Upuść duplikat pandy oprócz NAN

df[(~df.duplicated()) | (df['col'].isnull())]
Victorious Vulture