Training using tf.Dataset in TensorFlow 2.0

I'm having difficulty training my TensorFlow model using a tf.Dataset rather than, say, a pd.DataFrame (which works fine).


I have created a dummy example below that I would expect to work on given what I have read online/on the TensorFlow Certification.


!pip install tensorflow==2.0.0 > /dev/null

import numpy as np
import tensorflow as tf

features, target = np.random.rand(100, 30), np.random.randint(0, 2, 100)
dataset = tf.data.Dataset.from_tensor_slices((features, target))

model = tf.keras.Sequential([
    tf.keras.layers.Dense(30, activation='relu', input_shape=(30,)),
    tf.keras.layers.BatchNormalization(),
    tf.keras.layers.Dropout(0.5),

    tf.keras.layers.Dense(30, activation='relu'),
    tf.keras.layers.BatchNormalization(),
    tf.keras.layers.Dropout(0.5),

    tf.keras.layers.Dense(1, activation='sigmoid')
])

model.compile(
    optimizer='adam',
    loss='binary_crossentropy',
    metrics=['accuracy']
)

model.fit(
    dataset,
    epochs=10,
)

which returns the following error message

...

ValueError: Error when checking input: expected dense_input to have shape (30,) but got array with shape (1,)

Is there anything obviously wrong in the above? Why is TensorFlow grabbing an input with shape (1,)?