Package Base Details: python-torchvision

Git Clone URL: https://aur.archlinux.org/python-torchvision.git (read-only, click to copy)
Submitter: flacks
Maintainer: hottea (lilac)
Last Packager: lilac
Votes: 17
Popularity: 0.124768
First Submitted: 2018-01-31 10:46
Last Updated: 2020-12-02 13:14

Pinned Comments

hottea commented on 2020-05-02 09:00

You could install python-torchvision or python-torchvision-cuda from ArchLinux CN repo. I have no plan to provide two PKGBUILD for pytorch-torchvision and python-torchvision-cuda. However, pull request is welcome.

Latest Comments

« First ‹ Previous 1 2 3 4 5 Next › Last »

d_fajardo commented on 2020-08-20 12:17

@rhsperry Here is my self-made PKGBUILD. I modified the package name so as not to conflict with this AUR package. I have installed it and tested and it works. Hope it helps but you probably have to adjust the formatting since this box won't keep it.

Maintainer: Danny Fajardo br1ghtch1p@gmail.com

pkgname=python-pytorchvision _pkgname=torchvision pkgver=0.7.0 pkgrel=1 pkgdesc="the torchvision package consists of popular datasets, model architectures, and common image transformations for computer vision" arch=('any') url="https://github.com/pytorch/vision" license=("Apache License 2.0") depends=('python' 'python-pytorch') makedepends=('python-setuptools') conflicts=("${pkgname}" "${pkgname}-git") source=("https://github.com/pytorch/vision/archive/v$pkgver.tar.gz") sha256sums=('fa0a6f44a50451115d1499b3f2aa597e0092a07afce1068750260fa7dd2c85cb')

build() { cd "$srcdir/vision-$pkgver" python setup.py build }

package() { cd "$srcdir/vision-$pkgver" python setup.py install --skip-build --root="$pkgdir/" --optimize=1 }

rhysperry111 commented on 2020-08-13 12:34

@d_fajardo Could you please provide the PKBUILD that you used. I really dont want to have to install 1G of cuda (considering i dont even have an nvidia card)

hottea commented on 2020-08-13 00:54

@nylocx report this issue to upstream, not here. It seems relate to ffmpeg version.

nylocx commented on 2020-08-12 18:53

For me the current build segfaults during a test. Here is the relevant snippet, sorry for german locales.

test/test_utils.py::Tester::test_save_image_single_pixel_file_object PASSED [ 95%] test/test_video_reader.py::TestVideoReader::test_compare_read_video_from_memory_and_file Fatal Python error: Segmentation fault

Thread 0x00007fba87dff700 (most recent call first): <no Python frame>

Current thread 0x00007fbc40d1c740 (most recent call first): File "/home/nylocx/.cache/pikaur/build/python-torchvision/src/vision-0.7.0/test/test_video_reader.py", line 858 in test_compare_read_video_from_memory_and_file File "/usr/lib/python3.8/unittest/case.py", line 633 in _callTestMethod File "/usr/lib/python3.8/unittest/case.py", line 676 in run File "/usr/lib/python3.8/unittest/case.py", line 736 in call

If anyone has an idea what could cause that please let me know. The old Version 0.6.1 was building fine for me.

-- Edit Maybe I should say that I build against community/python-pytorch-opt-cuda 1.6.0-2 as my pytorch version.

d_fajardo commented on 2020-08-11 16:42

Never mind. I just wrote my own PKGBUILD and renamed the package slightly differently so as not to conflict with yours. Better than installing 1G+ of cuda.

rwd2 commented on 2020-08-11 16:42

build fails with "AttributeError: 'module' object has no attribute 'test_container' "

full output at: https://pastebin.com/raw/7NyvcNmQ

d_fajardo commented on 2020-08-09 07:11

Why are you not able to split this package into python-pytorch and python-pytorch-cuda versions? It's particularly important for non-cuda users as installing cuda itself demands a big storage.

trougnouf commented on 2020-07-10 20:51

Can you remove 'python-pytorch-cuda' and cuda from the global dependencies?

elanglois commented on 2020-05-02 18:49

Edit: Sorry, I misunderstood split packages and thought it would be possible to separate build(), check() and makedepends within a single PKGBUILD. It's unfortunate that it doesn't seem possible to have variable reuse with useable building for non-CUDA users.

Original comment:

@hottea You don't have to provide two PKGBUILDs but the code for build() and check() is already split into separate commands for the gpu and cpu versions so why not separate those into different function names so that no large unnecessary packages have to be installed for cpu users? Regardless of whether its possible to download the package elsewhere if this PKGBUILD is claiming to provide python-torchvision (which would be mainly used by non-CUDA users) then I think it makes sense to allow building python-torchvision without CUDA.

hottea commented on 2020-05-02 09:00

You could install python-torchvision or python-torchvision-cuda from ArchLinux CN repo. I have no plan to provide two PKGBUILD for pytorch-torchvision and python-torchvision-cuda. However, pull request is welcome.