views_to_mesh.py exits without printing any exception right after "LRM forward time:"

[realisticdreamer114514]

I tried running the script with python main.py --image_prompt [valid image path here] --save_folder ./outputs/test/ --max_faces_num 90000 --do_texture_mapping --save_memory and the output folder only has 2 images, 1 transparent and 1 multi-view image. It apparently didn't do sparse-view reconstruction and exited without printing any errors. What do I miss?

[realisticdreamer114514]

OK, running it again in the gradio UI and the console ends in
./outputs/app_output/0
=====> LRM forward time: 168.7118854522705
and nothing else.
It shows up in the UI as a big red "error" in the tabs after multi-view (obj and glb tabs), which also happened in the HF demo hosted by gokaygokay (https://huggingface.co/spaces/gokaygokay/Hunyuan3D-1.0) I tried before cloning your repo. Doesn't seem like a user problem (I use Windows 10)

[realisticdreamer114514]

With my 6GB GPU it can complete the 'image to view' step with 12GB (~6GB VRAM, ~6GB RAM) in about 35min (standard model), and then it would exit without any OOM warning when the 'views to mesh' step should start. (32GB RAM total so 16GB "shared VRAM") Not sure if I'm trying to pull the impossible with my hardware constraints

[realisticdreamer114514]

Tested infer/views_to_mesh.py after the new commit, same behavior, at least I'm sure the bug is at this step and it isn't a VRAM issue (I can run infer/image_to_views.py, the part with highest VRAM requirement, slowly)

[Quantuman134]

meet the same problem

[1999XIAOZHANG]

遇到同样的问题

你是怎么解决的？我也遇到了这个 问题