Spleeter: [Discussion] Remove Error ModuleNotFoundError: No module named 'numba.decorators'

[hashblen]

Hello,
I've had a problem with spleeter, installed with pip. Fortunately this was quite an easy fix. Error says :

Traceback (most recent call last):
  File "D:\Program Files\Python\Python37\lib\runpy.py", line 193, in _run_module_as_main
    "__main__", mod_spec)
  File "D:\Program Files\Python\Python37\lib\runpy.py", line 85, in _run_code
    exec(code, run_globals)
  File "D:\Program Files\Python\Python37\lib\site-packages\spleeter\__main__.py", line 58, in <module>
    entrypoint()
  File "D:\Program Files\Python\Python37\lib\site-packages\spleeter\__main__.py", line 54, in entrypoint
    main(sys.argv)
  File "D:\Program Files\Python\Python37\lib\site-packages\spleeter\__main__.py", line 40, in main
    from .commands.separate import entrypoint
  File "D:\Program Files\Python\Python37\lib\site-packages\spleeter\commands\separate.py", line 15, in <module>
    from ..separator import Separator
  File "D:\Program Files\Python\Python37\lib\site-packages\spleeter\separator.py", line 23, in <module>
    from librosa.core import stft, istft
  File "D:\Program Files\Python\Python37\lib\site-packages\librosa\__init__.py", line 12, in <module>
    from . import core
  File "D:\Program Files\Python\Python37\lib\site-packages\librosa\core\__init__.py", line 125, in <module>
    from .time_frequency import *  # pylint: disable=wildcard-import
  File "D:\Program Files\Python\Python37\lib\site-packages\librosa\core\time_frequency.py", line 11, in <module>
    from ..util.exceptions import ParameterError
  File "D:\Program Files\Python\Python37\lib\site-packages\librosa\util\__init__.py", line 77, in <module>
    from .utils import *  # pylint: disable=wildcard-import
  File "D:\Program Files\Python\Python37\lib\site-packages\librosa\util\utils.py", line 15, in <module>
    from .decorators import deprecated
  File "D:\Program Files\Python\Python37\lib\site-packages\librosa\util\decorators.py", line 9, in <module>
    from numba.decorators import jit as optional_jit
ModuleNotFoundError: No module named 'numba.decorators'

So just go and modify this file : (D:\Program Files\Python)Python37\lib\site-packages\librosa\util\decorators.py and change in the import numba.decoratorsto numba.core.decoratorsbecause it changed names :)
Save and launch spleeter and voilà.

[athpap]

I have the same exact problem. However, I'm a complete noob at this so do you mind explaining how I modify this file exactly?

[angristan]

Same issue here

[angristan]

Fixed with:

pip install numba==0.43.0
pip install llvmlite==0.32.1

[hashblen]

@angristan yes reverting to an old numba (<=0.48.x) could work but I had only errors when installing them
@athpap Go where python 3.6.x or 3.7.x is installed, then lib, site-packages, librosa, util, and open decorators.py with notepad, notepad++, python ide or basically any text programm and change between from import,
numba.decorators to numba.core.decorators. then save and you can use spleeter.

[waffletower]

Had this issue with a 3rd party pytorch notebook. But I see that I will have to fix my spleeter install too. python devs don't seem to know what the word downstream actually means.

[mmoussallam]

Hi everyone

The dependency issue is actually in librosa a package we rely on for fast STFT/ISTFT. The quickest fix is just to downgrade your numba install with:

pip install numba<=0.48

[AKJelane]

So just go and modify this file : (D:\Program Files\Python)Python37\lib\site-packageslibrosa\utildecorators.py and change in the import numba.decorators to numba.core.decorators because it changed names :)
Save and launch

thats worked for me

[romi1502]

We added an explicit dependency on numba 0.48 in spleeter 1.5.3. This should then fix the issue.

[waffletower]

Hi everyone

The dependency issue is actually in librosa a package we rely on for fast STFT/ISTFT. The quickest fix is just to downgrade your numba install with:

pip install numba<=0.48

This doesn't work easily in every circumstance. A better solution would be for python developers to learn how to pin dependencies as they learned decades ago in the Java ecosystem. https://en.wikiquote.org/wiki/George_Santayana
But pragmatically, I found the decorators.py fix suggested by hashblen to be more helpful.

[alejosandu]

Fixed with:

pip install numba==0.43.0
pip install llvmlite==0.32.1

Had issues with this, but adding the --user parameter at the end worked fine