Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd
Upload
384 views

Cheat Sheets of Python Libraries Tensorflow

This document contains a summary of key concepts in TensorFlow including placeholders, variables, initialization, saving and restoring variables in a session. It discusses placeholders as inputs that must be fed values, variables that can be initialized and updated, and the need to initialize variables before running other ops. It also covers how to save and restore a session's variables using a tf.train.Saver object and choosing specific variables to save.

Uploaded by

gprasadatvu
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
384 views

Cheat Sheets of Python Libraries Tensorflow

This document contains a summary of key concepts in TensorFlow including placeholders, variables, initialization, saving and restoring variables in a session. It discusses placeholders as inputs that must be fed values, variables that can be initialized and updated, and the need to initialize variables before running other ops. It also covers how to save and restore a session's variables using a tf.train.Saver object and choosing specific variables to save.

Uploaded by

gprasadatvu
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

11/4/2017 cheat_sheets_of_python_libraries/tensorflow.

md at master · jalajthanaki/cheat_sheets_of_python_libraries · GitHub

jalajthanaki / cheat_sheets_of_python_libraries

Branch: master cheat_sheets_of_python_libraries / 8_tensorflow / tensorflow.md Find file Copy path

jalajthanaki cheat_sheets_added a344a28 3 days ago

1 contributor

121 lines (92 sloc) 3.7 KB

Cheatsheet for TensorFlow Credit goes to Cristian Cardellino


for making this cheat sheet.

Importing
The line to import tensorflow by convention is as follows:

import tensorflow as tf

Placeholders (inputs)
A placeholder is a value to input on a TensorFlow computation. It cannot be evaluated, as it must be fed a value using a
feed_dict .

x = tf.placeholder(tf.float32, shape=(1024, 1024))


y = tf.matmul(x, x) # Matrix Multiplication (or Dot product)

with tf.Session() as session:


print(session.run(y)) # Error: x wasn't fed any value

rand_array = np.random.rand(1024, 1024)


print(session.run(y, feed_dict={x: rand_array}) # Success

Variables
Variables are useful to hold and update parameters. Unlike placeholders, variables are initialized with some value and can be
updated through the computation. They are in-memory buffers containing tensors.

# create two variables


w = tf.Variable(tf.random_normal([784, 200], stddev=0.35),
name='w')
b = tf.Variable(tf.zeros([200]), name='b')

Initialization
A variable initializer must be run before other ops in the model can be run. Exists an operation that runs all variable
initilizations: tf.initialize_all_variables() .

# create two variables


w = tf.Variable(tf.random_normal([784, 200], stddev=0.35),
name='w')

https://github.com/jalajthanaki/cheat_sheets_of_python_libraries/blob/master/8_tensorflow/tensorflow.md 1/3
11/4/2017 cheat_sheets_of_python_libraries/tensorflow.md at master · jalajthanaki/cheat_sheets_of_python_libraries · GitHub
b = tf.Variable(tf.zeros([200]), name='b')
...
# add an op to initialize the variables
init_op = tf.initialize_all_variables()

# later, when launching the model


with tf.Session() as sess:
# run the init operation
sess.run(init_op)
...
# use the model
...

Initialization from another variable

# create a variable with a random value.


w1 = tf.Variable(tf.random_normal([784, 200], stddev=0.35),
name='w1')
# create another variable with the same value as 'w1'.
w2 = tf.Variable(w1.initialized_value(), name='w1')
# Create another variable with twice the value of 'w1'
w1_twice = tf.Variable(w1.initialized_value() * 2.0, name="w1_twice")

Saving a session's variables


Use a tf.train.Saver() object to save the session's variables (all or some) and restore it later.

# Create some variables.


v1 = tf.Variable(..., name="v1")
v2 = tf.Variable(..., name="v2")
...
# Add an op to initialize the variables.
init_op = tf.initialize_all_variables()

# Add ops to save and restore all the variables.


saver = tf.train.Saver()

# Later, launch the model, initialize the variables, do some work, save the
# variables to disk.
with tf.Session() as sess:
sess.run(init_op)
# Do some work with the model.
..
# Save the variables to disk.
save_path = saver.save(sess, "/tmp/model.ckpt")
print("Model saved in file: %s" % save_path)

Restoring a session's variables


How to restore variables:

# Create some variables.


v1 = tf.Variable(..., name="v1")
v2 = tf.Variable(..., name="v2")
...
# Add ops to save and restore all the variables.
saver = tf.train.Saver()

# Later, launch the model, use the saver to restore variables from disk, and
# do some work with the model.
with tf.Session() as sess:
# Restore variables from disk.
saver.restore(sess, "/tmp/model.ckpt")
print("Model restored.")

https://github.com/jalajthanaki/cheat_sheets_of_python_libraries/blob/master/8_tensorflow/tensorflow.md 2/3
11/4/2017 cheat_sheets_of_python_libraries/tensorflow.md at master · jalajthanaki/cheat_sheets_of_python_libraries · GitHub
# Do some work with the model
...

Choose variables to save in a session

# Create some variables.


v1 = tf.Variable(..., name="v1")
v2 = tf.Variable(..., name="v2")
...
# Add ops to save and restore only 'v2' using the name "my_v2"
saver = tf.train.Saver({"my_v2": v2})
# Use the saver object normally after that.
...

https://github.com/jalajthanaki/cheat_sheets_of_python_libraries/blob/master/8_tensorflow/tensorflow.md 3/3

You might also like