NVIDIAGameWorks / NVIDIAGameWorks/kaolin
Potentially unnecessary CUDA dependency
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 5.2k
- Forks
- 629
- Avg merge
- 53m
- Merged PRs (30d)
- 2
Description
In point_to_mesh_distance, a call to torch.cuda.synchronize causes a fatal crash on systems without NVIDIA GPUs installed.
This synchronize call seems redundant since the _unbatched_naive_point_to_mesh_distance only gets called if the inputs are not on a CUDA device.
We encountered this when trying to use the CPU wheel at https://nvidia-kaolin.s3.us-east-2.amazonaws.com/torch-2.8.0_cpu.html in a CI pipeline on a machine without a GPU.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start in kaolin/metrics/trianglemesh.py at point_to_mesh_distance and compare the CPU path with the torch.cuda.synchronize call around line 232. Reproduce the failure using the CPU wheel on a machine without an NVIDIA GPU, then verify that CPU inputs complete successfully without the fatal CUDA dependency.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, pytorch
- Domain
- machine-learning
- Issue type
- Bug
- Difficulty
- 2/5
- Estimated time
- 1-3 hours
- Activity status
- Active
- Clarity
- Clearly specified
- Newbie friendliness
- 84/100