Digital Life Server setup

Repo Base Env Conda Torch related Others ASR faster-whisper-large-v3 (top choice)Whisper-large-v3 Distilled Whisper-large-v3 (monolingual)TTS xTTS-v2

Repo


git clone git@github.com:BenSiso/Digital_Life_Server.git
cd Digital_Life_Server

If you want to use VITS for TTS service instead of xTTS

run following to pull down .TTS/vits as well


git submodule update --init --recursive

and refer to original repo for installation instructions

Base Env

Conda


conda create -n DLS python==3.10
conda activate DLS

Torch related


pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121

# or if you only use cpu
pip install torch torchvision torchaudio

Others


pip install requirements_linux.txt

ASR

faster-whisper-large-v3 (top choice)

test case

Model


cd [where/you/want/to/put/models]
git lfs install
git clone https://huggingface.co/Systran/faster-whisper-large-v3

Env


pip install requirements_faster_whisper.txt

(Optional) If encountered Please make sure libcudnn_ops_infer.so.8 is in your library path! issue, fix it as following


python -c 'import os; import nvidia.cublas.lib; import nvidia.cudnn.lib; print(os.path.dirname(nvidia.cublas.lib.__file__) + ":" + os.path.dirname(nvidia.cudnn.lib.__file__))'

then copy the printed paths, prepend them to LD_LIBRARY_PATH


export LD_LIBRARY_PATH="[printed paths]":$LD_LIBRARY_PATH

~/.bashrc

Don’t forget to make it effective


source ~/.bashrc
echo $LD_LIBRARY_PATH
conda activate DLS

Whisper-large-v3

test case

Flash Attention

test case

Distilled Whisper-large-v3 (monolingual)

test case

TTS

xTTS-v2

Model


cd [where/you/want/to/put/models]
git lfs install
git clone https://huggingface.co/coqui/XTTS-v2

Env


pip install TTS