The model was trained on Lyriel (v2.0 - Reliberate), and I advise you to use it.
I don't know how Lora works with the others.
Recommendations:
Lora - 0.6-0.7
CFG Scale - 5.5-6.5 (7-7.5 for v2.0)
25/27 steps.
DMP++ 2M
Description
The whole style has been changed. It looks different now.
I suggest using Reliberate, majicMIX_realistic.