Does cache() counter the effects of shuffling input data between epochs

Question

Hopefully I've understood the process enough that this quesiton actually makes sense.

For training my model (time series) I do my preprocessing outside of tf.dataset functions. I save these as numpy arrays split into several files (eg. train_1.npy,.. train_n.npy.) and load them to model.fit using a generator.

train_dataset = tf.data.Dataset.from_generator(generator=sequence_generator,args= ['train'], output_types = (tf.float32, tf.float32))

---

train_dataset = train_dataset.cache().batch(BATCH).prefetch(tf.data.AUTOTUNE)

validation_dataset = validation_dataset.cache().batch(BATCH).prefetch(tf.data.AUTOTUNE)

test_dataset = test_dataset.batch(64)

---

The generator yields a random file from the set of train/val/test .npy files to help with avoiding overfitting.

My question is, between epochs, does the .cache() part of my code nullify the effect of randomising the .npy files yielded.