UMX with flac output

[Mixerrog]

🐛 Bug

To Reproduce

Steps to reproduce the behavior:

1. 1. 1.

Expected behavior

Environment

Please add some information about your environment

• PyTorch Version (e.g., 1.2):
• OS (e.g., Linux):
• torchaudio loader (y/n):
• Python version:
• CUDA/cuDNN version:
• Any other relevant information:

If unsure you can paste the output from the pytorch environment collection script (or fill out the checklist below manually).

You can get that script and run it with:

wget https://raw.githubusercontent.com/pytorch/pytorch/master/torch/utils/collect_env.py
# For security purposes, please check the contents of collect_env.py before running it.
python collect_env.py

Additional context

[Mixerrog]

I had asked previously about flac output options & you said --- "this is already supported in the stempeg backend. E.g. run umx --ext .flac --audio-backend stempeg input.wav "

I did this & it runs but the stems are not the same separation quality as the UMX 32 bit wav version & only take 1/2 the time to produce. Looking at the flac stem spectrums, they are not cutoff at 16khz also so I think this is running the older version of open-unmix with this command instead of the new UMX with larger training tables. Is this "umx --ext .flac --audio-backend stempeg input.wav " the correct command line for the new stems version as the stem separations are the same as the older Open-Unmix which I also have on my computer using Windows 10?

Thanks for your help on this, Rog

[faroit]

@Mixerrog I don't understand what you mean by "versions". Each version of the python package produces identical results. it all depends on the pre-trained model you are using.

umx --ext .flac --audio-backend stempeg input.wav 

uses umxhq by default

you can changes this to umxl using umx --model umxl --ext .flac --audio-backend stempeg input.wav

[Mixerrog]

Faroit,

I should have referred to my question about a flac output to my # 91 issue question. Since my heading at that #91 issue was "UMX-L 1.2 installation", I thought the answer you gave me at that time "umx --ext .flac --audio-backend stempeg input.wav" was for a output from the new UMX version & that is why I used it.

Thanks for the correction.

One more thing, for the new UMX version, how do you trigger the GPU use instead of CPU use?

Rog

[faroit]

@Mixerrog i just released a new version of the python package. So you run pip install openunmix --upgrade you should get version 1.2.1. This version will use umxl by default :-)

One more thing, for the new UMX version, how do you trigger the GPU use instead of CPU use?

GPU is used automatically when available. If its not used, you need to check your python environment and maybe install a pytorch version with gpu support (see pytorch guides)

[Mixerrog]

pip install openunmix --update did not work, got error saying "no such update: --update"

Thanks for your help, Rog

[faroit]

pip install openunmix --update did not work, got error saying "no such update: --update"

Thanks for your help, Rog

sorry, its upgrade not update

[Mixerrog]

That installed version 1.2.1 & runs but still getting a errors of No module named 'pywin32_bootstrap' This is my dos command line when I run my bat file for umx --model umxl --ext .flac "audio input file" --audio-backend stempeg --outdir "output file location" I did notice you added a progress bar but it just sets at 0% till the program ends & then jumps to 100% instead of jumping in 25% steps like the old Open-Unmix did -- could this be updated to show true progress of the program?

Here are the DOS lines I am getting below

B:\Audio Files(0000) Artists Music (Flacs)\Beatles\2009 Stereo re-masters(01) Please Please Me (2009 Stereo Remaster)>REM @echo off

B:\Audio Files(0000) Artists Music (Flacs)\Beatles\2009 Stereo re-masters(01) Please Please Me (2009 Stereo Remaster)>call C:\Users\User\Anaconda3\Scripts\activate.bat C:\Users\User\Anaconda3 Error processing line 7 of C:\Users\User\Anaconda3\lib\site-packages\pywin32.pth:

Traceback (most recent call last): File "C:\Users\User\Anaconda3\lib\site.py", line 168, in addpackage exec(line) File "", line 1, in ModuleNotFoundError: No module named 'pywin32_bootstrap'

Remainder of file ignored

(base) B:\Audio Files(0000) Artists Music (Flacs)\Beatles\2009 Stereo re-masters(01) Please Please Me (2009 Stereo Remaster)>umx --model umxl --ext .flac "B:\Audio Files(0000) Artists Music (Flacs)\Beatles\2009 Stereo re-masters(01) Please Please Me (2009 Stereo Remaster)\08 Love Me Do.flac" --audio-backend stempeg --outdir "E:\Open-Unmix UMX" Error processing line 7 of c:\users\user\anaconda3\lib\site-packages\pywin32.pth:

Traceback (most recent call last): File "c:\users\user\anaconda3\lib\site.py", line 168, in addpackage exec(line) File "", line 1, in ModuleNotFoundError: No module named 'pywin32_bootstrap'

Remainder of file ignored 0%| | 0/1 [00:00<?, ?it/s]Error processing line 7 of c:\users\user\anaconda3\lib\site-packages\pywin32.pth:

Traceback (most recent call last): File "c:\users\user\anaconda3\lib\site.py", line 168, in addpackage exec(line) File "", line 1, in ModuleNotFoundError: No module named 'pywin32_bootstrap'

Remainder of file ignored Error processing line 7 of c:\users\user\anaconda3\lib\site-packages\pywin32.pth:

Error processing line 7 of c:\users\user\anaconda3\lib\site-packages\pywin32.pth:

Traceback (most recent call last): File "c:\users\user\anaconda3\lib\site.py", line 168, in addpackage exec(line) File "", line 1, in ModuleNotFoundError: No module named 'pywin32_bootstrap'

Remainder of file ignored Traceback (most recent call last): File "c:\users\user\anaconda3\lib\site.py", line 168, in addpackage exec(line) File "", line 1, in ModuleNotFoundError: No module named 'pywin32_bootstrap'

Remainder of file ignored Error processing line 7 of c:\users\user\anaconda3\lib\site-packages\pywin32.pth:

Traceback (most recent call last): File "c:\users\user\anaconda3\lib\site.py", line 168, in addpackage exec(line) File "", line 1, in ModuleNotFoundError: No module named 'pywin32_bootstrap'

Remainder of file ignored 100%|███████████████████████████████████████████████████████████████████████████████████| 1/1 [03:36<00:00, 216.43s/it]

Thanks for your help, Rog

[faroit]

@Mixerrog For your windows related error, i can't provide help... See https://stackoverflow.com/questions/58989806/anaconda-terminal-error-pywin32-bootstrap

The progress bar in the new version is one step per file now, not one per source. This is because since version 1.1.0, the model is end-to-end and we don't want to add a progress bar inside the model.forward path