Tensorflow: tf.data.Dataset, Cannot batch tensors with different shapes in component 0

Question

I have the following error in my input pipeline:

tensorflow.python.framework.errors_impl.InvalidArgumentError: Cannot batch tensors with different shape

Answer 1

What worked for me with tensorflow 2 :

1) setting the repeat() function when creating the dataset, therefore replacing

data = tf.data.Dataset.from_tensor_slices(x)

by

data = tf.data.Dataset.from_tensor_slices(x).repeat()

B) Passing the steps/epoch parameter to the fit method, therefore replacing

history = model.fit(dataset, epochs=EPOCHS, callbacks=[checkpoint_callback])

by

history = model.fit(dataset, epochs=EPOCHS, steps_per_epoch=data[0]/BUFFER_SIZE, callbacks=[checkpoint_callback])