innnky / so-vits-svc

基于vits与softvc的歌声音色转换模型
GNU Affero General Public License v3.0
3.57k stars 1 forks source link

Possibly add torchvision==0.11.0+cu113 and librosa==0.8.1 to requirements.txt. #168

Closed zz-xx closed 1 year ago

zz-xx commented 1 year ago

From my experience in setting up on Windows 11 these requirements are must to run. Especially any version after librosa==0.8.1 is incompatible. Can possibly be added to requirements.txt to save future users from errors that happen while executing resample.py and preprocess_hubert_f0.py.

Anomyous1 commented 1 year ago

current .txt looks like below and i use this online with 11.4 which seems to work sometimes !pip install torch==1.12.1+cu113 torchvision==0.13.1+cu113 torchaudio==0.12.1 --extra-index-url https://download.pytorch.org/whl/cu113
also seems to work so im not sure if it depends on cuda version or what. in theory it should be possible to generate based on your environment but i've nver seen that. im not even sure if it works on 11.3 since ive been using 11.6

Flask Flask_Cors gradio numpy==1.22.4 pyworld==0.3.2 scipy==1.7.3 SoundFile==0.12.1 torch==1.13.1 torchaudio==0.13.1 tqdm scikit-maad praat-parselmouth onnx onnxsim onnxoptimizer fairseq==0.12.2 librosa==0.8.1

zz-xx commented 1 year ago

Closing this as 4.0 branch has separate "requirements_win.txt" which solves problem. This issue is for 3.0 branch on windows if someone comes across this.