About sparsification code

xiph / LPCNet

Efficient neural speech synthesis

BSD 3-Clause "New" or "Revised" License

1.12k stars 295 forks source link

About sparsification code #48

Closed HanzhengLi closed 5 years ago

HanzhengLi commented 5 years ago

sHi, I find that when doing sparsification on GRU_A during training process, the sparsified weights (with shape (384, 1152)) are vertical strips of 1. However, when dumping the model, the weight matrix (with shape (384,1152)) are horizontal strips of non-zero values. Why is this happening? This might be a silly question, but it troubled me the entire day. Thanks!

HanzhengLi commented 5 years ago

OK, there's difference in the way of CudnnGRU/GRU storing weights.

weixsong commented 4 years ago

is there any source code could explain the difference of CuDnnGRU & GRU?