Machine Learning and Smart Search Immich on Truenas Scale
Truenas Scale ElectricEel-24.10.1 (latest)
on Minisforum BD790i (64GB memory, AMD Ryzen 9 7945HX, 2TB Nvme OS, 4 16TB HDs in full 2x2 mirror and an RTC 3060)
Tech level.. noob to NAS and linux now but have some 30year old background in using various unix shell stuff which seems to be coming back a bit. Like I used VI to edit something.. shocker.
DETAILS in the attachment
Any help may prevent a suicide. self-harm via alcohol consumption. Successful help may result in a different sort of alcohol consumption.
9 Replies
:wave: Hey @markthorpe,
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.I have read the checklist and provided clips of all the relevant logs, afaik. Immich does function. I can go to 192.168.50.199:30041 and get the web interface. I have used tailscale and duckdns.org to access from the app on mobile.. though, mobile app on my local net cannot get to teh 198. local ipaddress in the app.. it does work in a browser on mobile. I will tackle the whole HTTPS issue later.. but I need smart search and machine leaning to work.
Can you share what actual issue you’re facing?
You say smart search and ML
they do not work... nor does face detection which requires smart search to run first
for example.. I first uploaded just a few images.. some people and dogs.. yet I saw nowhere to identify the people after smartsearch allegedly completed. And searching for "dog" or "dogs" in context failed to search. Search fails. I think the clip of the machine learning log shows that machine learning, and thus smart search, is failing on every image.. therefore there is no database of identified images to search.. that is a guess
Zeus.. let me ask you this.. how do you know smart search is working? How do I know face detection has worked? As far as I can tell from looking this up.. I should be able to go to INFO on an image and identify the face in the image... then, presumably, it will update teh database of photos with that same face. No such interface presents itself to me.
Right now Smart Search thinks it is running, going through 47K images 1 at a time (I read something about someone failing with it set to 2).. but teh machine leaning log shows this:
[01/03/25 20:38:16] DEBUG Could not load ANN shared libraries, using ONNX:
libmali.so: cannot open shared object file: No such file or directory
libmali.so: cannot open shared object file: No such file or directory
I can't view all the logs etc on mobile, but this sounds like your hardware acceleration setup is wrong
I just watched a video that showed off Explore-People... I only see Places under Explore option. Now, I know smart search is still running.. this time around. I have reinstalled immich from scratch and had the same results last time after smart search and everything else "completed". It took many days to complete.. way too many images uploaded before I realized something was wrong. I think the process to machine learn every image is called upon.. and fails each time. So more images just adds time before you realize it did nothing. So I am hoping someone can see what's wrong in the setup and suggest how to fix it.
one thing I was unclear on was the "bypass" setting on install.. it said (non-nvidia gpus).. now.. there is an AMD APU on the mobo but I also have a 3060 that shows up as recognized. I was unsure what to do there. But that might not be the issue. There is also a setting for whether to use Cuda and also a setting somewhere for video for the NVENC.. I aussmed I wanted Cuda and NVENC and. I think I see in the log that it sets execution providers to ['CUDAExecutionProvider', 'CPUExecutionProvider'], after loading the model 'ViT-B-32__openai'
As long as you see
CUDAExecutionProvider, you're goldenQuick FYI:
After numerous attempts with Microsoft Copilot to debug and help, it mostly concluded with Copilot suggesting some kind of package installation for NVIDIA support that TrueNAS ElectricEel prevents. Additionally, I noticed that the logs showing requests to "http://machine-learning:32002/" just weren't happening, which I assume is a setup issue that I couldn’t overcome.
As a result, I switched the GPU selection to an AMD APU and changed the CUDA option in the Immich setup to the default CPU. Now, the machine-learning logs seem to be functioning properly. The AMD chip is performing well; the Minisforum BD790i essentially has a nice laptop motherboard in a mini-ITX form. With 64GB of memory, it has a significant cache, and I’ve assigned 4 of the 8 cores to Immich.
If this setup proves successful, I may opt to ditch the GPU for a proper SATA extension card instead of the odd 6-SATA port NVMe adapter. This would indicate that the issues with my setup were related to the NVIDIA GPU. I even joined the HexOS beta, only to discover that they don’t support GPUs and machine learning yet. I’ll keep an eye on their progress for the future.
yes.. changing the GPU to the APU and using the default CPU-based smartsearch works. Sad... but that failed setup trying to use the RTX 3060 (which is still physically installed) and CUDA for smart search/machine learning took forever to run because it was failing on every image. This was relatively quick for 47K images and I now see PEOPLE as a choice in EXPLORE in Immich. I can now start identifying people and, I expect, other search terms will work also.