def load_npy_data(data_file, name):
path = get_dir(osp.join(data_file, name + '.content'))
idx_features_labels = np.genfromtxt(path, dtype=np.dtype(str))
features = sp.csr_matrix(idx_features_labels[:, 1:-1], dtype=np.float32)
labels = encode_onehot(idx_features_labels[:, -1])
# build graph
idx = np.array(idx_features_labels[:, 0], dtype=np.int32)
idx_map = {j: i for i, j in enumerate(idx)}
path = get_dir(osp.join(data_file, name + '.cites'))
edges_unordered = np.genfromtxt(path, dtype=np.int32)
edges = np.array(list(map(idx_map.get, edges_unordered.flatten())),
dtype=np.int32).reshape(edges_unordered.shape)
adj = sp.coo_matrix((np.ones(edges.shape[0]), (edges[:, 0], edges[:, 1])),
shape=(labels.shape[0], labels.shape[0]),
dtype=np.float32)
# build symmetric adjacency matrix
adj = adj + adj.T.multiply(adj.T > adj) - adj.multiply(adj.T > adj)
print('Dataset has {} nodes, {} edges, {} features.'.format(
adj.shape[0], edges.shape[0], features.shape[1]))
return features.todense(), adj, labels
def read_edgelist_label_data(folder, prefix, save_path):
graph_path = get_dir(osp.join(folder, '{}.ungraph'.format(prefix)))
cmty_path = get_dir(osp.join(folder, '{}.cmty'.format(prefix)))
G = nx.read_edgelist(graph_path, nodetype=int, create_using=nx.Graph())
num_node = G.number_of_nodes()
nodes = np.array(list(G.nodes()))
print('node number: ', num_node)
with open(graph_path) as f:
context = f.readlines()
print('edge number: ', len(context))
# tow line, each line has two corresponding node, total len(context) edges
edge_index = np.zeros((2, len(context)))
for i, line in enumerate(context):
edge_index[:, i] = list(map(int, line.strip().split('\t')))
with open(cmty_path) as f:
context = f.readlines()
print('class number: ', len(context))
# if node m belongs to class k, thus (m, k) = 1
label = np.zeros((num_node, len(context)))
for i, line in enumerate(context):
line = map(int, line.strip().split('\t'))
for node in line:
label[node, i] = 1
np.savez(osp.join(save_path, prefix), nodes=nodes, edge_index=edge_index, label=label)
def get_data(self, name):
if not osp.exists(get_dir(osp.join(self.processed, name + '.npz'))):
# raw_file_names = ['{}.{}'.format(s, f) for s, f in product([name], ['ungraph', 'cmty'])]
# for file_name in raw_file_names:
# download_url('{}/{}'.format(self.url, file_name), self.raw_dir)
read_edgelist_label_data(self.raw_dir, name, self.processed)
npzfile = np.load(get_dir(osp.join(self.processed, name + '.npz')))
return npzfile['nodes'], npzfile['edge_index'], npzfile['label']
def __init__(self, root):
super(Build_dataset, self).__init__()
self.root = root
self.raw_dir = osp.join(root, 'raw')
self.processed = osp.join(root, 'processed')
if not osp.exists(self.processed):
os.makedirs(get_dir(self.processed))
self.url = 'https://github.com/THUDM/ProNE/raw/master/data'
from cogdl.models import Build_model
from para_config import Config
from cogdl.tasks import Build_task
from cogdl.datasets import Build_dataset
from cogdl.data import get_dir
if __name__ == '__main__':
args = Config()
if args.gpu:
os.environ["CUDA_VISIBLE_DEVICES"] = args.device_id
gpus = tf.config.experimental.list_physical_devices('GPU')
tf.config.experimental.set_memory_growth(gpus[0], True)
random.seed(args.seed[0])
np.random.seed(args.seed[0])
args.set_model('gcn')
args.set_dataset('cora')
args.set_task('node_classification')
task = Build_task(args).build()
result = task.train()
print(result)
result_file = get_dir(
osp.join(args.save_dir, args.dataset + '_' + args.model + '.json'))
json_str = json.dumps(result)
with open(result_file, 'w') as json_file:
json_file.write(json_str)