NOTE: if on runpod remember to set HF_HOME to workspace cache (and HF_TOKEN too)
WARN: on run it will download "meta-llama/Meta-Llama-3-8B-Instruct" if its not present in current HF_HOME
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txtI include https://github.com/vgel/repeng/ in this repo for the repeng library
start the server and show the web interface
uvicorn main:app --reloadvisit http://localhost:8000 or http://localhost:8000/docs
can load and test vectors or submit new ones to the training queue
python run_worker.pyworker will process the queue and train new vectors



