def __init__(self, data):
super(Model, self).__init__()
self.train_data = data
self.num_node_feat = data.num_node_feat
# Create layers.
# for _ in range(n_outputs): TODO: finish real multiple-choice learning
self.layers = create_layers(self, 'layer', FLAGS.layer_num)
assert (len(self.layers) > 0)
self._print_layers(None, self.layers)
# Create layers for branches
# (except for the main branch which has been created above).
bnames = get_branch_names()
if bnames:
for bname in bnames:
blayers = create_layers(
self, bname,
get_flag('{}_layer_num'.format(bname), check=True))
setattr(self, bname, blayers) # so that print(model) can print
self._print_layers(bname, getattr(self, bname))
# Map layer object to output.
# Some layers may not register its output to this dictionary.
# Currently only NodeEmbedding layers store its output.
# This dictionary does NOT get cleaned up after each iteration.
self.layer_output = {}
def _build(self):
# Create layers according to FLAGS.
self.layers = create_layers(self)
assert (len(self.layers) > 0)
print('Created {} layers: {}'.format(
len(self.layers), ', '.join(l.get_name() for l in self.layers)))
# Build the siamese model for train and val_test, respectively,
for tvt in ['train', 'val_test']:
print(tvt)
# Go through each layer except the last one.
acts = [self._get_ins(self.layers[0], tvt)]
outs = None
for k, layer in enumerate(self.layers):
ins = self._proc_ins(acts[-1], k, layer, tvt)
print(layer.name)
outs = layer(ins)
outs = self._proc_outs(outs, k, layer, tvt)
acts.append(outs)
if tvt == 'train':
self.train_outputs = outs
self.train_acts = acts
else:
self.val_test_output = outs
self.val_test_pred_score = self._val_test_pred_score()
self.val_test_acts = acts
self.node_embeddings = self._get_last_gcn_layer_outputs('val_test')
# Store model variables for easy access.
variables = tf.get_collection(tf.GraphKeys.GLOBAL_VARIABLES)
self.vars = {var.name: var for var in variables}
def _build(self):
if logging_enabled == True:
print("- Entered model::Model::_build Private Method")
# Create layers
self.layers = create_layers(self, 'layer', ec.layer_num)
assert (len(self.layers) > 0)
print('\ninfo: {} layers have been Created:\n\t{}\n'.format(
len(self.layers), ',\n\t'.join(l.get_name() for l in self.layers)))
self.gembd_layer_id = self._gemb_layer_id()
print('info: index (0-based) of the Graph Embedding Layer: {}'.format(
self.gembd_layer_id))
# Build the siamese model for train and val_test, respectively,
for tvt in ['train', 'val_test']:
# Go through each layer except the last one.
acts = [self._get_ins(self.layers[0], tvt)]
outs = None
if tvt == 'train':
print('\n* Training:')
else:
print('\n* Validation Testing:')
for k, layer in enumerate(self.layers):
print("\t", layer.name)
ins = self._proc_ins(acts[-1], k, layer, tvt)
outs = layer(ins)
outs = self._proc_outs(outs, k, layer, tvt)
acts.append(outs)
if tvt == 'train':
self.train_outputs = outs
self.train_acts = acts
print("train_outputs::outs:\n", self.train_outputs)
print("train_outputs::acts:\n", self.train_acts)
else:
self.val_test_output = outs
self.val_test_pred_score = self._val_test_pred_score()
self.val_test_acts = acts
print("val_test_output::outs:\n", self.val_test_output)
print("val_test_pred_score::acts:\n", self.val_test_pred_score)
print("val_test_acts::acts:\n", self.val_test_acts)
self.node_embeddings = self._get_all_gcn_layer_outputs('val_test')
# Store model variables for easy access.
variables = tf.compat.v1.get_collection(
tf.compat.v1.GraphKeys.GLOBAL_VARIABLES)
self.vars = {var.name: var for var in variables}