diegovalsesia / ran-gnn-molpcba

RAN-GNN code for molpcba open graph benchmark

ogbg-molpcba

To train baselines with FLAG in the default setup, run

GIN+FLAG, the baseline model here.

python main_pyg.py --dataset ogbg-molpcba --gnn gin --step-size 8e-3

GIN+V+FLAG, the baseline model here.

python main_pyg.py --dataset ogbg-molpcba --gnn gin-virtual --step-size 8e-3

To train baselines with FLAG and random architecture in the default setup, run

RAN-GIN+FLAG, the baseline model here.

python main_pyg.py --dataset ogbg-molpcba --gnn randomgin --step-size 8e-3 --num_layer 12 --emb_dim 200

RAN-GIN+V+FLAG, the baseline model here.

python main_pyg.py --dataset ogbg-molpcba --gnn randomgin-virtual --step-size 8e-3 --num_layer 12 --emb_dim 248

THis code is derived from https://github.com/devnkong/FLAG