Training stages of extended LDL. We first trained the base LDL parameters on the large volumes available for the fiducial model (CAMELS/CV) at z = 0.21. We then conditioned the LDL parameters on the cosmological and astrophysical parameters using the numerous boxes from the CAMELS/LH set. While the n_e model performs equally at all redshifts, the T emulator has to be retrained separately at all available redshifts.