How is the map function in tensorflow working over batched data?

Question

This is an example from the tensorflow overfit tutorial:

def pack_row(*row):
  label = row[0]
  features = tf.stack(row[1:],1)
  return features, label

packed_ds