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

Remove ununsed code

This commit is contained in:
Edresson Casanova 2023-11-14 13:39:37 -03:00
parent b231ca8878
commit f5fd31f8e8
1 changed files with 0 additions and 31 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

31
TTS/tts/models/xtts.py
View File

@ -7,7 +7,6 @@ import torch.nn.functional as F
import torchaudio import torchaudio
from coqpit import Coqpit from coqpit import Coqpit
from TTS.tts.layers.tortoise.audio_utils import wav_to_univnet_mel
from TTS.tts.layers.xtts.gpt import GPT from TTS.tts.layers.xtts.gpt import GPT
from TTS.tts.layers.xtts.hifigan_decoder import HifiDecoder from TTS.tts.layers.xtts.hifigan_decoder import HifiDecoder
from TTS.tts.layers.xtts.stream_generator import init_stream_support from TTS.tts.layers.xtts.stream_generator import init_stream_support
@ -308,26 +307,6 @@ class Xtts(BaseTTS):
cond_latent = self.gpt.get_style_emb(mel.to(self.device)) cond_latent = self.gpt.get_style_emb(mel.to(self.device))
return cond_latent.transpose(1, 2) return cond_latent.transpose(1, 2)
@torch.inference_mode()
def get_diffusion_cond_latents(self, audio, sr):
from math import ceil
diffusion_conds = []
CHUNK_SIZE = 102400
audio_24k = torchaudio.functional.resample(audio, sr, 24000)
for chunk in range(ceil(audio_24k.shape[1] / CHUNK_SIZE)):
current_sample = audio_24k[:, chunk * CHUNK_SIZE : (chunk + 1) * CHUNK_SIZE]
current_sample = pad_or_truncate(current_sample, CHUNK_SIZE)
cond_mel = wav_to_univnet_mel(
current_sample.to(self.device),
do_normalization=False,
device=self.device,
)
diffusion_conds.append(cond_mel)
diffusion_conds = torch.stack(diffusion_conds, dim=1)
diffusion_latent = self.diffusion_decoder.get_conditioning(diffusion_conds)
return diffusion_latent
@torch.inference_mode() @torch.inference_mode()
def get_speaker_embedding(self, audio, sr): def get_speaker_embedding(self, audio, sr):
audio_16k = torchaudio.functional.resample(audio, sr, 16000) audio_16k = torchaudio.functional.resample(audio, sr, 16000)
@ -575,16 +554,6 @@ class Xtts(BaseTTS):
return_attentions=False, return_attentions=False,
return_latent=True, return_latent=True,
) )
silence_token = 83
ctokens = 0
for k in range(gpt_codes.shape[-1]):
if gpt_codes[0, k] == silence_token:
ctokens += 1
else:
ctokens = 0
if ctokens > 8:
gpt_latents = gpt_latents[:, :k]
break
if length_scale != 1.0: if length_scale != 1.0:
gpt_latents = F.interpolate( gpt_latents = F.interpolate(