# functiongemma-270m-function-calling for Flutter Gemma

MediaPipe `.task` conversion of `phattrandeveloper/functiongemma-270m-function-calling` for `flutter_gemma` mobile inference.

## Artifact

- Task bundle: `functiongemma-270m-function-calling.task`
- Flutter Gemma file type: `ModelFileType.task`
- Flutter Gemma model type: `ModelType.functionGemma`
- Quantization: `dynamic_int8`
- Prefill sequence length: `2048`
- KV cache max length: `4096`
- Prompt template embedded: `False`
- SHA-256: `068757b3c6fe9ec8182fe11c2eb612423a0d77ff42cae20142476aff10f99820`

Download URL:

```text
https://huggingface.co/phattrandeveloper/functiongemma-270m-function-calling/resolve/main/functiongemma-270m-function-calling.task
```

## Flutter Gemma usage

```dart
await FlutterGemma.installModel(modelType: ModelType.functionGemma)
  .fromNetwork('https://huggingface.co/phattrandeveloper/functiongemma-270m-function-calling/resolve/main/functiongemma-270m-function-calling.task')
  .install();
```

Use the banking tool schema and `systemInstruction` from this project when creating the chat.

## Conversion references

- Google Gemma HF SafeTensors to MediaPipe Task guide: https://ai.google.dev/gemma/docs/conversions/hf-to-mediapipe-task
- Google MediaPipe LLM Inference guide: https://ai.google.dev/edge/mediapipe/solutions/genai/llm_inference
- Flutter Gemma docs: https://pub.dev/packages/flutter_gemma

## License

This converted artifact follows the license and use restrictions of the source Gemma/FunctionGemma model. Review the source model card and Gemma terms before redistribution or production use.