Question about evaluation scripts for quantitative metrics (PSNR / SSIM / LPIPS)

Hi, thank you for the great work and for releasing the code and pretrained checkpoints.

I have a question regarding the evaluation of the quantitative metrics reported in the paper, such as PSNR, SSIM, and LPIPS. I was wondering:

Do you have any plan to release the official evaluation / testing scripts used to compute these metrics?

Or alternatively, could you provide some clarification on the exact evaluation protocol?

I have implemented my own evaluation pipeline following common practices. However, when evaluating the official checkpoint, I consistently obtain PSNR values that are about 2 dB lower than those reported in the paper. This makes me suspect that there might be some differences in the evaluation settings.

Any guidance on how these metrics were computed would be greatly appreciated. Thank you very much for your time and help!

No files, tests, or entry points are named. Start by locating the repository's evaluation scripts and checkpoint-testing entry point, then compare their PSNR, SSIM, and LPIPS settings with the paper's protocol; done means the official scripts or exact protocol explains how to reproduce the reported metrics and the approximately 2 dB PSNR difference.