notes

Useful techniques and methods that you can use while you are developing machine learning models.

Stop training process if the trained model reaches the desired accuracy (or loss)

DESIRED_ACCURACY = 0.98

class myCallback(tf.keras.callbacks.Callback):
  def on_epoch_end(self, epoch, logs = {}):
    if(logs.get('acc') > DESIRED_ACCURACY):
      print(f"\nReached {DESIRED_ACCURACY * 100}% accuracy so cancelling training!")
      self.model.stop_training = True

callbacks = myCallback()

How to use?

We can make use of tf.keras.callbacks.Callback to stop training if it reaches a desired accuracy (or loss). Simply create a myCallback class and call it. After that, don't forget to add [callbacks].

model.fit(training_data, training_labels, epochs=100, callbacks=[callbacks])

If you would like to use loss instead of accuracy, use logs.get('loss').