NurseCitizenDeveloper/relational-intelligence-unsloth-medgemma

🏥 Relational Intelligence MedGemma

A fine-tuned medical language model specialized in person-centred nursing care and practice development.

🎯 Model Description

This model extends Google's MedGemma 4B with specialized knowledge in:

• Person-centred care principles and practice
• Practice development methodology (IPDJ framework)
• Relational care in nursing contexts
• Critical creativity and emancipatory practice
• Human flourishing in healthcare settings

Training Data

Fine-tuned on 15,000+ question-answer pairs derived from:

• International Practice Development Journal (IPDJ)
• Foundation of Nursing Studies (FoNS) resources
• Practice development literature and frameworks

🚀 Quick Start

Using with Transformers

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# Load base model
base_model = AutoModelForCausalLM.from_pretrained(
    "google/medgemma-4b-it",
    torch_dtype=torch.float16,
    device_map="auto",
    token="YOUR_HF_TOKEN"  # MedGemma requires auth
)

# Load LoRA adapter
model = PeftModel.from_pretrained(
    base_model,
    "NurseCitizenDeveloper/relational-intelligence-unsloth-medgemma"
)

# Load tokenizer
tokenizer = AutoTokenizer.from_pretrained(
    "NurseCitizenDeveloper/relational-intelligence-unsloth-medgemma"
)

# Generate
prompt = "What is person-centred care in nursing?"
formatted = f"<start_of_turn>user\n{prompt}<end_of_turn>\n<start_of_turn>model\n"

inputs = tokenizer(formatted, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=256, temperature=0.7)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response)

Using with Unsloth (Recommended for Speed)

from unsloth import FastLanguageModel

model, tokenizer = FastLanguageModel.from_pretrained(
    "NurseCitizenDeveloper/relational-intelligence-unsloth-medgemma",
    max_seq_length=2048,
    load_in_4bit=True,
)

FastLanguageModel.for_inference(model)

# Use tokenizer.tokenizer for text-only (MedGemma has multimodal processor)
prompt = "Explain relational care in nursing practice."
formatted = f"<start_of_turn>user\n{prompt}<end_of_turn>\n<start_of_turn>model\n"
inputs = tokenizer.tokenizer(formatted, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.tokenizer.decode(outputs[0], skip_special_tokens=True))

💡 Example Prompts

Topic	Example Prompt
Person-centred care	"What are the key principles of person-centred nursing?"
Practice development	"How does practice development improve patient outcomes?"
Relational care	"Explain the importance of relational care in healthcare."
Critical creativity	"What is critical creativity in nursing practice?"
Emancipatory practice	"How can nurses develop emancipatory practice?"

📊 Training Details

Parameter	Value
Base Model	google/medgemma-4b-it
Training Framework	Unsloth
LoRA Rank	16
LoRA Alpha	16
Learning Rate	2e-4
Epochs	1
Batch Size	2 (×4 gradient accumulation)
Max Sequence Length	2048
Precision	4-bit quantization
Optimizer	AdamW 8-bit

Target Modules

• q_proj, k_proj, v_proj, o_proj
• gate_proj, up_proj, down_proj

🎓 Intended Use

This model is designed for:

✅ Educational purposes - Learning about person-centred care
✅ Research - Exploring AI applications in nursing practice development
✅ Practice reflection - Supporting nurses in reflective practice
✅ Knowledge exploration - Understanding IPDJ and FONS frameworks

Limitations

⚠️ Not for clinical decision-making - This is a research prototype
⚠️ Requires validation - Outputs should be verified by qualified professionals
⚠️ May hallucinate - Like all LLMs, may generate plausible but incorrect information

🤝 Acknowledgments

• Foundation of Nursing Studies (FoNS) - For championing person-centred practice development
• International Practice Development Journal (IPDJ) - Source of practice development knowledge
• Google - For MedGemma base model
• Unsloth - For efficient fine-tuning framework

👨‍⚕️ About the Developer

Lincoln Gombedza - Nurse Citizen Developer

This project demonstrates how nurses can leverage AI and open-source tools to advance person-centred care. Built with the philosophy that technology should serve relational practice, not replace it.

📚 Related Resources

• Foundation of Nursing Studies
• International Practice Development Journal
• Unsloth Documentation
• MedGemma Model Card

📜 License

This model is licensed under Creative Commons Attribution-NonCommercial 3.0 (CC BY-NC 3.0).

You are free to:

• Share — copy and redistribute the material
• Adapt — remix, transform, and build upon the material

Under the following terms:

• Attribution — Credit must be given to the creator
• NonCommercial — Not for commercial purposes

Note: Please also comply with MedGemma's terms of use for the base model.

---

"Democratising nursing work for the common good through technology that speaks nursing's language."