90 lines
2.4 KiB
Markdown
90 lines
2.4 KiB
Markdown
|
---
|
||
|
|
license: mit
|
||
|
|
datasets:
|
||
|
|
- FreedomIntelligence/medical-o1-reasoning-SFT
|
||
|
|
base_model:
|
||
|
|
- Qwen/Qwen3-4B
|
||
|
|
tags:
|
||
|
|
- medical
|
||
|
|
---
|
||
|
|
|
||
|
|
# 🧠 Qwen3-4B-MedicalDataset
|
||
|
|
|
||
|
|
A fine-tuned version of [Qwen3-4B](https://huggingface.co/Qwen/Qwen3-4B) on a curated **medical dataset**, designed to assist with medical Q&A, clinical documentation, and healthcare-related reasoning. Developed by **XformAI-India**.
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## 📌 Model Details
|
||
|
|
|
||
|
|
- **Base Model**: Qwen3-4B (from Qwen by Alibaba)
|
||
|
|
- **Architecture**: Transformer Decoder (GPT-like)
|
||
|
|
- **Parameters**: 4 Billion
|
||
|
|
- **Precision**: bfloat16 / float16
|
||
|
|
- **Fine-tuning method**: Supervised fine-tuning (SFT)
|
||
|
|
- **Dataset**: FreedomIntelligence/medical-o1-reasoning-SFT
|
||
|
|
---
|
||
|
|
|
||
|
|
## 🧪 Intended Use
|
||
|
|
|
||
|
|
This model is intended for **research** and **educational purposes** in the following domains:
|
||
|
|
|
||
|
|
- Medical question answering
|
||
|
|
- Summarizing patient records
|
||
|
|
- Medical reasoning and triage
|
||
|
|
- Chatbot integration for healthcare support (non-diagnostic)
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## 🚫 Limitations & Disclaimer
|
||
|
|
|
||
|
|
> ⚠️ **Not for clinical use.** This model is not a replacement for professional medical advice, diagnosis, or treatment.
|
||
|
|
|
||
|
|
- May hallucinate or produce incorrect medical information.
|
||
|
|
- Trained only on publicly available or synthetic datasets — not real patient data.
|
||
|
|
- Should not be used in emergency or high-stakes settings.
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## 💡 Example Usage
|
||
|
|
|
||
|
|
\`\`\`python
|
||
|
|
from transformers import AutoTokenizer, AutoModelForCausalLM
|
||
|
|
|
||
|
|
tokenizer = AutoTokenizer.from_pretrained("XformAI-india/Qwen3-4B-medicaldataset", trust_remote_code=True)
|
||
|
|
model = AutoModelForCausalLM.from_pretrained("XformAI-india/Qwen3-4B-medicaldataset", trust_remote_code=True)
|
||
|
|
|
||
|
|
prompt = "Patient presents with chest pain and shortness of breath. What are possible differential diagnoses?"
|
||
|
|
inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
|
||
|
|
outputs = model.generate(**inputs, max_new_tokens=200)
|
||
|
|
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
|
||
|
|
\`\`\`
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## 🏗 Training Configuration
|
||
|
|
|
||
|
|
- **Epochs**: 3
|
||
|
|
- **Batch Size**: 8
|
||
|
|
- **Optimizer**: AdamW
|
||
|
|
- **Learning Rate**: 2e-5
|
||
|
|
- **Gradient Accumulation**: 4
|
||
|
|
- **Framework**: PyTorch, Hugging Face Transformers
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
|
||
|
|
## 🧠 Citation
|
||
|
|
|
||
|
|
If you use this model, please cite:
|
||
|
|
|
||
|
|
|
||
|
|
```bibtex
|
||
|
|
@misc{qwen3medical2025,
|
||
|
|
title={Qwen3-4B-MedicalDataset: A Fine-Tuned Transformer for Medical AI Research},
|
||
|
|
author={XformAI-India},
|
||
|
|
year={2025},
|
||
|
|
url={https://huggingface.co/XformAI-india/Qwen3-4B-medicaldataset}
|
||
|
|
}
|
||
|
|
```
|
||
|
|
|
||
|
|
---
|