README_Enhanced_MedicalChatBot.md

# 🩺 MedicalChatBot

**MedicalChatBot** is a medical domain-focused chatbot fine-tuned using **LoRA (Low-Rank Adaptation)** on top of [`mistralai/Mistral-7B-Instruct`](https://huggingface.co/mistralai/Mistral-7B-Instruct).  
It is designed for health education, medical Q&A, and research use only.

---

## 📌 Overview

- 🧠 Based on Mistral-7B-Instruct, a powerful instruction-following LLM
- 🔧 Fine-tuned using [PEFT](https://github.com/huggingface/peft) + LoRA on a medical dataset
- 📚 Trained on: [`kberta2014/medical-chat-dataset`](https://huggingface.co/datasets/kberta2014/medical-chat-dataset)
- ⚡ Efficient: Only trains adapter layers instead of the full model
- 📦 Deployment-ready: Compatible with Hugging Face `transformers`, `Gradio`, and Spaces

---

## 🧠 Prompt Format

Use the model in the following format:

```
### Instruction:
<Your question>

### Input:
<Optional additional context>

### Response:
```

Example:
```
### Instruction:
What are the symptoms of high blood pressure?

### Input:


### Response:
```

---

## 💬 Example Usage

```python
from transformers import pipeline

pipe = pipeline("text-generation", model="kberta2014/MedicalChatBot", tokenizer="kberta2014/MedicalChatBot")

prompt = '''### Instruction:
What are common symptoms of diabetes?

### Input:


### Response:
'''

output = pipe(prompt, max_new_tokens=200, temperature=0.7)
print(output[0]["generated_text"])
```

---

## 🤖 Gradio Chatbot Interface

```python
import gradio as gr
from transformers import pipeline

pipe = pipeline("text-generation", model="kberta2014/MedicalChatBot", tokenizer="kberta2014/MedicalChatBot")

def chat(instruction, input_text=""):
    prompt = f"### Instruction:\n{instruction}\n\n### Input:\n{input_text}\n\n### Response:\n"
    return pipe(prompt, max_new_tokens=200, temperature=0.7)[0]["generated_text"]

gr.Interface(fn=chat,
             inputs=["text", "text"],
             outputs="text",
             title="🩺 MedicalChatBot",
             description="Ask medical questions and get responses from a fine-tuned LLM"
).launch()
```

---

## 🏋️ Training Configuration

- **Model**: `mistralai/Mistral-7B-Instruct`
- **Dataset**: [`kberta2014/medical-chat-dataset`](https://huggingface.co/datasets/kberta2014/medical-chat-dataset)
- **Framework**: Hugging Face `transformers`, `peft`, `datasets`
- **PEFT Config**:
  - `r=8`, `lora_alpha=16`, `target_modules=["q_proj", "v_proj"]`
  - `lora_dropout=0.05`, `bias="none"`, `task_type="CAUSAL_LM"`
- **Training Time**: ~3 epochs on Colab T4
- **Batch Size**: 2
- **Learning Rate**: 2e-4
- **Precision**: bf16 / float16

---

## 📊 Training Metrics (Sample)

| Metric            | Value       |
|-------------------|-------------|
| Training loss     | ~1.02       |
| Eval loss         | ~0.94       |
| Perplexity        | ~2.6        |
| Epochs            | 3           |
| Trainable params  | ~7M (LoRA)  |

---

## 🧾 Citation

If you use this model in your research or application, please cite:

```bibtex
@misc{medicalchatbot2025,
  title={MedicalChatBot: A LoRA Fine-Tuned Mistral-7B Model for Medical QA},
  author={kberta2014},
  year={2025},
  url={https://huggingface.co/kberta2014/MedicalChatBot},
  note={Hugging Face model repository}
}
```

---

## ⚠️ Disclaimer

This model is intended for **research and educational purposes only**.  
It is **not a replacement for professional medical advice or diagnosis**.  
Always consult a licensed healthcare provider for real medical concerns.

---

## 📄 License

Apache 2.0 — same as the base model `mistralai/Mistral-7B-Instruct`.