西西

Tutorial 3 : Aligment Model not the same as the paper

2

This is the alignment model in paper which has three single layer MLP: v, W and U, but in Tutorial 3 implementations，we cat vector *S* and encoder output h, and...

One-to-Many

1

can it use as an "one-to-many" conversion model? I have few unpaired datasets, 1 hour for each, 1k+ sentences, SNR > 40dB. I want to know if I can make...

Samples on reverberation set

1

Model seems only remove noise on samples but not reverberation, is that right? If the model can remove both noise and reverb?

word preprocess

1

Thanks for this awesome project ! I found a preprocess issue: ``` print(phoney.phonize("$234.009")) print(phoney.phonize("$234.09")) #T UW1 HH AH1 N D R AH0 D AH0 N D TH ER1 D IY2...

shaollow copy

1

this is a shaollow copy, makes the "_x" and "x" totally the same one? https://github.com/hyunwoongko/transformer/blob/0e5ce57589d7307cf76b53241cc523841ff67655/models/blocks/encoder_layer.py#L27

how can I get the traning data?

NS and BE(SR) in one model design

1

Is it possible to change the multiplicative design of the mask for noise suppression/dereverb to a residual design like the band-width extension, e.g. replace the activation function of the last...

**Describe the bug** The registration function seems to be malfunctioning, as the phone does not receive the verification code. However, if registered manually through a browser, the code is received.