Paper | Dataset | Install PyTorch Geometric
$ python train.py --help
usage: train.py [-h] [--model {mlp,gcn,gat,sage,tagcn}] [--hidden HIDDEN]
[--epoch EPOCH] [--lr LR] [--step STEP] [--gamma GAMMA]
[--bs BS] [--outpath OUTPATH] [--dataset_file DATASET_FILE]
optional arguments:
-h, --help show this help message and exit
--model {mlp,gcn,gat,sage,tagcn}
Type of model (default: sage)
--hidden HIDDEN Number of hidden/messsage passing layers (default: 2)
--epoch EPOCH Number of epochs to train (default: 100)
--lr LR Learning rate (default: 0.004)
--step STEP Step size for exponential learning rate scheduling
(default: 10)
--gamma GAMMA Decay rate for exponential learning rate scheduling
(default: 0.8)
--bs BS Batch size for training (default: 128)
--outpath OUTPATH Path to save results (default: ./results)
--dataset_file DATASET_FILE
House-GAN dataset .npy file path (default:
./data/housegan_clean_data.npy)