Awesome
Code for the synthetic continuous task in the Continuous GFlowNets paper
To run the code:
conda create -n contgfn python=3.10
conda activate contfgn
pip install numpy matplotlib scikit-learn scipy torch tqdm
python main.py
and enjoy playing with the parameters. For examples, you could run
python main.py --loss tb --lr 1e-3 --delta 0.1 --PB tied --gamma_scheduler 0.5 --scheduler_milestone 2500 --n_components_s0 2 --n_components 1
If you want to get some nice plots, use wandb
. Don't forget to set USE_WANDB
to True
in the header of main.py
.