1
它來自Udacity深度學習基礎課程。它似乎爲他們工作。但它在我的電腦中不起作用。請看一看。感謝您的幫助!Tensorflow保存和恢復變量不一樣
講座和我的電腦的tensorflow版本都是1.0.0。
import tensorflow as tf
# The file path to save the data
save_file = './model.ckpt'
# Two Tensor Variables: weights and bias
weights = tf.Variable(tf.truncated_normal([2, 3]))
bias = tf.Variable(tf.truncated_normal([3]))
# Class used to save and/or restore Tensor Variables
saver = tf.train.Saver()
with tf.Session() as sess:
# Initialize all the Variables
sess.run(tf.global_variables_initializer())
# Show the values of weights and bias
print('Weights:')
print(sess.run(weights))
print('Bias:')
print(sess.run(bias))
# Save the model
saver.save(sess, save_file)
# Remove the previous weights and bias
tf.reset_default_graph()
# Two Variables: weights and bias
weights = tf.Variable(tf.truncated_normal([2, 3]))
bias = tf.Variable(tf.truncated_normal([3]))
# Class used to save and/or restore Tensor Variables
saver = tf.train.Saver()
with tf.Session() as sess:
# Load the weights and bias
saver.restore(sess, save_file)
# Show the values of weights and bias
print('Weight:')
print(sess.run(weights))
print('Bias:')
print(sess.run(bias))
你的代碼適用於我的最新TensorFlow。你能否更新到最新的TensorFlow版本並重試?如果它仍然不起作用,會出現什麼問題?你有錯誤還是錯誤的輸出? –