def eval_archs_rgcn(dataset,
conv,
channel_size,
dropout,
lr,
wd,
runs,
splits,
train_examples,
val_examples,
models=[MonoRGCN]):
return eval_gnn(
dataset,
conv,
channel_size,
dropout,
lr,
wd,
heads=1,
attention_dropout=0.3, # dummy values for heads and attention_dropout
models=models,
num_runs=runs,
num_splits=splits,
test_score=True,
train_examples=train_examples,
val_examples=val_examples)
def eval_archs_gat(dataset,
channel_size,
dropout,
lr,
wd,
heads,
attention_dropout,
runs,
splits,
train_examples,
val_examples,
models=[MonoGAT],
isDirected=False):
if isDirected:
models = [MonoGAT]
return eval_gnn(dataset,
GATConv,
channel_size,
dropout,
lr,
wd,
heads=heads,
attention_dropout=attention_dropout,
models=models,
num_runs=runs,
num_splits=splits,
test_score=True,
train_examples=train_examples,
val_examples=val_examples)
def eval_archs_gcn(dataset,conv,channel_size,dropout,lr,wd,models=[MonoModel]):
if isDirected:
models = [MonoModel]
if conv == APPNP:
models = [MonoAPPNPModel]
return eval_gnn(dataset,conv,channel_size,dropout,lr,wd,heads=1,
models=models,num_runs=num_runs,num_splits=num_splits,
train_examples = args.train_examples, val_examples = args.val_examples)
def eval_archs_gat(dataset, channel_size, dropout, lr, wd, heads, attention_dropout=0.3, models=[MonoGAT]):
if isDirected:
models = [MonoGAT]
return eval_gnn(dataset, GATConv, channel_size, dropout, lr, wd, heads=heads, attention_dropout=attention_dropout,
models=models, num_runs=args.runs, num_splits=args.splits,
train_examples = args.train_examples, val_examples = args.val_examples)