You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi,
this is not an issue ora bug, just a info request.
I would like to measure the Vosk CPU usage at runtime, especially for the transcript calls (after the initial model creation).
More in general can I have some info about Vosk CPU usage? It use multi cores if available, right?
Unfortunately pidstat allows a poll, not less than every second. Anyway I see the process allocates a variable number of CPUs (from 2 to 7) and in percentage I see a medium value of 100%-123% CPU usage.
Is that mean that when I transcript I get a spike of at list 100% of a core?
I'm asking to understand how to manage a suitable server architecture in nodejs, to allow multiple concurrent requests. My final question is if/when I have to manage nodejs thread workers.
Any suggestion welcome
Thanks
Giorgio
The text was updated successfully, but these errors were encountered:
I'm asking to understand how to manage a suitable server architecture in nodejs, to allow multiple concurrent requests. My final question is if/when I have to manage nodejs thread workers.
You can check python server implementation. If you have multiple requests you need to create recognizer per request and run each recognizer in own thread or in a shared thread pool.
Hi,
this is not an issue ora bug, just a info request.
I would like to measure the Vosk CPU usage at runtime, especially for the transcript calls (after the initial model creation).
More in general can I have some info about Vosk CPU usage? It use multi cores if available, right?
BTW, I just realized a simple nodejs wrapper to the Vosk API nodejs binding:
https://github.com/solyarisoftware/voskJs
To transcript an audio file follwing a give model, I run the command line program:
Now, I want to measure the CPU usage of the above process. I tried to use linux
pidstat
command, with these result:Unfortunately pidstat allows a poll, not less than every second. Anyway I see the process allocates a variable number of CPUs (from 2 to 7) and in percentage I see a medium value of 100%-123% CPU usage.
Is that mean that when I transcript I get a spike of at list 100% of a core?
I'm asking to understand how to manage a suitable server architecture in nodejs, to allow multiple concurrent requests. My final question is if/when I have to manage nodejs thread workers.
Any suggestion welcome
Thanks
Giorgio
The text was updated successfully, but these errors were encountered: