Directions

Set up Environment

Create a virtual environment: python3 -m venv .env

Load the virtual environment: source .env/bin/activate

Install dependencies: pip install -r requirements.txt

Set HF token export HF_HOME=<directory> export HUGGINGFACE_HUB_TOKEN=<token>

Simple LoRA finetuning run with Llama-2-7b on SciQ

python sciq.py

Set up Accelerate for Multi-GPU training

Type accelerate config in terminal and configure your environment

Run with Accelerate

To launch a run across multiple gpus, type accelerate launch sciq.py