mirror of https://github.com/coqui-ai/TTS.git
Fix return gpt_latents
This commit is contained in:
parent
d64cafccd2
commit
83988c922a
|
@ -568,11 +568,12 @@ class Xtts(BaseTTS):
|
|||
mode="linear"
|
||||
).transpose(1, 2)
|
||||
|
||||
gpt_latents_list.append(gpt_latents.cpu())
|
||||
wavs.append(self.hifigan_decoder(gpt_latents, g=speaker_embedding).cpu().squeeze())
|
||||
|
||||
return {
|
||||
"wav": torch.cat(wavs, dim=0).numpy(),
|
||||
"gpt_latents": torch.cat(gpt_latents_list, dim=0).numpy(),
|
||||
"gpt_latents": torch.cat(gpt_latents_list, dim=1).numpy(),
|
||||
"speaker_embedding": speaker_embedding,
|
||||
}
|
||||
|
||||
|
|
Loading…
Reference in New Issue