How to use the yasa.numba._rms function in yasa

sp_abs = np.zeros(n_sp)
    sp_rel = np.zeros(n_sp)
    sp_sta = np.zeros(n_sp)

    # Number of oscillations (= number of peaks separated by at least 60 ms)
    # --> 60 ms because 1000 ms / 16 Hz = 62.5 ms, in other words, at 16 Hz,
    # peaks are separated by 62.5 ms. At 11 Hz, peaks are separated by 90 ms.
    distance = 60 * sf / 1000

    for i in np.arange(len(sp))[good_dur]:
        # Important: detrend the signal to avoid wrong peak-to-peak amplitude
        sp_x = np.arange(data[sp[i]].size, dtype=np.float64)
        sp_det = _detrend(sp_x, data[sp[i]])
        # sp_det = signal.detrend(data[sp[i]], type='linear')
        sp_amp[i] = np.ptp(sp_det)  # Peak-to-peak amplitude
        sp_rms[i] = _rms(sp_det)  # Root mean square
        sp_rel[i] = np.median(rel_pow[sp[i]])  # Median relative power

        # Hilbert-based instantaneous properties
        sp_inst_freq = inst_freq[sp[i]]
        sp_inst_pow = inst_pow[sp[i]]
        sp_abs[i] = np.median(np.log10(sp_inst_pow[sp_inst_pow > 0]))
        sp_freq[i] = np.median(sp_inst_freq[sp_inst_freq > 0])

        # Number of oscillations
        peaks, peaks_params = signal.find_peaks(sp_det,
                                                distance=distance,
                                                prominence=(None, None))
        sp_osc[i] = len(peaks)

        # For frequency and amplitude, we can also optionally use these
        # faster alternatives. If we use them, we do not need to compute the