Wired Depth Image of Free Viewport Rendering

[JuntingLee]

Describe the bug
I run the commond for free viewport rendering as the usage guide.
The depth of result is just a red image for each view, while the color, normal and voxel look good.

To Reproduce
python render.py Synthetic_NSVF/Wineholder/ --user-dir fairnr --task single_object_rendering --path logs/checkpoint_best.pt --model-overrides '{"chunk_size":512,"raymarching_tolerance":0.01}' --render-beam 1 --render-angular-speed 3 --render-num-frames 15 --render-save-fps 24 --render-resolution "800x800" --render-path-style "circle" --render-path-args "{'radius': 3, 'h': 2, 'axis': 'z', 't0': -2, 'r':-1}" --render-output logs/output --render-output-types "color" "depth" "voxel" "normal" --render-combine-output

Expected behavior
The right depth image as the example in usage guide.

Screenshots

Desktop:

• OS: Ubuntu20.04
• GPU: RTX3090
• CUDA: 11.1
• python: 3.8
• pytorch: 1.12