nvcc fatal : Unknown option 'Wl,--no-as-needed,/app/install/libtorch/lib/libtorch.so'

[Edwardmark]

nvcc fatal : Unknown option 'Wl,--no-as-needed,/app/install/libtorch/lib/libtorch.so'

To Reproduce

docker pull nvcr.io/nvidia/tensorrt:19.02-py3
docker run -ti nvcr.io/nvidia/tensorrt:19.02-py3 bash
mkdir /app
cd /app
git clone https://github.com/pytorch/vision.git
wget https://download.pytorch.org/libtorch/cu92/libtorch-cxx11-abi-shared-with-deps-1.4.0%2Bcu92.zip
unzip libtorch to /app/
Steps to reproduce the behavior:

1. cd /app/vision/
2. mkdir build
3. change line 3 "set(CMAKE_CXX_STANDARD 14)" to

add_definitions(-std=c++11)
set(CMAKE_CXX_STANDARD 11)

4. cd build
5. cmake -DWITH_CUDA=on -DCMAKE_PREFIX_PATH=/app/libtorch/ ..
6. make

Expected behavior

build successfully.

Environment

Collecting environment information...
PyTorch version: 1.2.0+cu92
Is debug build: No
CUDA used to build PyTorch: 9.2.148

OS: Ubuntu 16.04.5 LTS
GCC version: (Ubuntu 5.4.0-6ubuntu1~16.04.11) 5.4.0 20160609
CMake version: version 3.12.1

Python version: 3.5
Is CUDA available: Yes
CUDA runtime version: 10.0.130
GPU models and configuration:
GPU 0: TITAN X (Pascal)
GPU 1: TITAN X (Pascal)

Nvidia driver version: 430.26
cuDNN version: /usr/lib/x86_64-linux-gnu/libcudnn.so.7.4.2

Versions of relevant libraries:
[pip] Could not collect
[conda] Could not collect

output

cmake output

cd build && cmake -DWITH_CUDA=on -DCMAKE_PREFIX_PATH=/app/libtorch/ ..
-- The C compiler identification is GNU 5.4.0
-- The CXX compiler identification is GNU 5.4.0
-- Check for working C compiler: /usr/bin/cc
-- Check for working C compiler: /usr/bin/cc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Check for working CXX compiler: /usr/bin/c++
-- Check for working CXX compiler: /usr/bin/c++ -- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- The CUDA compiler identification is NVIDIA 10.0.130
-- Check for working CUDA compiler: /usr/local/cuda/bin/nvcc
-- Check for working CUDA compiler: /usr/local/cuda/bin/nvcc -- works
-- Detecting CUDA compiler ABI info
-- Detecting CUDA compiler ABI info - done
-- Found Python3: /usr/lib/python3.5/config-3.5m-x86_64-linux-gnu/libpython3.5m.so (found version "3.5.2") found components: Development
-- Looking for pthread.h
-- Looking for pthread.h - found
-- Looking for pthread_create
-- Looking for pthread_create - not found
-- Looking for pthread_create in pthreads
-- Looking for pthread_create in pthreads - not found
-- Looking for pthread_create in pthread
-- Looking for pthread_create in pthread - found
-- Found Threads: TRUE
-- Found CUDA: /usr/local/cuda (found version "10.0")
-- Caffe2: CUDA detected: 10.0
-- Caffe2: CUDA nvcc is: /usr/local/cuda/bin/nvcc
-- Caffe2: CUDA toolkit directory: /usr/local/cuda
-- Caffe2: Header version is: 10.0
-- Found CUDNN: /usr/lib/x86_64-linux-gnu/libcudnn.so
-- Found cuDNN: v7.4.2 (include: /usr/include, library: /usr/lib/x86_64-linux-gnu/libcudnn.so)
-- Autodetected CUDA architecture(s): 6.1 6.1
-- Added CUDA NVCC flags for: -gencode;arch=compute_61,code=sm_61
-- Found torch: /app/libtorch/lib/libtorch.so
-- Configuring done
-- Generating done
-- Build files have been written to: /app/vision/build

make ouptuts

make Scanning dependencies of target torchvision [ 4%] Building CXX object CMakeFiles/torchvision.dir/torchvision/csrc/models/alexnet.cpp.o [ 8%] Building CXX object CMakeFiles/torchvision.dir/torchvision/csrc/models/densenet.cpp.o [ 12%] Building CXX object CMakeFiles/torchvision.dir/torchvision/csrc/models/googlenet.cpp.o [ 16%] Building CXX object CMakeFiles/torchvision.dir/torchvision/csrc/models/inception.cpp.o [ 20%] Building CXX object CMakeFiles/torchvision.dir/torchvision/csrc/models/mnasnet.cpp.o [ 24%] Building CXX object CMakeFiles/torchvision.dir/torchvision/csrc/models/mobilenet.cpp.o [ 28%] Building CXX object CMakeFiles/torchvision.dir/torchvision/csrc/models/resnet.cpp.o [ 32%] Building CXX object CMakeFiles/torchvision.dir/torchvision/csrc/models/shufflenetv2.cpp.o [ 36%] Building CXX object CMakeFiles/torchvision.dir/torchvision/csrc/models/squeezenet.cpp.o [ 40%] Building CXX object CMakeFiles/torchvision.dir/torchvision/csrc/models/vgg.cpp.o [ 44%] Building CXX object CMakeFiles/torchvision.dir/torchvision/csrc/cpu/DeformConv_cpu.cpp.o [ 48%] Building CXX object CMakeFiles/torchvision.dir/torchvision/csrc/cpu/PSROIAlign_cpu.cpp.o [ 52%] Building CXX object CMakeFiles/torchvision.dir/torchvision/csrc/cpu/PSROIPool_cpu.cpp.o [ 56%] Building CXX object CMakeFiles/torchvision.dir/torchvision/csrc/cpu/ROIAlign_cpu.cpp.o [ 60%] Building CXX object CMakeFiles/torchvision.dir/torchvision/csrc/cpu/ROIPool_cpu.cpp.o [ 64%] Building CXX object CMakeFiles/torchvision.dir/torchvision/csrc/cpu/nms_cpu.cpp.o [ 68%] Building CUDA object CMakeFiles/torchvision.dir/torchvision/csrc/cuda/DeformConv_cuda.cu.o [ 72%] Building CUDA object CMakeFiles/torchvision.dir/torchvision/csrc/cuda/PSROIAlign_cuda.cu.o [ 76%] Building CUDA object CMakeFiles/torchvision.dir/torchvision/csrc/cuda/PSROIPool_cuda.cu.o [ 80%] Building CUDA object CMakeFiles/torchvision.dir/torchvision/csrc/cuda/ROIAlign_cuda.cu.o [ 84%] Building CUDA object CMakeFiles/torchvision.dir/torchvision/csrc/cuda/ROIPool_cuda.cu.o [ 88%] Building CUDA object CMakeFiles/torchvision.dir/torchvision/csrc/cuda/nms_cuda.cu.o [ 92%] Building CXX object CMakeFiles/torchvision.dir/torchvision/csrc/vision.cpp.o [ 96%] Linking CUDA device code CMakeFiles/torchvision.dir/cmake_device_link.o nvcc fatal : Unknown option 'Wl,--no-as-needed,/app/install/libtorch/lib/libtorch.so' CMakeFiles/torchvision.dir/build.make:428: recipe for target 'CMakeFiles/torchvision.dir/cmake_device_link.o' failed make[2]: *** [CMakeFiles/torchvision.dir/cmake_device_link.o] Error 1 CMakeFiles/Makefile2:72: recipe for target 'CMakeFiles/torchvision.dir/all' failed make[1]: *** [CMakeFiles/torchvision.dir/all] Error 2 Makefile:129: recipe for target 'all' failed make: *** [all] Error 2

Please help me out? Why is that?

[fmassa]

@bmanga could you have a look?

[bmanga]

Hi @Edwardmark, what is the version of cmake being used there?

[Edwardmark]

@bmanga cmake 3.12.1

[Edwardmark]

@bmanga
Hi, bmanga,
Did you reproduce the error? And could you tell me how to fix it? Thanks.
Bests,
Edward.

[Edwardmark]

@bmanga Upgrading CMake to 3.13.2 works. Closed the issure.