fatchord / WaveRNN

WaveRNN Vocoder + TTS

Home Page:https://fatchord.github.io/model_outputs/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

(Solved, but can be useful to someone) Problems getting the project working for the first time

forestrf opened this issue · comments

Trying to set up the project for the first time by following the instructions didn't work, some dependencies were incorrect. After messing with them I eventually found a combination that works, so I'm leaving this here for those who may encounter the same problem.
I needed to use specific version of the required packages (many of them would have worked in a range of versions, but I just selected specific ones that worked to make my life easier) and also needed python 3.7 due to a package requiring a constant that was removed on python 3.8.

I replaced the contents of the requirements.txt file with:
(I just noticed another set of requirements here that seem more generic than mine https://github.com/fatchord/WaveRNN/pull/238/files)

numpy==1.16.2
librosa==0.6.3
scipy==1.5.4
scikit-learn==1.0.2
numba==0.40.1
matplotlib==3.4.3
unidecode
inflect==4.1.1
nltk==3.6.2
llvmlite==0.32.1

Then installed Python 3.7 Like so:

sudo apt update
sudo apt install software-properties-common
sudo add-apt-repository ppa:deadsnakes/ppa
sudo apt install python3.7
sudo apt install python3.7-distutils

Finally installed the pip requirements, including the CUDA drivers;

python3.7 -m pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu113
python3.7 -m pip install -r requirements.txt

And with that, the project works... at least for me.

If this is an actual issue then it would be a documentation issue, but feel free to close it otherwise.
Thank you.