Splitting refers to the process of segregating subsets of data and designating them for distinct purposes. Typically, three main splits are created: train (provided to the model for learning), valid (used for evaluating the model during the training process), and test (reserved until the final stage to assess how well the model generalizes to unseen data).
