windows+CUDA+pyenv平台部署Whisper-WebUI、CosyVoice

快速运行命令

运行

jhj0517/Whisper-WebUI

激活 pyenv 环境

cd C:/w1/Whisper-WebUI; .venv/Scripts/activate; python app.py --faster_whisper_model_dir C:/w1/ai/models/whisper--share --server_name 0.0.0.0 --server_port 7860

激活 conda 环境

cd C:/w1/Whisper-WebUI; conda activate whisper; python app.py --faster_whisper_model_dir C:/w1/ai/models/whisper --share --server_name 0.0.0.0 --server_port 7860

FunAudioLLM/CosyVoice

替换环境变量，激活 CUDA12.1

$Env:PATH = ("C:\_env\nvidia\cuda-12.1\bin;C:\_env\nvidia\cuda-12.1\bin;C:\_env\nvidia\cuda-12.1\libnvvp;" + $Env:PATH -split ";" | Get-Unique) -join ";"

激活 conda 环境

cd C:/w1/CosyVoice; conda activate cosyvoice; python webui.py --model_dir C:/w1/ai/models/cosyvoice/CosyVoice2-0.5B --port 8000

克隆

jhj0517/Whisper-WebUI

git clone https://github.com/jhj0517/Whisper-WebUI.git

FunAudioLLM/CosyVoice

git clone --recursive https://github.com/FunAudioLLM/CosyVoice.git
cd CosyVoice
git submodule update --init --recursive

下载

下载 faster-whisper-large-v3 模型

$Env:HF_ENDPOINT="https://hf-mirror.com"; python -c "import huggingface_hub as hub; hub.snapshot_download('Systran/faster-whisper-large-v3', local_dir='C:/w1/ai/models/whisper/faster-whisper-large-v3', local_dir_use_symlinks=False, resume_download=True)"

下载 CosyVoice2-0.5B 模型

python -c "import modelscope as sc; sc.snapshot_download('iic/CosyVoice2-0.5B', local_dir='C:/w1/ai/models/cosyvoice/CosyVoice2-0.5B', local_dir_use_symlinks=False, resume_download=True, token=os.environ.get('HF_TOKEN'))"

Whipser-WebUI（py>=3.9）

没有版本要求，可支持 py312-cu124-torch251

部署

环境准备（pyenv）

$Env:PYENV_ROOT="C:/_env/pyenv"; $Env:PATH+=";$Env:PYENV_ROOT/pyenv-win/bin;$Env:PYENV_ROOT/pyenv-win/shims"
pyenv versions
pyenv local 3.12.6
pip install virtualenv
virtualenv .venv
.venv/Scripts/activate

环境准备（conda）

conda create -n whisper -y python=3.12.6
cd C:/w1/Whisper-WebUI; conda activate whisper
# conda deactivate
# conda env remove -n whisper -y

安装依赖

requirement.txt 中无 torch 版本要求

# pip install torch --extra-index-url https://download.pytorch.org/whl/cu124
pip install C:/_env/torch-2.5.1+cu124-cp312-cp312-win_amd64.whl
pip install torch==2.5.1 -r requirements.txt

运行

下载模型

参考

执行

python app.py

常见错误

onnx

ImportError: DLL load failed while importing onnx_cpp2py_export: 动态链接库(DLL)初始化例程失败。

pip uninstall onnx
pip install onnx==1.16.1

模型无法下载

使用镜像站点

$Env:HF_ENDPOINT="https://hf-mirror.com"

CosyVoice（conda:py310+torch231+cu121+tensor10.1）

部署

环境准备（仅conda）

conda create -n cosyvoice -y python=3.10.11
cd C:/w1/CosyVoice; conda activate cosyvoice;
# conda deactivate
# conda env remove -n cosyvoice -y

安装依赖

conda install -y -c conda-forge pynini==2.1.5
# -i https://mirrors.aliyun.com/pypi/simple/ --trusted-host=mirrors.aliyun.com
pip install C:/_env/torch-2.3.1+cu121-cp310-cp310-win_amd64.whl -r requirements.txt

# If you encounter sox compatibility issues
# ubuntu
sudo apt-get install sox libsox-dev
# centos
sudo yum install sox sox-devel

下载模型

参考

• 完整文件

from modelscope import snapshot_download
snapshot_download('iic/CosyVoice2-0.5B', local_dir='pretrained_models/CosyVoice2-0.5B')
snapshot_download('iic/CosyVoice-300M', local_dir='pretrained_models/CosyVoice-300M')
snapshot_download('iic/CosyVoice-300M-25Hz', local_dir='pretrained_models/CosyVoice-300M-25Hz')
snapshot_download('iic/CosyVoice-300M-SFT', local_dir='pretrained_models/CosyVoice-300M-SFT')
snapshot_download('iic/CosyVoice-300M-Instruct', local_dir='pretrained_models/CosyVoice-300M-Instruct')
snapshot_download('iic/CosyVoice-ttsfrd', local_dir='pretrained_models/CosyVoice-ttsfrd')

• 也可以用GIT

# git模型下载，请确保已安装git lfs
mkdir -p pretrained_models
git clone https://www.modelscope.cn/iic/CosyVoice2-0.5B.git pretrained_models/CosyVoice2-0.5B
git clone https://www.modelscope.cn/iic/CosyVoice-300M.git pretrained_models/CosyVoice-300M
git clone https://www.modelscope.cn/iic/CosyVoice-300M-25Hz.git pretrained_models/CosyVoice-300M-25Hz
git clone https://www.modelscope.cn/iic/CosyVoice-300M-SFT.git pretrained_models/CosyVoice-300M-SFT
git clone https://www.modelscope.cn/iic/CosyVoice-300M-Instruct.git pretrained_models/CosyVoice-300M-Instruct
git clone https://www.modelscope.cn/iic/CosyVoice-ttsfrd.git pretrained_models/CosyVoice-ttsfrd

错误处理

modelscope AttributeError

'ClassDef' object has no attribute 'typed_params'

pip uninstall modelscope
pip install modelscope # 1.22.1 的最新版正常的，1.14下载卡死，1.20找不到~/.cache

No module named 'onnxruntime' 报错

pip install onnxruntime==1.18.0

No module named 'tensorrt' 报错

打开 https://developer.nvidia.com/nvidia-tensorrt-download
内含 whl 文件

failed to import ttsfrd, use WeTextProcessing instead

首次运行非常非常的慢，有这个报错但不影响，继续等待即可

FileNotFoundError: [WinError 2] 系统找不到指定的文件

安装 FFMPEG...

运行

# 可以不带参数，以下是默认参数
python webui.py --port 8000 --model_dir pretrained_models/CosyVoice2-0.5B