best approach for fitting final model after tuning

Your understanding about set.seed() is good. However, initial_split() uses random numbers so setting the seed after that won't control the randomness of the split (so everything downstream will be different).

1 Like