how to define condition statement in tensorflow?

Question

I have three functions in file fe_extraction.py

def rms_value(x):
    return tf.sqrt(tf.reduce_mean(tf.square(x)))
def meanabs(x):
    return tf.reduce_mean(tf.abs(x))

def req_value(x,y,Thersh):
    z = tf.cond(y>Thersh,rms_freq(x),peak_value(x))
return z

I want to simply apply a condition if y > thershold perform rms_freq(x) or else peak_value(x) and return that value. y is the value obtained from another function.

# given values
# Thershold = 10.69 
# x is defined as tf.Variable , dtype tf.float64
# y = 45.34 obtained from function
....
z = fe_extraction.req_value(x,y,Thershold)

I get error as TypeError:fn1 must be callable.

nessuno · Accepted Answer · 2017-01-22 09:39:11Z

2

With rms_freq(x) and peak_value(x) you're calling the function rms_freq and peak_value respectively, passing x as argument.

Instead, you have to pass a callable or, in other words, a function, that tf.cond can execute.

Since you want x as a parameter for your functions, you can wrap them in a lambda that defines a callable object that captures the outside scope and thus sees the parameter x.

z = tf.cond(y>Thersh,lambda: rms_freq(x) ,lambda: peak_value(x))

answered Jan 22, 2017 at 9:39

nessuno

27.1k5 gold badges87 silver badges78 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

Raady Over a year ago

it gives me an error again 'TypeError: pred must not be a Python bool'

nessuno Over a year ago

y>Thersh is a python boolean, this means that both y and Thersh are python variables, so tf.cond is useless in this case, you can just use a python if statement. By the way, cond should be a tensorflow boolean, you must use tf operations like tf.greater(y, Thersh, name=None)

Collectives™ on Stack Overflow

how to define condition statement in tensorflow?

1 Answer 1

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related