来自tf.session.run的网络输出与使用keras.Model.predict获得的输出大不相同

Question

我正试图通过Tensorflow会话使用Keras模型.但结果形式model.predict和sess.run不同.有没有办法通过Tensorflow会话使用Kers模型？

Tensorflow版本:1.4.0 Keras
版本:2.1.1

from sklearn.datasets.samples_generator import make_circles
from keras.models import Sequential
from keras.layers import Dense
from keras.optimizers import SGD
import numpy as np
import tensorflow as tf
from keras import backend as K

sess = tf.Session()
K.tensorflow_backend.set_session(sess)

X, y = make_circles(n_samples=1000,
                    noise=0.1,
                    factor=0.2,
                    random_state=0)

model = Sequential()
model.add(Dense(4, input_shape=(2,), activation='tanh'))
model.add(Dense(1, activation='sigmoid'))
model.compile(SGD(lr=0.5), 'binary_crossentropy', metrics=['accuracy'])
model.fit(X, y, epochs=20)

print("Keras model. First prediction: " + str(model.predict(np.c_[0, 0])))
print("Keras model. Second prediction: " + str(model.predict(np.c_[1.5, 1.5])))

with sess.as_default():

    y_tensor = model.outputs[0]
    x_tensor = model.inputs[0]
    sess.run(tf.global_variables_initializer())

    print("TF model. First prediction: " + str(sess.run(y_tensor, feed_dict={x_tensor: np.c_[0, 0]} )))
    print("TF model. Second prediction: " + str(sess.run(y_tensor, feed_dict={x_tensor: np.c_[1.5, 1.5]} )))

Answer 1

好的,K.set_session(s)不是K.tensorflow_backend.set_session(s).

第二:sess.run(tf.global_variables_initializer())使用各自的初始化程序重置所有变量,包括网络权重(xavier默认情况下使用初始化程序).

所以你是:

1. 训练keras模型
2. 打印keras模型的预测
3. 重置为随机权重
4. 打印同一模型的预测

评论sess.run(tf.global_variables_initializer())修复了问题:

Keras model. First prediction: [[ 0.99195099]]
Keras model. Second prediction: [[ 0.03110269]]
TF model. First prediction: [[ 0.99195099]]
TF model. Second prediction: [[ 0.03110269]]