Package Details: python-dlib-cuda 19.24.6-1

Git Clone URL: https://aur.archlinux.org/python-dlib.git (read-only, click to copy)
Package Base: python-dlib
Description: Dlib is a general purpose cross-platform C++ library designed using contract programming and modern C++ techniques.
Upstream URL: http://www.dlib.net/
Licenses: boost
Conflicts: python-dlib
Provides: python-dlib
Submitter: u1bmoW1r16
Maintainer: petronny (AutoUpdateBot)
Last Packager: petronny
Votes: 16
Popularity: 0.25
First Submitted: 2016-10-27 15:05 (UTC)
Last Updated: 2024-08-10 03:07 (UTC)

Required by (6)

Sources (1)

Pinned Comments

petronny commented on 2024-03-30 13:11 (UTC) (edited on 2024-03-30 13:13 (UTC) by petronny)

Set _build_cpu and _build_cuda to 1 and 0 to build or not build the CPU or CUDA version.

If options=(!lto) has no effect, please check if it's still enabled in /etc/makepkg.conf.

petronny commented on 2023-11-28 04:57 (UTC)

Prebuilt binaries of this package can be found in arch4edu.

Latest Comments

1 2 3 4 5 6 Next › Last »

envolution commented on 2024-11-12 06:41 (UTC)

https://github.com/boltgolt/howdy/issues/928 is the issue I was referring to - I had moved to the aur/python-dlib-git version since and have not had the issue. without having to read that full thread - dlib was broken on systems using numpy 2, and required a patch to support a later pybind - I had assumed there would have been a new dlib release included it, but it seems there hasn't been.

I do understand how PKGBUILD files work, I support a few multi-packages myself - I just find it confusing that it has to be manually edited by the installing user when the package name itself could be used instead. See https://aur.archlinux.org/cgit/aur.git/tree/PKGBUILD?h=python-dlib-cuda-git https://aur.archlinux.org/cgit/aur.git/tree/PKGBUILD?h=python-dlib-git - surely an approach like this would solve the need for manual edits?

petronny commented on 2024-11-12 05:40 (UTC) (edited on 2024-11-12 05:41 (UTC) by petronny)

since you have a -cuda package id as well as one without, what's the purpose of setting _build_cpu and _build_cuda? Shouldn't the installation of the appropriate package suffice?

@envolution These are two packages within the same pkgbase, which means they share the same PKGBUILD. By default, the same PKGBUILD would result in both CPU and CUDA versions being built. However, most users typically require only one version. To address this, I've added the _build_cpu and _build_cuda variables, providing users with the convenience of selectively building the version they need.

the current $pkgversion also has issues with aur/howdy preventing it from working which has been fixed for months in a new upstream release.

Please post the link to the new upstream release.

petronny commented on 2024-11-12 05:26 (UTC)

@halan The PKGBUILD works fine with extra-x86_64-build for both the CPU and CUDA version.

halan commented on 2024-11-08 18:53 (UTC) (edited on 2024-11-08 18:53 (UTC) by halan)

Python 3.12 remove distutils, so I cannot build this package anymore.

==> Starting build()...
/home/kai/.cache/yay/python-dlib/src/dlib-19.24.6/setup.py:234: SyntaxWarning: invalid escape sequence '\('
  major = re.findall("set\(CPACK_PACKAGE_VERSION_MAJOR.*\"(.*)\"", open(cmake_file).read())[0]
/home/kai/.cache/yay/python-dlib/src/dlib-19.24.6/setup.py:235: SyntaxWarning: invalid escape sequence '\('
  minor = re.findall("set\(CPACK_PACKAGE_VERSION_MINOR.*\"(.*)\"", open(cmake_file).read())[0]
/home/kai/.cache/yay/python-dlib/src/dlib-19.24.6/setup.py:236: SyntaxWarning: invalid escape sequence '\('
  patch = re.findall("set\(CPACK_PACKAGE_VERSION_PATCH.*\"(.*)\"", open(cmake_file).read())[0]
Traceback (most recent call last):
  File "/home/kai/.cache/yay/python-dlib/src/dlib-19.24.6/setup.py", line 35, in <module>
    from distutils import log
ModuleNotFoundError: No module named 'distutils'

envolution commented on 2024-10-23 06:05 (UTC)

since you have a -cuda package id as well as one without, what's the purpose of setting _build_cpu and _build_cuda? Shouldn't the installation of the appropriate package suffice?

zyldragoon commented on 2024-09-22 14:54 (UTC)

Due to the recent libjxl 0.11 version update, you need to recompile this package to use it

@petronny maybe considering update the pkgrel number to remind user update

SelfRef commented on 2024-07-21 17:59 (UTC) (edited on 2024-07-21 18:03 (UTC) by SelfRef)

The _build_cpu and _build_cuda could be gotten from some global variables if set, and fallback to package default values. It's inconvenient to set the values every time in PKGBUILD.

Or could be split into two packages, depends on need of CUDA.

brianc commented on 2024-07-01 22:05 (UTC) (edited on 2024-07-01 22:07 (UTC) by brianc)

The CUDA version cannot be built as of cuda 12.5.0-1 that remove the symlinks hardcoded for CUDA compilers in the PKGBUILD.

See: https://gitlab.archlinux.org/archlinux/packaging/packages/cuda/-/issues/8

I got around the issue by removing the hardcoded environment variables from lines 44-48 of the PKGBUILD

8bitrosso commented on 2024-06-18 00:27 (UTC) (edited on 2024-06-18 00:29 (UTC) by 8bitrosso)

Came here from howdy install errors. Had to build and install python-dlib manually. Do yay -G python-dlib and edit PKGBUILD.

Then set _build_cpu to 1 and _build_cuda to 0 (unless you have an nvidia gpu then set to 1). Also edit options to options=(... !lto). Then compile with makepkg -si.

Hope this helps anyone.

VictorGamerLOL commented on 2024-04-03 23:16 (UTC)

For some reason most recent version wants to compile it with CUDA even tho I am on an AMD system. I have python-dlib not python-dlib-cuda installed.

1 2 3 4 5 6 Next › Last »