yayitazale

You have some config issues, I reccomend you to startup with a minimal config...

You'll need to reinstall the app from scratch to use the new template, since many changes have been made. You can do this using your current folder and database, so you won't lose anything, but there are significant changes regarding ports (port 5000 is no longer used) and other changes that require the use of the updated template.

Sorry for the late reply; I was on vacation. The logs do not show the loaded model; if it starts up without errors, you should be using the model described. You will be able to observe changes in system metrics, inference time, and resource consumption.

Okay, this is useful for a very small percentage of deployments. In any case, if this is really that useful, I think it would be better for you to contribute these features to the main Frigate development branch—not as a separate project—since that’s what a Free Software project created and maintained by a community is for. https://github.com/blakeblackshear/frigate

This is pretty much useless in an UnRAID environment, where the OS already has a built-in system for periodically moving files from the cache pool (designed for use with SSDs) to the disk array (designed for use with HDDs)...

I think the problem resides on the Unraid side, not on the container side.

Did you created a V9 model?

Not sure but I think you don't need this: providers: - CUDAExecutionProvider cuda_graphs: false https://github.com/blakeblackshear/frigate/discussions/23546#discussioncomment-17412679

https://docs.frigate.video/frigate/installation#ports 8971 Authenticated UI and API access without TLS. Reverse proxies should use this port. EDIT: You're right, https://docs.frigate.video/configuration/tls, so I just modified the template so that future new installations will use HTTPS by default for the web UI. Thanks

I cannot help you without more info, can you share a screenshot of the configuration of the container and the configuration file (config.yaml) created during the initial startup?

From your logs, I notice that Frigate is starting successfully, but I don't see any specific detector initialization logs that would indicate whether the YOLOv9 model is loading correctly. The logs show the detector process starting but don't show model compilation or loading details. Check your config file and check if the model file is in the correct place.

What do you see in the logs after restarting the container?

Yes, the GTX 1080 Ti is perfectly suitable because, thanks to its 11 GB of VRAM, it has more than enough memory to run the YOLOv9 models and handle multiple video streams simultaneously and it’s more than capable of keeping model inference times low. In the first step of installing the instance, you must select the NVIDIA branch: Next, as you mentioned, you need to add “--runtime=nvidia” to the additional parameters and configure the NVIDIA entries in the configuration form. As for models, I recommend using YOLO V9: From your Unraid machine's console (not the container one), navigate to the folder where you want to save the model, preferably in the default path (model_cache): cd /mnt/user/appdata/frigate/model_cache/Run the command listed in the documentation at https://docs.frigate.video/configuration/object_detectors/#yolov9-for-other-detectors; if you have many cameras—more than 6, for example—start with a “tiny” (T) size and IMG_SIZE=320 to ensure acceptable inference times. Otherwise, with fewer than 6 cameras, you can try IMG_SIZE=640 and S or M sizes, but keep in mind that the larger the model, the longer the inference time and the higher the power consumption will be. Verify that your ONNX file was created correctly in the specified folder Configure detection with the ONNX detector pointing to your ONNX file (if you placed it in model_cache, you don’t need to change anything) https://docs.frigate.video/configuration/object_detectors#yolo-v3-v4-v7-v9-2 Restart Frigate and check the logs to see if the model has loaded correctly. In addition, if you're also going to perform hardware-accelerated video decoding using the NVIDIA GPU itself, you should use the NVIDIA decoder preset: https://docs.frigate.video/configuration/hardware_acceleration_video#setup-decoder

You should try a minimal setup, without any cameras configured or anything like that... just as described in the documentation. https://docs.frigate.video/guides/getting_started#configuring-frigate

You must read the docs before posting: https://docs.frigate.video/configuration/authentication#onboarding