Is it necessary to load volume in ckpt?

[zinsama]

I noticed in https://github.com/apchenstu/mvsnerf/blob/1fdf6487389d0872dade614b3cea61f7b099406e/train_mvs_nerf_finetuning_pl.py#L66 the mvs model loaded volume feature during finetuning, but volume feature tend to be scene-dependent so it should not be loaded. And actually pre-trained checkpint mvsnerf-v0.tar also does not have volume key. So should I load the volume during finetuning on our own dataset?