Spleeter: [Discussion] download model archive on ecach process ?

Created on 8 Mar 2020  路  11Comments  路  Source: deezer/spleeter

is it normal that spleeter lkeeps downloading the model archive on each process ?
I anytime get :
INFO:spleeter:Downloading model archive https://github.com/deezer/spleeter/releases/download/v1.4.0/2stems.tar.gz

Cheers Marc

question

Most helpful comment

the model should be downloaded in a folder called pretrained_models by default.
You can change that (and actually avoid re-downloading) by explicitely setting the environment variable MODEL_PATH before calling spleeter.

All 11 comments

Well, I have the same problem after trying to
pip install spleeter
and do
spleeter separate -i spleeter/audio_example.mp3 -p spleeter:2stems -o output
After checking the process:
I found
96251 ttys010 0:04.12 /Users/siyunhe/anaconda3/bin/python /Users/siyunhe/anaconda3/bin/spleeter separate -i spleeter/audio_example.mp3 -p sple 96254 ttys010 0:00.04 /Users/siyunhe/anaconda3/bin/python -c from multiprocessing.semaphore_tracker import main;main(7) 96255 ttys010 0:00.00 /Users/siyunhe/anaconda3/bin/python /Users/siyunhe/anaconda3/bin/spleeter separate -i spleeter/audio_example.mp3 -p sple 96256 ttys010 0:00.00 /Users/siyunhe/anaconda3/bin/python /Users/siyunhe/anaconda3/bin/spleeter separate -i spleeter/audio_example.mp3 -p sple 96257 ttys010 0:00.00 /Users/siyunhe/anaconda3/bin/python /Users/siyunhe/anaconda3/bin/spleeter separate -i spleeter/audio_example.mp3 -p sple 96258 ttys010 0:00.00 /Users/siyunhe/anaconda3/bin/python /Users/siyunhe/anaconda3/bin/spleeter separate -i spleeter/audio_example.mp3 -p sple 96259 ttys010 0:00.00 /Users/siyunhe/anaconda3/bin/python /Users/siyunhe/anaconda3/bin/spleeter separate -i spleeter/audio_example.mp3 -p sple 96260 ttys010 0:00.00 /Users/siyunhe/anaconda3/bin/python /Users/siyunhe/anaconda3/bin/spleeter separate -i spleeter/audio_example.mp3 -p sple 96261 ttys010 0:00.00 /Users/siyunhe/anaconda3/bin/python /Users/siyunhe/anaconda3/bin/spleeter separate -i spleeter/audio_example.mp3 -p sple 96262 ttys010 0:00.00 /Users/siyunhe/anaconda3/bin/python /Users/siyunhe/anaconda3/bin/spleeter separate -i spleeter/audio_example.mp3 -p sple 96263 ttys010 0:00.00 /Users/siyunhe/anaconda3/bin/python /Users/siyunhe/anaconda3/bin/spleeter separate -i spleeter/audio_example.mp3 -p sple 96264 ttys010 0:00.00 /Users/siyunhe/anaconda3/bin/python /Users/siyunhe/anaconda3/bin/spleeter separate -i spleeter/audio_example.mp3 -p sple 96265 ttys010 0:00.00 /Users/siyunhe/anaconda3/bin/python /Users/siyunhe/anaconda3/bin/spleeter separate -i spleeter/audio_example.mp3 -p sple 96266 ttys010 0:00.00 /Users/siyunhe/anaconda3/bin/python /Users/siyunhe/anaconda3/bin/spleeter separate -i spleeter/audio_example.mp3 -p sple
Does anyone know how long it takes for the model to download?

is it normal that spleeter lkeeps downloading the model archive on each process ?
I anytime get :
INFO:spleeter:Downloading model archive https://github.com/deezer/spleeter/releases/download/v1.4.0/2stems.tar.gz

Cheers Marc

Actually it is faster than I thought, around 6min with 300kb/s, better to have a process bar to
show the download procedure.

Hi @daslicht @Franceshe

It should definitely not need to download the model more than once. This probably means it gets saved in a temporary folder that is cleaned after your execution. What's your setup ? are you executing in a docker ?

Is there a way to point spleeter to the folder of pretrained_models instead of having it download it and extract it every time?

nope just macos mojave here. anyway to determine where it gets downloaded ?

the model should be downloaded in a folder called pretrained_models by default.
You can change that (and actually avoid re-downloading) by explicitely setting the environment variable MODEL_PATH before calling spleeter.

Ok, it seams to download that folder in the same folder execute the process. For any conversion I get an output and that pretrained_models folder. What is the best way to set the MODEL_PATH variable ?

something like export MODEL_PATH=<path/to/pretrained_models/>

and this will be permanent or just for the session ?

Just for the session but you can also set it up permanently. Look on google how to depending on your OS and which terminal/shell you're using.

Got it, Thank you very much !

Was this page helpful?
0 / 5 - 0 ratings

Related issues

daslicht picture daslicht  路  4Comments

geraldoramos picture geraldoramos  路  4Comments

ateammeech picture ateammeech  路  3Comments

danielkorg picture danielkorg  路  4Comments

brunobulgaron picture brunobulgaron  路  5Comments