AaltoML
dynamic ADC error
"When training on the Phototourism dataset with the --pipeline.model.use_adc True option, an error like below occurs during the training process. How can I resolve this?"
File "/mnt/ssd1/taeyoung/nerfstudio/nerfstudio/models/base_model.py", line 143, in forward return self.get_outputs(ray_bundle) File "/mnt/ssd1/taeyoung/desplat/desplat/desplat_model.py", line 1196, in get_outputs render_2d, alpha_2d, info_2d = rasterization( File "/opt/conda/lib/python3.10/site-packages/gsplat/rendering.py", line 190, in rasterization assert means.shape == (N, 3), means.shape AssertionError: torch.Size([1, 1, 3])
Hi! Thanks for your attention. I reran the script but didn’t encounter the bug on my end. Could you please provide a screenshot of the error when it occurs? How many steps into the run does the issue appear?
Additionally, could you let me know which versions of Nerfstudio and gsplat you’re using? That will help me reproduce the issue more accurately.
Thanks for your response!
my nerfstudio version is 1.1.3 & gsplat version is 1.0.0 and script is ns-train desplat --output-dir "/mnt/ssd1/taeyoung/desplat/outputs/bra" --viewer.quit-on-train-completion True --steps_per_save 200000 --max_num_iterations 200000 --pipeline.model.stop_split_at 100000 --pipeline.model.enable_appearance True --pipeline.model.app_per_gauss True phototourism-data --data /mnt/wild/PhotoTourism/brandenburg_gate/dense/
I'm encountering the same error. I'm using nerfstudio v1.1.4 and gsplat v1.0.0. The command was the same, but with my own dataset. Could you share any insights or solutions you've found? Thanks!
My temporary solution was to add "continue_cull_post_densification: bool = False" to the desplat config.