Noob Coder - Struggling to pass args at transcode Call

[miles-du]

I have a working python script using Whisper to do some voice transcribing. I am doing some refinements.
I want to use:
--threads THREADS number of threads used by torch for CPU inference; supercedes MKL_NUM_THREADS/OMP_NUM_THREADS
(default: 0)

To set the number of threads used to 4 when the script calls Whisper. I have tried all sorts of approaches and syntax trying to make this work with no luck. My last attempt was to set it up first and append it to the transcribe request:

%Initialize Whisper models
tiny_model = whisper.load_model('tiny')
base_model = whisper.load_model('base')

%Create a DecodingOptions object with the threads property set to 4
options = whisper.DecodingOptions(threads=4)

            result = tiny_model.transcribe('wake_detect.wav', options=options)

This does't work. I'm sure I'm just getting something simple wrong. Please help!

[phineas-pta]

not all options go through .transcribe()

use torch.set_num_threads(4)

[phineas-pta]

yes you need to import torch

whuy do you want to avoid it?

[miles-du]

Elegance. Probably my lack of experience, but puzzled why can't just pass in-built arg direct to whisper? Is it because it's a script vs CLI?

[phineas-pta]

import a package doesn't make things ugly

why can't just pass in-built arg direct to whisper?

because that's how whisper was written, take a look at source code! (file whisper/transcribe.py)

[miles-du]

Ok, many thanks and of course I'll take your advice. Cheers.

[miles-du]

Worked perfect! Thanks phineas.