2

I am trying to train model with image data-augmentation generators on TensorFlow 2.0, after downloading Kaggle's cats_vs_dogs dataset using below code.

train_datagen = ImageDataGenerator(rescale=1. / 255,
                                   rotation_range=40,
                                   width_shift_range=0.2,
                                   height_shift_range=0.2,
                                   shear_range=0.2,
                                   zoom_range=0.2,
                                   horizontal_flip=True)

test_datagen = ImageDataGenerator(rescale=1. / 255)

train_generator = train_datagen.flow_from_directory(train_dir,  
                                                    target_size=(150, 150), 
                                                    batch_size=32,
                                                    class_mode='binary')

validation_generator = test_datagen.flow_from_directory(validation_dir,  
                                                    target_size=(150, 150), 
                                                    batch_size=32,
                                                    class_mode='binary')

history = model.fit_generator(train_generator,
                              steps_per_epoch=100,
                              epochs=100,
                              validation_data=validation_generator,
                              validation_steps=50)

But on first epoch, getting this error: 

Found 2000 images belonging to 2 classes.
Found 1000 images belonging to 2 classes.
WARNING:tensorflow:From <ipython-input-18-e571f2719e1b>:27: Model.fit_generator (from tensorflow.python.keras.engine.training) is deprecated and will be removed in a future version.
Instructions for updating:
Please use Model.fit, which supports generators.
WARNING:tensorflow:sample_weight modes were coerced from
  ...
    to  
  ['...']
WARNING:tensorflow:sample_weight modes were coerced from
  ...
    to  
  ['...']
Train for 100 steps, validate for 50 steps
Epoch 1/100
 63/100 [=================>............] - ETA: 59s - loss: 0.7000 - accuracy: 0.5000 WARNING:tensorflow:Your input ran out of data; interrupting training. Make sure that your dataset or generator can generate at least `steps_per_epoch * epochs` batches (in this case, 10000 batches). You may need to use the repeat() function when building your dataset.

How should I modify the above code base for TensorFlow 2?

B--rian
  • 5,578
  • 10
  • 38
  • 89
rahiakela
  • 21
  • 4

1 Answers1

0

The kaggle dataset contain 25000 training examples. The error message states that:

Blockquote tensorflow:Your input ran out of data; interrupting training. Make sure that your dataset or generator can generate at least steps_per_epoch * epochs batches (in this case, 10000 batches). You may need to use the repeat() function when building your dataset.

Which means the data generator needs to generate at least 10000 batches. But with the current batch size of 32 the generator would produce only 25000 / 32 is approximately equal to 781 batches. My suggestion is try to reduce the steps_per_epoch or epochs and try.

You can get rid of the deprecation message by passing the generator object to model.fit(...) instead of model.fit_generator

Manik Tharaka
  • 298
  • 3
  • 9
  • 1
    Thanks..it is working after reducing epoch value but my purpose was generating more training data using data augmentation from existing training samples, by augmenting the samples via a number of random transformations that yield believable-looking images. – rahiakela Feb 21 '20 at 08:04
  • What makes you think that the training data was not increased. Try visualizing the generated images. Check out this [tutorial](https://machinelearningmastery.com/how-to-configure-image-data-augmentation-when-training-deep-learning-neural-networks/). This way you can make sure the data augmentation is actually working. – Manik Tharaka Feb 22 '20 at 14:14