AMD 8845HS w/ Radeon 780m -- Machine Learning not working out of the box.
I'm using latest Immich under TrueNAS 25.04.2.1 and had to do quite a lot to get machine learning working..
first, /dev/kfd was not passed into the container. was able to fix that on TrueNAS shell with checking config:
midclt call app.config immich
gave following output:
and used following input to pass /dev/kfd to the container:
midclt call -job app.update immich '{"values": {"resources": {"gpus": {"kfd_device_exists": true, "nvidia_gpu_selection": {}, "use_all_gpus": true}, "limits": { "cpus": 12, "memory": 8192 }}}}'
Second the rocm hip architecture files gfx1103 is missing, making it work with radeon 780m (RDNA 3).
workaround was to run immich as root (0) instead of apps (568) and run:
cp /opt/rocm/lib/rocblas/library/TensileLibrary_lazy_gfx1102.dat /opt/rocm/lib/rocblas/library/TensileLibrary_lazy_gfx1103.dat
since i was not able to SUDO and copy the file.
please add the gfx1103 file like with the patch, so immich can revert to non-privilaged container.
machine-learning/patches/0002-target-gfx900-gfx1102.patch9 Replies
:wave: Hey @GentleS,
Thanks for reaching out to us. Please carefully read this message and follow the recommended actions. This will help us be more effective in our support effort and leave more time for building Immich :immich:.
References
- Container Logs:
docker compose logs docs
- Container Status: docker ps -a docs
- Reverse Proxy: https://immich.app/docs/administration/reverse-proxy
- Code Formatting https://support.discord.com/hc/en-us/articles/210298617-Markdown-Text-101-Chat-Formatting-Bold-Italic-Underline#h_01GY0DAKGXDEHE263BCAYEGFJA
Checklist
I have...
1. :ballot_box_with_check: verified I'm on the latest release(note that mobile app releases may take some time).
2. :ballot_box_with_check: read applicable release notes.
3. :ballot_box_with_check: reviewed the FAQs for known issues.
4. :ballot_box_with_check: reviewed Github for known issues.
5. :ballot_box_with_check: tried accessing Immich via local ip (without a custom reverse proxy).
6. :ballot_box_with_check: uploaded the relevant information (see below).
7. :ballot_box_with_check: tried an incognito window, disabled extensions, cleared mobile app cache, logged out and back in, different browsers, etc. as applicable
(an item can be marked as "complete" by reacting with the appropriate number)
Information
In order to be able to effectively help you, we need you to provide clear information to show what the problem is. The exact details needed vary per case, but here is a list of things to consider:
- Your docker-compose.yml and .env files.
- Logs from all the containers and their status (see above).
- All the troubleshooting steps you've tried so far.
- Any recent changes you've made to Immich or your system.
- Details about your system (both software/OS and hardware).
- Details about your storage (filesystems, type of disks, output of commands like fdisk -l and df -h).
- The version of the Immich server, mobile app, and other relevant pieces.
- Any other information that you think might be relevant.
Please paste files and logs with proper code formatting, and especially avoid blurry screenshots.
Without the right information we can't work out what the problem is. Help us help you ;)
If this ticket can be closed you can use the /close command, and re-open it later if needed.gfx1102 is not the same as gfx1103 and you can't just copy the library like that
I never meant that they are the same. but copying and renaming gfx1102 to gfx1103 made it work. my request is to add the original gfx1103, that is missing.
I can't really understand or explain why it's only working that way.. but if I use the original file with HSA_OVERRIDE_GFX_VERSION=11.0.2 than the machinelearning will exit with the message GPU Hang.
but using the same file as mentioned, renaming it to gfx1103.dat and adding HSA_OVERRIDE_GFX_VERSION=11.0.3 instead made it start working.
Well I'd guess it's because you have a gfx1103 card...? 😛
yes^^ thats the point, but then why can't i just simply use 1102 instead, when its just a naming thing, the content must be the same ^^'
nevertheless, it would be great if the gfx1103 was added. currently boards like mine (Miniroute N7) with the Ryzen 7 8845HS seem to become popular for nas builts and feature the Radeon 780m chip..
It's not just a naming thing, it's a similar but decidedly different architecture. You will probably run into some hard crashes trying to do it like this
we'll see, for no its working.. unfortunately as a full privilaged container :S
and i have to copy and rename the file for every new start :/
you were right 😢 it did something but not recognizing faces.. or even learn intelligent search.
so my guess is, that i'll habe to wait until gfx1103 is implemented..