Medix-AI/coqui-tts
Medix-AI
/
coqui-tts
mirror of https://github.com/coqui-ai/TTS.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Turn more clear the VITS loss function

This commit is contained in:
Edresson 2021-11-22 08:48:56 -03:00 committed by Eren Gölge
parent 6fc3b9e679
commit 8c22d5ac49
1 changed files with 6 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
TTS/tts/layers/losses.py
View File

@ -586,6 +586,11 @@ class VitsGeneratorLoss(nn.Module):
l = kl / torch.sum(z_mask) l = kl / torch.sum(z_mask)
return l return l
@staticmethod
def cosine_similarity_loss(gt_spk_emb, syn_spk_emb):
l = - torch.nn.functional.cosine_similarity(gt_spk_emb, syn_spk_emb).mean()
return l
def forward( def forward(
self, self,
waveform, waveform,
@ -632,9 +637,7 @@ class VitsGeneratorLoss(nn.Module):
loss = loss_kl + loss_feat + loss_mel + loss_gen + loss_duration loss = loss_kl + loss_feat + loss_mel + loss_gen + loss_duration
if use_speaker_encoder_as_loss: if use_speaker_encoder_as_loss:
loss_se = ( loss_se = self.cosine_similarity_loss(gt_spk_emb, syn_spk_emb) * self.spk_encoder_loss_alpha
-torch.nn.functional.cosine_similarity(gt_spk_emb, syn_spk_emb).mean() * self.spk_encoder_loss_alpha
)
loss += loss_se loss += loss_se
return_dict["loss_spk_encoder"] = loss_se return_dict["loss_spk_encoder"] = loss_se