As described in the title, when I use scipy.optimize.curve_fit to fit some experimental data to a theory curve, with the methods 'lm' vs 'trf', I got comparable fitted parameters but covariance ...