In [ ]:

    

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import math
import os

import numpy as np
import pylab
import tensorflow as tf

%matplotlib inline


    

In [ ]:

    

# Create a directory to write summary data
# ./graphs will be created in the same directory 
# from which you run this notebook.

# Tip: if you later run into trouble displaying results in
# tensorboard, delete the contents of this directory 
# and run the notebook again.
LOGDIR = './graphs'


    

In [ ]:

    

tf.reset_default_graph()
sess = tf.Session()


    

In [ ]:

    

x_train = np.random.rand(100).astype(np.float32)
noise = np.random.normal(scale=0.01, size=len(x_train))
y_train = x_train * 0.1 + 0.3 + noise


    

In [ ]:

    

x_eval = np.random.rand(len(x_train)).astype(np.float32)
noise = np.random.normal(scale=0.01, size=len(x_train))
y_eval = x_eval * 0.1 + 0.3 + noise


    

In [ ]:

    

with tf.name_scope('placeholders'):
    x = tf.placeholder(shape=[None], dtype=tf.float32, name='x')
    y_label = tf.placeholder(shape=[None], dtype=tf.float32, name='labels')
    
with tf.name_scope('model'):
    W = tf.Variable(tf.random_normal([1]), name='weights')
    b = tf.Variable(tf.random_normal([1]), name='bias')
    y = W * x + b


    

In [ ]:

    

with tf.name_scope('training'):
    with tf.name_scope('loss'):
        loss = tf.reduce_mean(tf.square(y - y_label))
    with tf.name_scope('optimizer'):
        optimizer = tf.train.GradientDescentOptimizer(0.5)
        train = optimizer.minimize(loss)


    

In [ ]:

    

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


    

In [ ]:

    

# Write the graph so we can look at it later in TensorBoard
writer = tf.summary.FileWriter(LOGDIR)
writer.add_graph(sess.graph)


    

In [ ]:

    

# Keep track how loss changes, so we can visualize later it in TensorBoard
tf.summary.scalar('loss', loss)

# We might as well keep track of how W and b change as well
# Later, you can see these values under 'histograms' and 'distributions'
tf.summary.histogram('W', W)
tf.summary.histogram('b', b)

# Create an op that will write all of our summary data at once
summary_op = tf.summary.merge_all()


    

In [ ]:

    

for step in range(201):
    summary_results, _ = sess.run([summary_op, train], feed_dict={x: x_train, y_label: y_train})
    writer.add_summary(summary_results, step)


    

In [ ]:

    

# To start TensorBoard, run this command from a terminal
# Note: you should run this from the same directory that contains this notebook
# or, provide absolute path to the 'graphs' directory
# $ tensorboard --logdir=graphs

# To open TensorBoard, point your browser to http://localhost:6006
# Then, click on the tabs for 'scalars', 'distributions', 'histograms', and 'graphs'
# to learn more.


    

In [ ]:

    

# While we're at it, let's demonstrate saving and restoring a model


    

In [ ]:

    

def predict(x_in): return sess.run(y, feed_dict={x: [x_in]})


    

In [ ]:

    

# Save the model
saver = tf.train.Saver()
saver.save(sess, os.path.join(LOGDIR,'my_checkpoint.ckpt'))


    

In [ ]:

    

# Current prediction
predict(3)


    

In [ ]:

    

# Reset the model by running the init op again
sess.run(init)


    

In [ ]:

    

# Prediction after variables reinitialized
predict(3)


    

In [ ]:

    

saver.restore(sess, os.path.join(LOGDIR,'my_checkpoint.ckpt'))


    

In [ ]:

    

# Predictions after variables restored
predict(3)