-
Notifications
You must be signed in to change notification settings - Fork 818
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
CPU at 100% with Mobilenet-gpu - 8 cores of i7-4770 and GeFORCE RTX 3060 #1283
Comments
It does look like CompreFace is using my GPU but it's still locking up my CPU. This is using the docker-compose.yml from the Master Branch
Is my CPU just inadequate? |
No, I have a 12900k in my homelab server and I noticed for some reason all the sudden compreface is putting my CPU on its knees and maxing every core even though its also putting my GPU to work. SOMETHING broke compreface, but since this project seems to be abandoned, its likely not going to get fixed. Its a shame, this was a cool bit of software too |
Perhaps the GPU is running out of memory? I don't use Mobilenet-gpu but arcface-r100-gpu as I find it a lot more accurate. On 1.1.0 there is a massive memory leak and I found 1.2.0 for the other components unreliable. This is the env setup I am using and find it the most stable on my GTX1650 4GB |
As far as I can tell from the ‘nvidia-smi’ output, that’s not the case.
366MiB / 6144MiB
Edit: now that you mention it and looking at the usage from your screenshots, it's almost like it's not using my GPU RAM at all.
|
Only 78MB of that is Compreface (uwsgi) that's probably just the base application. I am guessing it's loading up, then running out of memory and exiting. I'm assuming you only have 1 camera stream (ffmpeg) I'm using a dual pcie Google Coral so don't have any other Frigate processes, the frigate.embedding_manager showing on mine is the new GenAI stuff from 0.15.b Edit: |
Describe the bug
When I go to deploy the Mobilenet-gpu custom-build with
docker-compose up -d
, CPU usage goes to 100% and locks up the server.To Reproduce
Deploy Mobilenet-gpu
Expected behavior
I expect GPU usage to increase instead of CPU
Screenshots
If applicable, add screenshots to help explain your problem.
Desktop (please complete the following information):
Debian 11 LXC in Proxmox. This LXC concurrently runs Frigate with successful GPU access running TensorRT
Logs
Run those commands and attach result to the ticket:
docker ps
docker-compose logs
Had to cut out some log due to max character limits
Additional context
I'm giving the LXC 8 core of i7-4770 CPU @ 3.40GHz and the GPU is a GeFORCE RTX 3060. Is this enough?
Here's my config file, is there anything I can do to offload efforts to my GPU?
Add any other context about the problem here.
The text was updated successfully, but these errors were encountered: