Why use slice(lr) inside fit1cycle() after freeze?

To my understanding,
learn.fit_one_cycle(5, slice(lr))

the above code, slice(lr) assigns lr to the last layer group and all the other groups are lr/10

but freeze() method sets all other groups untrainable, so when calling fit_one_cycle() we can only train the last layer group.

Then there is no need to use slice(lr) in the code.

So my question is: why do we use slice(lr) here?

Is this just a mistake?