persona-db

This repository host the implementation of the persona databases project

Setup Instructions

1. Clone the repository and change to its working directory:

   git clone https://github.com/dice-group/persona-db.git
   cd persona-db
   

2. Create and activate a Conda environment:

   conda create -n .venv python=3.11.13 --no-default-packages
   conda activate .venv
   

3. Install pip and dependencies:

   conda install pip
   pip3 install -r requirements.txt
   

4. Configure your model and dataset split:

   • Open config.py.
     • Set MODEL_NAME to the path of your quantized model.
     • Set DATASET_SPLIT to the portion of the dataset you want to run. This helps avoid reprocessing data.
   • Examples:
     • DATASET_SPLIT = "train" # Runs on the entire training set
     • DATASET_SPLIT = "train[:100]" # Runs on the first 100 samples only
     • DATASET_SPLIT = "train[101:1001]" # Runs from sample 101 to 1000

5. Run the main script:

   python3 main.py
   

The link to the used dataset can be found in HuggingFace-PersonaHub-Persona