Chris Choy

Unfortunately, no. However there are some issues which people successfully used WSL for MinkowskiEngine.

For the linear interpolation to upsample features, please refer to the https://nvidia.github.io/MinkowskiEngine/source/MinkowskiEngine.html?highlight=interpolation#module-MinkowskiEngine.MinkowskiInterpolation

Hmm, it is strange that the cached allocator does not clean up when there is `MemoryError: std::bad_alloc: cudaErrorMemoryAllocation: out of memory`. Potentially this is coming from thrust functions.

Thanks for your contribution. However, this function requires creating two existing coords map A.C, B.C. Could you use get_union_map on https://github.com/StanfordVL/MinkowskiEngine/blob/master/MinkowskiEngine/MinkowskiCoords.py#L314 to remove the unnecessary coordinate creation?

It uses torch.cuda.is_available() to determine whether to use cuda compilation. Make sure the above command gets True.

The custom region has not been ported to v0.5 yet. Sorry for the delay. I'll try to finish it in a few weeks.

Yes this is expected since you need a GPU-driver to actually use a GPU. It is possible to overcome this, but I don't see any practical use case when you...

Thanks for the heads-up. This was due MinkowskiEngine function classes requiring non pytorch Tensors as a part of input. Specifically, we need to pass the CoordinateManager to function classes and...

There are some previous closed issues about people successfully using WSL for MinkowskiEngine. I haven't tried it myself, but I dont see why WSL would not work for compilation.

If you convolve a sparse tensor with some np.nan in its features, the output sparse tensor would have nans with all elements adjacent to the element with input nan features....