A fiber beam splitter with the same splitting ratio as a bulk beam splitter will behave exactly the same in a Hong-Ou-Mandel interference experiment.
Linear-optical elements are fully determined by their action on a single photon, and a balanced beam splitter takes (by definition)
$$
|1\rangle_a|0\rangle_b \mapsto \alpha|1\rangle_a|0\rangle_b + \beta|0\rangle_a|1\rangle_b, \qquad |\alpha|^2 = |\beta|^2 = \frac{1}{2}.
$$
The only free parameters determined by the physical implementation are the global and relative phases of the amplitudes $\alpha$ and $\beta$. However, in a Hong-Ou-Mandel effect the input state is
$$
|\psi\rangle = |1\rangle_a|1\rangle_b,
$$
and any phase shifts in the input modes simply become global phases that can't affect the output probabilities. Similarly, phases in the two output modes don't couple the two modes, and can't change the photon number. The bunching effect therefore doesn't depend on these phases at all, only on the splitting ratio.
From a more practical point of view, a realistic description of the HOM effect requires a multi-mode treatment, and HOM visibility is often limited by imperfect mode matching. Since single-mode fibers, and fiber beam splitters, only allow for a single spatial mode, it is comparatively easier to achieve high HOM interference visibility. However, additional care needs to be taken to ensure that the polarisation of the photons are the same at the beam splitter, since the fibers will induce random polarisation rotations and in general the exact state of polarisation at the beam splitter will be unknown.
