Tutorials

[Anjok]

PS C:\Users\Robert\Documents\vocal-remover-2.1.0> python inference.py --absolute.mp3 --gpu 0
C:\Users\Robert\AppData\Local\Programs\Python\Pyth on37\lib\site-packages\librosa\util\decorators.py:9: NumbaDeprecationWarning: An import was requested from a module that has moved location.
Import requested from: 'numba.decorators', please update to use 'numba.core.decorators' or pin to Numba version 0.48.0. This alias will not be present in Numba version 0.50.0.
from numba.decorators import jit as optional_jit
C:\Users\Robert\AppData\Local\Programs\Python\Pyth on37\lib\site-packages\librosa\util\decorators.py:9: NumbaDeprecationWarning: An import was requested from a module that has moved location.
Import of 'jit' requested from: 'numba.decorators', please update to use 'numba.core.decorators' or pin to Numba version 0.48.0. This alias will not be present in Numba version 0.50.0.
from numba.decorators import jit as optional_jit
usage: inference.py [-h] [--gpu GPU] [--model MODEL] --input INPUT [--sr SR]
[--hop_length HOP_LENGTH] [--window_size WINDOW_SIZE]
[--out_mask] [--postprocess]
inference.py: error: the following arguments are required: --input/-i
PS C:\Users\Robert\Documents\vocal-remover-2.1.0

tried it just thru cpu every way
tried every cmd every way as gpu
unless im missing something lol

tried installing numba
PS C:\> pip install numba Requirement already satisfied: numba in c:\users\robert\appdata\local\programs\python\pyth on37\lib\site-packages (0.49.0)
Requirement already satisfied: numpy>=1.15 in c:\users\robert\appdata\local\programs\python\pyth on37\lib\site-packages (from numba) (1.18.3)
Requirement already satisfied: llvmlite<=0.33.0.dev0,>=0.31.0.dev0 in c:\users\robert\appdata\local\programs\python\pyth on37\lib\site-packages (from numba) (0.32.0)
Requirement already satisfied: setuptools in c:\users\robert\appdata\local\programs\python\pyth on37\lib\site-packages (from numba) (41.2.0)

You're so close! Looks like you actually got it all working, I just see one problem with your command

You're using this -

python inference.py --absolute.mp3 --gpu 0

When it should be this -

python inference.py --input absolute.mp3 --gpu 0

[rkeane]

still the same my friend hopefully it can get figured out.....i am very interested in the results


PS C:\Users\Robert\Documents\vocal-remover-2.1.0> python inference.py --input absolute.mp3 --gpu 0
C:\Users\Robert\AppData\Local\Programs\Python\Pyth on37\lib\site-packages\librosa\util\decorators.py:9: NumbaDeprecationWarning: An import was requested from a module that has moved location.
Import requested from: 'numba.decorators', please update to use 'numba.core.decorators' or pin to Numba version 0.48.0. This alias will not be present in Numba version 0.50.0.
from numba.decorators import jit as optional_jit
C:\Users\Robert\AppData\Local\Programs\Python\Pyth on37\lib\site-packages\librosa\util\decorators.py:9: NumbaDeprecationWarning: An import was requested from a module that has moved location.
Import of 'jit' requested from: 'numba.decorators', please update to use 'numba.core.decorators' or pin to Numba version 0.48.0. This alias will not be present in Numba version 0.50.0.
from numba.decorators import jit as optional_jit
loading model... Traceback (most recent call last):
File "inference.py", line 92, in
main()
File "inference.py", line 31, in main
model.load_state_dict(torch.load(args.model, map_location=device))
File "C:\Users\Robert\AppData\Local\Programs\Python\Pyth on37\lib\site-packages\torch\serialization.py", line 419, in load
f = open(f, 'rb')
FileNotFoundError: [Errno 2] No such file or directory: 'models/baseline.pth'

[Anjok]

still the same my friend hopefully it can get figured out.....i am very interested in the results


PS C:\Users\Robert\Documents\vocal-remover-2.1.0> python inference.py --input absolute.mp3 --gpu 0
C:\Users\Robert\AppData\Local\Programs\Python\Pyth on37\lib\site-packages\librosa\util\decorators.py:9: NumbaDeprecationWarning: An import was requested from a module that has moved location.
Import requested from: 'numba.decorators', please update to use 'numba.core.decorators' or pin to Numba version 0.48.0. This alias will not be present in Numba version 0.50.0.
from numba.decorators import jit as optional_jit
C:\Users\Robert\AppData\Local\Programs\Python\Pyth on37\lib\site-packages\librosa\util\decorators.py:9: NumbaDeprecationWarning: An import was requested from a module that has moved location.
Import of 'jit' requested from: 'numba.decorators', please update to use 'numba.core.decorators' or pin to Numba version 0.48.0. This alias will not be present in Numba version 0.50.0.
from numba.decorators import jit as optional_jit
loading model... Traceback (most recent call last):
File "inference.py", line 92, in
main()
File "inference.py", line 31, in main
model.load_state_dict(torch.load(args.model, map_location=device))
File "C:\Users\Robert\AppData\Local\Programs\Python\Pyth on37\lib\site-packages\torch\serialization.py", line 419, in load
f = open(f, 'rb')
FileNotFoundError: [Errno 2] No such file or directory: 'models/baseline.pth'

Seems the model is missing from the "models" folder? There should be a file in the "models" directory called "baseline.pth", that's the model.

Download this folder and do your conversions using the AI I tailored here, if it's missing - https://www.mediafire.com/file/3slcb...mover.rar/file

[Anjok]

Will do - excited to try this out!

How has your experience been with this?

[misterbeatzit]

Anjok: Wouldn't it feasibly be possible to create training models using something like a studio acapella layered over an official instrumental to a completely different song, or a vinyl rip of acapella/inst combined, even though those wouldn't sound good to be heard? The AI would still "learn" from those, wouldn't they?

[rkeane]

well got it going once i installed baseline.pth
the link for mediafire didn't work
taking about 15-30 mins depending on song length
still get an error but dunno what difference it would make

PS C:\Users\Robert\Documents\vocal-remover-2.1.0> python inference.py --input absolute.mp3 --gpu 0
C:\Users\Robert\AppData\Local\Programs\Python\Pyth on37\lib\site-packages\librosa\util\decorators.py:9: NumbaDeprecationWarning: An import was requested from a module that has moved location.
Import requested from: 'numba.decorators', please update to use 'numba.core.decorators' or pin to Numba version 0.48.0. This alias will not be present in Numba version 0.50.0.
from numba.decorators import jit as optional_jit
C:\Users\Robert\AppData\Local\Programs\Python\Pyth on37\lib\site-packages\librosa\util\decorators.py:9: NumbaDeprecationWarning: An import was requested from a module that has moved location.
Import of 'jit' requested from: 'numba.decorators', please update to use 'numba.core.decorators' or pin to Numba version 0.48.0. This alias will not be present in Numba version 0.50.0.
from numba.decorators import jit as optional_jit

but as i said it works

[Anjok]

well got it going once i installed baseline.pth
the link for mediafire didn't work
taking about 15-30 mins depending on song length
still get an error but dunno what difference it would make

PS C:\Users\Robert\Documents\vocal-remover-2.1.0> python inference.py --input absolute.mp3 --gpu 0
C:\Users\Robert\AppData\Local\Programs\Python\Pyth on37\lib\site-packages\librosa\util\decorators.py:9: NumbaDeprecationWarning: An import was requested from a module that has moved location.
Import requested from: 'numba.decorators', please update to use 'numba.core.decorators' or pin to Numba version 0.48.0. This alias will not be present in Numba version 0.50.0.
from numba.decorators import jit as optional_jit
C:\Users\Robert\AppData\Local\Programs\Python\Pyth on37\lib\site-packages\librosa\util\decorators.py:9: NumbaDeprecationWarning: An import was requested from a module that has moved location.
Import of 'jit' requested from: 'numba.decorators', please update to use 'numba.core.decorators' or pin to Numba version 0.48.0. This alias will not be present in Numba version 0.50.0.
from numba.decorators import jit as optional_jit

but as i said it works

Sorry about the link! That's because I updated the main thread and package. Here's the new link - https://www.mediafire.com/file/97r5s...mover.rar/file

If you have a Nvidia GPU and install the compatible Cuda drivers, the conversions will be 100x faster.

[Anjok]

To give everyone an update - I'm training a few new models for this AI and will be sharing them as they complete (I'm training using my GPU). According to the GitHub page, the baseline model was trained on 150 pairs of anime songs. The model I'm currently training is on 305 pairs consisting of multiple genres for generalization, so it should be better than the baseline model already provided and it should also be fine-tunable.

[algernonblack]

How has your experience been with this?

Couldn't even figure out how to get the code to run, unfortunately. I think for most people looking to try this out, some sort of standard executable with a simple GUI is going to be needed, because for the layperson running code is going to be far too difficult.

[rkeane]

pretty much just a copy and paste scenario........no need to type all that code


Open Python ----- check Add Python to path and install
Unzip Vocal Remover to Documents


Note: By this point, Python should be installed, the vocal-remover folder should be extracted to your documents.
Go to Document folder and open vocal-remover-2.1.0 folder. Press Shift and Right Click > Open PowerShell window here


Copy the following command hit return and wait for the installations to complete - pip install torch==1.3.0 torchvision==0.4.0 -f https://download.pytorch.org/whl/torch_stable.html
Copy the following command hit return and wait for the installations to complete - pip install -r requirements.txt
Copy the following command hit return and wait for the installations to complete - pip install soundstretch
Copy the following command hit return and wait for the installations to complete - pip install numba==0.48.0


Running Conversions

This AI comes with a pretty good pre-trained baseline model that you can do test conversions on. I want to make this as easy as possible to use.

1. Drop the file you want to convert into the main "vocal-remover" folder. Rename songs to a shorter text ( Somewhere Over The Rainbow ) to rainbow etc
2. Run one of the following commands:

-To perform the conversions on your CPU, copy and paste the following -
python inference.py --input YOURSONG.mp3 (or flac on cmd line below also remember to change song title)
-To perform the conversions on your GPU (faster), copy and paste the following -
python inference.py --input YOURSONG.mp3 --gpu 0
If you have a Nvidia GPU and install the compatible Cuda drivers, the conversions will be 100x faster.