We could also just insist that there's something commendable about believing in things that aren't knowable. But the inference there is really hard to figure out. Not everything we can do as humans -- that is, not every option open to us, without regressing to some non-rational state -- is necessarily a good idea, something worth praising or recommending or admiring.
The best option I know to make that inference work is something called "fideism". ("Fides" is Latin for "faith".) Fideists believe that reason is inadequate to reach truth, and all actual knowledge ultimately rests on assumptions which are held as a matter of faith. In other words, faith is an epistemological attitude like belief or committment or acceptance, but taken towards axioms -- things that can't be questioned because they form the basis of one's set of knowledge. And that might seem to work, at least at first glance. It's worth believing in things that aren't knowable because you have to believe in things that aren't knowable. That's how your knowledge gets started: you have assumptions, and knowledge proceeds from them.
There's two problems here. First, fideism is pretty untenable. It's hard to find anyone who really believes it any more. Kierkegaard's account of faith fits in here, and it's illustrative of just how weird the idea is. According to him, believing in Christ -- the perfect, absolute god in limited, imperfect flesh -- requires a leap of faith. It is a paradox, an impossibility, and to accept it thus requires doing something frightening and absurd, believing what you really shouldn't believe at all.
So, on the fideist view, faith is just accepting things, even if they're things that don't make sense, even if they're actively crazy. If that's supposed to make faith appear attractive and worth of commendation, it's a pretty dismal failure.
Second, it's not clear that knowledge works by building up from a set of axioms. In fact, that's pretty much only the way knowledge works in some very limited, very formal domains. Mathematics is the obvious example. Starting with basic principles -- which, ultimately, aren't proven, only assumed -- and derive consequences from them. Of course, this example shows how poor this view of knowledge really is. You can't defend the axioms, because they are where defenses start.
But, practicing mathematicians will (probably) tell you that this isn't actually true. They can and do defend their choice of axiom sets. And they do it by appealing to the consequences derived from them.
Here's an example of how that works. When you're doing geometric proofs, you can assume, following Euclid, that parallel lines never meet. Or you can assume, following the non-Euclidean elliptic geometry, that they do. Whichever set of axioms you're using, you can defend them by looking at the consequences of the axioms and what sorts of problems you're trying to solve. The set of axioms which works best with the problems you're interested in is the one that you're likely to use.
What this shows is that it's a caricature of knowledge to say that is does -- or, really, that it should -- proceed by deducing things from axioms, whether those axioms are proven by reason or accepted on faith. Knowledge actually works as more of a set of interconnected claims. Sometimes axioms provide support for consequences; but sometimes consequences provide support for axioms. (Support can also go laterally, as axioms support each other, but that's tangential.)
And if fideism depends on caricaturing how knowledge works, then it's not a position people should adopt. And if that's the alternative to Kantian rational faith, then Kantian rational faith is the best version of faith available.
Which means that there's nothing worthwhile about faith. It's an option. It's not crazy to have faith. But it's completely unnecessary, and certainly not admirable.