revise wording

tensorflow/python/data/ops/dataset_ops.py

@@ -1406,10 +1406,10 @@ def shuffle(
     maintaining the 1,000 element buffer.
 
     `reshuffle_each_iteration` controls whether the shuffle order should be
-    different for each epoch. However you should avoid using `shuffle` with
-    `reshuffle_each_iteration=True`, then `take` and `skip` to split a dataset
-    into training and test set, which would lead to data leakage (as the
-    entire dataset would be re-shuffled then re-split after each epoch).
+    different for each epoch. However you should avoid using
+    `shuffle(reshuffle_each_iteration=True)`, then `take` and `skip` to split
+    a dataset into training and test sets, which would lead to data leakage (as
+    the entire dataset would be re-shuffled then re-split after each epoch).
     Please use the `tf.keras.utils.split_dataset` method instead. In TF 1.X,
     the idiomatic way to create epochs was through the `repeat` transformation: