Precise TFLite benchmarking

Looking great!

I’d be interested to see if you notice any difference between the resource usage of the OVOS and Mycroft Precise TFlite plugin if you have time too?

I was not aware about this one :smiley:
I’ll run some tests to see what the resources consumption are.

The CPU consumption is much higher than with precise-engine :scream: and compare to the OVOS plugin it consumes more in every aspects (except the disk IOPS).

I got nothing wrong in the logs and there is no extra activity. After pip install the container has been restarted to clear the possible artifacts from previous precise-engine-lite load.

CPU usage for mycroft_voice container (40 minutes sample):

Memory usage for mycroft_voice container (40 minutes sample):

Network usage for mycroft_voice container (40 minutes sample):

1 Like

Hey thanks for doing that - I split this out as I didn’t want to hijack your thread on Docker with lots of chatter about TFlite benchmarking.

I was pretty surprised since the plugin only runs a TFlite runtime, no extra precise engine. However it looks like there’s some kind of feedback loop happening that is reasonably quickly taking CPU usage (and memory to a lesser degree) through the roof.

1 Like

For anyone else reading along this was split from:

which contains benchmarking for the OVOS-precise-lite plugin.