Typeerror: '>' Not Supported Between Instances Of 'nonetype' And 'float'

I have this code and it raise an error in python 3 and such a comparison can work on python 2 how can I change it? import tensorflow as tf def train_set(): class MyCallBacks(t

Solution 1:

Tensorflow 2.0

DESIRED_ACCURACY = 0.979classmyCallback(tf.keras.callbacks.Callback):
    defon_epoch_end(self, epochs, logs={}) :
        if(logs.get('acc') isnotNoneand logs.get('acc') >= DESIRED_ACCURACY) :
            print('\nReached 99.9% accuracy so cancelling training!')
            self.model.stop_training = True

callbacks = myCallback()

Solution 2:

it seems that your error is similar to Exception with Callback in Keras - Tensorflow 2.0 - Python try replacing logs.get('acc') with logs.get('accuracy')

Solution 3:

It works in Python2 because in Python2 you can compare None with float but this is not possible in Python3.

This line

logs.get('acc')

returns None and there is your problem.

Quick solution would be to replace the condition with

if logs.get('acc') isnot None and logs.get('acc') > 0.95:

If logs.get('acc') is None then the above condition will be short-circuited and the second part, logs.get('acc') > 0.95, will not be evaluated, therefore it will not cause the mentioned error.

Solution 4:

Inside your callback try this :

classmyCallback(tf.keras.callbacks.Callback):
      defon_epoch_end(self, epoch, logs={}):
        print("---",logs,"---")
        '''
        if(logs.get('acc')>=0.99):
          print("Reached 99% accuracy so cancelling training!")
        '''

It gave me this --- {'loss': 0.18487292938232422, 'acc': 0.94411665} ---

I had acc so I used, if had accuracy I would have used accuracy. So log and what do you have then use that.

TF goes through major changes all the time, so its ok to play safe, very safe.

Solution 5:

Use 'acc' instead of 'accuracy' and you don't need to change.