Closed 1Kris closed 6 months ago
Thanks for opening this issue @1Kris! All decimal places are now exported. An update has been pushed to Docker Hub.
It looks fabulous now, thanks! You can close this as you wish of course.
I see you didn't skimp on the digits.
frigate_detector_inference_speed_seconds{name="coral1"} 0.007019999999999999
It is interesting that Frigate Other Processes memory utilization jumped from 4.5% to 7.3% coincident with the change. Nothing big, but surprising that it changed at all.
My only other desire may be a Mean line for Inference time, but I am not sure that it is feasible.
That's great. I didn't notice the numbers were not getting read fully until you mentioned it so thanks for that! Much better now.
It is easy to add a mean line in Grafana. Need to add a new query dataset to the inference graph: avg_over_time(frigate_detector_inference_speed_seconds[30m])
. This does a rolling average over 30mins. Or actually I have added a mean line for the inference speed to my dashboard. The updated dashboard is here.
It appears as though the Frigate API is showing inference time as milliseconds with two decimal places, like: "inference_speed": 6.79,
However, the exporter is showing it as seconds, with insufficient decimal places.
frigate_detector_inference_speed_seconds{name="coral1"} 0.007
It would be helpful to add two more decimal places to that value so we can see proper data granularity.
I see similar is true for Device Temperature, GPU Utilization, maybe others.
This is a great tool, thank you for that, and for your support with this.