AI Healthcare Bot

Our Deep Learning project

Run Locally

  git clone https://github.com/johnwparks03/ai_healthcare_bot.git

Go to the project directory

  cd ai_healthcare_bot

Training Data

Our human-annotated medical question and answer dataset is stored in HealthData.

Training the Model

We used LLMbot/train_model.py to fine-tune the model.

Fine-Tuned Model Weights

fine_tuned_medalpaca/checkpoint-1095/adapter_model.safetensors contains the LoRA weights. You can apply these weights on top of the base model to recreate the fine-tuned model.

model.tar.gz contains the complete model after merging the base model and the fine-tuned weights.

Experiment with the model

Start the backend

Ensure you have Python and pip installed

Go to the backend directory

  cd backend

Install the required libraries

  pip install -r requirements.txt

Change to the app directory

  cd app

Start the backend server and listen on port 8000

  uvicorn main:app --reload --port 8000

Start the frontend

Ensure you have Node (https://nodejs.org/en) and npm installed.

Verify the installion using

node -v
npm -v

Install Angular globally (v16 or later recommended)

npm install -g @angular/cli

Change to the Angular project directory from the root project directory

cd frontend/ai_healthcare_bot

Install dependencies

npm install

Start the web server

ng serve

You can now view the website at http://localhost:4200/