search

JonasSchult / Mask3D

Mask3D predicts accurate 3D semantic instances achieving state-of-the-art on ScanNet, ScanNet200, S3DIS and STPLS3D.
MIT License
547 stars 110 forks source link

Result Reproduction #63

Closed albjlee closed 1 year ago

albjlee commented 1 year ago

Hello! I'm just getting started with Mask3D, and I ran into some issues when trying to run main_instance_segmentation.py using the scannet_pretrained checkpoint. After installing the necessary dependencies and running setup.py, I tried running the following inference commands:

python main_instance_segmentation.py \ general.checkpoint='scannet_pretrained.ckpt' \ general.train_mode=false

However, I ran into the following error:

terminate called after throwing an instance of 'std::bad_alloc' what(): std::bad_alloc Fatal Python error: Aborted

Is there any known fix for this issue? Am I missing a crucial step? I am using Python 3.10.6 and Cuda 11.6. On top of this, I am using the following torch packages: torch 1.12.0+cu116
torch-scatter 2.1.1
torchmetrics 0.11.4
torchvision 0.13.0+cu116

Please let me know if there is any other information I can provide. Thank you so much!

JonasSchult commented 1 year ago

Hi!

Thanks a lot for your interest in our work! As it seems some newer dependencies are problematic for Mask3D. Try out the instructions in this pull request. Please let me know if you could install Mask3D in that way.

Best, Jonas