Error when calling MinkowskiPooling

[LinfengUP]

Describe the bug When I apply MinkowskiPooling, it returns \<unknown> error without any extra information.

---

To Reproduce

model.pooling = MinkowskiAvgPooling(
            kernel_size=2, stride=2, dimension=3)

with torch.no_grad():
    coordinates = ME.SparseTensor(
        features=raw_coordinates,
        coordinate_manager=aux[-1].coordinate_manager,
        coordinate_map_key=aux[-1].coordinate_map_key,
        device=aux[-1].device)
    coords = [coordinates]
    for _ in reversed(range(len(aux) - 1)):
        coords.append(model.pooling(coords[-1]))
    coords.reverse()

---

Expected behavior A clear and concise description of what you expected to happen.

---

Desktop (please complete the following information):

• OS: [e.g. Ubuntu 18.04]
• Python version: [3.10]
• Pytorch version: [1.12.1]
• CUDA version: [11.3]
• Minkowski Engine version [0.5.4]

[Jovendish]

I encountered the same problem. Has the problem been solved?

[LinfengUP]

I encountered the same problem. Has the problem been solved?

Oh, I realized I did not add CUDA_HOME in environment variables. Just check whether your nvcc -V can work well.