If you just want the frequencies corresponding to STFT bands, librosa.fft_frequencies() will compute that for you.
More generally, the discrete Fourier transform is defined by a collection of complex sinusoids (n_fft different frequencies) ranging from f=-sr/2 to f=+sr/2. If your signal is real-valued (as is the case for audio signals), then the negative frequencies are redundant, so we only keep the frequencies f=0, sr/n_fft, 2*sr/n_fft, ... sr/2. This results in a "real" Fourier transform with 1 + n_fft // 2 frequencies, in your case, 0 up to 5000 Hz in increments of 100 Hz (10000 / 100).