If I have a dataset with 1000 entries, can I only do a scikit test-train-split with Spyder (Python) or do I also need to have a separate test set to the training set which I