mapillary/inplace_abn

Can not install in GTX1080Ti, CUDA=10.1, PyTorch 1.4!

Closed this issue · 2 comments

linhduongtuan commented

Dear Mapillary,
First of all, your work is valuable and I trying to apply it. However, I can not install Inplace-ABN, following the instruction pip install git+https://github.com/mapillary/inplace_abn.git@v1.0.11 or the alternative one. My configuration is Ubuntu 18.04, Pytorch 1.4, GTX 1080Ti with CUDA ver. 10.1.
The message:

ERROR: Command errored out with exit status 1:
command: /home/linh/.conda/envs/CV/bin/python -u -c 'import sys, setuptools, tokenize; sys.argv[0] = '"'"'/tmp/pip-req-build-g6ls7n0j/setup.py'"'"'; file='"'"'/tmp/pip-req-build-g6ls7n0j/setup.py'"'"';f=getattr(tokenize, '"'"'open'"'"', open)(file);code=f.read().replace('"'"'\r\n'"'"', '"'"'\n'"'"');f.close();exec(compile(code, file, '"'"'exec'"'"'))' install --record /tmp/pip-record-g6p0jt8n/install-record.txt --single-version-externally-managed --compile
cwd: /tmp/pip-req-build-g6ls7n0j/
Complete output (20 lines):
running install
running build
running build_py
creating build
creating build/lib.linux-x86_64-3.6
creating build/lib.linux-x86_64-3.6/inplace_abn
copying inplace_abn/version.py -> build/lib.linux-x86_64-3.6/inplace_abn
copying inplace_abn/group.py -> build/lib.linux-x86_64-3.6/inplace_abn
copying inplace_abn/abn.py -> build/lib.linux-x86_64-3.6/inplace_abn
copying inplace_abn/init.py -> build/lib.linux-x86_64-3.6/inplace_abn
copying inplace_abn/functions.py -> build/lib.linux-x86_64-3.6/inplace_abn
running build_ext
building 'inplace_abn.backend' extension
creating build/temp.linux-x86_64-3.6
creating build/temp.linux-x86_64-3.6/src
/home/linh/.conda/envs/CV/bin/x86_64-conda_cos6-linux-gnu-cc -DNDEBUG -fwrapv -O2 -Wall -Wstrict-prototypes -march=nocona -mtune=haswell -ftree-vectorize -fPIC -fstack-protector-strong -fno-plt -O2 -ffunction-sections -pipe -DNDEBUG -D_FORTIFY_SOURCE=2 -O2 -fPIC -DWITH_CUDA=1 -Iinclude/ -I/home/linh/.conda/envs/CV/lib/python3.6/site-packages/torch/include -I/home/linh/.conda/envs/CV/lib/python3.6/site-packages/torch/include/torch/csrc/api/include -I/home/linh/.conda/envs/CV/lib/python3.6/site-packages/torch/include/TH -I/home/linh/.conda/envs/CV/lib/python3.6/site-packages/torch/include/THC -I/usr/local/cuda/include -I/home/linh/.conda/envs/CV/include/python3.6m -c src/inplace_abn.cpp -o build/temp.linux-x86_64-3.6/src/inplace_abn.o -O3 -DTORCH_API_INCLUDE_EXTENSION_H -DTORCH_EXTENSION_NAME=backend -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++
/usr/local/cuda/bin/nvcc -DWITH_CUDA=1 -Iinclude/ -I/home/linh/.conda/envs/CV/lib/python3.6/site-packages/torch/include -I/home/linh/.conda/envs/CV/lib/python3.6/site-packages/torch/include/torch/csrc/api/include -I/home/linh/.conda/envs/CV/lib/python3.6/site-packages/torch/include/TH -I/home/linh/.conda/envs/CV/lib/python3.6/site-packages/torch/include/THC -I/usr/local/cuda/include -I/home/linh/.conda/envs/CV/include/python3.6m -c src/inplace_abn_cuda.cu -o build/temp.linux-x86_64-3.6/src/inplace_abn_cuda.o -D__CUDA_NO_HALF_OPERATORS -D__CUDA_NO_HALF_CONVERSIONS_ -D__CUDA_NO_HALF2_OPERATORS__ --expt-relaxed-constexpr --compiler-options '-fPIC' -DTORCH_API_INCLUDE_EXTENSION_H -DTORCH_EXTENSION_NAME=_backend -D_GLIBCXX_USE_CXX11_ABI=0 -gencode=arch=compute_61,code=sm_61 -std=c++11
unable to execute '/usr/local/cuda/bin/nvcc': No such file or directory
error: command '/usr/local/cuda/bin/nvcc' failed with exit status 1
----------------------------------------
ERROR: Command errored out with exit status 1: /home/linh/.conda/envs/CV/bin/python -u -c 'import sys, setuptools, tokenize; sys.argv[0] = '"'"'/tmp/pip-req-build-g6ls7n0j/setup.py'"'"'; file='"'"'/tmp/pip-req-build-g6ls7n0j/setup.py'"'"';f=getattr(tokenize, '"'"'open'"'"', open)(file);code=f.read().replace('"'"'\r\n'"'"', '"'"'\n'"'"');f.close();exec(compile(code, file, '"'"'exec'"'"'))' install --record /tmp/pip-record-g6p0jt8n/install-record.txt --single-version-externally-managed --compile Check the logs for full command output.

Do you have any suggestion?
Best regards
Linh

ducksoup commented

This error indicates that the setup script is not able to automatically locate the position of the CUDA compiler: unable to execute '/usr/local/cuda/bin/nvcc': No such file or directory.
Please verify the following:

  • Does /usr/local/cuda/bin/nvcc exist on your machine?
  • Can you compile any other C++/CUDA Pyotorch extension, e.g. by following this tutorial: https://pytorch.org/tutorials/advanced/cpp_extension.html ?
  • In case you installed CUDA in a different location, can you please set the env variable CUDA_HOME={path to the actual cuda root directory on your machine}, and try again with the InPlace-ABN installation?
linhduongtuan commented

Hi Lorenzo,
I fixed it already. Now I'll try your great work with TResNet https://github.com/mrT23/TResNet.
Many thank for your support.
Linh