Some minor issues and loss=nan problem

Hi, there！Sorry to reply you so late, I have been reading your code and running experiments recently.

But I ran into some small problems as follows,

1. When running the test code, the following code will make an error

python eval.py  \
       --dataset_name blender_ray_patch_1image_rot3d  \
       --root_dir ./synthetic_SinNeRF/lego/  \
       --N_importance 64 --img_wh 400 400 --model nerf \
       --ckpt_path ./ckpts/lego_s6_4ft/last.ckpt \
       --timestamp test

When the default value of the --split is test, an error will be reported here. note that frame https://github.com/VITA-Group/SinNeRF/blob/6f101f924fe9ba7793df5a9bbc52b2c82423e251/datasets/blender_ray_patch_1image_rot3d.py#L540

---

2. The dtu file you uploaded is missing a part or the code is wrong? https://github.com/VITA-Group/SinNeRF/blob/6f101f924fe9ba7793df5a9bbc52b2c82423e251/datasets/dtu_proj.py#L433-L434

---

3. loss nan problem I am running the latest code on an RTX 3090, 24G, and environments are created using environment.yaml, but is still OOM, so I adjusted the patch_size, precision and --sH, --sW according to the README.

I set precision=16, and --sH, --sW remain the same(=6).

I found that the loss=nan problem will appear when the patch_size is too small, such as patch_size=8 or patch_size=16, even patch_size=32.

It works(without loss=nan) when patch_size=50, but that’s not a good number is it?

I would be grateful if you could provide advice on how to deal with this. Thank you !

[Ir1d]

@HannahHaensen For custom images, if the depth contains a lot of zeros, there are chances of nan as well. You can considering masking out the zero depth regions

[happysxpp]

@HannahHaensen Hi,

I set --patch_size 56 --sW 8 --sH 8, others remain the same.

Since then, I have not continued to study this paper in depth. So I can’t offer much help.