def load_ogb(name, dataset_dir):
if name[:4] == 'ogbn':
dataset = PygNodePropPredDataset(name=name, root=dataset_dir)
splits = dataset.get_idx_split()
split_names = ['train_mask', 'val_mask', 'test_mask']
for i, key in enumerate(splits.keys()):
mask = index2mask(splits[key], size=dataset.data.y.shape[0])
set_dataset_attr(dataset, split_names[i], mask, len(mask))
edge_index = to_undirected(dataset.data.edge_index)
set_dataset_attr(dataset, 'edge_index', edge_index,
edge_index.shape[1])
elif name[:4] == 'ogbg':
dataset = PygGraphPropPredDataset(name=name, root=dataset_dir)
splits = dataset.get_idx_split()
split_names = [
'train_graph_index', 'val_graph_index', 'test_graph_index'
]
for i, key in enumerate(splits.keys()):
id = splits[key]
set_dataset_attr(dataset, split_names[i], id, len(id))
elif name[:4] == "ogbl":
dataset = PygLinkPropPredDataset(name=name, root=dataset_dir)
splits = dataset.get_edge_split()
id = splits['train']['edge'].T
if cfg.dataset.resample_negative:
set_dataset_attr(dataset, 'train_pos_edge_index', id, id.shape[1])
# todo: applying transform for negative sampling is very slow
dataset.transform = neg_sampling_transform
else:
id_neg = negative_sampling(edge_index=id,
num_nodes=dataset.data.num_nodes[0],
num_neg_samples=id.shape[1])
id_all = torch.cat([id, id_neg], dim=-1)
label = get_link_label(id, id_neg)
set_dataset_attr(dataset, 'train_edge_index', id_all,
id_all.shape[1])
set_dataset_attr(dataset, 'train_edge_label', label, len(label))
id, id_neg = splits['valid']['edge'].T, splits['valid']['edge_neg'].T
id_all = torch.cat([id, id_neg], dim=-1)
label = get_link_label(id, id_neg)
set_dataset_attr(dataset, 'val_edge_index', id_all, id_all.shape[1])
set_dataset_attr(dataset, 'val_edge_label', label, len(label))
id, id_neg = splits['test']['edge'].T, splits['test']['edge_neg'].T
id_all = torch.cat([id, id_neg], dim=-1)
label = get_link_label(id, id_neg)
set_dataset_attr(dataset, 'test_edge_index', id_all, id_all.shape[1])
set_dataset_attr(dataset, 'test_edge_label', label, len(label))
else:
raise ValueError('OGB dataset: {} non-exist')
return dataset
def eval_with_partition(args):
model_load_path = args.model_load_path
print("Starting evaluating model stored at", model_load_path)
device = torch.device("cuda")
dataset = PygNodePropPredDataset(name=args.dataset, root=args.data_folder)
graph = dataset[0]
adj = SparseTensor(row=graph.edge_index[0], col=graph.edge_index[1])
if args.self_loop:
adj = adj.set_diag()
graph.edge_index = add_self_loops(edge_index=graph.edge_index,
num_nodes=graph.num_nodes)[0]
split_idx = dataset.get_idx_split()
evaluator = Evaluator(args.dataset)
args.in_channels = graph.x.size(-1)
args.num_tasks = dataset.num_classes
# print('%s' % args)
model = DeeperGCN(args).to(device)
ckpt = torch.load(model_load_path)
model.load_state_dict(ckpt['model_state_dict'])
res = test_with_partition(model,
graph,
adj,
split_idx,
num_clusters=args.eval_cluster_number,
partition_method=args.partition_method,
evaluator=evaluator,
device=device)
print(res)
return res
def eval_model(params):
model_load_path, args = params
if os.path.isdir(args.model_load_path):
model_load_dir = args.model_load_path
model_load_path = os.path.join(model_load_dir, model_load_path)
print("Starting evaluating model stored at", model_load_path)
dataset = PygNodePropPredDataset(name=args.dataset, root=args.data_folder)
graph = dataset[0]
if args.self_loop:
graph.edge_index = add_self_loops(edge_index=graph.edge_index,
num_nodes=graph.num_nodes)[0]
split_idx = dataset.get_idx_split()
evaluator = Evaluator(args.dataset)
args.in_channels = graph.x.size(-1)
args.num_tasks = dataset.num_classes
model = DeeperGCN(args)
ckpt = torch.load(model_load_path, map_location=torch.device('cpu'))
model.load_state_dict(ckpt['model_state_dict'])
test_res = test(model, graph.x, graph.edge_index, graph.y, split_idx,
evaluator)
test_res["model_load_path"] = model_load_path
return test_res
def main():
parser = argparse.ArgumentParser(description='OGBN-Arxiv (GNN)')
parser.add_argument('--device', type=int, default=0)
parser.add_argument('--log_steps', type=int, default=1)
parser.add_argument('--use_sage', action='store_true')
parser.add_argument('--num_layers', type=int, default=3)
parser.add_argument('--hidden_channels', type=int, default=256)
parser.add_argument('--dropout', type=float, default=0.5)
parser.add_argument('--lr', type=float, default=0.01)
parser.add_argument('--epochs', type=int, default=500)
parser.add_argument('--runs', type=int, default=10)
parser.add_argument('--kind', type=str, default="ReLU")
args = parser.parse_args()
print(args)
device = f'cuda:{args.device}' if torch.cuda.is_available() else 'cpu'
device = 'cpu'
device = torch.device(device)
dataset = PygNodePropPredDataset(name='ogbn-arxiv',
transform=T.ToSparseTensor())
data = dataset[0]
data.adj_t = data.adj_t.to_symmetric()
data = data.to(device)
split_idx = dataset.get_idx_split()
train_idx = split_idx['train'].to(device)
if args.use_sage:
model = SAGE(data.num_features, args.hidden_channels,
dataset.num_classes, args.num_layers,
args.dropout, kind=args.kind).to(device)
else:
model = GCN(data.num_features, args.hidden_channels,
dataset.num_classes, args.num_layers,
args.dropout, kind=args.kind).to(device)
evaluator = Evaluator(name='ogbn-arxiv')
logger = Logger(args.runs, args)
for run in range(args.runs):
model.reset_parameters()
optimizer = torch.optim.Adam(model.parameters(), lr=args.lr)
for epoch in range(1, 1 + args.epochs):
loss = train(model, data, train_idx, optimizer)
result = test(model, data, split_idx, evaluator)
logger.add_result(run, result)
if epoch % args.log_steps == 0:
train_acc, valid_acc, test_acc = result
print(f'Run: {run + 1:02d}, '
f'Epoch: {epoch:02d}, '
f'Loss: {loss:.4f}, '
f'Train: {100 * train_acc:.2f}%, '
f'Valid: {100 * valid_acc:.2f}% '
f'Test: {100 * test_acc:.2f}%')
logger.print_statistics(run)
logger.print_statistics()
def main():
parser = argparse.ArgumentParser(description="OGBN-Arxiv (MLP)")
parser.add_argument("--device", type=int, default=0)
parser.add_argument("--log_steps", type=int, default=1)
parser.add_argument("--use_node_embedding", action="store_true")
parser.add_argument("--num_layers", type=int, default=3)
parser.add_argument("--hidden_channels", type=int, default=256)
parser.add_argument("--dropout", type=float, default=0.5)
parser.add_argument("--lr", type=float, default=0.01)
parser.add_argument("--epochs", type=int, default=500)
parser.add_argument("--runs", type=int, default=10)
args = parser.parse_args()
print(args)
device = f"cuda:{args.device}" if torch.cuda.is_available() else "cpu"
device = torch.device(device)
dataset = PygNodePropPredDataset(name="ogbn-arxiv")
split_idx = dataset.get_idx_split()
data = dataset[0]
x = data.x
if args.use_node_embedding:
embedding = torch.load("embedding.pt", map_location="cpu")
x = torch.cat([x, embedding], dim=-1)
x = x.to(device)
y_true = data.y.to(device)
train_idx = split_idx["train"].to(device)
model = MLP(
x.size(-1),
args.hidden_channels,
dataset.num_classes,
args.num_layers,
args.dropout,
).to(device)
evaluator = Evaluator(name="ogbn-arxiv")
logger = Logger(args.runs, args)
for run in range(args.runs):
model.reset_parameters()
optimizer = torch.optim.Adam(model.parameters(), lr=args.lr)
for epoch in range(1, 1 + args.epochs):
loss = train(model, x, y_true, train_idx, optimizer)
result = test(model, x, y_true, split_idx, evaluator)
logger.add_result(run, result)
if epoch % args.log_steps == 0:
train_acc, valid_acc, test_acc = result
print(f"Run: {run + 1:02d}, "
f"Epoch: {epoch:02d}, "
f"Loss: {loss:.4f}, "
f"Train: {100 * train_acc:.2f}%, "
f"Valid: {100 * valid_acc:.2f}%, "
f"Test: {100 * test_acc:.2f}%")
logger.print_statistics(run)
logger.print_statistics()
def get_product_clusters():
dataset_name = "ogbn-products"
dataset = PygNodePropPredDataset(name=dataset_name)
print('The {} dataset has {} graph'.format(dataset_name, len(dataset)))
data = dataset[0]
print(data)
split_idx = dataset.get_idx_split()
train_idx = split_idx['train']
val_idx = split_idx['valid']
test_idx = split_idx['test']
train_mask = torch.zeros(data.num_nodes, dtype=torch.bool)
train_mask[train_idx] = True
data['train_mask'] = train_mask
val_mask = torch.zeros(data.num_nodes, dtype=torch.bool)
val_mask[val_idx] = True
data['valid_mask'] = val_mask
test_mask = torch.zeros(data.num_nodes, dtype=torch.bool)
test_mask[test_idx] = True
data['test_mask'] = test_mask
cluster_data = ClusterData(data, num_parts=15000, save_dir="dataset")
return cluster_data, dataset, data, split_idx
def main():
args = ArgsInit().args
dataset = PygNodePropPredDataset(name=args.dataset)
graph = dataset[0]
if args.self_loop:
graph.edge_index = add_self_loops(edge_index=graph.edge_index,
num_nodes=graph.num_nodes)[0]
split_idx = dataset.get_idx_split()
evaluator = Evaluator(args.dataset)
args.in_channels = graph.x.size(-1)
args.num_tasks = dataset.num_classes
print(args)
model = DeeperGCN(args)
print(model)
model.load_state_dict(torch.load(args.model_load_path)['model_state_dict'])
result = test(model, graph.x, graph.edge_index, graph.y, split_idx,
evaluator)
print(result)
model.print_params(final=True)
def main():
parser = argparse.ArgumentParser(description='OGBN-Proteins (MLP)')
parser.add_argument('--device', type=int, default=0)
parser.add_argument('--log_steps', type=int, default=1)
parser.add_argument('--use_node_embedding', action='store_true')
parser.add_argument('--num_layers', type=int, default=3)
parser.add_argument('--hidden_channels', type=int, default=256)
parser.add_argument('--dropout', type=float, default=0.5)
parser.add_argument('--lr', type=float, default=0.01)
parser.add_argument('--epochs', type=int, default=1000)
parser.add_argument('--eval_steps', type=int, default=5)
parser.add_argument('--runs', type=int, default=10)
args = parser.parse_args()
print(args)
device = f'cuda:{args.device}' if torch.cuda.is_available() else 'cpu'
device = torch.device(device)
dataset = PygNodePropPredDataset(name='ogbn-proteins')
split_idx = dataset.get_idx_split()
data = dataset[0]
x = scatter(data.edge_attr, data.edge_index[0], dim=0,
dim_size=data.num_nodes, reduce='mean').to('cpu')
if args.use_node_embedding:
embedding = torch.load('embedding.pt', map_location='cpu')
x = torch.cat([x, embedding], dim=-1)
x = x.to(device)
y_true = data.y.to(device)
train_idx = split_idx['train'].to(device)
model = MLP(x.size(-1), args.hidden_channels, 112, args.num_layers,
args.dropout).to(device)
evaluator = Evaluator(name='ogbn-proteins')
logger = Logger(args.runs, args)
for run in range(args.runs):
model.reset_parameters()
optimizer = torch.optim.Adam(model.parameters(), lr=args.lr)
for epoch in range(1, 1 + args.epochs):
loss = train(model, x, y_true, train_idx, optimizer)
if epoch % args.eval_steps == 0:
result = test(model, x, y_true, split_idx, evaluator)
logger.add_result(run, result)
if epoch % args.log_steps == 0:
train_rocauc, valid_rocauc, test_rocauc = result
print(f'Run: {run + 1:02d}, '
f'Epoch: {epoch:02d}, '
f'Loss: {loss:.4f}, '
f'Train: {100 * train_rocauc:.2f}%, '
f'Valid: {100 * valid_rocauc:.2f}% '
f'Test: {100 * test_rocauc:.2f}%')
logger.print_statistics(run)
logger.print_statistics()
def main():
parser = argparse.ArgumentParser(description='OGBN-Arxiv (GNN)')
parser.add_argument('--device', type=int, default=0)
parser.add_argument('--log_steps', type=int, default=1)
parser.add_argument('--use_sage', action='store_true')
parser.add_argument('--num_layers', type=int, default=3)
parser.add_argument('--hidden_channels', type=int, default=256)
parser.add_argument('--dropout', type=float, default=0.5)
parser.add_argument('--lr', type=float, default=0.01)
parser.add_argument('--epochs', type=int, default=3000)
parser.add_argument('--runs', type=int, default=1)
args = parser.parse_args()
print(args)
device = f'cuda:{args.device}' if torch.cuda.is_available() else 'cpu'
device = torch.device(device)
dataset = PygNodePropPredDataset(name='ogbn-arxiv',root='/mnt/ogbdata',
transform=T.ToSparseTensor())
data = dataset[0]
data.adj_t = data.adj_t.to_symmetric()
data = data.to(device)
split_idx = dataset.get_idx_split()
print(split_idx['train'].nonzero().size(0)/len(split_idx['train']),
split_idx['valid'].nonzero().size(0)/len(split_idx['train']),
split_idx['test'].nonzero().size(0)/len(split_idx['train']))
def main():
parser = argparse.ArgumentParser(description='OGBN-Products (MLP)')
parser.add_argument('--device', type=int, default=0)
parser.add_argument('--log_steps', type=int, default=1)
parser.add_argument('--use_node_embedding', action='store_true')
parser.add_argument('--num_layers', type=int, default=3)
parser.add_argument('--hidden_channels', type=int, default=256)
parser.add_argument('--dropout', type=float, default=0.0)
parser.add_argument('--lr', type=float, default=0.01)
parser.add_argument('--epochs', type=int, default=300)
parser.add_argument('--runs', type=int, default=10)
args = parser.parse_args()
print(args)
device = f'cuda:{args.device}' if torch.cuda.is_available() else 'cpu'
device = torch.device(device)
dataset = PygNodePropPredDataset(name='ogbn-products')
split_idx = dataset.get_idx_split()
data = dataset[0]
x = data.x
if args.use_node_embedding:
embedding = np.load('./embed_results/embeddings.npy')
embedding = torch.from_numpy(embedding).float()
x = torch.cat([x, embedding], dim=-1)
x = x.to(device)
y_true = data.y.to(device)
train_idx = split_idx['train'].to(device)
model = MLP(x.size(-1), args.hidden_channels, dataset.num_classes,
args.num_layers, args.dropout).to(device)
evaluator = Evaluator(name='ogbn-products')
logger = Logger(args.runs, args)
for run in range(args.runs):
model.reset_parameters()
optimizer = torch.optim.Adam(model.parameters(), lr=args.lr)
for epoch in range(1, 1 + args.epochs):
loss = train(model, x, y_true, train_idx, optimizer)
result = test(model, x, y_true, split_idx, evaluator)
logger.add_result(run, result)
if epoch % args.log_steps == 0:
train_acc, valid_acc, test_acc = result
print(f'Run: {run + 1:02d}, '
f'Epoch: {epoch:02d}, '
f'Loss: {loss:.4f}, '
f'Train: {100 * train_acc:.2f}%, '
f'Valid: {100 * valid_acc:.2f}%, '
f'Test: {100 * test_acc:.2f}%')
logger.print_statistics(run)
logger.print_statistics()
total_params = sum(p.numel() for p in model.parameters())
print(f'mlp total params are {total_params}')
def load_ogb_2(dataset):
## Load the dataset
## Setup PyTorch
device_name = 'cuda:0' if torch.cuda.is_available() else 'cpu'
device = torch.device(device_name)
dataset = PygNodePropPredDataset(name=dataset,
transform=T.ToSparseTensor())
ogb_data = dataset[0]
# TODO: Not sure how to format adj_t...
ogb_data.adj_t = ogb_data.adj_t.to_symmetric()
ogb_data = ogb_data.to(device)
split_idx = dataset.get_idx_split()
train_idx, valid_idx, test_idx = split_idx["train"], split_idx[
"valid"], split_idx["test"]
train_idx, valid_idx, split_idx = train_idx.numpy(), valid_idx.numpy(
), test_idx.numpy()
# Convert OGB's data split pytorch index vectors to Wang's data split numpy boolean masks
train_mask_2 = indexes2booleanvec(ogb_data.num_nodes, train_idx)
val_mask_2 = indexes2booleanvec(ogb_data.num_nodes, valid_idx)
test_mask_2 = indexes2booleanvec(ogb_data.num_nodes, test_idx)
# Add 1's down the diagonal of adj_t
adj_t = ogb_data.adj_t.to_torch_sparse_coo_tensor()
adj_t = adj_t + sparse_identity(adj_t.shape[0])
# Convert OGB's adjacency SparseTensor to Wang's adjacency index matrix (Nx2)
adj_2_0 = adj_t.coalesce().indices().numpy()
adj_2_0 = adj_2_0.T.astype('int32')
##adj_2_0 = np.vstack((adj_2_0, np.array([[i,i] for i in range(ogb_data.num_nodes)])))
adj_2_1 = adj_t.coalesce().values().numpy().astype('float64')
adj_2_2 = tuple(adj_t.size())
#TODO: Fix the adjacency matrix, bc it probably is symmetric with identity
adj_2 = (adj_2_0, adj_2_1, adj_2_2)
from sklearn.preprocessing import OneHotEncoder
labels_2 = ogb_data.y.numpy()
labels_2 = OneHotEncoder(sparse=False).fit_transform(labels_2)
#TODO: I don't know if this feature vector will work
# OGB used a skip-gram encoding,
# whereas Wang's Citeseer just used normalized rows with 1-0 for different words
x = ogb_data.x + 1.5
norm_x = np.apply_along_axis(np.linalg.norm, 1, x)
x = x / norm_x[:, None]
x = x.to_sparse()
features_2_0 = x.indices().numpy().T.astype('int32')
features_2_1 = x.values().numpy()
#features_2_1 = 1.5 + features_2_1
features_2_1 = features_2_1.astype('float64')
features_2_2 = tuple(x.size())
features_2 = features_2_0, features_2_1, features_2_2
data2 = features_2, labels_2, adj_2, train_mask_2, val_mask_2, test_mask_2
return data2
def main_fixed_mask(args):
device = torch.device("cuda:" + str(args.device))
dataset = PygNodePropPredDataset(name=args.dataset)
data = dataset[0]
split_idx = dataset.get_idx_split()
evaluator = Evaluator(args.dataset)
x = data.x.to(device)
y_true = data.y.to(device)
train_idx = split_idx['train'].to(device)
edge_index = data.edge_index.to(device)
edge_index = to_undirected(edge_index, data.num_nodes)
if args.self_loop:
edge_index = add_self_loops(edge_index, num_nodes=data.num_nodes)[0]
args.in_channels = data.x.size(-1)
args.num_tasks = dataset.num_classes
model = DeeperGCN(args).to(device)
pruning.add_mask(model, args.num_layers)
for name, param in model.named_parameters():
if 'mask' in name:
param.requires_grad = False
optimizer = torch.optim.Adam(model.parameters(), lr=args.lr)
results = {'highest_valid': 0, 'final_train': 0, 'final_test': 0, 'highest_train': 0, 'epoch': 0}
start_epoch = 1
for epoch in range(start_epoch, args.epochs + 1):
epoch_loss = train_fixed(model, x, edge_index, y_true, train_idx, optimizer, args)
result = test(model, x, edge_index, y_true, split_idx, evaluator)
train_accuracy, valid_accuracy, test_accuracy = result
if valid_accuracy > results['highest_valid']:
results['highest_valid'] = valid_accuracy
results['final_train'] = train_accuracy
results['final_test'] = test_accuracy
results['epoch'] = epoch
print(time.strftime("%Y-%m-%d %H:%M:%S", time.localtime()) + ' | ' +
'Baseline (FIX Mask) Epoch:[{}/{}]\t LOSS:[{:.4f}] Train :[{:.2f}] Valid:[{:.2f}] Test:[{:.2f}] | Update Test:[{:.2f}] at epoch:[{}]'
.format(epoch, args.epochs, epoch_loss, train_accuracy * 100,
valid_accuracy * 100,
test_accuracy * 100,
results['final_test'] * 100,
results['epoch']))
print("=" * 120)
print("syd final: Baseline, Train:[{:.2f}] Best Val:[{:.2f}] at epoch:[{}] | Final Test Acc:[{:.2f}]"
.format( results['final_train'] * 100,
results['highest_valid'] * 100,
results['epoch'],
results['final_test'] * 100))
print("=" * 120)
def main():
parser = argparse.ArgumentParser(description='OGBN-Products (Cluster-GCN)')
parser.add_argument('--device', type=int, default=0)
parser.add_argument('--log_steps', type=int, default=1)
parser.add_argument('--num_partitions', type=int, default=15000)
parser.add_argument('--num_workers', type=int, default=6)
parser.add_argument('--num_layers', type=int, default=3)
parser.add_argument('--hidden_channels', type=int, default=256)
parser.add_argument('--dropout', type=float, default=0.0)
parser.add_argument('--batch_size', type=int, default=256)
parser.add_argument('--lr', type=float, default=0.01)
parser.add_argument('--epochs', type=int, default=20)
parser.add_argument('--runs', type=int, default=10)
args = parser.parse_args()
print(args)
device = f'cuda:{args.device}' if torch.cuda.is_available() else 'cpu'
device = torch.device(device)
dataset = PygNodePropPredDataset(name='ogbn-products')
splitted_idx = dataset.get_idx_split()
data = dataset[0]
# Convert split indices to boolean masks and add them to `data`.
for key, idx in splitted_idx.items():
mask = torch.zeros(data.num_nodes, dtype=torch.bool)
mask[idx] = True
data[f'{key}_mask'] = mask
cluster_data = ClusterData(data,
num_parts=args.num_partitions,
recursive=False,
save_dir=dataset.processed_dir)
loader = ClusterLoader(cluster_data,
batch_size=args.batch_size,
shuffle=True,
num_workers=args.num_workers)
model = SAGE(data.x.size(-1), args.hidden_channels, 47, args.num_layers,
args.dropout).to(device)
evaluator = Evaluator(name='ogbn-products')
logger = Logger(args.runs, args)
for run in range(args.runs):
model.reset_parameters()
optimizer = torch.optim.Adam(model.parameters(), lr=args.lr)
for epoch in range(1, 1 + args.epochs):
loss = train(model, loader, optimizer, device)
if epoch % args.log_steps == 0:
print(f'Run: {run + 1:02d}, '
f'Epoch: {epoch:02d}, '
f'Loss: {loss:.4f}')
result = test(model, data, evaluator)
logger.add_result(run, result)
logger.print_statistics(run)
logger.print_statistics()
def main():
parser = argparse.ArgumentParser(description='OGBN-Products (SIGN)')
parser.add_argument('--device', type=int, default=0)
parser.add_argument('--log_steps', type=int, default=1)
parser.add_argument('--num_layers', type=int, default=3)
parser.add_argument('--hidden_channels', type=int, default=256)
parser.add_argument('--dropout', type=float, default=0.5)
parser.add_argument('--lr', type=float, default=0.01)
parser.add_argument('--epochs', type=int, default=200)
parser.add_argument('--runs', type=int, default=10)
args = parser.parse_args()
print(args)
device = f'cuda:{args.device}' if torch.cuda.is_available() else 'cpu'
device = torch.device(device)
dataset = PygNodePropPredDataset(name='ogbn-products')
split_idx = dataset.get_idx_split()
data = SIGN(args.num_layers)(dataset[0]) # This might take a while.
xs = [data.x] + [data[f'x{i}'] for i in range(1, args.num_layers + 1)]
xs_train = [x[split_idx['train']].to(device) for x in xs]
xs_valid = [x[split_idx['valid']].to(device) for x in xs]
xs_test = [x[split_idx['test']].to(device) for x in xs]
y_train_true = data.y[split_idx['train']].to(device)
y_valid_true = data.y[split_idx['valid']].to(device)
y_test_true = data.y[split_idx['test']].to(device)
model = MLP(data.x.size(-1), args.hidden_channels, dataset.num_classes,
args.num_layers, args.dropout).to(device)
evaluator = Evaluator(name='ogbn-products')
logger = Logger(args.runs, args)
for run in range(args.runs):
model.reset_parameters()
optimizer = torch.optim.Adam(model.parameters(), lr=args.lr)
for epoch in range(1, 1 + args.epochs):
loss = train(model, xs_train, y_train_true, optimizer)
train_acc = test(model, xs_train, y_train_true, evaluator)
valid_acc = test(model, xs_valid, y_valid_true, evaluator)
test_acc = test(model, xs_test, y_test_true, evaluator)
result = (train_acc, valid_acc, test_acc)
logger.add_result(run, result)
if epoch % args.log_steps == 0:
train_acc, valid_acc, test_acc = result
print(f'Run: {run + 1:02d}, '
f'Epoch: {epoch:02d}, '
f'Loss: {loss:.4f}, '
f'Train: {100 * train_acc:.2f}%, '
f'Valid: {100 * valid_acc:.2f}%, '
f'Test: {100 * test_acc:.2f}%')
logger.print_statistics(run)
logger.print_statistics()
def arxiv_data(root):
# keep the same data loading logic for all architectures
dataset = PygNodePropPredDataset(
name="ogbn-arxiv",
root=root, # transform=T.ToSparseTensor(),
)
data = dataset[0]
# data.adj_t = data.adj_t.to_symmetric()
data.edge_index = to_undirected(data.edge_index)
split_idx = dataset.get_idx_split()
return data, split_idx
def get_data(args):
dataset = PygNodePropPredDataset(name=args['dataset_name'], transform=T.ToSparseTensor())
evaluator = Evaluator(name=args['dataset_name'])
data = dataset[0]
data.adj_t = data.adj_t.to_symmetric()
split_idx = dataset.get_idx_split()
device = 'cuda' if torch.cuda.is_available() else 'cpu'
data = data.to(device)
for setname in ['train', 'valid', 'test']:
split_idx[setname] = split_idx[setname].to(device)
return data, dataset, split_idx, evaluator
def main():
args = ArgsInit().args
if args.use_gpu:
device = torch.device("cuda:" +
str(args.device)) if torch.cuda.is_available(
) else torch.device("cpu")
else:
device = torch.device('cpu')
dataset = PygNodePropPredDataset(name=args.dataset)
data = dataset[0]
split_idx = dataset.get_idx_split()
evaluator = Evaluator(args.dataset)
x = data.x.to(device)
y_true = data.y.to(device)
edge_index = data.edge_index.to(device)
edge_index = to_undirected(edge_index, data.num_nodes)
if args.self_loop:
edge_index = add_self_loops(edge_index, num_nodes=data.num_nodes)[0]
args.in_channels = data.x.size(-1)
args.num_tasks = dataset.num_classes
print(args)
model = DeeperGCN(args)
model.load_state_dict(torch.load(args.model_load_path)['model_state_dict'])
model.to(device)
result = test(model, x, edge_index, y_true, split_idx, evaluator)
train_accuracy, valid_accuracy, test_accuracy = result
print({
'Train': train_accuracy,
'Validation': valid_accuracy,
'Test': test_accuracy
})
model.print_params(final=True)
def load_data(args, datapath):
if args.dataset in ['arxiv'] and args.task == 'lp':
data = {}
dataset = PygNodePropPredDataset(name='ogbn-{}'.format(args.dataset),
root='/pasteur/u/jeffgu/hgcn/data')
split_idx = dataset.get_idx_split()
train_idx, valid_idx, test_idx = split_idx["train"], split_idx[
"valid"], split_idx["test"]
induced_edges_train, _ = subgraph(train_idx, dataset[0].edge_index)
induced_edges_valid, _ = subgraph(valid_idx, dataset[0].edge_index)
induced_edges_test, _ = subgraph(test_idx, dataset[0].edge_index)
neg_edges_train = negative_sampling(induced_edges_train)
neg_edges_valid = negative_sampling(induced_edges_valid)
neg_edges_test = negative_sampling(induced_edges_test)
data['adj_train'] = to_scipy_sparse_matrix(
dataset[0].edge_index).tocsr()
data['features'] = dataset[0].x
data['train_edges'], data[
'train_edges_false'] = induced_edges_train, neg_edges_train
data['val_edges'], data[
'val_edges_false'] = induced_edges_valid, neg_edges_valid
data['test_edges'], data[
'test_edges_false'] = induced_edges_test, neg_edges_test
elif args.task == 'nc':
data = load_data_nc(args.dataset, args.use_feats, datapath,
args.split_seed)
else:
data = load_data_lp(args.dataset, args.use_feats, datapath)
adj = data['adj_train']
if args.task == 'lp':
adj_train, train_edges, train_edges_false, val_edges, val_edges_false, test_edges, test_edges_false = mask_edges(
adj, args.val_prop, args.test_prop, args.split_seed)
data['adj_train'] = adj_train
data['train_edges'], data[
'train_edges_false'] = train_edges, train_edges_false
data['val_edges'], data[
'val_edges_false'] = val_edges, val_edges_false
data['test_edges'], data[
'test_edges_false'] = test_edges, test_edges_false
data['adj_train_norm'], data['features'] = process(data['adj_train'],
data['features'],
args.normalize_adj,
args.normalize_feats)
if args.dataset == 'airport':
data['features'] = augment(data['adj_train'], data['features'])
return data
def load_ogb_graph(dataset_name):
if not os.path.isfile('torch_geometric_data/dgl_' + dataset_name):
dataset = PygNodePropPredDataset(name="ogbn-" + dataset_name,
root='torch_geometric_data/')
split_idx = dataset.get_idx_split()
train_idx, valid_idx, test_idx = split_idx["train"], split_idx[
"valid"], split_idx["test"]
edge = dataset[0].edge_index
num_classes = len(np.unique(dataset[0].y))
print("Nodes: %d, edges: %d, features: %d, classes: %d. \n" %
(dataset[0].y.shape[0], len(edge[0]) / 2, len(
dataset[0].x[0]), num_classes))
graph = dgl.DGLGraph((edge[0], edge[1]))
graph.ndata['features'] = dataset[0].x
graph.ndata['labels'] = dataset[0].y
dgl.data.utils.save_graphs('torch_geometric_data/dgl_' + dataset_name,
graph)
torch.save(
train_idx, 'torch_geometric_data/ogbn_' + dataset_name +
'/train_' + dataset_name + '.pt')
torch.save(
valid_idx, 'torch_geometric_data/ogbn_' + dataset_name +
'/valid_' + dataset_name + '.pt')
torch.save(
test_idx, 'torch_geometric_data/ogbn_' + dataset_name + '/test_' +
dataset_name + '.pt')
labels = graph.ndata.pop('labels')
features = graph.ndata.pop('features')
features = torch.hstack([features, torch.ones([features.shape[0], 1])])
#print(features)
elif os.path.isfile('torch_geometric_data/dgl_' + dataset_name):
graph = dgl.data.utils.load_graphs('torch_geometric_data/dgl_' +
dataset_name)[0][0]
labels = graph.ndata.pop('labels')
features = graph.ndata.pop('features')
features = torch.hstack([features, torch.ones([features.shape[0], 1])])
train_idx = torch.load('torch_geometric_data/ogbn_' + dataset_name +
'/train_' + dataset_name + '.pt')
valid_idx = torch.load('torch_geometric_data/ogbn_' + dataset_name +
'/valid_' + dataset_name + '.pt')
test_idx = torch.load('torch_geometric_data/ogbn_' + dataset_name +
'/test_' + dataset_name + '.pt')
num_classes = len(torch.unique(labels))
return graph, features, labels, num_classes, train_idx, valid_idx, test_idx
def process_PygNodeDataset_hetero(
self,
dataset: PygNodePropPredDataset,
):
data = dataset[0]
self._name = dataset.name
self.edge_index_dict = data.edge_index_dict
self.num_nodes_dict = data.num_nodes_dict if hasattr(
data, "num_nodes_dict") else self.get_num_nodes_dict(
self.edge_index_dict)
if self.node_types is None:
self.node_types = list(self.num_nodes_dict.keys())
if hasattr(data, "x_dict"):
self.x_dict = data.x_dict
elif hasattr(data, "x"):
self.x_dict = {self.head_node_type: data.x}
else:
self.x_dict = {}
if hasattr(data, "y_dict"):
self.y_dict = data.y_dict
elif hasattr(data, "y"):
self.y_dict = {self.head_node_type: data.y}
else:
self.y_dict = {}
self.y_index_dict = {
node_type: torch.arange(self.num_nodes_dict[node_type])
for node_type in self.y_dict.keys()
}
if self.head_node_type is None:
if hasattr(self, "y_dict"):
self.head_node_type = list(self.y_dict.keys())[0]
else:
self.head_node_type = self.node_types[0]
self.metapaths = list(self.edge_index_dict.keys())
split_idx = dataset.get_idx_split()
self.training_idx, self.validation_idx, self.testing_idx = split_idx["train"][self.head_node_type], \
split_idx["valid"][self.head_node_type], \
split_idx["test"][self.head_node_type]
def load_data_nc(dataset, use_feats, data_path, split_seed):
if dataset in ['cora', 'pubmed']:
adj, features, labels, idx_train, idx_val, idx_test = load_citation_data(
dataset, use_feats, data_path, split_seed)
elif dataset == 'arxiv':
dataset = PygNodePropPredDataset(name='ogbn-arxiv',
root='/pasteur/u/jeffgu/hgcn/data')
split_idx = dataset.get_idx_split()
idx_train, idx_val, idx_test = split_idx["train"], split_idx[
"valid"], split_idx["test"]
adj = to_scipy_sparse_matrix(dataset[0].edge_index).tocsr()
features = dataset[0].x
labels = dataset[0].y
else:
if dataset == 'disease_nc':
adj, features, labels = load_synthetic_data(
dataset, use_feats, data_path)
val_prop, test_prop = 0.10, 0.60
elif dataset == 'airport':
adj, features, labels = load_data_airport(dataset,
data_path,
return_label=True)
val_prop, test_prop = 0.15, 0.15
else:
raise FileNotFoundError(
'Dataset {} is not supported.'.format(dataset))
idx_val, idx_test, idx_train = split_data(labels,
val_prop,
test_prop,
seed=split_seed)
labels = torch.LongTensor(labels)
data = {
'adj_train': adj,
'features': features,
'labels': labels,
'idx_train': idx_train,
'idx_val': idx_val,
'idx_test': idx_test
}
return data
def load_pyg_dataset(dataset_name, root='dataset/'):
from ogb.nodeproppred import PygNodePropPredDataset, Evaluator
source, name = dataset_name.split('-', maxsplit=1)
assert source in ['ogbn', 'pyg', 'custom']
if source == 'ogbn':
dataset = PygNodePropPredDataset(name=dataset_name, root=root)
return dataset, dataset.get_idx_split(), Evaluator(dataset_name)
elif source == 'pyg':
from torch_geometric.datasets import KarateClub, CoraFull
if name == "karate":
dataset = KarateClub()
elif name == "cora":
dataset = CoraFull(root)
else:
raise Exception("Dataset not recognized")
num_nodes = dataset[0].x.shape[0]
num_train = int(num_nodes * 0.8)
num_val = int(num_nodes * 0.1)
perm = np.arange(num_nodes, dtype=int)
np.random.shuffle(perm)
split_idx = {
'train': perm[:num_train],
'valid': perm[num_train:num_train + num_val],
'test': perm[num_train + num_val:]
}
return dataset, split_idx, Evaluator('ogbn-arxiv')
elif source == "custom":
from dataset import registry
dataset = registry[name]()
split_idx = {
'train': dataset[0].idx_train,
'valid': dataset[0].idx_val,
'test': dataset[0].idx_test
}
return dataset, split_idx, CustomEvaluator()
else:
raise Exception("Dataset not recognized")
def load_ogb(dataset_name, time_budget, zero_features=True, sample_num=None):
print("*" * 30, "Start!", "*" * 30)
dataset = PygNodePropPredDataset(name=dataset_name)
split_idx = dataset.get_idx_split()
train_idx, test_idx = split_idx["train"].numpy().tolist(), split_idx[
"valid"].numpy().tolist() + split_idx["test"].numpy().tolist()
print("Train rate {}, test rate {}".format(
len(train_idx) / (len(train_idx) + len(test_idx)),
len(test_idx) / (len(train_idx) + len(test_idx))))
graph = dataset[0] # pyg graph object
features, labels = graph.x.numpy(), graph.y.numpy()
edge_index = graph.edge_index.numpy()
edge_weight = graph.edge_attr
print(features.shape, labels.shape, edge_index.shape)
if zero_features:
features = np.zeros((features.shape[0], features.shape[1]),
dtype=np.float)
if edge_weight is None:
edge_weight = np.ones(edge_index.shape[1])
adj_matrix = sp.coo_matrix((edge_weight.reshape(-1), edge_index),
shape=(labels.shape[0], labels.shape[0]))
node_indexs = np.arange(labels.shape[0])
n_class = len(np.unique(labels))
# output directory control
output_dir = os.path.join(os.path.dirname(__file__) + '../data-offline')
os.makedirs(output_dir, exist_ok=True)
sample_num_str = "" if sample_num is None else str(sample_num)
data_dir = os.path.join(output_dir, dataset_name + sample_num_str)
if os.path.exists(data_dir):
shutil.rmtree(data_dir)
train_dir = os.path.join(data_dir, 'train.data')
os.makedirs(data_dir, exist_ok=True)
os.makedirs(train_dir, exist_ok=True)
transform_to_autograph_format(features, labels, adj_matrix, node_indexs,
n_class, data_dir, train_dir, time_budget)
print("*" * 30, "Finish!", "*" * 30)
def setup_ogb(self):
dataset = PygNodePropPredDataset(name='ogbn-arxiv',
root=self.root,
transform=T.ToSparseTensor())
data = dataset[0]
self.metric = 'Accuracy'
self.num_classes = dataset.num_classes
self.split_idx = dataset.get_idx_split()
self.x = data.x
self.y = data.y
self.adj_t = data.adj_t.to_symmetric()
self.num_nodes = data.num_nodes
if self.make_edge_index:
row = self.adj_t.storage.row()
col = self.adj_t.storage.col()
self.edge_index = torch.stack((row, col), dim=0)
self.criterion = torch.nn.CrossEntropyLoss()
def load_proteins_dataset():
dataset = PygNodePropPredDataset('ogbn-proteins', root='../data')
splitted_idx = dataset.get_idx_split()
data = dataset[0]
data.node_species = None
data.y = data.y.to(torch.float)
data.n_id = torch.arange(data.num_nodes)
# Initialize features of nodes by aggregating edge features.
row, col = data.edge_index
data.x = scatter(data.edge_attr,
col,
0,
dim_size=data.num_nodes,
reduce='add')
# Set split indices to masks.
for split in ['train', 'valid', 'test']:
mask = torch.zeros(data.num_nodes, dtype=torch.bool)
mask[splitted_idx[split]] = True
data[f'{split}_mask'] = mask
return data
# Reaches around 0.7945 ± 0.0059 test accuracy.
import os.path as osp
import torch
import torch.nn.functional as F
from torch.nn import Linear as Lin
from tqdm import tqdm
from ogb.nodeproppred import PygNodePropPredDataset, Evaluator
from torch_geometric.data import NeighborSampler
from torch_geometric.nn import GATConv
root = osp.join(osp.dirname(osp.realpath(__file__)), '..', 'data', 'products')
dataset = PygNodePropPredDataset('ogbn-products', root)
split_idx = dataset.get_idx_split()
evaluator = Evaluator(name='ogbn-products')
data = dataset[0]
train_idx = split_idx['train']
train_loader = NeighborSampler(data.edge_index,
node_idx=train_idx,
sizes=[10, 10, 10],
batch_size=512,
shuffle=True,
num_workers=12)
subgraph_loader = NeighborSampler(data.edge_index,
node_idx=None,
sizes=[-1],
batch_size=1024,
shuffle=False,
num_workers=12)
def main():
parser = argparse.ArgumentParser(description='gen_models')
parser.add_argument('--device', type=int, default=0)
parser.add_argument('--dataset', type=str, default='arxiv')
parser.add_argument('--log_steps', type=int, default=1)
parser.add_argument('--model', type=str, default='mlp')
parser.add_argument('--num_layers', type=int, default=3)
parser.add_argument('--hidden_channels', type=int, default=256)
parser.add_argument('--use_embeddings', action='store_true')
parser.add_argument('--dropout', type=float, default=0.5)
parser.add_argument('--lr', type=float, default=0.01)
parser.add_argument('--epochs', type=int, default=300)
parser.add_argument('--runs', type=int, default=10)
args = parser.parse_args()
print(args)
device = f'cuda:{args.device}' if torch.cuda.is_available() else 'cpu'
device = torch.device(device)
dataset = PygNodePropPredDataset(name=f'ogbn-{args.dataset}',
transform=T.ToSparseTensor())
data = dataset[0]
data.adj_t = data.adj_t.to_symmetric()
x = data.x
split_idx = dataset.get_idx_split()
preprocess_data = PygNodePropPredDataset(name=f'ogbn-{args.dataset}')[0]
if args.dataset == 'arxiv':
embeddings = torch.cat([
preprocess(preprocess_data, 'diffusion', post_fix=args.dataset),
preprocess(preprocess_data, 'spectral', post_fix=args.dataset)
],
dim=-1)
elif args.dataset == 'products':
embeddings = preprocess(preprocess_data,
'spectral',
post_fix=args.dataset)
if args.use_embeddings:
x = torch.cat([x, embeddings], dim=-1)
if args.dataset == 'arxiv':
x = (x - x.mean(0)) / x.std(0)
if args.model == 'mlp':
model = MLP(x.size(-1), args.hidden_channels, dataset.num_classes,
args.num_layers, 0.5,
args.dataset == 'products').to(device)
elif args.model == 'linear':
model = MLPLinear(x.size(-1), dataset.num_classes).to(device)
elif args.model == 'plain':
model = MLPLinear(x.size(-1), dataset.num_classes).to(device)
elif args.model == 'sgc':
model = SGC(x.size(-1), dataset.num_classes).to(device)
x = x.to(device)
y_true = data.y.to(device)
train_idx = split_idx['train'].to(device)
model_dir = prepare_folder(f'{args.dataset}_{args.model}', model)
evaluator = Evaluator(name=f'ogbn-{args.dataset}')
logger = Logger(args.runs, args)
for run in range(args.runs):
import gc
gc.collect()
print(sum(p.numel() for p in model.parameters()))
model.reset_parameters()
optimizer = torch.optim.Adam(model.parameters(), lr=args.lr)
best_valid = 0
best_out = None
for epoch in range(1, args.epochs):
loss = train(model, x, y_true, train_idx, optimizer)
result, out = test(model, x, y_true, split_idx, evaluator)
train_acc, valid_acc, test_acc = result
if valid_acc > best_valid:
best_valid = valid_acc
best_out = out.cpu().exp()
if (epoch % 10 == 0):
print(f'Run: {run + 1:02d}, '
f'Epoch: {epoch:02d}, '
f'Loss: {loss:.4f}, '
f'Train: {100 * train_acc:.2f}%, '
f'Valid: {100 * valid_acc:.2f}% '
f'Test: {100 * test_acc:.2f}%')
logger.add_result(run, result)
logger.print_statistics(run)
torch.save(best_out, f'{model_dir}/{run}.pt')
logger.print_statistics()
def main():
parser = argparse.ArgumentParser(description='OGBN-Products (GraphSAINT)')
parser.add_argument('--device', type=int, default=0)
parser.add_argument('--log_steps', type=int, default=1)
parser.add_argument('--inductive', action='store_true')
parser.add_argument('--num_layers', type=int, default=3)
parser.add_argument('--hidden_channels', type=int, default=256)
parser.add_argument('--dropout', type=float, default=0.5)
parser.add_argument('--batch_size', type=int, default=20000)
parser.add_argument('--walk_length', type=int, default=3)
parser.add_argument('--lr', type=float, default=0.01)
parser.add_argument('--num_steps', type=int, default=30)
parser.add_argument('--epochs', type=int, default=20)
parser.add_argument('--eval_steps', type=int, default=2)
parser.add_argument('--runs', type=int, default=10)
args = parser.parse_args()
print(args)
device = f'cuda:{args.device}' if torch.cuda.is_available() else 'cpu'
device = torch.device(device)
dataset = PygNodePropPredDataset(
name='ogbn-products', root='/srv/scratch/ogb/datasets/nodeproppred')
split_idx = dataset.get_idx_split()
data = dataset[0]
# Convert split indices to boolean masks and add them to `data`.
for key, idx in split_idx.items():
mask = torch.zeros(data.num_nodes, dtype=torch.bool)
mask[idx] = True
data[f'{key}_mask'] = mask
# We omit normalization factors here since those are only defined for the
# inductive learning setup.
sampler_data = data
if args.inductive:
sampler_data = to_inductive(data)
loader = GraphSAINTRandomWalkSampler(sampler_data,
batch_size=args.batch_size,
walk_length=args.walk_length,
num_steps=args.num_steps,
sample_coverage=0,
save_dir=dataset.processed_dir)
model = SAGE(data.x.size(-1), args.hidden_channels, dataset.num_classes,
args.num_layers, args.dropout).to(device)
subgraph_loader = NeighborSampler(data.edge_index,
sizes=[-1],
batch_size=4096,
shuffle=False,
num_workers=12)
evaluator = Evaluator(name='ogbn-products')
logger = Logger(args.runs, args)
for run in range(args.runs):
model.reset_parameters()
optimizer = torch.optim.Adam(model.parameters(), lr=args.lr)
for epoch in range(1, 1 + args.epochs):
loss = train(model, loader, optimizer, device)
if epoch % args.log_steps == 0:
print(f'Run: {run + 1:02d}, '
f'Epoch: {epoch:02d}, '
f'Loss: {loss:.4f}')
if epoch > 9 and epoch % args.eval_steps == 0:
result = test(model, data, evaluator, subgraph_loader, device)
logger.add_result(run, result)
train_acc, valid_acc, test_acc = result
print(f'Run: {run + 1:02d}, '
f'Epoch: {epoch:02d}, '
f'Train: {100 * train_acc:.2f}%, '
f'Valid: {100 * valid_acc:.2f}% '
f'Test: {100 * test_acc:.2f}%')
logger.add_result(run, result)
logger.print_statistics(run)
logger.print_statistics()
def main():
parser = argparse.ArgumentParser(description='OGBN-Products (Cluster-GCN)')
parser.add_argument('--device', type=int, default=0)
parser.add_argument('--log_steps', type=int, default=1)
parser.add_argument('--num_partitions', type=int, default=15000)
parser.add_argument('--num_workers', type=int, default=12)
parser.add_argument('--num_layers', type=int, default=3)
parser.add_argument('--hidden_channels', type=int, default=256)
parser.add_argument('--dropout', type=float, default=0.5)
parser.add_argument('--batch_size', type=int, default=32)
parser.add_argument('--lr', type=float, default=0.001)
parser.add_argument('--epochs', type=int, default=50)
parser.add_argument('--eval_steps', type=int, default=5)
parser.add_argument('--runs', type=int, default=10)
args = parser.parse_args()
print(args)
device = f'cuda:{args.device}' if torch.cuda.is_available() else 'cpu'
device = torch.device(device)
dataset = PygNodePropPredDataset(name='ogbn-products')
split_idx = dataset.get_idx_split()
data = dataset[0]
# Convert split indices to boolean masks and add them to `data`.
for key, idx in split_idx.items():
mask = torch.zeros(data.num_nodes, dtype=torch.bool)
mask[idx] = True
data[f'{key}_mask'] = mask
cluster_data = ClusterData(data, num_parts=args.num_partitions,
recursive=False, save_dir=dataset.processed_dir)
loader = ClusterLoader(cluster_data, batch_size=args.batch_size,
shuffle=True, num_workers=args.num_workers)
subgraph_loader = NeighborSampler(data.edge_index, sizes=[-1],
batch_size=1024, shuffle=False,
num_workers=args.num_workers)
model = GCN(data.x.size(-1), args.hidden_channels, dataset.num_classes,
args.num_layers, args.dropout).to(device)
evaluator = Evaluator(name='ogbn-products')
logger = Logger(args.runs, args)
logger_orig = Logger(args.runs, args)
adj = process_adj(data)
for run in range(args.runs):
model.reset_parameters()
optimizer = torch.optim.Adam(model.parameters(), lr=args.lr)
best_valid = 0
best_out = None
for epoch in range(1, 1 + args.epochs):
loss, train_acc = train(model, loader, optimizer, device)
if epoch % args.log_steps == 0:
print(f'Run: {run + 1:02d}, '
f'Epoch: {epoch:02d}, '
f'Loss: {loss:.4f}, '
f'Approx Train Acc: {train_acc:.4f}')
if epoch > 19 and epoch % args.eval_steps == 0:
out, result = test(model, data, evaluator, subgraph_loader, device)
logger_orig.add_result(run, result)
train_acc, valid_acc, test_acc = result
print(f'Run: {run + 1:02d}, '
f'Epoch: {epoch:02d}, '
f'Train: {100 * train_acc:.2f}%, '
f'Valid: {100 * valid_acc:.2f}% '
f'Test: {100 * test_acc:.2f}%')
logger.print_statistics(run)
logger.print_statistics()
logger_orig.print_statistics()
def main():
parser = argparse.ArgumentParser(description='OGBN-papers100M (MLP)')
parser.add_argument('--data_root_dir', type=str, default='../../dataset')
parser.add_argument('--num_propagations', type=int, default=3)
parser.add_argument('--dropedge_rate', type=float, default=0.4)
parser.add_argument('--node_emb_path', type=str, default=None)
parser.add_argument('--output_path', type=str, required=True)
args = parser.parse_args()
# SGC pre-processing ######################################################
dataset = PygNodePropPredDataset(name='ogbn-papers100M',
root=args.data_root_dir)
split_idx = dataset.get_idx_split()
data = dataset[0]
x = None
if args.node_emb_path:
x = np.load(args.node_emb_path)
else:
x = data.x.numpy()
N = data.num_nodes
print('Making the graph undirected.')
### Randomly drop some edges to save computation
data.edge_index, _ = dropout_adj(data.edge_index,
p=args.dropedge_rate,
num_nodes=data.num_nodes)
data.edge_index = to_undirected(data.edge_index, data.num_nodes)
print(data)
row, col = data.edge_index
print('Computing adj...')
adj = SparseTensor(row=row, col=col, sparse_sizes=(N, N))
adj = adj.set_diag()
deg = adj.sum(dim=1).to(torch.float)
deg_inv_sqrt = deg.pow(-0.5)
deg_inv_sqrt[deg_inv_sqrt == float('inf')] = 0
adj = deg_inv_sqrt.view(-1, 1) * adj * deg_inv_sqrt.view(1, -1)
adj = adj.to_scipy(layout='csr')
train_idx, valid_idx, test_idx = split_idx['train'], split_idx[
'valid'], split_idx['test']
all_idx = torch.cat([train_idx, valid_idx, test_idx])
mapped_train_idx = torch.arange(len(train_idx))
mapped_valid_idx = torch.arange(len(train_idx),
len(train_idx) + len(valid_idx))
mapped_test_idx = torch.arange(
len(train_idx) + len(valid_idx),
len(train_idx) + len(valid_idx) + len(test_idx))
sgc_dict = {}
sgc_dict['label'] = data.y.data[all_idx].to(torch.long)
sgc_dict['split_idx'] = {
'train': mapped_train_idx,
'valid': mapped_valid_idx,
'test': mapped_test_idx
}
print('Start SGC processing')
for _ in tqdm(range(args.num_propagations)):
x = adj @ x
sgc_dict['sgc_embedding'] = torch.from_numpy(x[all_idx]).to(torch.float)
torch.save(sgc_dict, args.output_path)
