if __name__ == '__main__':
nmt_parser = argparse.ArgumentParser()
add_arguments(nmt_parser)
FLAGS, unparsed = nmt_parser.parse_known_args()
hparams = create_hparams(FLAGS)
# loading the data from a file
adj, features, edges = load_data(hparams.graph_file, hparams.nodes)
num_nodes = adj[0].shape[0]
num_features = features[0].shape[1]
#print("Debug", num_nodes, adj[0][0])
# Training
model = VAEG(hparams, placeholders, num_nodes, num_features, edges)
# model.restore(hparams.out_dir)
model.initialize()
model.train(placeholders, hparams, adj, features)
# Test code
'''
model2 = VAEG(hparams, placeholders, 30, 1)
model2.restore(hparams.out_dir)
hparams.sample = True
i = 0
G_good = load_embeddings(hparams.z_dir+'train0.txt')
G_bad = load_embeddings(hparams.z_dir+'test_11.txt')
#model2.sample_graph_slerp(hparams, placeholders, 5, G_good, G_bad, num=29)
while i < 10:
G_bad = model2.sample_graph_slerp(hparams, placeholders, i, G_good, G_bad, num=29)
sample=flags.sample,
neg_sample_size=flags.neg_sample_size,
node_sample=flags.node_sample,
bfs_sample=flags.bfs_sample
)
if __name__ == '__main__':
nmt_parser = argparse.ArgumentParser()
add_arguments(nmt_parser)
FLAGS, unparsed = nmt_parser.parse_known_args()
hparams = create_hparams(FLAGS)
# loading the data from a file
adj, weight, weight_bin, features, edges, neg_edges, features1, = load_data_new(hparams.graph_file, hparams.nodes, hparams.node_sample, hparams.bfs_sample, hparams.bin_dim)
num_nodes = adj[0].shape[0]
num_features = features[0].shape[1]
lenedges = [len(edge[0]) for edge in edges]
lenweight_bin = [len(weight_b[0]) for weight_b in weight_bin]
print("Len edges", lenedges, lenweight_bin)
print("Num features", num_features)
print("Num examples", len(adj))
#print("Neg_index", neg_index)
e = max([len(edge) for edge in edges])
log_fact_k = log_fact(e)
# Training
#'''
model = VAEG(hparams, placeholders, num_nodes, num_features,log_fact_k, len(adj))
model.restore(hparams.out_dir)
model.train(placeholders, hparams, adj, weight, weight_bin, features, edges, neg_edges, features1)