An experiment you cannot rerun is not a result. Seeding the shuffle is the difference between "the new model is better" and "the new model got a luckier split".
Write split_indices(n, test_fraction, seed).
Return a tuple (train_idx, test_idx) of two integer arrays that together contain every index from 0 to n - 1 exactly once, shuffled, with the test set holding floor(n * test_fraction) of them.
You have read the opening. The rest is free with a Bezen account, and no card is needed.
Behind the sign-in