Performance reduce before and after adding Feature extraction

[imnotwannafire]

Hi Mr.Dbourke,

I'm doing 04. Transfer Learning in TensorFlow Part 1: Feature Extraction Exercises with mobileNetv2 base model
I'm very confused when after adding data augmentation layer to the model, the performance become worse.
Without data augmentation, accuracy is 0.8
After data augmentation, accuracy is only 0.32
Here is the code I'm writing for augmentation data

ata_augmentation_layer = tf.keras.Sequential([
    layers.RandomFlip("horizontal"),
    layers.RandomRotation(0.2),
    layers.RandomZoom(0.2),
    layers.RandomHeight(0.2),
    layers.RandomWidth(0.2),
    layers.Resizing(224, 224),# add resizing to fit the output of data augmentation layer with feature extraction input
    layers.Rescaling(1./255)
], name="data_augmented_layer")

mobilenetv2_model_augmented = tf.keras.Sequential([
    # layers.Input(shape=(224,224,3), name="input_layer"),
    data_augmentation_layer,
    feature_extractor_layer,
    layers.Dense(10, activation="softmax", name="output_layer")
])