Search Criteria
Package Details: waifu2x-ncnn-vulkan-git 20210521.r0.g4981ffc-1
Package Actions
Git Clone URL: | https://aur.archlinux.org/waifu2x-ncnn-vulkan-git.git (read-only, click to copy) |
---|---|
Package Base: | waifu2x-ncnn-vulkan-git |
Description: | waifu2x converter ncnn version, runs fast on intel / amd / nvidia GPU with vulkan |
Upstream URL: | https://github.com/nihui/waifu2x-ncnn-vulkan |
Licenses: | MIT |
Conflicts: | waifu2x-ncnn-vulkan, waifu2x-ncnn-vulkan-bin |
Provides: | waifu2x-ncnn-vulkan |
Submitter: | salviati |
Maintainer: | PumpkinCheshire |
Last Packager: | PumpkinCheshire |
Votes: | 7 |
Popularity: | 0.000091 |
First Submitted: | 2019-07-20 00:31 (UTC) |
Last Updated: | 2021-05-28 03:18 (UTC) |
Dependencies (7)
- libwebp (libwebp-git)
- vulkan-icd-loader (vulkan-icd-loader-git)
- cmake (cmake-git) (make)
- git (git-git, git-vfs, git-run-command-patch-git) (make)
- glslang-git (make)
- ncnn (ncnn-git) (make)
- vulkan-headers (vulkan-headers-git) (make)
Required by (2)
- quickmedia-git (requires waifu2x-ncnn-vulkan) (optional)
- video2x (requires waifu2x-ncnn-vulkan) (optional)
Latest Comments
PumpkinCheshire commented on 2020-11-01 12:01 (UTC)
Hi, I think I solved the build dependency problem. Just add -DGLSLANG_TARGET_DIR="/usr/lib/cmake/" to specify the glslangTarget.cmake path. This should work for ncnn and all packages with submodule of ncnn. In this way, you can remove the submodules downloading and use 'ncnn-git' as make dependency.
Thank you for your package. I made dain-ncnn-vulkan and cain-ncnn-vulkan and fixed the load_module issue, hope that will be helpful for you too.
katt commented on 2020-08-24 16:35 (UTC)
If that happens maybe simply enforce the -git variant of the dependency?
Still, at the very least follow the guidelines for submodules so that it can get cached. https://wiki.archlinux.org/index.php/VCS_package_guidelines#Git_submodules
ipha commented on 2020-08-24 16:01 (UTC)
I'll switch to system webp/ncnn on the stable package, but I'm hesitant to use system ncnn on this -git package. It depends on a specific version of ncnn that is sometimes between the latest release and head.
SR-G commented on 2020-08-24 15:53 (UTC) (edited on 2020-08-24 15:54 (UTC) by SR-G)
That same problem with "Whoops! Connection failed. Please try again. If the problem persists, contact our customer support." is back. So 3.7.4 has NOT resolved the problem (as i'm still running that exact 3.7.4 binary) and the problem was related to something else.
In the end, NordVPN is just unusable for me in any way, there are so many different errors over time. - This ARCH package does not seem very active / no real help here (maybe because no one has any knowledge about the potential issues) - NordVPN support is not willing to help in any way (because i'm not using one of the packaged created by NordVPN) - There are 0 detailled logs attached to that NordVPN binary and that could lead to a possible problem / a possible solution
Anyway i'm done with nordvpn in this state, and as soon as my subscription is going to be expired, i'll check another VPN provider.
katt commented on 2020-08-24 12:38 (UTC) (edited on 2020-08-24 14:31 (UTC) by katt)
This comment was originally about submodule guidelines, but after looking at the CMakeLists.txt I found out you can use system ncnn and libwebp.
Could you make it use system ncnn and libwebp instead of submodules? Would make re-compiling faster and done properly instead of re-downloading (and re-compiling) it each time.
ipha commented on 2020-05-22 17:24 (UTC)
Thanks, I updated the pkgbuild for the ncnn submodule.
kescherAUR commented on 2020-05-22 15:39 (UTC)
Had to add
to the prepare function after the cd command.
OrakMoya commented on 2020-05-10 09:28 (UTC)
Now it works, good work
ipha commented on 2020-05-05 13:15 (UTC)
I've added a ncnn package that follows releases instead of git master. Try removing ncnn-git then rebuilding.
PowerMan2206 commented on 2020-05-05 10:08 (UTC)
Like the others said, the output is always a black image (just sayin').
OrakMoya commented on 2020-05-01 23:18 (UTC)
Tried your value and the minimum, 32. Both returned black image. I will start an issue on the github page.
ipha commented on 2020-05-01 23:07 (UTC) (edited on 2020-05-01 23:07 (UTC) by ipha)
There used to be some bugs that caused black images if you ran out of vram, but should have been fixed and throw an error instead.
You could try setting
-t 200
to reduce the tile size and see if it helps. If not you should start an issue upstream: https://github.com/nihui/waifu2x-ncnn-vulkan/issuesOrakMoya commented on 2020-05-01 22:45 (UTC)
Installed latest ncnn-git and recompiled the package, it didn't help.
Not sure how to check the vulkan driver, but I have a GTX 1060 running nvidia driver version 440.82
ipha commented on 2020-05-01 22:30 (UTC)
It works for me with: ncnn-git: 20200413.r10.gbd55ddcf-1 waifu2x-ncnn-vulkan-git: 20200414.r1.g23a679c-1
Try updating ncnn, then recompiling this package. If that doesn't help, what's your hardware and vulkan driver you're using? It may be an upstream issue.
OrakMoya commented on 2020-05-01 22:09 (UTC) (edited on 2020-05-01 22:13 (UTC) by OrakMoya)
No matter what picture I give it the output is always a completely black 4k image. Edit : should probably add that the program reports no errors with the verbose tag
ipha commented on 2020-04-03 20:28 (UTC)
ncnn changed something that breaks this. Upstream will need to fix it, but until then I've pinned ncnn-git at a commit that works. You'll need to rebuild ncnn-git then this should compile.
PowerMan2206 commented on 2020-04-03 19:03 (UTC)
This error shows up when compiling
idk where to mention this except here so ¯_(ツ)_/¯
ipha commented on 2020-01-15 01:14 (UTC)
good catch, fixed.
Iglu47 commented on 2020-01-13 16:38 (UTC)
was 'add upconv_7_photo model' https://github.com/nihui/waifu2x-ncnn-vulkan/commit/b373ce2ec7184959e3b9bea985a90ab1964d1f87
but it's not pick into package after build - pls fix it
ipha commented on 2019-08-21 15:38 (UTC)
Fixed!
lilydjwg commented on 2019-08-21 14:17 (UTC)
Could you update the dependencies?
depends=('vulkan-icd-loader' 'vulkan-driver') makedepends=('git' 'cmake' 'ncnn-git')
ncnn-git is only needed while building, and vulkan-driver must be installed for it to actually work.
katt commented on 2019-08-01 11:57 (UTC)
Models are installed to /usr/share/models instead of /usr/share/waifu2x-ncnn-vulkan