When you train a model, you use variables to hold and update parameters. Variables are in-memory buffers containing tensors. They must be explicitly initialized and can be saved to disk during and after training. You can later restore saved values to exercise or analyse the model.
This document references the following TensorFlow classes. Follow the links to their reference manual for a complete description of their API:
tf.Variable
class.tf.train.Saver
class.When you create a Variable you pass a
Tensor
as its initial value to the Variable()
constructor. TensorFlow
provides a collection of ops that produce tensors often used for initialization
from constants or random values.
Note that all these ops require you to specify the shape of the tensors. That shape automatically becomes the shape of the variable. Variables generally have a fixed shape, but TensorFlow provides advanced mechanisms to reshape variables.
# Create two variables.
weights = tf.Variable(tf.random_normal([784, 200], stddev=0.35),
name="weights")
biases = tf.Variable(tf.zeros([200]), name="biases")
Calling tf.Variable()
adds several ops to the graph:
variable
op that holds the variable value.tf.assign
op.zeros
op for the biases
variable in the example are also added to the graph.The value returned by tf.Variable()
value is an instance of the Python class
tf.Variable
.
Variable initializers must be run explicitly before other ops in your model can be run. The easiest way to do that is to add an op that runs all the variable initializers, and run that op before using the model.
You can alternatively restore variable values from a checkpoint file, see below.
Use tf.initialize_all_variables()
to add an op to run variable initializers.
Only run that op after you have fully constructed your model and launched it in
a session.
# Create two variables.
weights = tf.Variable(tf.random_normal([784, 200], stddev=0.35),
name="weights")
biases = tf.Variable(tf.zeros([200]), name="biases")
...
# 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
...
You sometimes need to initialize a variable from the initial value of another
variable. As the op added by tf.initialize_all_variables()
initializes all
variables in parallel you have to be careful when this is needed.
To initialize a new variable from the value of another variable use the other
variable's initialized_value()
property. You can use the initialized value
directly as the initial value for the new variable, or you can use it as any
other tensor to compute a value for the new variable.
# Create a variable with a random value.
weights = tf.Variable(tf.random_normal([784, 200], stddev=0.35),
name="weights")
# Create another variable with the same value as 'weights'.
w2 = tf.Variable(weights.initialized_value(), name="w2")
# Create another variable with twice the value of 'weights'
w_twice = tf.Variable(weights.initialized_value() * 0.2, name="w_twice")
The convenience function tf.initialize_all_variables()
adds an op to
initialize all variables in the model. You can also pass it an explicit list
of variables to initialize. See the
Variables Documentation for more options,
including checking if variables are initialized.
The easiest way to save and restore a model is to use a tf.train.Saver
object.
The constructor adds save
and restore
ops to the graph for all, or a
specified list, of the variables in the graph. The saver object provides
methods to run these ops, specifying paths for the checkpoint files to write to
or read from.
Variables are saved in binary files that, roughly, contain a map from variable names to tensor values.
When you create a Saver
object, you can optionally choose names for the
variables in the checkpoint files. By default, it uses the value of the
Variable.name
property for
each variable.
Create a Saver
with tf.train.Saver()
to manage all variables in
the model.
# 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: ", save_path
The same Saver
object is used to restore variables. Note that when you
restore variables from a file you do not have to initialize them beforehand.
# 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."
# Do some work with the model
...
If you do not pass any argument to tf.train.Saver()
the saver handles all
variables in the graph. Each one of them is saved under the name that was
passed when the variable was created.
It is sometimes useful to explicitly specify names for variables in the
checkpoint files. For example, you may have trained a model with a variable
named "weights"
whose value you want to restore in a new variable named
"params"
.
It is also sometimes useful to only save or restore a subset of the variables used by a model. For example, you may have trained a neural net with 5 layers, and you now want to train a new model with 6 layers, restoring the parameters from the 5 layers of the previously trained model into the first 5 layers of the new model.
You can easily specify the names and variables to save by passing to the
tf.train.Saver()
constructor a Python dictionary: keys are the
names to use, values are the variables to manage.
Notes:
You can create as many saver objects as you want if you need to save and
restore different subsets of the model variables. The same variable can be
listed in multiple saver objects, its value is only changed when the saver
restore()
method is run.
tf.initialize_variables()
for more information.# 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.
...