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]
#Test code
#''' interpolation
model2 = VAEG(hparams, placeholders, hparams.nodes, 1, edges)
model2.restore(hparams.out_dir)
#hparams.sample = True
i = 0
'''
# getting embeddings
sample_1 = model2.getembeddings(hparams, placeholders, adj[i], features[i])
'''
'''
sample_1 = model2.getembeddings(hparams, placeholders, adj[0], features[0])
sample_2 = model2.getembeddings(hparams, placeholders, adj[1], features[1])
while i < 1:
model2.sample_graph_slerp(hparams, placeholders, i,sample_1, sample_2, "slerp", (i+1)*0.1, num=70)
model2.sample_graph_slerp(hparams, placeholders, i,sample_1, sample_2, "lerp", (i+1)*0.1, num=70)
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, hde = load_data(hparams.graph_file, hparams.nodes, hparams.bin_dim)
#Test code
#'''
e = max([len(edge) for edge in edges])
n_f = len(features[0][0])
log_fact_k = log_fact(e)
model2 = VAEG(hparams, placeholders, hparams.nodes, n_f, edges, log_fact_k, hde)
model2.restore(hparams.out_dir)
latent_points = []
'''
for i1 in range(len(adj)):
sample1 = model2.getembeddings(hparams, placeholders, adj[i1], features[i1], weight_bin[i1], weight[i1])
latent_points.append(np.reshape(np.array(sample1), -1))
'''
#np.savetxt("latent_features.txt", np.array(latent_points))
#sample
i = 0
while i < 1:
model2.sample_graph(hparams, placeholders,adj, features, weight, weight_bin, i+hparams.offset, hde, hparams.nodes, hparams.edges)
i += 1
sample=flags.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, 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)
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, smiles = load_data_new(
hparams.graph_file, hparams.nodes, 1, 1, hparams.bin_dim)
#Test code
e = max([len(edge) for edge in edges])
n_f = len(features[0][0])
log_fact_k = log_fact(e)
model2 = VAEG(hparams, placeholders, hparams.nodes, n_f, log_fact_k,
len(adj))
model2.restore(hparams.out_dir)
while i < 100:
smiles = []
smiles_new = model2.sample_graph(hparams, placeholders, adj, features,
features1, weight, weight_bin, edges,
i)
for s in smiles_new:
if s != 'None':
smiles.append(s)
i += 1
print smiles
with open(hparams.sample_file + "smiles.txt", "a") as f:
for s in smiles:
f.write(s + "\n")
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)