PLEASE NOTE
The original maintainer Thomas Eizinger no longer maintains photoprism from June 2024. After I took over maintenance, I made some adjustments to it. I have split original photoprism into multiple packages, because except photoprism itself, the other dependency packages that were split almost do not need to be updated. And now it's dependent package photoprism-tensorflow
will not conflict with tensorflow/tensorflow-cuda/tensorflow-opt/tensorflow-opt-cuda.
WHICH PACKAGE SHOULD I CHOOSE
-
x86_64: There are 3 packages to provide
photoprism-tensorflow
. As long as your CPU supports it, the recommended order of selection is: photoprism-tensorflow-avx2-bin > photoprism-tensorflow-avx-bin > photoprism-tensorflow-cpu-bin. Runlscpu | grep avx2
to detect whether your CPU supports AVX2. If the previous command has output, it means CPU supports AVX2. Runlscpu | grep 'avx '
to detect whether your CPU supports AVX. If the previous command has output, it means CPU supports AVX. -
aarch64: There are 2 packages to provide
photoprism-tensorflow
. If your graphics card is NVIDIA Jetson Nano, you can choose photoprism-tensorflow-nvidia-jetson-bin, otherwise choose photoprism-tensorflow-cpu-bin. -
armv7h: You can only choose photoprism-tensorflow-cpu-bin.
UPGRADE FAILS
I have added replaces
var in photoprism-tensorflow-avx2-bin
, photoprism-tensorflow-avx-bin
, photoprism-tensorflow-cpu-bin
and photoprism-tensorflow-nvidia-jetson-bin
, it will be automatically replaced tensorflow115
during the upgrade photoprism
. If it still doesn't work, please uninstall old photoprism
and tensorflow115
first.
WHICH SERVICE TO CHOOSE
/usr/lib/systemd/system/photoprism.service
is run by system userphotoprism
, it can be enabled by runningsudo systemctl enable --now photoprism.service
. It's default config yaml file is/etc/photoprism/defaults.yml
, and it's data is saved in/var/lib/photoprism
./usr/lib/systemd/user/photoprism.service
is run by current normal user, it can be enabled by runningsystemctl enable --user --now photoprism.service
. It's default config yaml file is$XDG_CONFIG_HOME/photoprism/defaults.yml
, but config need to set by youself.
HOW TO CONFIG
- If you run photoprism using system user, please create
/var/lib/photoprism/config/options.yml
for configuaration according to https://docs.photoprism.app/getting-started/config-files. Settings in/var/lib/photoprism/config/options.yml
will override settings in/etc/photoprism/defaults.yml
, so please do not change/etc/photoprism/defaults.yml
. And I have not setbackup
var inPKGBUILD
. - If you run photoprism using current normal user, please set in
$XDG_CONFIG_HOME/photoprism/defaults.yml
or$XDG_CONFIG_HOME/photoprism/options.yml
according to https://docs.photoprism.app/getting-started/config-files.
Pinned Comments
shtrophic commented on 2025-04-25 14:03 (UTC) (edited on 2025-04-25 14:04 (UTC) by shtrophic)
Please note: since upstream switched to tensorflow-2.x, there is no need anymore to vendor a compatible tensorflow version as the
photoprism-tensorflow-*
packages.I also chose to vendor the model weights again, to simplify everything. (effectively un-doing the efforts of the previous maintainer, who chose to create a new package for every zip/tarball. It is definitely an opinionated change, but it should have no effect on end-users).
Therefore, this package now depends on
tensorflow
(which can be either-opt,-cuda
and so on) and conflicts/replacesphoroprism-{nasnet,nsfw,facenet}
.evine commented on 2024-09-04 01:58 (UTC) (edited on 2024-09-24 04:22 (UTC) by evine)
Do not use nodejs 22.7.0/22.8.0 to make.