def train_sequence(self, index):
# if the graph is changed?
labels = self.labels_onehot[index]
sequence = FullBatchNodeSequence(
[self.feature_inputs, self.structure_inputs, index], labels, device=self.device)
return sequence
def train_sequence(self, index):
labels = self.graph.labels[index]
sequence = FullBatchNodeSequence(
[self.feature_inputs, self.structure_inputs, index],
labels,
device=self.device)
return sequence
def train_sequence(self, index):
labels = self.graph.node_label[index]
sequence = FullBatchNodeSequence(
[self.feature_inputs, self.structure_inputs, index],
labels,
device=self.device,
escape=type(self.structure_inputs))
return sequence
def train_sequence(self, index):
index = gf.astensor(index)
labels = self.graph.node_label[index]
feature_inputs = tf.gather(self.feature_inputs, index)
sequence = FullBatchNodeSequence(feature_inputs,
labels,
device=self.device)
return sequence
def train_sequence(self, index):
index = T.astensor(index)
labels = self.graph.labels[index]
if self.kind == "T":
feature_inputs = tf.gather(self.feature_inputs, index)
else:
feature_inputs = self.feature_inputs[index]
sequence = FullBatchNodeSequence(feature_inputs, labels, device=self.device)
return sequence
def train_sequence(self, index, batch_size=np.inf):
mask = T.indices2mask(index, self.graph.n_nodes)
index = get_indice_graph(self.structure_inputs, index, batch_size)
while index.size < self.K:
index = get_indice_graph(self.structure_inputs, index)
structure_inputs = self.structure_inputs[index][:, index]
feature_inputs = self.feature_inputs[index]
mask = mask[index]
labels = self.graph.labels[index[mask]]
sequence = FullBatchNodeSequence(
[feature_inputs, structure_inputs, mask],
labels,
device=self.device)
return sequence
def train(self,
idx_train,
idx_val=None,
pre_train_epochs=100,
epochs=100,
early_stopping=None,
verbose=None,
save_best=True,
weight_path=None,
as_model=False,
monitor='val_acc',
early_stop_metric='val_loss'):
histories = []
index_all = tf.range(self.graph.n_nodes, dtype=self.intx)
# pre train model_q
self.model = self.model_q
history = super().train(idx_train,
idx_val,
epochs=pre_train_epochs,
early_stopping=early_stopping,
verbose=verbose,
save_best=save_best,
weight_path=weight_path,
as_model=True,
monitor=monitor,
early_stop_metric=early_stop_metric)
histories.append(history)
label_predict = self.predict(index_all).argmax(1)
label_predict[idx_train] = self.graph.labels[idx_train]
label_predict = tf.one_hot(label_predict, depth=self.graph.n_classes)
# train model_p fitst
train_sequence = FullBatchNodeSequence(
[label_predict, self.structure_inputs, index_all],
label_predict,
device=self.device)
if idx_val is not None:
val_sequence = FullBatchNodeSequence(
[label_predict, self.structure_inputs, idx_val],
self.labels_onehot[idx_val],
device=self.device)
else:
val_sequence = None
self.model = self.model_p
history = super().train(train_sequence,
val_sequence,
epochs=epochs,
early_stopping=early_stopping,
verbose=verbose,
save_best=save_best,
weight_path=weight_path,
as_model=as_model,
monitor=monitor,
early_stop_metric=early_stop_metric)
histories.append(history)
# then train model_q again
label_predict = self.model.predict_on_batch(
T.astensors(label_predict,
self.structure_inputs,
index_all,
device=self.device))
label_predict = softmax(label_predict)
if tf.is_tensor(label_predict):
label_predict = label_predict.numpy()
label_predict[idx_train] = self.labels_onehot[idx_train]
self.model = self.model_q
train_sequence = FullBatchNodeSequence(
[self.feature_inputs, self.structure_inputs, index_all],
label_predict,
device=self.device)
history = super().train(train_sequence,
idx_val,
epochs=epochs,
early_stopping=early_stopping,
verbose=verbose,
save_best=save_best,
weight_path=weight_path,
as_model=as_model,
monitor=monitor,
early_stop_metric=early_stop_metric)
histories.append(history)
return histories
def train_sequence(self, index):
labels = self.graph.node_label[index]
with tf.device(self.device):
sequence = FullBatchNodeSequence(
[self.feature_inputs, self.structure_inputs, index], labels)
return sequence