-
I'm trying to tune my transcription output. I've got a CPU that indicates it has 12 processors. Can I use I searched for discussion threads on "tuning" here on GitHub but didn't see any. If i missed them, please send it on. This is what cpuinfo shows:
|
Beta Was this translation helpful? Give feedback.
Replies: 1 comment
-
The A rule of thumb is that |
Beta Was this translation helpful? Give feedback.
The
-p
argument ofmain
is not related to the CPU processor.These are "Whisper" processors. It's a strategy that we explored to split the input audio into
-p
chunks and process the chunks in parallel. Sometimes it can speed-up the process, but the transcription result at the splits will likely be worse due to partial words and losing context.A rule of thumb is that
-p
*-t
should be less or equal to the number of CPU cores that you have.