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

Merge pull request #520 from SanjaESC/patch-1 

fix taco2 speaker-embeddings dimension during inference
pull/523/head
Eren Gölge 2021-06-02 09:32:24 +02:00 committed by GitHub
parent f53bb9ab1b c1eb9bdcca
commit ea31215087
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 2 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
TTS/tts/models/tacotron2.py
View File

@ -255,6 +255,7 @@ class Tacotron2(TacotronAbstract):
if self.num_speakers > 1:
if not self.embeddings_per_sample:
speaker_embeddings = self.speaker_embedding(speaker_ids)[:, None]
speaker_embeddings = torch.unsqueeze(speaker_embeddings, 0).transpose(1, 2)
encoder_outputs = self._concat_speaker_embedding(encoder_outputs, speaker_embeddings)
decoder_outputs, alignments, stop_tokens = self.decoder.inference(encoder_outputs)
@ -277,6 +278,7 @@ class Tacotron2(TacotronAbstract):
if self.num_speakers > 1:
if not self.embeddings_per_sample:
speaker_embeddings = self.speaker_embedding(speaker_ids)[:, None]
speaker_embeddings = torch.unsqueeze(speaker_embeddings, 0).transpose(1, 2)
encoder_outputs = self._concat_speaker_embedding(encoder_outputs, speaker_embeddings)
mel_outputs, alignments, stop_tokens = self.decoder.inference_truncated(encoder_outputs)