Keras predict memory leak

from __future__ import print_function

import os
import gc
from time import sleep

import keras.backend as K
import numpy as np
import psutil
from keras.models import Sequential
from keras.layers import Dense, Activation

# set parameters:
train_size = 512
test_size = 128
num_features = 400
batch_size = 32
hidden_size = 1024 * 8
epochs = 1

print("Generate data...")
np.random.seed(42)
x_train = np.random.rand(train_size, num_features)
y_train = np.random.randint(0, 1, size=train_size)
x_test = np.random.rand(test_size, num_features)
y_test = np.random.randint(0, 1, size=test_size )
print('x_train shape:', x_train.shape)
print('x_test shape:', x_test.shape)

print('Build model...')
model = Sequential()
model.add(Dense(hidden_size, input_shape=(num_features,)))
model.add(Dense(1))
model.add(Activation('sigmoid'))

model.compile(loss='binary_crossentropy',
              optimizer='adam',
              metrics=['accuracy'])

# print("Train...")
# model.fit(x_train, y_train,
#           batch_size=batch_size,
#           epochs=epochs)

print("Predict...")
K.set_learning_phase(0)
process = psutil.Process(os.getpid())
while True:
    print("rss={}MB\tvms={}MB".format(
        process.memory_info().rss // 10 ** 6,
        process.memory_info().vms // 10 ** 6
    ))
    sleep(1)
    y_pred = model.predict(x_test, batch_size=256)
    del y_pred
    gc.collect()

In Keras==2.1.6, tensorflow-gpu==1.8.0, on GTX 1070 I observed constant memory usage (rss=1108MB vms=26964MB). But I'm still hunting a leak in our prediction code.