AttributeError: module 'umap' has no attribute 'UMAP' on macOS Big Sur

[emonigma]

I tried a clean installation of the repo on macOS with Python3.8:

brew install python@3.8
mkdir venv
python3.8 -m venv ./venv
source venv/bin/activate 
python3 -m pip install --upgrade pip
# packages missing from requirements.txt
python3 -m pip install torch torchvision torchaudio unidecode librosa inflect sounddevice umap PyQt5
python3 -m pip install -r requirements.txt
#python3 demo_cli.py
python3 demo_toolbox.py

I upload 4 utterances (30-45s long), and after the last one, I get this error:

Traceback (most recent call last):
  File "~/Real-Time-Voice-Cloning/toolbox/__init__.py", line 114, in <lambda>
    func = lambda: self.synthesize() or self.vocode()
  File "~/Real-Time-Voice-Cloning/toolbox/__init__.py", line 311, in vocode
    self.ui.draw_umap_projections(self.utterances)
  File "~/Real-Time-Voice-Cloning/toolbox/ui.py", line 118, in draw_umap_projections
    reducer = umap.UMAP(int(np.ceil(np.sqrt(len(embeds)))), metric="cosine")
AttributeError: module 'umap' has no attribute 'UMAP'

I try to generate speech, but the result is completely garbled and I get another of the errors above.

I try the CLI with python3 demo_cli.py and all tests pass. I pass a single voice file to clone a voice and I get no speech either; instead, it sounds like the recording of someone running.

I use Python 3.8 because I am sure that the program installs and runs on that version. I tried Python 3.12 and get this error:

python3 -m pip install torch torchvision torchaudio unidecode librosa inflect sounddevice umap PyQt5
ERROR: Could not find a version that satisfies the requirement torch (from versions: none)

Since this is a clean install in a new environment, I'm not sure what I'm doing wrong, so I submit an issue.