xiph / LPCNet

Efficient neural speech synthesis
BSD 3-Clause "New" or "Revised" License
1.14k stars 295 forks source link

More on model optimization. #114

Closed zhangjh915 closed 3 years ago

zhangjh915 commented 4 years ago

I wonder if anyone else is also working on model optimization to achieve faster speed and lower RTF of LPCNet. I have trained LPCNet with different parameters, and tried to add sparsity in the kernel matrix of RGU_A, but none of them gave me satisfying result. Can anyone share some ideas and solutions?

alexdemartos commented 4 years ago

https://arxiv.org/abs/2005.05551

justforhome commented 4 years ago

@zhangjh915 what's your email address ? I wanna discuss this with you.