def build_default_args_for_unsupervised_node_classification(dataset):
args = {
"hidden_size": 128,
"num_shuffle": 5,
"cpu": True,
"enhance": None,
"save_dir": ".",
"seed": [0, 1, 2],
"walk_length": 80,
"walk_num": 40,
"window_size": 5,
"worker": 10,
"iteration": 10,
"task": "unsupervised_node_classification",
"model": "deepwalk",
"dataset": dataset,
}
args = get_extra_args(args)
return build_args_from_dict(args)
def build_default_args_for_heterogeneous_node_classification(dataset):
cpu = not torch.cuda.is_available()
args = {
"hidden_size": 128,
"patience": 100,
"max_epoch": 500,
"cpu": cpu,
"device_id": [0],
"lr": 0.005,
"weight_decay": 0.001,
"seed": [0, 1, 2],
"num_layers": 2,
"num_channels": 2,
"task": "heterogeneous_node_classification",
"model": "gtn",
"dataset": dataset,
}
args = get_extra_args(args)
return build_args_from_dict(args)
def build_default_args_for_node_classification(dataset):
cpu = not torch.cuda.is_available()
args = {
"lr": 0.01,
"weight_decay": 5e-4,
"max_epoch": 500,
"patience": 100,
"cpu": cpu,
"device_id": [0],
"seed": [0, 1, 2],
"dropout": 0.5,
"hidden_size": 64,
"num_layers": 2,
"filter_size": 5,
"task": "node_classification",
"model": "chebyshev",
"dataset": dataset
}
return build_args_from_dict(args)
def build_default_args():
cpu = not torch.cuda.is_available()
args = {
"lr": 0.01,
"weight_decay": 5e-4,
"max_epoch": 500,
"patience": 100,
"cpu": cpu,
"device_id": [0],
"seed": [0, 1, 2],
"dropout": 0.5,
"hidden_size": 128,
"num_layers": 2,
"sample_size": [10, 10],
"task": "node_classification",
"model": "graphsage",
"dataset": "cora"
}
return build_args_from_dict(args)
def get_unsupervised_nn_args():
default_dict = {
"hidden_size": 16,
"num_layers": 2,
"lr": 0.01,
"dropout": 0.0,
"patience": 1,
"max_epoch": 1,
"cpu": not torch.cuda.is_available(),
"weight_decay": 5e-4,
"num_shuffle": 2,
"save_dir": ",",
"enhance": None,
"device_id": [
0,
],
"task": "unsupervised_node_classification",
}
return build_args_from_dict(default_dict)
def build_default_args_for_multiplex_node_classification(dataset):
cpu = not torch.cuda.is_available()
args = {
"hidden_size": 128,
"cpu": cpu,
"enhance": None,
"save_dir": ".",
"seed": [0, 1, 2],
"lr": 0.025,
"walk_length": 80,
"walk_num": 40,
"batch_size": 1000,
"hop": 2,
"negative": 5,
"epochs": 1,
"task": "multiplex_node_classification",
"model": "hin2vec",
"dataset": dataset,
}
return build_args_from_dict(args)
def build_default_args_for_unsupervised_node_classification(dataset):
args = {
"hidden_size": 128,
"num_shuffle": 5,
"cpu": True,
"enhance": None,
"save_dir": ".",
"seed": [0, 1, 2],
"walk_length": 80,
"walk_num": 20,
"negative": 5,
"batch_size": 1000,
"alpha": 0.025,
"order": 3,
"task": "unsupervised_node_classification",
"model": "line",
"dataset": dataset,
}
args = get_extra_args(args)
return build_args_from_dict(args)
def build_default_args_for_node_classification(dataset):
cpu = not torch.cuda.is_available()
args = {
"lr": 0.01,
"weight_decay": 5e-4,
"max_epoch": 500,
"patience": 100,
"cpu": cpu,
"device_id": [0],
"seed": [0, 1, 2],
"dropout": 0.5,
"hidden_size": 64,
"propagation_type": 'appnp', #can also be 'ppnp'
"alpha": 0.1,
"num_iterations": 10,
"task": "node_classification",
"model": "ppnp",
"dataset": dataset
}
return build_args_from_dict(args)
def build_default_args_for_multiplex_node_classification(dataset):
args = {
"hidden_size": 128,
"cpu": True,
"enhance": None,
"save_dir": ".",
"seed": [0, 1, 2],
"walk_length": 80,
"walk_num": 40,
"window_size": 5,
"worker": 10,
"iteration": 10,
"schema": "No",
"task": "multiplex_node_classification",
"model": "metapath2vec",
"dataset": dataset
}
return build_args_from_dict(args)
def get_kg_default_args():
default_dict = {
"max_epoch": 2,
"num_bases": 5,
"num_layers": 2,
"hidden_size": 40,
"penalty": 0.001,
"sampling_rate": 0.001,
"dropout": 0.3,
"evaluate_interval": 2,
"patience": 20,
"lr": 0.001,
"weight_decay": 0,
"negative_ratio": 3,
"cpu": True,
"checkpoint": False,
"save_dir": ".",
"device_id": [0],
}
return build_args_from_dict(default_dict)
def get_default_args():
cuda_available = torch.cuda.is_available()
default_dict = {
"task": "unsupervised_graph_classification",
"gamma": 0.5,
"device_id": [0 if cuda_available else "cpu"],
"num_shuffle": 1,
"save_dir": ".",
"dropout": 0.5,
"patience": 1,
"epoch": 2,
"cpu": not cuda_available,
"lr": 0.001,
"weight_decay": 5e-4,
"checkpoint": False,
"activation": "relu",
"residual": False,
"norm": None,
}
return build_args_from_dict(default_dict)
def build_default_args_for_node_classification(dataset, missing_rate=0, num_layers=40):
cpu = not torch.cuda.is_available()
args = {
"lr": 0.005,
"weight_decay": 5e-4,
"max_epoch": 1000,
"patience": 1000,
"cpu": cpu,
"device_id": [0],
"seed": [0, 1, 2, 3, 4],
"missing_rate": missing_rate,
"norm_mode": "PN",
"norm_scale": 10,
"dropout": 0.6,
"num_layers": num_layers,
"task": "node_classification",
"model": "sgcpn",
"dataset": dataset,
}
return build_args_from_dict(args)
def get_default_args():
cuda_available = torch.cuda.is_available()
default_dict = {
'task': 'graph_classification',
'hidden_size': 64,
'dropout': 0.5,
'patience': 1,
'max_epoch': 2,
'cpu': not cuda_available,
'lr': 0.001,
'kfold': False,
'seed': [0],
'weight_decay': 5e-4,
'gamma': 0.5,
'train_ratio': 0.7,
'test_ratio': 0.1,
'device_id': [0 if cuda_available else 'cpu'],
'degree_feature': False
}
return build_args_from_dict(default_dict)
def build_default_args_for_unsupervised_node_classification(dataset):
args = {
"hidden_size": 128,
"num_shuffle": 5,
"cpu": True,
"enhance": None,
"save_dir": ".",
"seed": [0, 1, 2],
"lr": 0.001,
"max_epoch": 500,
"hidden_size1": 1000,
"hidden_size2": 128,
"noise": 0.2,
"alpha": 0.1,
"step": 10,
"task": "unsupervised_node_classification",
"model": "dngr",
"dataset": dataset,
}
return build_args_from_dict(args)
def get_default_args():
default_dict = {
"hidden_size": 16,
"dropout": 0.5,
"patience": 2,
"device_id": [0],
"max_epoch": 3,
"sampler": "none",
"num_layers": 2,
"cpu": True,
"lr": 0.01,
"weight_decay": 5e-4,
"missing_rate": -1,
"task": "node_classification",
"dataset": "cora",
"checkpoint": False,
"sampler": "none",
"auxiliary_task": "none",
}
return build_args_from_dict(default_dict)
def build_default_args_for_node_classification(dataset):
cpu = not torch.cuda.is_available()
args = {
"lr": 0.01,
"weight_decay": 5e-4,
"max_epoch": 200,
"patience": 100,
"cpu": cpu,
"device_id": [0],
"seed": [0, 1, 2],
"dropout": 0.5,
"hidden_size": 16,
"eps": 0.01,
"update_freq": 128,
"alpha": None,
"gamma": 100,
"task": "node_classification",
"model": "ssp",
"dataset": dataset
}
return build_args_from_dict(args)
def build_default_args_for_node_classification(dataset):
cpu = not torch.cuda.is_available()
args = {
"lr": 0.005,
"weight_decay": 5e-4,
"max_epoch": 1000,
"patience": 100,
"cpu": cpu,
"device_id": [0],
"seed": [
72,
],
"dropout": 0.6,
"hidden_size": 8,
"alpha": 0.2,
"nheads": 8,
"task": "node_classification",
"model": "gat",
"dataset": dataset,
}
return build_args_from_dict(args)
def get_default_args():
cuda_available = torch.cuda.is_available()
default_dict = {
"task": "graph_classification",
"hidden_size": 32,
"dropout": 0.5,
"patience": 1,
"max_epoch": 2,
"cpu": not cuda_available,
"lr": 0.001,
"kfold": False,
"seed": [0],
"weight_decay": 5e-4,
"gamma": 0.5,
"train_ratio": 0.7,
"test_ratio": 0.1,
"device_id": [0 if cuda_available else "cpu"],
"sampler": "none",
"degree_feature": False,
}
return build_args_from_dict(default_dict)
def get_gnn_link_prediction_args():
args = {
"hidden_size": 32,
"dataset": "cora",
"model": "gcn",
"task": "link_prediction",
"lr": 0.005,
"weight_decay": 5e-4,
"max_epoch": 60,
"patience": 2,
"num_layers": 2,
"evaluate_interval": 1,
"cpu": True,
"device_id": [0],
"dropout": 0.5,
"checkpoint": False,
"save_dir": ".",
"activation": "relu",
"residual": False,
"norm": None,
}
return build_args_from_dict(args)
def get_strategies_for_pretrain_args():
cuda_available = torch.cuda.is_available()
args = {
"dataset": "test_bio",
"model": "stpgnn",
"task": "pretrain",
"batch_size": 32,
"num_layers": 2,
"JK": "last",
"hidden_size": 32,
"num_workers": 2,
"finetune": False,
"dropout": 0.5,
"lr": 0.001,
"cpu": not cuda_available,
"device_id": [0],
"weight_decay": 5e-4,
"max_epoch": 3,
"patience": 2,
"output_model_file": "./saved"
}
return build_args_from_dict(args)
def build_default_args_for_node_classification(dataset):
cpu = not torch.cuda.is_available()
args = {
"lr": 0.001,
"weight_decay": 0,
"max_epoch": 1000,
"max_epochs": 1000,
"patience": 20,
"cpu": cpu,
"device_id": [0],
"seed": [0],
"num_shuffle": 5,
"dropout": 0.0,
"hidden_size": 512,
"num_layers": 2,
"sampler": "none",
"task": "unsupervised_node_classification",
"model": "dgi",
"dataset": dataset,
"save_dir": "./saved",
"enhance": None,
}
return build_args_from_dict(args)
def build_default_args_for_multiplex_link_prediction(dataset):
cpu = not torch.cuda.is_available()
args = {
"hidden_size": 200,
"cpu": cpu,
"eval_type": "all",
"seed": [0, 1, 2],
"walk_length": 10,
"walk_num": 10,
"window_size": 5,
"worker": 10,
"epoch": 20,
"batch_size": 256,
"edge_dim": 10,
"att_dim": 20,
"negative_samples": 5,
"neighbor_samples": 10,
"schema": None,
"task": "multiplex_link_prediction",
"model": "gatne",
"dataset": dataset,
}
return build_args_from_dict(args)
def build_default_args_for_node_classification(dataset):
cpu = not torch.cuda.is_available()
args = {
"lr": 0.01,
"weight_decay": 5e-4,
"max_epoch": 1000,
"max_epochs": 1000,
"patience": 100,
"cpu": cpu,
"device_id": [0],
"seed": [42],
"dropout": 0.5,
"hidden_size": 256,
"num_layers": 32,
"lmbda": 0.5,
"wd1": 0.001,
"wd2": 5e-4,
"alpha": 0.1,
"task": "node_classification",
"model": "gcnii",
"dataset": dataset,
}
return build_args_from_dict(args)
def build_default_args_for_unsupervised_node_classification(dataset):
args = {
"hidden_size": 128,
"num_shuffle": 5,
"cpu": True,
"enhance": None,
"save_dir": ".",
"seed": [0, 1, 2],
"lr": 0.001,
"max_epoch": 500,
"hidden_size1": 1000,
"hidden_size2": 128,
"droput": 0.5,
"alpha": 0.1,
"beta": 5,
"nu1": 1e-4,
"nu2": 1e-3,
"task": "unsupervised_node_classification",
"model": "sdne",
"dataset": dataset,
}
args = get_extra_args(args)
return build_args_from_dict(args)
def get_unsupervised_nn_args():
default_dict = {
"hidden_size": 16,
"num_layers": 2,
"lr": 0.01,
"dropout": 0.0,
"patience": 1,
"max_epoch": 1,
"cpu": not torch.cuda.is_available(),
"weight_decay": 5e-4,
"num_shuffle": 2,
"save_dir": "./embedding",
"enhance": None,
"device_id": [
0,
],
"task": "unsupervised_node_classification",
"checkpoint": False,
"load_emb_path": None,
"sampling": False,
"sample_size": 20,
"training_percents": [0.1],
}
return build_args_from_dict(default_dict)
def build_default_args_for_graph_classification(dataset):
cpu = not torch.cuda.is_available()
args = {
"hidden_size": 128,
"dropout": 0.0,
"pooling": 0.5,
"batch_size": 64,
"train_ratio": 0.8,
"test_ratio": 0.1,
"lr": 0.001,
"weight_decay": 0.001,
"patience": 100,
"max_epoch": 500,
"sample_neighbor": True,
"sparse_attention": True,
"structure_learning": True,
"cpu": cpu,
"device_id": [0],
"seed": [777],
"task": "graph_classification",
"model": "hgpsl",
"dataset": dataset,
}
return build_args_from_dict(args)
def build_default_args_for_node_classification(dataset):
cpu = not torch.cuda.is_available()
args = {
"lr": 0.01,
"weight_decay": 0.0005,
"max_epoch": 1000,
"patience": 100,
"cpu": cpu,
"device_id": [0],
"seed": [1],
"n_dropout": 0.90,
"adj_dropout": 0.05,
"hidden_size": 128,
"aug_adj": False,
"improved": False,
"n_pool": 4,
"pool_rate": [0.7, 0.5, 0.5, 0.4],
"activation": "relu",
"task": "node_classification",
"model": "unet",
"dataset": dataset,
"missing_rate": -1,
}
return build_args_from_dict(args)
def default_parameter():
args = {
"seed": [0, 1, 2],
}
return build_args_from_dict(args)
def test_cora():
args = build_args_from_dict({'dataset': 'cora'})
assert args.dataset == 'cora'
cora = build_dataset(args)
assert cora.data.num_nodes == 2708
assert cora.data.num_edges == 10556
def get_diff_args(args1, args2):
d1 = copy.deepcopy(args1.__dict__)
d2 = args2.__dict__
for k in d2.keys():
d1.pop(k, None)
return build_args_from_dict(d1)
