practice.py

import tensorflow as tf
import numpy as np
import time

#create data
x_data = np.random.rand(100).astype(np.float32)
y_data = x_data*0.1+0.3

#create tensorflow structure start#
Weights = tf.Variable(tf.random_uniform([1], -1.0, 1.0))
biases = tf.Variable(tf.zeros([1]))

y = Weights*x_data+biases

loss=tf.reduce_mean(tf.square(y-y_data))

optimizer=tf.train.GradientDescentOptimizer(0.5)
#optimizer = tf.compat.v1.train.GradientDescentOptimizer(0.5)
train=optimizer.minimize(loss)

#initialize variable
init=tf.initialize_all_variables()
#create tensorflow structure end#

sess=tf.Session()
sess.run(init)

if __name__ == '__main__':
    localtime=time.time()
    for step in range(201):

        sess.run(train)
        if step % 20 == 0:
            print(step,sess.run(Weights),sess.run(biases))

    print('time_cost:',(time.time() - localtime))