Installation error on Ubuntu 18.04 with PyTorch 0.4.1, CUDA 9.1, GCC 7.3.0

[jbohnslav]

I have PyTorch installed already.

python --version: Python 3.7.0
nvcc --version: release 9.1, V9.1.85
gcc --version: 7.3.0

There are pages and pages of errors, but here is the beginning of the output:

running install
running bdist_egg
running egg_info
writing neural_renderer.egg-info/PKG-INFO
writing dependency_links to neural_renderer.egg-info/dependency_links.txt
writing requirements to neural_renderer.egg-info/requires.txt
writing top-level names to neural_renderer.egg-info/top_level.txt
reading manifest file 'neural_renderer.egg-info/SOURCES.txt'
writing manifest file 'neural_renderer.egg-info/SOURCES.txt'
installing library code to build/bdist.linux-x86_64/egg
running install_lib
running build_py
running build_ext
building 'neural_renderer.cuda.load_textures' extension
gcc -pthread -B /home/jim/anaconda3/envs/pytorch/compiler_compat -Wl,--sysroot=/ -Wsign-compare -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -fPIC -I/home/jim/anaconda3/envs/pytorch/lib/python3.7/site-packages/torch/lib/include -I/home/jim/anaconda3/envs/pytorch/lib/python3.7/site-packages/torch/lib/include/TH -I/home/jim/anaconda3/envs/pytorch/lib/python3.7/site-packages/torch/lib/include/THC -I/usr/include -I/home/jim/anaconda3/envs/pytorch/include/python3.7m -c neural_renderer/cuda/load_textures_cuda.cpp -o build/temp.linux-x86_64-3.7/neural_renderer/cuda/load_textures_cuda.o -DTORCH_EXTENSION_NAME=load_textures -D_GLIBCXX_USE_CXX11_ABI=0 -std=c++11
cc1plus: warning: command line option ‘-Wstrict-prototypes’ is valid for C/ObjC but not for C++
In file included from /home/jim/anaconda3/envs/pytorch/lib/python3.7/site-packages/torch/lib/include/pybind11/cast.h:16:0,
                 from /home/jim/anaconda3/envs/pytorch/lib/python3.7/site-packages/torch/lib/include/pybind11/attr.h:13,
                 from /home/jim/anaconda3/envs/pytorch/lib/python3.7/site-packages/torch/lib/include/pybind11/pybind11.h:43,
                 from /home/jim/anaconda3/envs/pytorch/lib/python3.7/site-packages/torch/lib/include/torch/torch.h:5,
                 from neural_renderer/cuda/load_textures_cuda.cpp:1:
/home/jim/anaconda3/envs/pytorch/lib/python3.7/site-packages/torch/lib/include/pybind11/detail/internals.h:82:34: warning: ‘int PyThread_create_key()’ is deprecated [-Wdeprecated-declarations]
     decltype(PyThread_create_key()) tstate = 0; // Usually an int but a long on Cygwin64 with Python 3.x

There are a bunch of warnings that int PyThread_create_key() is deprecated, then...

/usr/bin/nvcc -I/home/jim/anaconda3/envs/pytorch/lib/python3.7/site-packages/torch/lib/include -I/home/jim/anaconda3/envs/pytorch/lib/python3.7/site-packages/torch/lib/include/TH -I/home/jim/anaconda3/envs/pytorch/lib/python3.7/site-packages/torch/lib/include/THC -I/usr/include -I/home/jim/anaconda3/envs/pytorch/include/python3.7m -c neural_renderer/cuda/load_textures_cuda_kernel.cu -o build/temp.linux-x86_64-3.7/neural_renderer/cuda/load_textures_cuda_kernel.o -DTORCH_EXTENSION_NAME=load_textures -D_GLIBCXX_USE_CXX11_ABI=0 --compiler-options '-fPIC' -std=c++11
/usr/include/c++/6/tuple: In instantiation of ‘static constexpr bool std::_TC<<anonymous>, _Elements>::_MoveConstructibleTuple() [with _UElements = {std::tuple<at::Tensor, at::Tensor, at::Tensor>}; bool <anonymous> = true; _Elements = {at::Tensor, at::Tensor, at::Tensor}]’:
/usr/include/c++/6/tuple:626:248:   required by substitution of ‘template<class ... _UElements, typename std::enable_if<(((std::_TC<(sizeof... (_UElements) == 1), at::Tensor, at::Tensor, at::Tensor>::_NotSameTuple<_UElements ...>() && std::_TC<(1ul == sizeof... (_UElements)), at::Tensor, at::Tensor, at::Tensor>::_MoveConstructibleTuple<_UElements ...>()) && std::_TC<(1ul == sizeof... (_UElements)), at::Tensor, at::Tensor, at::Tensor>::_ImplicitlyMoveConvertibleTuple<_UElements ...>()) && (3ul >= 1)), bool>::type <anonymous> > constexpr std::tuple< <template-parameter-1-1> >::tuple(_UElements&& ...) [with _UElements = {std::tuple<at::Tensor, at::Tensor, at::Tensor>}; typename std::enable_if<(((std::_TC<(sizeof... (_UElements) == 1), at::Tensor, at::Tensor, at::Tensor>::_NotSameTuple<_UElements ...>() && std::_TC<(1ul == sizeof... (_UElements)), at::Tensor, at::Tensor, at::Tensor>::_MoveConstructibleTuple<_UElements ...>()) && std::_TC<(1ul == sizeof... (_UElements)), at::Tensor, at::Tensor, at::Tensor>::_ImplicitlyMoveConvertibleTuple<_UElements ...>()) && (3ul >= 1)), bool>::type <anonymous> = <missing>]’
/home/jim/anaconda3/envs/pytorch/lib/python3.7/site-packages/torch/lib/include/ATen/TensorMethods.h:646:36:   required from here
/usr/include/c++/6/tuple:483:67: error: mismatched argument pack lengths while expanding ‘std::is_constructible<_Elements, _UElements&&>’
       return __and_<is_constructible<_Elements, _UElements&&>...>::value;
                                                                   ^~~~~
/usr/include/c++/6/tuple:484:1: error: body of constexpr function ‘static constexpr bool std::_TC<<anonymous>, _Elements>::_MoveConstructibleTuple() [with _UElements = {std::tuple<at::Tensor, at::Tensor, at::Tensor>}; bool <anonymous> = true; _Elements = {at::Tensor, at::Tensor, at::Tensor}]’ not a return-statement
     }
 ^
/usr/include/c++/6/tuple: In instantiation of ‘static constexpr bool std::_TC<<anonymous>, _Elements>::_ImplicitlyMoveConvertibleTuple() [with _UElements = {std::tuple<at::Tensor, at::Tensor, at::Tensor>}; bool <anonymous> = true; _Elements = {at::Tensor, at::Tensor, at::Tensor}]’:
/usr/include/c++/6/tuple:626:362:   required by substitution of ‘template<class ... _UElements, typename std::enable_if<(((std::_TC<(sizeof... (_UElements) == 1), at::Tensor, at::Tensor, at::Tensor>::_NotSameTuple<_UElements ...>() && std::_TC<(1ul == sizeof... (_UElements)), at::Tensor, at::Tensor, at::Tensor>::_MoveConstructibleTuple<_UElements ...>()) && std::_TC<(1ul == sizeof... (_UElements)), at::Tensor, at::Tensor, at::Tensor>::_ImplicitlyMoveConvertibleTuple<_UElements ...>()) && (3ul >= 1)), bool>::type <anonymous> > constexpr std::tuple< <template-parameter-1-1> >::tuple(_UElements&& ...) [with _UElements = {std::tuple<at::Tensor, at::Tensor, at::Tensor>}; typename std::enable_if<(((std::_TC<(sizeof... (_UElements) == 1), at::Tensor, at::Tensor, at::Tensor>::_NotSameTuple<_UElements ...>() && std::_TC<(1ul == sizeof... (_UElements)), at::Tensor, at::Tensor, at::Tensor>::_MoveConstructibleTuple<_UElements ...>()) && std::_TC<(1ul == sizeof... (_UElements)), at::Tensor, at::Tensor, at::Tensor>::_ImplicitlyMoveConvertibleTuple<_UElements ...>()) && (3ul >= 1)), bool>::type <anonymous> = <missing>]’
/home/jim/anaconda3/envs/pytorch/lib/python3.7/site-packages/torch/lib/include/ATen/TensorMethods.h:646:36:   required from here
/usr/include/c++/6/tuple:489:65: error: mismatched argument pack lengths while expanding ‘std::is_convertible<_UElements&&, _Elements>’
       return __and_<is_convertible<_UElements&&, _Elements>...>::value;
                                                                 ^~~~~

There are many pages of similar errors, which I've omitted here for length.

Does anyone know how to fix these errors? I assume it's some versioning problem with CUDA, GCC, etc.

[jbohnslav]

I fixed it. To those who have similar bugs: the problem was my CUDA 9.1 installation. Apparently this is a CUDA bug, fixed in 9.2. To install CUDA 9.2 on a Ubuntu machine, here is one useful guide and here is another