If Bayesianism replaces the categorical belief of former analyses of knowledge with a quantitative notion of partial belief, does Bayesianism have a concept of 'knowledge' at all?
Well, the notion of "knowledge" as an absolute certainty is flawed from the start. There is always some way you could be wrong. Even if you are dealing with something perfectly deductive, such as mathematics, there is always some chance of a mistake. Humans are fallible.
So if you interpret "knowledge" to mean "absolute certainty," you will find no places where you can legitimately use the word.
But "knowledge" is not a bad word. It's useful to describe high levels of justified confidence in something true, in which the justification is not based on a false picture of affairs (i.e. not Gettier problems). What level of justified confidence makes something knowledge? That's rather subjective. But it's not 100%. Maybe it is 95% or 99%. Bear in mind that people typically overestimate their confidence in things, so a threshold of 95% properly justified confidence is higher than it might sound.
Anyway. The point is, a notion of "knowledge" as "perfect certainty" is unattainable, so if we want to keep using the word "knowledge" - and it is a word useful in practice, so we should keep using it - we ought to define the word in a different way to reflect some high, though still imperfect, level of confidence.