Medical-Diagnosis-COT-DeepSeek

🏥 AI Doctor V1

<div align="center">

</div>

📋 Overview

AI Doctor V1 is an advanced medical language model fine-tuned for clinical reasoning, diagnostic support, and treatment planning. Built on state-of-the-art transformer architecture and trained on medical reasoning datasets, this model demonstrates strong capabilities in understanding complex medical scenarios and providing structured clinical insights.

Key Capabilities

• 🔍 Medical question answering
• 🩺 Clinical knowledge base
• 💊 Accurate medical Q&A
• 📊 Medical Q&A
• 🧬 Evidence-based responses

🛠️ Model Details

• Base Architecture: Transformer-based causal language model
• Fine-tuning Framework: Unsloth + TRL (Transformer Reinforcement Learning)
• Training Method: Supervised Fine-Tuning (SFT)
• Training Dataset: FreedomIntelligence/medical-o1-reasoning-SFT
• Domain Specialization: Medical/Healthcare/Clinical
• Language: English
• Model Type: Causal Language Model for Medical Question Answering

📦 Installation

Required Dependencies

# Core libraries

!pip install unsloth # install unsloth
!pip install --force-reinstall --no-cache-dir --no-deps git+https://github.com/unslothai/unsloth.git # Also get the latest version Unsloth!

!pip install torch torchaudio transformers accelerate
!pip install trl datasets huggingface_hub wandb

# Quantization and optimization
# Install bitsandbytes
!pip install -U bitsandbytes

# Optional: For better performance
!pip install sentencepiece protobuf

🚀 Usage

Google Colab Setup

Step1: Import necessary libraries

from unsloth import FastLanguageModel
import torch
from trl import SFTTrainer
from unsloth import is_bfloat16_supported
from huggingface_hub import login
from transformers import TrainingArguments
from datasets import load_dataset
import wandb

Step2: Check HF token

from google.colab import userdata
hf_token = userdata.get('HF_TOKEN')
login(hf_token)

Optional: Check CUDA And GPU availability

import torch
print("CUDA available:", torch.cuda.is_available())
print("GPU device:", torch.cuda.get_device_name(0) if torch.cuda.is_available() else "No GPU")

Step3: Setup The Model

model_name = "NeoAivara/AI_Doctor_V1"
max_sequence_length = 2048
dtype = None
load_in_4bit = True

model, tokenizer = FastLanguageModel.from_pretrained(
    model_name = model_name,
    max_seq_length = max_sequence_length,
    dtype = dtype,
    load_in_4bit = load_in_4bit,
    token = hf_token
)

Step4: Setup system prompt

prompt = f"### Question: {question}\n### Answer:"

Step5: Run Inference on the model

# Define a test question
question = """What are the common symptoms of pneumonia?"""

FastLanguageModel.for_inference(model_lora)

# Tokenize the input
inputs = tokenizer([prompt_style.format(question, "")], return_tensors="pt").to("cuda")

# Generate a response
outputs = model_lora.generate (
    input_ids = inputs.input_ids,
    attention_mask = inputs.attention_mask,
    max_new_tokens = 1200,
    use_cache = True
)

# Decode the response tokens back to text
response = tokenizer.batch_decode(outputs)
answer = response.split("### Answer:")[-1].strip()

print(answer)

🎯 Example Use Cases

1. Symptom Analysis

question = "A patient presents with fever, cough, and night sweats for 3 weeks. What should be considered?"

2. Treatment Recommendations

question = "What is the first-line treatment for uncomplicated community-acquired pneumonia in adults?"

3. Diagnostic Reasoning

question = "Explain the diagnostic approach for a patient with suspected acute appendicitis."

4. Medical Education

question = "Explain the pathophysiology of type 2 diabetes mellitus."

⚙️ Generation Parameters

Recommended parameters for different use cases:

⚠️ Important Disclaimers

Medical Disclaimer

🚨 CRITICAL: FOR RESEARCH AND EDUCATIONAL PURPOSES ONLY 🚨

This AI model is designed for:

• ✅ Medical education and training
• ✅ Research and development
• ✅ Clinical decision support as an assistive tool
• ✅ Medical knowledge exploration

This AI model is NOT intended for:

• ❌ Direct patient care without physician oversight
• ❌ Replacement of professional medical judgment
• ❌ Emergency medical decisions
• ❌ Self-diagnosis or self-treatment

Key Limitations

1. Not a Licensed Healthcare Provider: This model cannot legally practice medicine or provide medical advice
2. Potential for Errors: May generate incorrect, incomplete, or outdated medical information
3. No Clinical Validation: Not validated in real-world clinical settings or approved by regulatory bodies
4. Bias and Training Data: Limited by the scope and potential biases in training data
5. No Patient Context: Cannot account for individual patient factors, allergies, drug interactions, or complete medical history
6. Liability: Users assume all responsibility for clinical decisions made using this tool

Recommended Usage

• Always verify information with current medical literature and guidelines
• Use in conjunction with clinical expertise and professional judgment
• Consult qualified healthcare professionals for all medical decisions
• Follow institutional protocols and regulatory requirements
• Document AI assistance appropriately in medical records per institutional policy

📊 Performance Notes

• Inference Speed: Optimized for GPU acceleration (CUDA)
• Memory Requirements: ~8-16GB GPU memory recommended for fp16
• Quantization: Supports 8-bit quantization via bitsandbytes for lower memory usage

🤝 Contributing

Contributions to improve the model, documentation, or use cases are welcome. Please ensure all contributions maintain the educational and research focus of this project.

📄 License

This model is released under the Apache 2.0 License. See LICENSE file for details.

🙏 Acknowledgments

• Training Framework: Unsloth and TRL
• Dataset: FreedomIntelligence/medical-o1-reasoning-SFT
• Community: Hugging Face and the open-source medical AI community

📧 Contact & Support

For questions, issues, or collaboration inquiries, please open an issue on the model repository.

---

<div align="center">

Remember: AI is a tool to augment, not replace, human medical expertise.

Always prioritize patient safety and professional medical judgment.

</div>