Package Base Details: python-onnxruntime

Git Clone URL: https://aur.archlinux.org/python-onnxruntime.git (read-only, click to copy)
Submitter: yan12125_2
Maintainer: yan12125 (lilac)
Last Packager: lilac
Votes: 0
Popularity: 0.000000
First Submitted: 2019-07-12 11:06
Last Updated: 2021-10-05 08:38

Pinned Comments

yan12125 commented on 2021-01-26 08:35

Prebuilt packages for python-onnxruntime and dependencies are available in the unofficial [archlinuxcn] repository [1]. python-onnxruntime in that repository is also maintained by me.

[1] https://github.com/archlinuxcn/repo/

Latest Comments

1 2 Next › Last »

nikisalli commented on 2021-10-01 21:24

actually I installed it from pacman but I guess some other manually installed software installed another copy of it there by itself. anyways, it should work just fine with /opt/cuda/bin/nvcc. thanks for pointing out!

yan12125 commented on 2021-10-01 05:11

Thanks for sharing the finding! From the path /usr/local/cuda/bin/nvcc, I guess you installed CUDA from an installer downloaded from NVIDIA? I use Arch's package https://archlinux.org/packages/community/x86_64/cuda/. Maybe only the latter works out-of-box with Arch's cmake.

nikisalli commented on 2021-10-01 02:00

I managed to make it work. for anyone having the same problem, just add "-DCMAKE_CUDA_COMPILER:PATH=/usr/local/cuda/bin/nvcc \" after "_build build-cuda \" in the pkgbuild

yan12125 commented on 2021-09-26 15:22

It's related to CUDA...I'm afraid I cannot give any more useful advice with my quite-limited knowledge about CUDA compilers. Anyway, it looks more like a general CUDA issue instead of a package-specific one, so asking on mailing lists or forums may be better.

nikisalli commented on 2021-09-26 12:51

@yan12125 I managed to fix that error but now it keeps failing due to cmake errors:


CMake Error at /usr/share/cmake-3.21/Modules/CMakeDetermineCompilerId.cmake:706 (message):
  Compiling the CUDA compiler identification source file
  "CMakeCUDACompilerId.cu" failed.

  Compiler: CMAKE_CUDA_COMPILER-NOTFOUND

  Build flags:

  Id flags: -v



  The output was:

  No such file or directory





Call Stack (most recent call first):
  /usr/share/cmake-3.21/Modules/CMakeDetermineCompilerId.cmake:6 (CMAKE_DETERMINE_COMPILER_ID_BUILD)
  /usr/share/cmake-3.21/Modules/CMakeDetermineCompilerId.cmake:48 (__determine_compiler_id_test)
  /usr/share/cmake-3.21/Modules/CMakeDetermineCUDACompiler.cmake:283 (CMAKE_DETERMINE_COMPILER_ID)
  CMakeLists.txt:1533 (enable_language)


-- Configuring incomplete, errors occurred!

yan12125 commented on 2021-09-15 08:27

nikisalli: mind to share your build logs? Missing flake8 only causes a warning. Most likely the error is from something else.

nikisalli commented on 2021-09-14 23:20

build fails due to flake8 errors, please remove all the "include(flake8.cmake)" in the cmakefiles

yan12125 commented on 2021-07-23 12:16

After trying out LTO, I disabled it as it does not work. If someone finds a way to make LTO work, I will remove '!lto' even if it takes almost 34GB of RAM.

cobaltspace commented on 2021-07-23 07:53

using lto causes linking to use > 32GB. The following line may need to be added:

options=('!lto')

yan12125 commented on 2021-01-26 08:39

snarfies: Thanks for further testing. The only cases I know for such issues are insufficient disk space or insufficient memory size, and I don't have time to dig it deeper in near future. Alternatively, prebuilt packages mentioned in the recently pinned message may be useful.