Normally not at all, since the trained trades have no stop and TP. But often distant stops are still used for preventing rogue trades. The distance is then determined by average volatility.

Of course, if you want, you can optimize stops and TPs with a case 1 combination, after training the ML model.