ImportError for libcudart.so.11.0 when Running launch.py in Google Colab

[markuryy]

Environment:

• Google Colab
• CUDA version: 12.2 (as obtained from nvcc --version in Colab)

Issue Description:

I'm encountering an ImportError when trying to run cell 3 launch.py from the Wonder3D repository in a Google Colab environment. Despite following the installation instructions and ensuring dependencies are installed, the script fails with a missing CUDA runtime library error.

Steps to Reproduce:

1. Cloned the Google Colab environment.
2. Ran cell 3:

   %cd /content/Wonder3D/instant-nsr-pl
   !python launch.py --config configs/neuralangelo-ortho-wmask.yaml --gpu 0 --train dataset.root_dir=/content/Wonder3D/outputs/cropsize-192-cfg3.0 dataset.scene=dragonfly

3. Encountered the following error:

   /content/Wonder3D/instant-nsr-pl
   Traceback (most recent call last):
   File "/content/Wonder3D/instant-nsr-pl/launch.py", line 125, in <module>
   main()
   File "/content/Wonder3D/instant-nsr-pl/launch.py", line 38, in main
   import datasets
   File "/content/Wonder3D/instant-nsr-pl/datasets/__init__.py", line 16, in <module>
   from . import blender, colmap, dtu, ortho
   File "/content/Wonder3D/instant-nsr-pl/datasets/blender.py", line 14, in <module>
   from models.ray_utils import get_ray_directions
   File "/content/Wonder3D/instant-nsr-pl/models/__init__.py", line 16, in <module>
   from . import nerf, neus, geometry, texture
   File "/content/Wonder3D/instant-nsr-pl/models/nerf.py", line 9, in <module>
   from models.utils import chunk_batch
   File "/content/Wonder3D/instant-nsr-pl/models/utils.py", line 10, in <module>
   import tinycudann as tcnn
   File "/usr/local/lib/python3.10/dist-packages/tinycudann/__init__.py", line 9, in <module>
   from tinycudann.modules import free_temporary_memory, NetworkWithInputEncoding, Network, Encoding
   File "/usr/local/lib/python3.10/dist-packages/tinycudann/modules.py", line 51, in <module>
   _C = importlib.import_module(f"tinycudann_bindings._{cc}_C")
   File "/usr/lib/python3.10/importlib/__init__.py", line 126, in import_module
   return _bootstrap._gcd_import(name[level:], package, level)
   ImportError: libcudart.so.11.0: cannot open shared object file: No such file or directory

Attempts to Resolve:

• Installed tiny-cuda-nn as per the maintainer's instructions:

  !pip install ninja git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch

• Restarted the Colab runtime and re-ran the code.
• Checked for any updates or branches of Wonder3D that might be compatible with CUDA 12.2.

Additional Context:

It appears that tiny-cuda-nn requires CUDA 11.0, but Google Colab's environment is using CUDA 12.2. I'm unsure how to resolve this version mismatch or if there's a workaround available.

I'd appreciate any guidance or suggestions on how to proceed with running Wonder3D in Google Colab. Thank you!

[camenduru]

thanks for the info ❤ please try again