Can't build on ubuntu 18

[urimerhav]

I followed the install procedure as outlined in the docs with python 3.7. On mac it just works. On a ubuntu 18 machine with gpu (CUDA Version: 11.0), it fails with the following logs. Are the "type traits" errors the reason for failure, or something else? Been bashing my head at this for hours now and not really making any progress.

-- pybind11 v2.6.0 dev
-- Using pybind11: (version "2.6.0" dev)
-- Build with CUDA support
INFO Building without TensorFlow support (not found)
-- Configuring done
-- Generating done
-- Build files have been written to: /home/ubuntu/Code/diffvg/build/temp.linux-x86_64-3.7
[ 14%] Building NVCC (Device) object CMakeFiles/diffvg.dir/diffvg_generated_scene.cpp.o
[ 28%] Building NVCC (Device) object CMakeFiles/diffvg.dir/diffvg_generated_diffvg.cpp.o
/home/ubuntu/anaconda3/envs/pytorch_latest_p37/x86_64-conda-linux-gnu/include/c++/9.3.0/type_traits(148): error: "__or_v" is not a function or static data member

/home/ubuntu/anaconda3/envs/pytorch_latest_p37/x86_64-conda-linux-gnu/include/c++/9.3.0/type_traits(148): error: "constexpr" is not valid here

/home/ubuntu/anaconda3/envs/pytorch_latest_p37/x86_64-conda-linux-gnu/include/c++/9.3.0/type_traits(150): error: "__and_v" is not a function or static data member
... more of these warnings...
/home/ubuntu/anaconda3/envs/pytorch_latest_p37/x86_64-conda-linux-gnu/include/c++/9.3.0/type_traits(2885): error: "is_scalar_v" is not a function or static data member
    self.run_command('build_ext')
  File "/home/ubuntu/anaconda3/envs/pytorch_latest_p37/lib/python3.7/distutils/cmd.py", line 313, in run_command
    self.distribution.run_command(command)
  File "/home/ubuntu/anaconda3/envs/pytorch_latest_p37/lib/python3.7/distutils/dist.py", line 985, in run_command
    cmd_obj.run()
  File "setup.py", line 31, in run
    super().run()
  File "/home/ubuntu/anaconda3/envs/pytorch_latest_p37/lib/python3.7/site-packages/setuptools/command/build_ext.py", line 79, in run
    _build_ext.run(self)
  File "/home/ubuntu/anaconda3/envs/pytorch_latest_p37/lib/python3.7/site-packages/Cython/Distutils/old_build_ext.py", line 186, in run
    _build_ext.build_ext.run(self)
  File "/home/ubuntu/anaconda3/envs/pytorch_latest_p37/lib/python3.7/distutils/command/build_ext.py", line 340, in run
    self.build_extensions()
  File "/home/ubuntu/anaconda3/envs/pytorch_latest_p37/lib/python3.7/site-packages/Cython/Distutils/old_build_ext.py", line 195, in build_extensions
    _build_ext.build_ext.build_extensions(self)
  File "/home/ubuntu/anaconda3/envs/pytorch_latest_p37/lib/python3.7/distutils/command/build_ext.py", line 449, in build_extensions
    self._build_extensions_serial()
  File "/home/ubuntu/anaconda3/envs/pytorch_latest_p37/lib/python3.7/distutils/command/build_ext.py", line 474, in _build_extensions_serial
    self.build_extension(ext)
  File "setup.py", line 65, in build_extension
    subprocess.check_call(['cmake', '--build', '.'] + build_args, cwd=self.build_temp)
  File "/home/ubuntu/anaconda3/envs/pytorch_latest_p37/lib/python3.7/subprocess.py", line 363, in check_call
    raise CalledProcessError(retcode, cmd)
subprocess.CalledProcessError: Command '['cmake', '--build', '.', '--config', 'Release', '--', '-j8']' returned non-zero exit status 2.

[urimerhav]

@BachiLi Trying to follow up on this. If you could point out any Amazon machine image or setup instructions for a GPU machine that works (what CUDA version, hardware, pytorch version etc) - I'd really appreciate it!

[XiaoleiDi]

I have same problem,
cuda version 10.1
torch version 1.7.1
hardware amd R7-5800h RTX 3060
i tried to run CLIPasso at my computer with ubuntu 22.04
sorry about the grammar my English is not good

running install_lib
running build_py
running build_ext
-- pybind11 v2.6.0 dev
CMake Warning (dev) at /home/dxl/anaconda3/envs/CLP/share/cmake-3.22/Modules/CMakeDependentOption.cmake:84 (message):
  Policy CMP0127 is not set: cmake_dependent_option() supports full Condition
  Syntax.  Run "cmake --help-policy CMP0127" for policy details.  Use the
...
    subprocess.check_call(['cmake', '--build', '.'] + build_args, cwd=self.build_temp)
  File "/home/dxl/anaconda3/envs/CLP/lib/python3.7/subprocess.py", line 363, in check_call
    raise CalledProcessError(retcode, cmd)
subprocess.CalledProcessError: Command '['cmake', '--build', '.', '--config', 'Release', '--', '-j8']' returned non-zero exit status 2.

[zhulifengsheng]

I have same problem too.
raise CalledProcessError(retcode, cmd) subprocess.CalledProcessError: Command '['cmake', '--build', '.', '--config', 'Release', '--', '-j8']' returned non-zero exit status 2.

[fxcdl]

I solved this problem on ubuntu20.04 cuda11.3 torch1.12.1 cmake3.22.1(it's important)

[yudianzheng]

For me,
conda create -n diffvg python=3.7.5 tensorflow=1.15
, and the same as ubuntu20.04 cuda11.3 torch1.12.1.

[LittleQBerry]

same problem

[yonhar]

I solved the problem by downgrading gcc. It turns out that cmake build doesn't work well with versions later than 8.
following this link I solved it by:
sudo apt remove gcc
sudo apt-get install gcc-7 g++-7 -y
sudo ln -s /usr/bin/gcc-7 /usr/bin/gcc
sudo ln -s /usr/bin/g++-7 /usr/bin/g++
sudo ln -s /usr/bin/gcc-7 /usr/bin/cc
sudo ln -s /usr/bin/g++-7 /usr/bin/c++

[AnCoSONG]

I solved this problem on ubuntu20.04 cuda11.3 torch1.12.1 cmake3.22.1(it's important)

My environment is a little different from yours(cuda11.6). In my case, I could run cmake --build ... script successfully in shell but the subprocess.check_call always return with 2. So I changed it to subprocess.call. Everything looks good so far.

[fibarrola]

I solved the problem by downgrading gcc. It turns out that cmake build doesn't work well with versions later than 8. following this link I solved it by: sudo apt remove gcc sudo apt-get install gcc-7 g++-7 -y sudo ln -s /usr/bin/gcc-7 /usr/bin/gcc sudo ln -s /usr/bin/g++-7 /usr/bin/g++ sudo ln -s /usr/bin/gcc-7 /usr/bin/cc sudo ln -s /usr/bin/g++-7 /usr/bin/c++

This one allowed me to run in Ubuntu 22.04 by doing

sudo nano /etc/apt/sources.list

add the line: deb [arch=amd64] http://archive.ubuntu.com/ubuntu focal main universe

sudo apt update

and then doing your thing

[yuchen-xiyue]

I am using Ubuntu20.02(WSL2) + cuda12.2 and meet the same problem:
Command '['cmake', '--build', '.', '--config', 'Release', '--', '-j8']' returned non-zero exit status 2.
But I found there is no problem with cmake command.
So I just simply tried:
cmake --build . --config Debug -- -j8
Comment Line 65 in setup.py and then
python setup.py install
conda install -c conda-forge gcc=12.1.0
I successfully run the demos.

[Yi-Lynn]

I have same problem, cuda version 10.1 torch version 1.7.1 hardware amd R7-5800h RTX 3060 i tried to run CLIPasso at my computer with ubuntu 22.04 sorry about the grammar my English is not good

running install_lib
running build_py
running build_ext
-- pybind11 v2.6.0 dev
CMake Warning (dev) at /home/dxl/anaconda3/envs/CLP/share/cmake-3.22/Modules/CMakeDependentOption.cmake:84 (message):
  Policy CMP0127 is not set: cmake_dependent_option() supports full Condition
  Syntax.  Run "cmake --help-policy CMP0127" for policy details.  Use the
...
    subprocess.check_call(['cmake', '--build', '.'] + build_args, cwd=self.build_temp)
  File "/home/dxl/anaconda3/envs/CLP/lib/python3.7/subprocess.py", line 363, in check_call
    raise CalledProcessError(retcode, cmd)
subprocess.CalledProcessError: Command '['cmake', '--build', '.', '--config', 'Release', '--', '-j8']' returned non-zero exit status 2.

Hi, I am also running the CLIPasso project on Ubuntu 18.04 with cuda version 11.3 and torch version 1.11. Have you solved this problem? Thanks!

[Yi-Lynn]

I am also running the CLIPasso project on Ubuntu 18.04 with cuda version 11.3 and torch version 1.11

Just solved this problem using the solution from @yuchen-xiyue. Thanks a lot!

[CrystalLiu1]

I solved this problem on ubuntu20.04 cuda11.3 torch1.12.1 cmake3.22.1(it's important)

My environment is a little different from yours(cuda11.6). In my case, I could run cmake --build ... script successfully in shell but the subprocess.check_call always return with 2. So I changed it to subprocess.call. Everything looks good so far.

Thank you very much, this is very useful to me.

[yuchen-xiyue]

I am also running the CLIPasso project on Ubuntu 18.04 with cuda version 11.3 and torch version 1.11

Just solved this problem using the solution from @yuchen-xiyue. Thanks a lot!

Moreover, I just successfully installed Diffvg in a newer Python version (3.11) on WSL2. It took me long time to troubleshoot everything and I believe it is worthy to let everybody know the following things:

1. Make sure your GCC and CUDA version strictly match each other: gcc-10 with cuda-11.5; gcc-8 with cuda-10.2
2. Change everything in CMakeLists.txt, if(DIFFVG_CUDA) ... endif() block as like:

    find_package(CUDA REQUIRED)

    set(CMAKE_CUDA_STANDARD 17)
    set(CMAKE_CUDA_STANDARD_REQUIRED ON)

    set(CUDA_HOST_COMPILER /usr/bin/gcc-10)
    set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -ccbin=${CUDA_HOST_COMPILER}")

    message(STATUS "CUDA NVCC Flags: ${CUDA_NVCC_FLAGS}")
else()
    message(STATUS "Build without CUDA support")
    find_package(Thrust REQUIRED)

It seems way too different from the origin one but it works.
3. After installation, if this error message raised "ImportError: libstdc++.so.6: version GLIBCXX_3.4.30' not found" when you try to import pydiffvg, then paste your system dlib in find /usr -name "libstdc++.so.6"` to where your env's (like "/home/username/anaconda3/envs/diffvg310/lib/libstdc++.so.6*").

At the end, Docker could always be the plan Z.