Context search from text much slower then search of similar photos
Searching images from text from the browser takes ~5 seconds with my current setup, but searching "similar photos" produces results instantaneously. Both the database and thumbnails are unfortunately on HDDs. My Immich instance is running on TrueNAS with no changes to the .env or compose file.
Is this behaviour expected or is there something wrong in my setup?
12 Replies
:wave: Hey @SuperMarcomen,
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. :blue_square: 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.
Successfully submitted, a tag has been added to inform contributors. :white_check_mark:Similar photos search happens entirely in the database. Text search needs a call to the ML service, that's probably what's slow
My server is running on a Xeon E5-2640. Is it normal for the ML service (plus the rest of the search) to need 5 seconds to run?
That depends on a bunch of things and is hard to say outright. But that CPU is getting on in years and I wouldn't be surprised if it's really not well suited to these ML tasks, they're pretty intensive
Is only the first search small, or are consecutive searches (IIRC within 5min) equally slow?
I too thought of this but forgot to test it. Consecutive searches are faster than the first one. I think it has to do with the CLIP model not being in RAM (?). Is there a way to always keep it loaded?
Perfect, thank you! Other than setting
MACHINE_LEARNING_MODEL_TTL to 0, I should also set MACHINE_LEARNING_PRELOAD__CLIP__TEXTUAL, right? Do you recommend to set MACHINE_LEARNING_MODEL_INTRA_OP_THREADS to something higher than 2?Yes and I have no idea lol
I just discovered why sometimes the search felt REALLY slow and took about ~15 seconds: my ML container did not have persistent storage. After every restart/update, the model had to be downloaded again, while I was waiting for the results of the search to apperar 🙃
Then I'm honestly impressed it wasn't much slower lmao
My server has a 1Gbit/s connection and it takes ~16 seconds to download the ~2GB of the ML model. Only the first search after an update would take round about 16 seconds, but this would only happen if I would search something after an update and before uploading any new images (Since that would also cause the model to be downloaded)