use Docker

docker model run hf.co/rbrmafort/functiongemma-270m-event-commands-gguf:Q4_K_M

Llama new
LM Studio
Jan

vLLM

How to use rbrmafort/functiongemma-270m-event-commands-gguf with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "rbrmafort/functiongemma-270m-event-commands-gguf"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "rbrmafort/functiongemma-270m-event-commands-gguf",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/rbrmafort/functiongemma-270m-event-commands-gguf:Q4_K_M

Ollama
How to use rbrmafort/functiongemma-270m-event-commands-gguf with Ollama:
```
ollama run hf.co/rbrmafort/functiongemma-270m-event-commands-gguf:Q4_K_M
```
Unsloth Desktop

How to use rbrmafort/functiongemma-270m-event-commands-gguf with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf rbrmafort/functiongemma-270m-event-commands-gguf:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "rbrmafort/functiongemma-270m-event-commands-gguf:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use rbrmafort/functiongemma-270m-event-commands-gguf with Docker Model Runner:
```
docker model run hf.co/rbrmafort/functiongemma-270m-event-commands-gguf:Q4_K_M
```

Lemonade

How to use rbrmafort/functiongemma-270m-event-commands-gguf with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull rbrmafort/functiongemma-270m-event-commands-gguf:Q4_K_M

Run and chat with the model

lemonade run user.functiongemma-270m-event-commands-gguf-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use rbrmafort/functiongemma-270m-event-commands-gguf with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf rbrmafort/functiongemma-270m-event-commands-gguf:Q4_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default rbrmafort/functiongemma-270m-event-commands-gguf:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use rbrmafort/functiongemma-270m-event-commands-gguf with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf rbrmafort/functiongemma-270m-event-commands-gguf:Q4_K_M

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "rbrmafort/functiongemma-270m-event-commands-gguf:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

FunctionGemma 270M — event commands, Q4_K_M

Selected fine-tune (epoch-2), merged with the pinned upstream model and quantized for offline CPU use in an Android application. This is a modified derivative, not the stock upstream model.

Intended use

Extract baby-tracking requests into proposed function calls. It is not a conversational assistant, medical adviser, memory-navigation model or storytelling model. Validate names, arguments, dates, quantities, permissions and record IDs in application code. Require user review before writes. No calibrated confidence score is provided.

Artifact

File: q4_k_m.gguf
Size: 261,329,984 bytes
SHA-256: 007e3e5e6857652f0ccea8e205ad47467328817b4f7db48111506ab24106c88d
Base: google/functiongemma-270m-it, revision 39eccb091651513a5dfb56892d3714c1b5b8276c
Export: LoRA merge, GGUF, Q4_K_M; pinned llama.cpp 3d65c90d04d337e88f2b1f7f0061f40a5324e662.

App protocol

Use the functiongemma-native chat template, the extraction instruction and context layout in prompt_protocol.json, and the thirteen supported functions in command_tools.json. Those files are technical protocol definitions, not family data. JSON call envelopes are used for Granite/Qwen; FunctionGemma emits its native <start_function_call> format. An ordinary chat prompt or a broad application planning schema is not equivalent. The app converts complete, validated calls into reviewable plans; it does not fabricate confidence.

The tokenizer and chat template are embedded in GGUF. The tested mobile configuration uses an 8,192-token context, greedy command decoding and a 512-token output cap. JSON syntax constraints do not establish factual correctness; reject partial/truncated outputs.

Evaluation and limits

The earlier BF16 desktop checkpoint scored 74.5% exact tool/argument matches across 440 synthetic matched cases in ten languages, with separate weak-language Amharic stress tests. This is not a mobile-Q4 accuracy claim or native-speaker validation. The desktop tool catalog was broader than the thirteen-tool app protocol. Quantization, different prompts/catalogs and hardware can change results. Errors can include wrong quantities, dates, IDs and unwanted writes despite valid syntax. English and Brazilian Portuguese were the training languages.

No dataset, family conversations, logs, credentials or app database is included in this release. See MODEL_MANIFEST.json for provenance and NOTICE for modifications. The model weights remain subject to the upstream license in LICENSE.

Gemma use terms

Use and redistribution are governed by the included Gemma Terms of Use, including Section 3.2 and the Gemma Prohibited Use Policy. Gemma is provided under and subject to the Gemma Terms of Use found at https://ai.google.dev/gemma/terms.

Mobile integration check

Eight synthetic app-catalog requests (six English and two Portuguese) through the actual Android runtime on Samsung SM-S901E, Q4_K_M, CPU, 8K context. Exact complete tool/argument matches: 7/8. Peak whole Flutter debug app RSS 1.34 GiB, maximum request-baseline increase 622 MiB, median 21.5 seconds. This tiny integration sample is not a production accuracy estimate, model-only RAM measurement, or validation on smaller phones.

The published prompt examples omit application branding; the integration measurements used the original corresponding prompts.

Downloads last month: 353

GGUF

Model size

0.3B params

Architecture

gemma3

Hardware compatibility

4-bit

Model tree for rbrmafort/functiongemma-270m-event-commands-gguf

Base model

google/functiongemma-270m-it

Finetuned

(472)

this model