Is it necessary to load volume in ckpt?

[zinsama]

I noticed in

mvsnerf/train_mvs_nerf_finetuning_pl.py

Line 66 in 1fdf648

volume_feature = ckpts['volume']['feat_volume']

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?