cntk.contrib.crosstalk.crosstalk_tensorflow module

class TensorFlowCrosstalk

Bases: cntk.contrib.crosstalk.Crosstalk

TensorFlow implementation for crosstalk

is_trainable(name)

Check if variable with name is a trainable

Parameters:name (str) – Variable name to check

load_all_trainables()

Load all trainables from files in working directory

save_all_trainables()

Save all trainables to files in working directory

set_data(sess, data)

Set session and mapped data for setter/getters

Parameters:

• sess – The tensorflow session
• data – The input data feed dict for eval

find_trainable(name, scope=None)

Find a single trainable variable in a function by name when the function has multiple parameters.

Parameters:

• func – The function to search
• name (str) – The name of the parameter
• scope (str) – The scope of the search

Returns:

The trainable variable that is found