OOM for Raw-to-Raw dataset when inference using a single A100 GPU

[RenYangSCU]

Thanks for your great work. I encountered out-of-memory mistakes during inference in the RAW-to-RAW dataset using a single 40G A100 GPU. However, as described in your paper, the results are reported on an NVIDIA GeForce RTX 4090 GPU, which has only 24 GB of GPU memory. Are there any mistakes or special procedures during inference? I'm looking forward to hearing back from you. Thanks!