Current approach: training a hybrid CNN>GNN model for predicting properties of PETase variants.
Quickstart (in dev) 1 - create venv & install requirements:
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
2 - EDA & preprocess data + conduct demo training?
3 - training with GPU
python3 src/train.py --epochs 100 --batch_size 32 --device cuda
4 - evaluate
python3 src/evaluate.py --model