kelsbeans/qwen3-1.7b-digestive-coach-n195

TEXT GENERATIONPricing:Input $0.32 / Cached $0.064 / Output $1.6Concurrent Unit Cost:1Model Size:2BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 20, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

kelsbeans/qwen3-1.7b-digestive-coach-n195 is a 2 billion parameter Qwen3-1.7B model fine-tuned by kelsbeans using Unsloth QLoRA. This model is specifically optimized as a digestive wellness coach, demonstrating high parse rates and schema validity for health-related queries. It is designed as a research artifact for data-to-behavior fine-tuning projects, focusing on structured responses in a specialized domain.

Loading preview...

Overview

This model, kelsbeans/qwen3-1.7b-digestive-coach-n195, is a specialized fine-tune of the Qwen/Qwen3-1.7B base model, utilizing Unsloth QLoRA. It was trained on 195 rows from the kelsbeans/digestive-coach-dataset, focusing on digestive wellness scenarios. The model is provided in a merged fp16 format, allowing direct loading with transformers without requiring PEFT.

Key Capabilities

  • High Parse Rate: Achieves a 97.5% parse rate on a held-out evaluation set, significantly outperforming the base model's 12.5%.
  • Schema Validity: Demonstrates 97.5% schema validity, compared to 2.5% for the base model, indicating strong adherence to structured output requirements.
  • Specialized Domain: Optimized for generating responses related to digestive health, functioning as a "digestive wellness coach."
  • Research Artifact: Developed as part of a data-to-behavior fine-tuning research project.

Good For

  • Specialized Chatbots: Ideal for applications requiring structured, domain-specific responses in digestive health.
  • Research & Development: Suitable for researchers exploring fine-tuning techniques for specific behavioral outcomes.
  • Structured Data Generation: Excels in scenarios where output needs to conform to a predefined schema, as evidenced by its high schema validity.

Limitations

  • Not Medical Advice: Explicitly stated as a research artifact and not intended for providing medical advice.
  • Limited Spec Adherence: While improved over the base model, its spec adherence is 10.0%, suggesting room for further refinement in complex instruction following.

Finetunes
2 models