Skip to content

wqt2019/wavernn_c_mb

Repository files navigation

wavernn_multi-band c++(eigen3+mkl) inference

在c++下使用eigen3+mkl矩阵库实现wavernn_multi-band(cpu)的推理服务.先将模型权重保存为bin文件(非稀疏模型),参考convert_model.py.

编译

解压eigen3,安装mkl库

修改CMakeLists.txt中的eigen3 mkl的路径,修改wavernn.h net_impl.h中的参数,与训练时的参数相匹配

mkdir cmake-build-debug

cd cmake-build-debug

cmake -D PYTHON_EXECUTABLE=/home/wqt/work/anaconda3/envs/python37/bin/python ../

make -j4

运行

./cmake-build-debug/vocoder 或者 python test_wavernnvocoder.py

About

cpp implementation for mb-wavernn inference

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published