error checking compiler version for cl --> Cuda error

[jnsLs]

📚 Installation

the dependencies torch-scatter, torch-sparse and torch cluster are not installed correctly
I ran the command for pytorch scatter and get the output (same for sparse and cluster):

running install
running bdist_egg
running egg_info
writing torch_scatter.egg-info\PKG-INFO
writing dependency_links to torch_scatter.egg-info\dependency_links.txt
writing top-level names to torch_scatter.egg-info\top_level.txt
reading manifest file 'torch_scatter.egg-info\SOURCES.txt'
reading manifest template 'MANIFEST.in'
writing manifest file 'torch_scatter.egg-info\SOURCES.txt'
installing library code to build\bdist.win-amd64\egg
running install_lib
running build_py
running build_ext
C:\Users\Jonas\Anaconda3\envs\schnetenv\lib\site-packages\torch\utils\cpp_extension.py:184: UserWarning: Error checking compiler version for cl: 'utf-8' codec can't decode byte 0x81 in position 62: invalid start byte
warnings.warn('Error checking compiler version for {}: {}'.format(compiler, error))
creating build\bdist.win-amd64\egg
creating build\bdist.win-amd64\egg\test
copying build\lib.win-amd64-3.7\test\test_backward.py -> build\bdist.win-amd64\egg\test
copying build\lib.win-amd64-3.7\test\test_forward.py -> build\bdist.win-amd64\egg\test
copying build\lib.win-amd64-3.7\test\test_max_min.py -> build\bdist.win-amd64\egg\test
copying build\lib.win-amd64-3.7\test\test_multi_gpu.py -> build\bdist.win-amd64\egg\test
copying build\lib.win-amd64-3.7\test\test_std.py -> build\bdist.win-amd64\egg\test
copying build\lib.win-amd64-3.7\test\utils.py -> build\bdist.win-amd64\egg\test
copying build\lib.win-amd64-3.7\test_init_.py -> build\bdist.win-amd64\egg\test
creating build\bdist.win-amd64\egg\torch_scatter
copying build\lib.win-amd64-3.7\torch_scatter\add.py -> build\bdist.win-amd64\egg\torch_scatter
copying build\lib.win-amd64-3.7\torch_scatter\div.py -> build\bdist.win-amd64\egg\torch_scatter
copying build\lib.win-amd64-3.7\torch_scatter\max.py -> build\bdist.win-amd64\egg\torch_scatter
copying build\lib.win-amd64-3.7\torch_scatter\mean.py -> build\bdist.win-amd64\egg\torch_scatter
copying build\lib.win-amd64-3.7\torch_scatter\min.py -> build\bdist.win-amd64\egg\torch_scatter
copying build\lib.win-amd64-3.7\torch_scatter\mul.py -> build\bdist.win-amd64\egg\torch_scatter
copying build\lib.win-amd64-3.7\torch_scatter\scatter_cpu.cp37-win_amd64.pyd -> build\bdist.win-amd64\egg\torch_scatter
copying build\lib.win-amd64-3.7\torch_scatter\scatter_cuda.cp37-win_amd64.pyd -> build\bdist.win-amd64\egg\torch_scatter
copying build\lib.win-amd64-3.7\torch_scatter\std.py -> build\bdist.win-amd64\egg\torch_scatter
copying build\lib.win-amd64-3.7\torch_scatter\sub.py -> build\bdist.win-amd64\egg\torch_scatter
creating build\bdist.win-amd64\egg\torch_scatter\utils
copying build\lib.win-amd64-3.7\torch_scatter\utils\ext.py -> build\bdist.win-amd64\egg\torch_scatter\utils
copying build\lib.win-amd64-3.7\torch_scatter\utils\gen.py -> build\bdist.win-amd64\egg\torch_scatter\utils
copying build\lib.win-amd64-3.7\torch_scatter\utils_init_.py -> build\bdist.win-amd64\egg\torch_scatter\utils
copying build\lib.win-amd64-3.7\torch_scatter_init_.py -> build\bdist.win-amd64\egg\torch_scatter
byte-compiling build\bdist.win-amd64\egg\test\test_backward.py to test_backward.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\test\test_forward.py to test_forward.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\test\test_max_min.py to test_max_min.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\test\test_multi_gpu.py to test_multi_gpu.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\test\test_std.py to test_std.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\test\utils.py to utils.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\test_init_.py to init.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\torch_scatter\add.py to add.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\torch_scatter\div.py to div.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\torch_scatter\max.py to max.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\torch_scatter\mean.py to mean.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\torch_scatter\min.py to min.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\torch_scatter\mul.py to mul.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\torch_scatter\std.py to std.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\torch_scatter\sub.py to sub.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\torch_scatter\utils\ext.py to ext.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\torch_scatter\utils\gen.py to gen.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\torch_scatter\utils_init_.py to init.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\torch_scatter_init_.py to init.cpython-37.pyc
creating stub loader for torch_scatter\scatter_cpu.cp37-win_amd64.pyd
creating stub loader for torch_scatter\scatter_cuda.cp37-win_amd64.pyd
byte-compiling build\bdist.win-amd64\egg\torch_scatter\scatter_cpu.py to scatter_cpu.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\torch_scatter\scatter_cuda.py to scatter_cuda.cpython-37.pyc
creating build\bdist.win-amd64\egg\EGG-INFO
copying torch_scatter.egg-info\PKG-INFO -> build\bdist.win-amd64\egg\EGG-INFO
copying torch_scatter.egg-info\SOURCES.txt -> build\bdist.win-amd64\egg\EGG-INFO
copying torch_scatter.egg-info\dependency_links.txt -> build\bdist.win-amd64\egg\EGG-INFO
copying torch_scatter.egg-info\top_level.txt -> build\bdist.win-amd64\egg\EGG-INFO
writing build\bdist.win-amd64\egg\EGG-INFO\native_libs.txt
zip_safe flag not set; analyzing archive contents...
torch_scatter.pycache.scatter_cpu.cpython-37: module references file
torch_scatter.pycache.scatter_cuda.cpython-37: module references file
creating 'dist\torch_scatter-1.2.0-py3.7-win-amd64.egg' and adding 'build\bdist.win-amd64\egg' to it
removing 'build\bdist.win-amd64\egg' (and everything under it)
Processing torch_scatter-1.2.0-py3.7-win-amd64.egg
creating c:\users\jonas\anaconda3\envs\schnetenv\lib\site-packages\torch_scatter-1.2.0-py3.7-win-amd64.egg
Extracting torch_scatter-1.2.0-py3.7-win-amd64.egg to c:\users\jonas\anaconda3\envs\schnetenv\lib\site-packages
Adding torch-scatter 1.2.0 to easy-install.pth file

Installed c:\users\jonas\anaconda3\envs\schnetenv\lib\site-packages\torch_scatter-1.2.0-py3.7-win-amd64.egg
Processing dependencies for torch-scatter==1.2.0
Finished processing dependencies for torch-scatter==1.2.0

Environment

• OS: Windows 10
• Python version: 3.7
• PyTorch version: 1.1.0
• CUDA/cuDNN version: 10.1, V10.1.168
• GCC version: gcc (MinGW.org GCC-8.2.0-3) 8.2.0
• How you tried to install PyTorch Geometric and its extensions (pip, source): source using
• Any other relevant information: using visual studio 2019

Checklist

• [x ] I followed the installation guide.
• [x ] I cannot find my error message in the FAQ.
• [ x] I set up CUDA correctly and can compile CUDA code via nvcc.
• [x ] I have cloned the repository and tried a manual installation from source.
• I do have multiple CUDA versions on my machine.
• [x ] I checked if the official extension example runs on my machine.
• The offical extension example runs on my machine.

##additional info:

running the first example works

import torch
from torch_geometric.data import Data

edge_index = torch.tensor([[0, 1, 1, 2],
[1, 0, 2, 1]], dtype=torch.long)
x = torch.tensor([[-1], [0], [1]], dtype=torch.float)

data = Data(x=x, edge_index=edge_index)

but transfering to the GPU

device = torch.device('cuda')
data = data.to(device)

results in the error:
Traceback (most recent call last):
File "", line 1, in
File "C:\Users\Jonas\Anaconda3\envs\schnetenv\lib\site-packages\torch_geometric\data\data.py", line 247, in to
return self.apply(lambda x: x.to(device), *keys)
File "C:\Users\Jonas\Anaconda3\envs\schnetenv\lib\site-packages\torch_geometric\data\data.py", line 233, in apply
self[key] = func(item)
File "C:\Users\Jonas\Anaconda3\envs\schnetenv\lib\site-packages\torch_geometric\data\data.py", line 247, in
return self.apply(lambda x: x.to(device), *keys)
File "C:\Users\Jonas\Anaconda3\envs\schnetenv\lib\site-packages\torch\cuda_init_.py", line 163, in _lazy_init
torch._C._cuda_init()
RuntimeError: CUDA error: unknown error

Using cuda with torch but without importing torch_geometric, works. It seems that importing torch_geometric is corrupting the use of cuda. I also tried with cuda 10.0 but it did not make a difference.

Thank you in advance
Jonas

[rusty1s]

I'm sadly not sure, as I've never tried to install the packages on Windows, but I'm definitely sure that it's possible. It is maybe wisely to also ask for help here.

[kevin-keraudren]

@jnsLs : you need to load the Visual Studio environment variables before running pip install for torch-scatter. In particular, this will put the Visual Studio compiler tools on your PATH.

In my case, something like this works:

call "%VS2017INSTALLDIR%\VC\Auxiliary\Build\vcvarsall.bat" x64 10.0.17763.0 && set DISTUTILS_USE_SDK=1 && set PY_VCRUNTIME_REDIST=No thanks && set MSSdk=1 &&  pip install --verbose --no-cache-dir torch-scatter

I could not compile with CUDA 10.1 but it worked with CUDA 10.0, so you might want to downgrade your CUDA installation to match with pytorch.

[jnsLs]

Thank you both. However, unfortunately, that didn't solve my problem. Even though torch-scatter is installed successfully (only with the warning: "WARNING: Subprocess output does not appear to be encoded as cp1252"), I can not send tensors to the GPU:
----code-----------------------------------------------

import torch
from torch_geometric.data import Data
edge_index = torch.tensor([[0, 1, 1, 2],
[1, 0, 2, 1]], dtype=torch.long)
x = torch.tensor([[-1], [0], [1]], dtype=torch.float)
data = Data(x=x, edge_index=edge_index)
device = torch.device('cuda')
data = data.to(device)

---

Traceback (most recent call last):
File "", line 1, in
File "C:\Users\Jonas\Anaconda3\envs\schnet\lib\site-packages\torch_geometric\data\data.py", line 246, in to
return self.apply(lambda x: x.to(device), *keys)
File "C:\Users\Jonas\Anaconda3\envs\schnet\lib\site-packages\torch_geometric\data\data.py", line 232, in apply
self[key] = func(item)
File "C:\Users\Jonas\Anaconda3\envs\schnet\lib\site-packages\torch_geometric\data\data.py", line 246, in
return self.apply(lambda x: x.to(device), *keys)
File "C:\Users\Jonas\Anaconda3\envs\schnet\lib\site-packages\torch\cuda_init_.py", line 163, in _lazy_init
torch._C._cuda_init()
RuntimeError: CUDA error: unknown error

[rusty1s]

Related pytorch/pytorch#11989?