Hi, recently I've tried to reproduce the results of Lumina-Image-2.0 on DPG-Bench. However, I finally got a score much lower than that reported in your paper (mine is 80.93 but the reported one is 87.20). I supposed that your parameter setting may vary from the default setting of your given example in HF repo. Could you please share your parameter setting for DPG-Bench testing?