[Bug]: Package conflict with numpy

[Sewlell]

OS version

Windows 11 22H2

GPU

NVIDIA MX130

Python version

Python 3.10.6

PyTorch version

Torch 1.13.1

Branch of sovits

4.0(Default)

Dataset source (Used to judge the dataset quality)

UVR-processed

Where thr problem occurs or what command you executed

python inference_main.py -m "logs/44k/G_16000.pth" -c "logs/44k/config.json" -cm "/logs/44k/kmeans_10000.pt" -cr 0.5 -n "2.wav" -t 0 -s "Ado20mdataset" -f0p rmvpe

Situation description

First thing first, somehow there is no choice of choosing 4.1 Stable branch of sovits at that section? I am using the 4.1 Stable, not the 4.0.

Anyway, I tried to edit the cluster infer ratio, but I always got the Error 1 ( Look at the Log section ). It said I am missing cluster_model, so I go pip install cluster_model.

After installing cluster_model, I run the same command again and this time it show up the Error 2, which is SystemError: initialization of _internal failed without raising an exception.

I have check the stack overflow, and it said that it is the numpy issues. So I exchange it from the cluster_model's 1.25.1 to version below 1.24 because I see some compability issues show up. And it have return back Error 1 after the changing.

Also it could work without any issues if there is no -cr 0.5 command, I think there is a trouble here.

Log

https://pastebin.com/W5kFS1vx

^ Error 1

https://pastebin.com/u3rtvewA

^ Error 2

ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
espnet 202304 requires protobuf<=3.20.1, but you have protobuf 3.20.3 which is incompatible.
numba 0.56.4 requires numpy<1.24,>=1.18, but you have numpy 1.25.1 which is incompatible.
torchcrepe 0.0.20 requires librosa==0.9.1, but you have librosa 0.9.2 which is incompatible.

^ When installing numpy-1.25.1

Supplementary description

No response