How to use the torchdiffeq._impl.misc._scaled_dot_product function in torchdiffeq

def _interp_fit_dopri5(y0, y1, k, dt, tableau=_DORMAND_PRINCE_SHAMPINE_TABLEAU):
    """Fit an interpolating polynomial to the results of a Runge-Kutta step."""
    dt = dt.type_as(y0[0])
    y_mid = tuple(y0_ + _scaled_dot_product(dt, DPS_C_MID, k_) for y0_, k_ in zip(y0, k))
    f0 = tuple(k_[0] for k_ in k)
    f1 = tuple(k_[-1] for k_ in k)
    return _interp_fit(y0, y1, y_mid, f0, f1, dt)

def _interp_fit_adaptive_heun(y0, y1, k, dt, tableau=_ADAPTIVE_HEUN_TABLEAU):
    """Fit an interpolating polynomial to the results of a Runge-Kutta step."""
    dt = dt.type_as(y0[0])
    y_mid = tuple(y0_ + _scaled_dot_product(dt, AH_C_MID, k_) for y0_, k_ in zip(y0, k))
    f0 = tuple(k_[0] for k_ in k)
    f1 = tuple(k_[-1] for k_ in k)
    return _interp_fit(y0, y1, y_mid, f0, f1, dt)