Ra-Is/medical-gen-small-CoT

Medical Generation Model (CoT Fine-Tuned)

Overview

This repository contains Ra-Is/medical-gen-small-CoT, a fine-tuned version of Ra-Is/medical-gen-small. This model incorporates Complex Chain of Thought (CoT) reasoning, improving medical diagnosis generation by enhancing logical and step-by-step reasoning in clinical scenarios.

Fine-tuned on structured medical datasets, this model is optimized to provide more contextually aware and clinically relevant responses, making it useful for medical professionals and AI-assisted healthcare solutions.

Model Details

• Base Model: Ra-Is/medical-gen-small
• Fine-tuning Technique: Complex Chain of Thought (CoT)
• Tokenizer: T5 tokenizer
• Training Data: Clinical scenarios, structured medical datasets
• Use Case: Medical diagnosis and treatment recommendation

Installation

To use this model, install the required libraries with pip:

pip install transformers
pip install tensorflow


# Load the fine-tuned model and tokenizer
from transformers import T5Tokenizer, TFT5ForConditionalGeneration

model_id = "Ra-Is/medical-gen-small-CoT"
model = TFT5ForConditionalGeneration.from_pretrained(model_id)
tokenizer = T5Tokenizer.from_pretrained(model_id)

# Prepare a sample input prompt
input_prompt = ("A 35-year-old female presents with a 2-week history of "
                "persistent cough, shortness of breath, and fatigue. She has "
                "a history of asthma and has recently been exposed to a sick "
                "family member with a respiratory infection. Chest X-ray shows "
                "bilateral infiltrates. What is the likely diagnosis, and what "
                "should be the treatment?")

# Tokenize the input
input_ids = tokenizer(input_prompt, return_tensors="tf").input_ids

# Generate the output (diagnosis)
outputs = model.generate(
        input_ids,
        max_length=512,
        num_beams=5,
        temperature=1,
        top_k=50,
        top_p=0.9,
        do_sample=True,  # Enable sampling
        early_stopping=True
    )

# Decode and print the output
generated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(generated_text)