Why do they matter?
If you choose the wrong hyperparameters, your model will either:
Underfit (perform poorly because it's too simple).
Overfit (memorize the training data but fail on new data).
Take forever to train.
Fail to converge (never find a good solution).