Qwen3-1.7B Coffee Order Assistant (Traditional Chinese)

This model, developed by Ren-He Huang, is a 1.7 billion parameter language model based on the Qwen3 architecture. It has undergone Full Parameter SFT (Supervised Fine-Tuning) specifically for traditional Chinese coffee ordering scenarios.

Key Capabilities

• Understands user intent for coffee orders.
• Handles multi-turn dialogues to refine orders.
• Confirms order details including drink type, temperature (iced/hot), and espresso shots.
• Processes order modifications and cancellations.
• Supports a fixed menu of Americano, Latte, Oat Milk Latte, and Fresh Milk, with options for temperature and an extra espresso shot.

Training Details

• Base Model: Qwen/Qwen3-1.7B.
• Dataset: Trained on over 2900 multi-turn dialogues from the renhehuang/coffee-order-zhtw dataset, exclusively in traditional Chinese (Taiwan).
• Hardware: Trained on Apple Silicon (MPS).

Limitations

• Specialized: Exclusively trained for coffee ordering; not suitable for general conversation.
• Fixed Menu: Cannot process items outside its predefined menu.
• Order Constraints: Limited to one extra espresso shot per drink, and all drinks are assumed to be large size.

Should I use this for my use case?

This model is ideal if you need a highly specialized, efficient, and accurate AI assistant for traditional Chinese coffee ordering. Its focused training makes it perform exceptionally well within this narrow domain, offering precise understanding and interaction for specific menu items and modifications. Do not use it for general conversational AI or tasks outside of coffee ordering.