jiqizaisikao

It seems that the prosody not transferd

1

I have checked your sample result of 420k trainning,and tried align the referr sound and the target sound,it seems that they are much different. So i really confused how the...

Hi,is there any guarantee not only a few of embeddings being selected?

3

Hi,i have run your source code,it seems that it works well,but i have one question,for the weights of the embeddings are inited randomly,and in tranning it always select the nearest...

only noise generated

1

I generated the sine wav with pitch 200HZ of 10 S wav file and trained wavenet with 300 steps ,and the loss down to 0.001,but when i geneated the wave...

It is greate for your sharing LSTM codes.I have tried to run your programe,but there maybe some bugs in it,when i change the value of the MOMENTUM or the hidden...

How to use multiple length sequences for trainning

1

Hi,I have tried to modify the code for multiple length sequences trainning like this: df=BatchData(df,self.batch_size) How to modify code to support multiple length sequences？

position_ids not used

2

Hi,why the var position_ids calculated but not used ： if position_ids is None: --   | position_ids = torch.arange(   | past_length,   | input_shape[-1] + past_length,   | dtype=torch.long,   | device=device,   | )  ...

Hi,i got nan loss the same as you

4

I implemented the code with some other source codes,Im sure that the parts are right beacase i checkd them Independently,but i got the same nan loss as you,when trained 100...

Hi! sir, i have used dynamoRIO 、pin、 some other DBI so on.Your paper and codes is the first one i think run fastest because of your nearly perfect idea。 I...

Hi have you try your crossentropy loss

5

I really do not understand your code,i think that the crossentropy loss is like this: loss=-p*(log(Q/P))+(-(1-P)*log((1-Q)/(1-P))) when P=Q the loss is zero if loss=-P*(log(Q/P)) when P=Q the loss is not...