My thoughts exactly. A DL for 100 row dataset seems like overkill (and likely overfits).