Windows 10 installation problem

[Timothychance]

I am sorry to bother you, and i have noticed that there are some similar issues, but I still can't fix problems according to instructions and i relly need help!

1.pytorch1.7+NVIDA cuda11.0+cudnn8.04 +set CUB_HOME + VS2019 16.9

2.in '' pytorch3d\csrc\iou_box3d\iou_utils.cuh'' and '' pytorch3d\csrc\iou_box3d\iou_utils.h'', use '!' to replace 'not'
deleted const auto kEpsilon and replaced all of its references with its value (1e-4) in pytorch3d\csrc\iou_box3d/iou_utils.cuh

3.set DISTUTILS_USE_SDK=1
python setup.py install

4 . Complete error recording：
error_1.txt

I am very grateful for your help！

[bottler]

I'm afraid I can't read the main error in the log, due to language and encoding problems. I see

D:\Anaconda\lib\site-packages\torch\include\ATen/cuda/Exceptions.h(54): error C2373: ¡°cusparseGetErrorString¡±: Öض¨Ò壻²»Í¬µÄÀàÐÍÐÞÊηû
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.0\include\cusparse.h(306): note: ²Î¼û¡°cusparseGetErrorString¡±µÄÉùÃ÷

or

D:\Anaconda\lib\site-packages\torch\include\ATen/cuda/Exceptions.h(54): error C2373: °∞cusparseGetErrorString°±: ÷ÿ∂®“£ª≤ªÕ¨µƒ¿‡–Õ–fi Œ∑˚
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.0\include\cusparse.h(306): note: ≤Œº˚°∞cusparseGetErrorString°±µƒ…˘√˜
◊

Can you see the error properly? Also, what versions of PyTorch and CUDA and VSCode are you using?

[Timothychance]

I'm afraid I can't read the main error in the log, due to language and encoding problems. I see

D:\Anaconda\lib\site-packages\torch\include\ATen/cuda/Exceptions.h(54): error C2373: ¡°cusparseGetErrorString¡±: Öض¨Ò壻²»Í¬µÄÀàÐÍÐÞÊηû
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.0\include\cusparse.h(306): note: ²Î¼û¡°cusparseGetErrorString¡±µÄÉùÃ÷

or

D:\Anaconda\lib\site-packages\torch\include\ATen/cuda/Exceptions.h(54): error C2373: °∞cusparseGetErrorString°±: ÷ÿ∂®“£ª≤ªÕ¨µƒ¿‡–Õ–fi Œ∑˚
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.0\include\cusparse.h(306): note: ≤Œº˚°∞cusparseGetErrorString°±µƒ…˘√˜
◊

Can you see the error properly? Also, what versions of PyTorch and CUDA and VSCode are you using?

Thank you so much for such a quick reply, I am sorry for such mistakes, and now I have changed encoding from ANSI to UTF-8, does it work this time?
error.txt

By the way, I use pytorch1.7, cuda11.0 and VS2019 16.9.6

[Timothychance]

I notice what you say in #920
I am using the latest main, so there is no problem with the code. I just make the change:
extra_compile_args = {"cxx": ["/std:c++14"]}

but I get the error report, and I did not find similar error in existing issues.

I don't know what's wrong, is there any good idea to fix the problems?Thank a lot! @bottler

[bottler]

Thank you so much for such a quick reply, I am sorry for such mistakes, and now I have changed encoding from ANSI to UTF-8, does it work this time? error.txt

There's no compiler error in this file, but the build just stopped. I think the parallel compilation ran out of memory. Can you retry the build setting the environment variable PYTORCH3D_NO_NINJA to 1 or MAX_JOBS to 1?

[hfpro]

hello. please tell me what the error is and how to fix
it when compiling an assembly in windows 10 and visual studio 2019, an error occurs.
error.txt

[bottler]

@hfpro I think you are not using the latest version of the code from the main branch. Your error refers to something which has been fixed since the last release. If you have any further questions, please start a new issue, rather than commenting on an existing issue.

[hfpro]

I tried to compile from the links and the main branch, but the errors are repeated.
please help
error.txt

[bottler]

No. Please open a new issue.

[Timothychance]

Hi, thank for your help, I try it again follow your suggestions, and now I get :
Finished processing dependencies for pytorch3d==0.6.0

it seems like a complete build.

Now I follow INSTALL.md to test:
python3 -m unittest discover -p *.py
but here seems to be something wrong:

FAIL: test_simple_sphere_batched (test_render_points.TestRenderPoints)

Traceback (most recent call last):
File "D:\Users\O0001974\Downloads\Compressed\pytorch3d-main\tests\pulsar..\test_render_points.py", line 327, in test_simple_sphere_batched
self.assertClose(rgb, image_ref)
File "D:\Users\O0001974\Downloads\Compressed\pytorch3d-main\tests\common_testing.py", line 183, in assertClose
self.fail(err)
AssertionError: Not close. Max diff 1.0. Shape (256, 256, 3). At (24, 148, 0).

FAIL: test_simple_sphere_pulsar (test_render_points.TestRenderPoints)

Traceback (most recent call last):
File "D:\Users\O0001974\Downloads\Compressed\pytorch3d-main\tests\pulsar..\test_render_points.py", line 129, in test_simple_sphere_pulsar
self.assertClose(rgb, image_ref, rtol=7e-3, atol=5e-3)
File "D:\Users\O0001974\Downloads\Compressed\pytorch3d-main\tests\common_testing.py", line 183, in assertClose
self.fail(err)
AssertionError: Not close. Max diff 1.0. Shape (256, 256, 3). At (19, 116, 0).

FAIL: test_is_eigen_gpu (test_symeig3x3.TestSymEig3x3)

Traceback (most recent call last):
File "D:\Users\O0001974\Downloads\Compressed\pytorch3d-main\tests\test_symeig3x3.py", line 47, in test_is_eigen_gpu
self._test_is_eigen(test_input)
File "D:\Users\O0001974\Downloads\Compressed\pytorch3d-main\tests\test_symeig3x3.py", line 65, in _test_is_eigen
rtol=rtol,
File "D:\Users\O0001974\Downloads\Compressed\pytorch3d-main\tests\common_testing.py", line 183, in assertClose
self.fail(err)
AssertionError: Not close. Max diff 0.00126725435256958. Max relative diff 1.540574550628662 Shape (1024, 3, 3). At (371, 1, 1).

FAIL: test_inverse (test_transforms.TestTransform)

Traceback (most recent call last):
File "D:\Users\O0001974\Downloads\Compressed\pytorch3d-main\tests\pulsar..\test_transforms.py", line 310, in test_inverse
self.assertTrue(torch.allclose(m, m5, atol=1e-3))
AssertionError: False is not true

FAIL: test_coord_grid_transforms (test_volumes.TestVolumes)

Traceback (most recent call last):
File "D:\Users\O0001974\Downloads\Compressed\pytorch3d-main\tests\pulsar..\test_volumes.py", line 380, in test_coord_grid_transforms
self.assertClose(grid_world, grid_world_2, atol=1e-5)
File "D:\Users\O0001974\Downloads\Compressed\pytorch3d-main\tests\common_testing.py", line 183, in assertClose
self.fail(err)
AssertionError: Not close. Max diff 0.01593017578125. Max relative diff 0.0009660190553404391 Shape (3, 4, 5, 7, 3). At (0, 0, 0, 6, 0).

Ran 919 tests in 325.243s

FAILED (failures=47, errors=160, skipped=16)

So, I wonder whether the build is finished? Thank you for help!

[bottler]

I don't know if the tests pass on Windows. There may be small numerical differences which break some. Can you attach a log with all the failures? It may be that your build of pulsar is not working but the rest of PyTorch3D is fine.

[LewisSean]

my first installation's compile error log is same with yours
then i try to command out Line 60 in setup.py
# "-std=c++14",
it works!

[github-actions]

This issue is stale because it has been open 30 days with no activity. Remove stale label or comment or this will be closed in 5 days.