I am currently running the tagged latest release 12.02.13 on the RPi 3B. However I see a rather high load caused by the precise-engine?!
Can’t remember that was like this before? Just noticed it because the CLI was so unresponsive and by figuring out what my Pi was doing the precise-engine is on top with top showing a rather high CPU and MEM usuage.
Do others see a similar thing? Is something changed on precise or is it something I did?
Hey there @j1nx - can you post a screenshot of
In my experience, Precise can use 40-50% of CPU
I can however not at my desk atm. But I see similar numbers with a spike to 70 a 80 here and there.
It is only now that I see a laggy cli? Not really sure why and if the load of precise always was that high. For sure not witnessed the laggy cli before though.
Can’t give numbers but I have the impression too that with the latest release my Mark-I reacts slower than before: responses take longer, CLI lags from time to time, ssh-shell has sometime lags too
EDIT: fired up top - even when Mycroft has nothing to do the precise-engine is 50%-70% CPU on my Mark-I
Exactly the behaviour I noticed as well.
Of course even with Mycroft not doing anything, precise is always listening. However I have the feeling the last precise-engine with most likely the latest and greatest training model is a bit more CPU/MEM hungry than before, slowing down the Raspberry Pi 3B. Or at least slowing it down enough to be noticable.
Anyone else? n=2 now.
I am playing with Mycroft on an Raspberry PI 3B+. I am also seeing these high loads. Not sure if this is expected, or did I do something wrong in the configuration. I see a process running one of my logical cpu’s amost full blast. It looks like there are 4 Tensorflow processes running as well. Has anyone been able to recompile Tensorflow with 2 processes as a restriction, and/or is there another better way to decrease this load??
Can anyone tell me if this number of processes looks correct? This is running on a rpi3+. Screencaps below are all from a single htop (just sorted by process name, then scrolled down). I have also almost the same number of processes for the .speech and .skills processes.