renhehuang/functiongemma-270m-it-coffee-robot-mcp

Model Overview

This model, renhehuang/functiongemma-270m-it-coffee-robot-mcp, is a specialized Function Calling model based on google/functiongemma-270m-it. Developed by renhehuang, it is fine-tuned for internal enterprise testing of a coffee robot delivery system, integrating with a point-of-sale API.

Key Capabilities

• Dedicated Function Calling: Specifically designed to convert natural language commands into structured function calls for coffee delivery tasks.
• Traditional Chinese Support: Processes natural language input in Traditional Chinese.
• Lightweight & Efficient: With only 270 million parameters, it is suitable for edge deployment scenarios.
• MCP Compatibility: Designed to integrate seamlessly with the Model Context Protocol (MCP).
• send_coffee Function: Supports a send_coffee function with parameters for baseDrink (americano, latte, oat_latte, milk), floor (1-11), temperature (hot, iced), addons (extra_espresso, paper_cup), and quantity (1-10).

Performance & Training

The model demonstrates a 100.00% Function Calling accuracy for its specific domain, showing a +4.05% improvement over the base model's 95.95% accuracy. It was trained for 3 epochs with a learning rate of 1e-05, using 625 training samples and 74 validation samples.

Limitations

• Domain-Specific: Exclusively designed for coffee delivery function calling; not suitable for other tasks.
• Language: Primarily supports Traditional Chinese; performance in other languages may be suboptimal.
• Floor & Drink Restrictions: Limited to floors 1-11 and only four specific drink types.