Attempt to install on Windows

[psx2]

I would like the developers of this repository to attempt to install on a fresh Windows.

One error leads to another, and another, and another...

The installation instructions provided must be severely outdated. I followed them precisely, matching the exact dependencies, and have unable to get anywhere.

[Caenorst]

Hi @psx2 , we are currently working on setting up the windows CI, but I can tell that we have been able to install Kaolin on a windows machine without issue.

Usually Windows + pytorch is finicky, but we can't help you if you don't share more information about your system.

[DocSavage]

Not OP but I can give you some data on my Win install that kind of worked though I'll have to wait to see if detected errors are roadblocks. This was a kaolin install within an attempt to install wisp using a new conda environment.

OS = Win 10 Enterprise, 21H2, OS build 19044.1889, running on VMWare vWS with nvidia licensing (VDI solution) GPU = 3 x GRID RTX6000P-24Q. Driver 472.39 Pytorch 1.12.1 via standard approach. However nvcc --version showed I was at CUDA 11.4 and only 11.3 was available via conda packages, which I presume can work with 11.4, so did cudatoolkit=11.3. Python 3.8 installed when creating conda environment. Went with what Wisp install suggested. Kaolin version HEAD (commit 54876e0). When I tried doing checkout to v0.12.0 ran into bug that seems to have been patched. Was syntax error in kaolin\csrc\ops\mesh\mesh_intersection_cuda.cu. Will note that there were a lot of warnings and output when running python setup.py develop for kaolin install.

On doing python -c "import kaolin; print(kaolin.__version__)":

C:\Users\X\Anaconda3\envs\wisp\lib\site-packages\scipy-1.7.2-py3.8-win-amd64.egg\scipy\__init__.py:146: UserWarning: A NumPy version >=1.16.5 and <1.23.0 is required for this version of SciPy (detected version 1.23.1
  warnings.warn(f"A NumPy version >={np_minversion} and <{np_maxversion}"
0.12.0

Installed what was suggested for exhaustive checks via pytest. For the most part it passed or skipped with a few exceptions:

tests\python\kaolin\io\test_modelnet.py EEEEEEEEEEEE                                                                                       [  0%]
...
tests\python\kaolin\io\test_shapenet.py EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE [  1%]
EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE [  2%]
EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE                                       [  3%]
tests\python\kaolin\io\test_shrec.py EEEEEEEEEEEEEEFFFFFFFEEEEEEEEEEEEEEFFFFFFFEEEEEEEEEEEEEEFFFFFFFEEEEEEEEEEEEEEFFFFFFF                  [  4%]
... (all other tests passed or were skipped)
... (lots of error dumps skipped)
________________________________________________ TestSHREC16.test_neg_index[-2-test-categories5] ________________________________________________

self = <test_shrec.TestSHREC16 object at 0x00000116DC9D6D60>, shrec16_dataset = <kaolin.io.shrec.SHREC16 object at 0x00000116802A5A60>, index = -2

    @pytest.mark.parametrize('index', [-1, -2])
    def test_neg_index(self, shrec16_dataset, index):

>       assert len(shrec16_dataset) > 0
E       assert 0 > 0
E        +  where 0 = len(<kaolin.io.shrec.SHREC16 object at 0x00000116802A5A60>)

tests\python\kaolin\io\test_shrec.py:88: AssertionError
________________________________________________ TestSHREC16.test_neg_index[-2-test-categories6] ________________________________________________

self = <test_shrec.TestSHREC16 object at 0x00000116DC9D6DF0>, shrec16_dataset = <kaolin.io.shrec.SHREC16 object at 0x0000011680311EB0>, index = -2

    @pytest.mark.parametrize('index', [-1, -2])
    def test_neg_index(self, shrec16_dataset, index):

>       assert len(shrec16_dataset) > 0
E       assert 0 > 0
E        +  where 0 = len(<kaolin.io.shrec.SHREC16 object at 0x0000011680311EB0>)

tests\python\kaolin\io\test_shrec.py:88: AssertionError
=============================================================== warnings summary ================================================================
...
ERROR tests/python/kaolin/io/test_shrec.py::TestSHREC16::test_neg_index[-2-val-categories5] - ValueError: Class 02691156 (['airplane', 'aeropla...
ERROR tests/python/kaolin/io/test_shrec.py::TestSHREC16::test_neg_index[-2-val-categories6] - ValueError: Class 02691156 (['airplane', 'aeropla...
=========================== 28 failed, 12596 passed, 1024 skipped, 108337 warnings, 404 errors in 1706.10s (0:28:26) ============================

[Caenorst]

Thanks @DocSavage for sharing your configs.

The errors are expected since you don't have the datasets at the specific location, if you set CI environment variable to 'true' those test should be ignored