Hello,
I am trying to compile from source for a fastai conda environtment… My OS is ubuntu 16.04. My CUDA version is 9.0. cudnn is v7.
When compiling I get the following error
~/Downloads/pytorch/aten/build ~/Downloads/pytorch/aten ~/Downloads/pytorch
- cmake … -DCMAKE_BUILD_TYPE=Release -DNO_CUDA=0 -DNO_NNPACK=0 -DCUDNN_INCLUDE_DIR=/usr/include/ -DCUDNN_LIB_DIR=/usr/lib/x86_64-linux-gnu/ -DCUDNN_LIBRARY=/usr/lib/x86_64-linux-gnu/libcudnn.so.7 -DNO_MKLDNN=1 -DMKLDNN_INCLUDE_DIR= -DMKLDNN_LIB_DIR= -DMKLDNN_LIBRARY= -DATEN_NO_CONTRIB=1 -DCMAKE_INSTALL_PREFIX=/home/user/Downloads/pytorch/torch/lib/tmp_install -DCMAKE_EXPORT_COMPILE_COMMANDS=1 -DCMAKE_C_FLAGS= -DCMAKE_CXX_FLAGS= -DCMAKE_EXE_LINKER_FLAGS= -DCMAKE_SHARED_LINKER_FLAGS= -DWITH_ROCM=0
– Could NOT find CUDA (missing: CUDA_INCLUDE_DIRS) (found suitable version “8.0”, minimum required is “5.5”)
CMake Error: The following variables are used in this project, but they are set to NOTFOUND.
Please set them or make sure they are set and tested correctly in the CMake files:
CUDA_TOOLKIT_INCLUDE
used as include directory in directory /home/user/Downloads/pytorch/aten/build/CMakeFiles/CMakeTmp
used as include directory in directory /home/user/Downloads/pytorch/aten/build/CMakeFiles/CMakeTmp
used as include directory in directory /home/user/Downloads/pytorch/aten/build/CMakeFiles/CMakeTmp
CMake Error at /home/user/Downloads/pytorch/cmake/Modules_CUDA_fix/FindCUDA/select_compute_arch.cmake:85 (try_run):
Failed to configure test project build system.
Call Stack (most recent call first):
/home/user/Downloads/pytorch/cmake/Modules_CUDA_fix/FindCUDA/select_compute_arch.cmake:126 (CUDA_DETECT_INSTALLED_GPUS)
CMakeLists.txt:160 (CUDA_SELECT_NVCC_ARCH_FLAGS)
I had previously installed cuda8.0, but I uninstalled it to do an upgrade. I tested the cuda installation with tensorflow. It seems to be fine.
Thanks,