Large Language Models (LLM)

Large Language Models (LLM)

This project implements a simple medical chatbot application using a pretrained GPT-2 model from the Hugging Face Transformers library. The chatbot generates human-like responses to user queries within a medical assistant context, showcasing the practical application of Large Language Models (LLMs).

Medical Chatbot Function:

• medical_chatbot(user_input, context) function

• Constructs a prompt combining

• A predefined context: "You are a helpful medical assistant."

• User input.

• Expected AI response section.

• Tokenizes the prompt and moves it to the appropriate device.

• Generates text using the model’s generate() method with:
  
  
  

  • • max_length=200

    • temperature=0.7

    • top_p=0.9

    • Sampling enabled (do_sample=True)

    • Decodes the generated output, isolates the AI's response, and returns it.
      
      

 Interactive Chat Loop:

• Continuously takes user input.

• Exits on typing 'exit' or 'quit'.

• Passes user queries to the chatbot function.

• Prints out AI-generated responses.
  
  

📈 Results:

• The chatbot generates coherent, context-aware medical replies based on the user's queries.

• Demonstrates GPT-2’s capability in domain-specific text generation without fine-tuning, simply by steering context.

Libraries Used:

• transformers

• torch

Notable Features:

✅ Real-time chatbot using a pretrained GPT-2 LLM
✅ Context-controlled text generation via prompt engineering
✅ Configurable sampling parameters for creativity control
✅ Simple interactive console-based chat loop

Conclusion:

This project effectively demonstrates how pretrained Large Language Models (LLMs) like GPT-2 can be leveraged for specialized applications such as a medical chatbot. By designing a clear prompt structure and adjusting generation parameters, the model produces sensible and context-aware responses, even without domain-specific fine-tuning.

Google colab Notebook page :