blackteay/GPT-SoVITS
1
0
Fork 0
mirror of https://github.com/RVC-Boss/GPT-SoVITS.git synced 2025-08-19 16:29:47 +08:00
Code Issues Packages Projects Releases Wiki Activity

适配v4并行推理（还没写完）

Browse Source
This commit is contained in:
RVC-Boss 2025-04-21 22:43:46 +08:00 committed by GitHub
parent 8b394a15bc
commit 839ff9ce5b
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 113 additions and 100 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

213
GPT_SoVITS/TTS_infer_pack/TTS.py
View File

@ -25,7 +25,7 @@ from AR.models.t2s_lightning_module import Text2SemanticLightningModule
from BigVGAN.bigvgan import BigVGAN
from feature_extractor.cnhubert import CNHubert
from module.mel_processing import mel_spectrogram_torch, spectrogram_torch
from module.models import SynthesizerTrn, SynthesizerTrnV3
from module.models import SynthesizerTrn, SynthesizerTrnV3,Generator
from peft import LoraConfig, get_peft_model
from process_ckpt import get_sovits_version_from_path_fast, load_sovits_new
from transformers import AutoModelForMaskedLM, AutoTokenizer
@ -66,6 +66,19 @@ mel_fn = lambda x: mel_spectrogram_torch(
"center": False,
},
)
mel_fn_v4 = lambda x: mel_spectrogram_torch(
x,
**{
"n_fft": 1280,
"win_size": 1280,
"hop_size": 320,
"num_mels": 100,
"sampling_rate": 32000,
"fmin": 0,
"fmax": None,
"center": False,
},
)
def speed_change(input_audio: np.ndarray, speed: float, sr: int):
@ -92,11 +105,12 @@ def speed_change(input_audio: np.ndarray, speed: float, sr: int):
resample_transform_dict = {}
def resample(audio_tensor, sr0, device):
def resample(audio_tensor, sr0,sr1, device):
global resample_transform_dict
if sr0 not in resample_transform_dict:
resample_transform_dict[sr0] = torchaudio.transforms.Resample(sr0, 24000).to(device)
return resample_transform_dict[sr0](audio_tensor)
key="%s-%s"%(sr0,sr1)
if key not in resample_transform_dict:
resample_transform_dict[key] = torchaudio.transforms.Resample(sr0, sr1).to(device)
return resample_transform_dict[key](audio_tensor)
class DictToAttrRecursive(dict):
@ -130,44 +144,6 @@ class DictToAttrRecursive(dict):
class NO_PROMPT_ERROR(Exception):
pass
# configs/tts_infer.yaml
"""
custom:
bert_base_path: GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large
cnhuhbert_base_path: GPT_SoVITS/pretrained_models/chinese-hubert-base
device: cpu
is_half: false
t2s_weights_path: GPT_SoVITS/pretrained_models/gsv-v2final-pretrained/s1bert25hz-5kh-longer-epoch=12-step=369668.ckpt
vits_weights_path: GPT_SoVITS/pretrained_models/gsv-v2final-pretrained/s2G2333k.pth
version: v2
default:
bert_base_path: GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large
cnhuhbert_base_path: GPT_SoVITS/pretrained_models/chinese-hubert-base
device: cpu
is_half: false
t2s_weights_path: GPT_SoVITS/pretrained_models/s1bert25hz-2kh-longer-epoch=68e-step=50232.ckpt
vits_weights_path: GPT_SoVITS/pretrained_models/s2G488k.pth
version: v1
default_v2:
bert_base_path: GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large
cnhuhbert_base_path: GPT_SoVITS/pretrained_models/chinese-hubert-base
device: cpu
is_half: false
t2s_weights_path: GPT_SoVITS/pretrained_models/gsv-v2final-pretrained/s1bert25hz-5kh-longer-epoch=12-step=369668.ckpt
vits_weights_path: GPT_SoVITS/pretrained_models/gsv-v2final-pretrained/s2G2333k.pth
version: v2
default_v3:
bert_base_path: GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large
cnhuhbert_base_path: GPT_SoVITS/pretrained_models/chinese-hubert-base
device: cpu
is_half: false
t2s_weights_path: GPT_SoVITS/pretrained_models/s1v3.ckpt
vits_weights_path: GPT_SoVITS/pretrained_models/s2Gv3.pth
version: v3
"""
def set_seed(seed: int):
seed = int(seed)
seed = seed if seed != -1 else random.randint(0, 2**32 - 1)
@ -220,6 +196,15 @@ class TTS_Config:
"cnhuhbert_base_path": "GPT_SoVITS/pretrained_models/chinese-hubert-base",
"bert_base_path": "GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large",
},
"v4": {
"device": "cpu",
"is_half": False,
"version": "v4",
"t2s_weights_path": "GPT_SoVITS/pretrained_models/s1v3.ckpt",
"vits_weights_path": "GPT_SoVITS/pretrained_models/s2Gv3.pth",
"cnhuhbert_base_path": "GPT_SoVITS/pretrained_models/chinese-hubert-base",
"bert_base_path": "GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large",
},
}
configs: dict = None
v1_languages: list = ["auto", "en", "zh", "ja", "all_zh", "all_ja"]
@ -255,7 +240,7 @@ class TTS_Config:
assert isinstance(configs, dict)
version = configs.get("version", "v2").lower()
assert version in ["v1", "v2", "v3"]
assert version in ["v1", "v2", "v3", "v4"]
self.default_configs[version] = configs.get(version, self.default_configs[version])
self.configs: dict = configs.get("custom", deepcopy(self.default_configs[version]))
@ -356,7 +341,7 @@ class TTS_Config:
def __eq__(self, other):
return isinstance(other, TTS_Config) and self.configs_path == other.configs_path
from inference_webui import v3v4set
class TTS:
def __init__(self, configs: Union[dict, str, TTS_Config]):
if isinstance(configs, TTS_Config):
@ -369,7 +354,7 @@ class TTS:
self.bert_tokenizer: AutoTokenizer = None
self.bert_model: AutoModelForMaskedLM = None
self.cnhuhbert_model: CNHubert = None
self.bigvgan_model: BigVGAN = None
self.vocoder_model = None
self.sr_model: AP_BWE = None
self.sr_model_not_exist: bool = False
@ -423,10 +408,11 @@ class TTS:
def init_vits_weights(self, weights_path: str):
self.configs.vits_weights_path = weights_path
version, model_version, if_lora_v3 = get_sovits_version_from_path_fast(weights_path)
path_sovits_v3 = self.configs.default_configs["v3"]["vits_weights_path"]
print(self.configs.default_configs)
path_sovits = self.configs.default_configs[model_version]["vits_weights_path"]
if if_lora_v3 == True and os.path.exists(path_sovits_v3) == False:
info = path_sovits_v3 + i18n("SoVITS V3 底模缺失，无法加载相应 LoRA 权重")
if if_lora_v3 == True and os.path.exists(path_sovits) == False:
info = path_sovits + i18n("SoVITS %s 底模缺失，无法加载相应 LoRA 权重"%model_version)
raise FileExistsError(info)
# dict_s2 = torch.load(weights_path, map_location=self.configs.device,weights_only=False)
@ -456,7 +442,7 @@ class TTS:
# print(f"model_version:{model_version}")
# print(f'hps["model"]["version"]:{hps["model"]["version"]}')
if model_version != "v3":
if model_version not in v3v4set:
vits_model = SynthesizerTrn(
self.configs.filter_length // 2 + 1,
self.configs.segment_size // self.configs.hop_length,
@ -465,14 +451,14 @@ class TTS:
)
self.configs.is_v3_synthesizer = False
else:
self.configs.is_v3_synthesizer = kwargs["version"]=model_version
vits_model = SynthesizerTrnV3(
self.configs.filter_length // 2 + 1,
self.configs.segment_size // self.configs.hop_length,
n_speakers=self.configs.n_speakers,
**kwargs,
)
self.configs.is_v3_synthesizer = True
self.init_bigvgan()
self.init_vocoder()
if "pretrained" not in weights_path and hasattr(vits_model, "enc_q"):
del vits_model.enc_q
@ -482,7 +468,7 @@ class TTS:
)
else:
print(
f"Loading VITS pretrained weights from {weights_path}. {vits_model.load_state_dict(load_sovits_new(path_sovits_v3)['weight'], strict=False)}"
f"Loading VITS pretrained weights from {weights_path}. {vits_model.load_state_dict(load_sovits_new(path_sovits)['weight'], strict=False)}"
)
lora_rank = dict_s2["lora_rank"]
lora_config = LoraConfig(
@ -521,20 +507,36 @@ class TTS:
if self.configs.is_half and str(self.configs.device) != "cpu":
self.t2s_model = self.t2s_model.half()
def init_bigvgan(self):
if self.bigvgan_model is not None:
def init_vocoder(self):
if self.vocoder_model is not None:
return
self.bigvgan_model = BigVGAN.from_pretrained(
"%s/GPT_SoVITS/pretrained_models/models--nvidia--bigvgan_v2_24khz_100band_256x" % (now_dir,),
use_cuda_kernel=False,
) # if True, RuntimeError: Ninja is required to load C++ extensions
# remove weight norm in the model and set to eval mode
self.bigvgan_model.remove_weight_norm()
self.bigvgan_model = self.bigvgan_model.eval()
if self.configs.is_half == True:
self.bigvgan_model = self.bigvgan_model.half().to(self.configs.device)
if self.configs.is_v3_synthesizer=="v3":
self.vocoder_model = BigVGAN.from_pretrained(
"%s/GPT_SoVITS/pretrained_models/models--nvidia--bigvgan_v2_24khz_100band_256x" % (now_dir,),
use_cuda_kernel=False,
) # if True, RuntimeError: Ninja is required to load C++ extensions
# remove weight norm in the model and set to eval mode
self.vocoder_model.remove_weight_norm()
self.vocoder_model = self.vocoder_model.eval()
else:
self.bigvgan_model = self.bigvgan_model.to(self.configs.device)
self.vocoder_model = Generator(
initial_channel=100,
resblock="1",
resblock_kernel_sizes=[3, 7, 11],
resblock_dilation_sizes=[[1, 3, 5], [1, 3, 5], [1, 3, 5]],
upsample_rates=[10, 6, 2, 2, 2],
upsample_initial_channel=512,
upsample_kernel_sizes=[20, 12, 4, 4, 4],
gin_channels=0, is_bias=True
)
self.vocoder_model.eval()
self.vocoder_model.remove_weight_norm()
state_dict_g = torch.load("%s/GPT_SoVITS/pretrained_models/gsv-v4-pretrained/vocoder.pth" % (now_dir,), map_location="cpu")
print("loading v4 vocoder", self.vocoder_model.load_state_dict(state_dict_g))
if self.configs.is_half == True:
self.vocoder_model = self.vocoder_model.half().to(self.configs.device)
else:
self.vocoder_model = self.vocoder_model.to(self.configs.device)
def init_sr_model(self):
if self.sr_model is not None:
@ -570,8 +572,8 @@ class TTS:
self.bert_model = self.bert_model.half()
if self.cnhuhbert_model is not None:
self.cnhuhbert_model = self.cnhuhbert_model.half()
if self.bigvgan_model is not None:
self.bigvgan_model = self.bigvgan_model.half()
if self.vocoder_model is not None:
self.vocoder_model = self.vocoder_model.half()
else:
if self.t2s_model is not None:
self.t2s_model = self.t2s_model.float()
@ -581,8 +583,8 @@ class TTS:
self.bert_model = self.bert_model.float()
if self.cnhuhbert_model is not None:
self.cnhuhbert_model = self.cnhuhbert_model.float()
if self.bigvgan_model is not None:
self.bigvgan_model = self.bigvgan_model.float()
if self.vocoder_model is not None:
self.vocoder_model = self.vocoder_model.float()
def set_device(self, device: torch.device, save: bool = True):
"""
@ -601,8 +603,8 @@ class TTS:
self.bert_model = self.bert_model.to(device)
if self.cnhuhbert_model is not None:
self.cnhuhbert_model = self.cnhuhbert_model.to(device)
if self.bigvgan_model is not None:
self.bigvgan_model = self.bigvgan_model.to(device)
if self.vocoder_model is not None:
self.vocoder_model = self.vocoder_model.to(device)
if self.sr_model is not None:
self.sr_model = self.sr_model.to(device)
@ -913,13 +915,13 @@ class TTS:
split_bucket = False
print(i18n("分段返回模式不支持分桶处理，已自动关闭分桶处理"))
if split_bucket and speed_factor == 1.0 and not (self.configs.is_v3_synthesizer and parallel_infer):
if split_bucket and speed_factor == 1.0 and not (self.configs.is_v3_synthesizer!=False and parallel_infer):
print(i18n("分桶处理模式已开启"))
elif speed_factor != 1.0:
print(i18n("语速调节不支持分桶处理，已自动关闭分桶处理"))
split_bucket = False
elif self.configs.is_v3_synthesizer and parallel_infer:
print(i18n("当开启并行推理模式时，SoVits V3模型不支持分桶处理，已自动关闭分桶处理"))
elif self.configs.is_v3_synthesizer!=False and parallel_infer:
print(i18n("当开启并行推理模式时，SoVits V3V4模型不支持分桶处理，已自动关闭分桶处理"))
split_bucket = False
else:
print(i18n("分桶处理模式已关闭"))
@ -936,7 +938,7 @@ class TTS:
if not no_prompt_text:
assert prompt_lang in self.configs.languages
if no_prompt_text and self.configs.is_v3_synthesizer:
if no_prompt_text and self.configs.is_v3_synthesizer!=False:
raise NO_PROMPT_ERROR("prompt_text cannot be empty when using SoVITS_V3")
if ref_audio_path in [None, ""] and (
@ -1044,7 +1046,12 @@ class TTS:
t_34 = 0.0
t_45 = 0.0
audio = []
output_sr = self.configs.sampling_rate if not self.configs.is_v3_synthesizer else 24000
if self.configs.is_v3_synthesizer==False:
output_sr = 32000
elif self.configs.is_v3_synthesizer == "v3":
output_sr = 24000
else:
output_sr = 48000 # v4
for item in data:
t3 = time.perf_counter()
if return_fragment:
@ -1144,7 +1151,7 @@ class TTS:
if parallel_infer:
print(f"{i18n('并行合成中')}...")
audio_fragments = self.v3_synthesis_batched_infer(
idx_list, pred_semantic_list, batch_phones, speed=speed_factor, sample_steps=sample_steps
idx_list, pred_semantic_list, batch_phones, speed=speed_factor, sample_steps=sample_steps,model_version=self.configs.is_v3_synthesizer
)
batch_audio_fragment.extend(audio_fragments)
else:
@ -1154,7 +1161,7 @@ class TTS:
pred_semantic_list[i][-idx:].unsqueeze(0).unsqueeze(0)
) # .unsqueeze(0)#mq要多unsqueeze一次
audio_fragment = self.v3_synthesis(
_pred_semantic, phones, speed=speed_factor, sample_steps=sample_steps
_pred_semantic, phones, speed=speed_factor, sample_steps=sample_steps,model_version=self.configs.is_v3_synthesizer
)
batch_audio_fragment.append(audio_fragment)
@ -1169,7 +1176,7 @@ class TTS:
speed_factor,
False,
fragment_interval,
super_sampling if self.configs.is_v3_synthesizer else False,
super_sampling if self.configs.is_v3_synthesizer=="v3" else False,
)
else:
audio.append(batch_audio_fragment)
@ -1190,7 +1197,7 @@ class TTS:
speed_factor,
split_bucket,
fragment_interval,
super_sampling if self.configs.is_v3_synthesizer else False,
super_sampling if self.configs.is_v3_synthesizer=="v3" else False,
)
except Exception as e:
@ -1273,7 +1280,7 @@ class TTS:
return sr, audio
def v3_synthesis(
self, semantic_tokens: torch.Tensor, phones: torch.Tensor, speed: float = 1.0, sample_steps: int = 32
self, semantic_tokens: torch.Tensor, phones: torch.Tensor, speed: float = 1.0, sample_steps: int = 32,model_version="v4"
):
prompt_semantic_tokens = self.prompt_cache["prompt_semantic"].unsqueeze(0).unsqueeze(0).to(self.configs.device)
prompt_phones = torch.LongTensor(self.prompt_cache["phones"]).unsqueeze(0).to(self.configs.device)
@ -1285,19 +1292,22 @@ class TTS:
ref_audio = ref_audio.to(self.configs.device).float()
if ref_audio.shape[0] == 2:
ref_audio = ref_audio.mean(0).unsqueeze(0)
if ref_sr != 24000:
ref_audio = resample(ref_audio, ref_sr, self.configs.device)
tgt_sr = 24000 if model_version == "v3" else 32000
if ref_sr != tgt_sr:
ref_audio = resample(ref_audio, ref_sr,tgt_sr, self.configs.device)
mel2 = mel_fn(ref_audio)
mel2 = mel_fn(ref_audio) if model_version == "v3" else mel_fn_v4(ref_audio)
mel2 = norm_spec(mel2)
T_min = min(mel2.shape[2], fea_ref.shape[2])
mel2 = mel2[:, :, :T_min]
fea_ref = fea_ref[:, :, :T_min]
if T_min > 468:
mel2 = mel2[:, :, -468:]
fea_ref = fea_ref[:, :, -468:]
T_min = 468
chunk_len = 934 - T_min
Tref = 468 if model_version == "v3" else 500
Tchunk = 934 if model_version == "v3" else 1000
if T_min > Tref:
mel2 = mel2[:, :, -Tref:]
fea_ref = fea_ref[:, :, -Tref:]
T_min = Tref
chunk_len = Tchunk - T_min
mel2 = mel2.to(self.precision)
fea_todo, ge = self.vits_model.decode_encp(semantic_tokens, phones, refer_audio_spec, ge, speed)
@ -1324,7 +1334,7 @@ class TTS:
cfm_res = denorm_spec(cfm_res)
with torch.inference_mode():
wav_gen = self.bigvgan_model(cfm_res)
wav_gen = self.vocoder_model(cfm_res)
audio = wav_gen[0][0] # .cpu().detach().numpy()
return audio
@ -1335,7 +1345,7 @@ class TTS:
semantic_tokens_list: List[torch.Tensor],
batch_phones: List[torch.Tensor],
speed: float = 1.0,
sample_steps: int = 32,
sample_steps: int = 32,model_version="v4"
) -> List[torch.Tensor]:
prompt_semantic_tokens = self.prompt_cache["prompt_semantic"].unsqueeze(0).unsqueeze(0).to(self.configs.device)
prompt_phones = torch.LongTensor(self.prompt_cache["phones"]).unsqueeze(0).to(self.configs.device)
@ -1347,19 +1357,22 @@ class TTS:
ref_audio = ref_audio.to(self.configs.device).float()
if ref_audio.shape[0] == 2:
ref_audio = ref_audio.mean(0).unsqueeze(0)
if ref_sr != 24000:
ref_audio = resample(ref_audio, ref_sr, self.configs.device)
tgt_sr = 24000 if model_version == "v3" else 32000
if ref_sr != tgt_sr:
ref_audio = resample(ref_audio, ref_sr,tgt_sr, self.configs.device)
mel2 = mel_fn(ref_audio)
mel2 = mel_fn(ref_audio) if model_version == "v3" else mel_fn_v4(ref_audio)
mel2 = norm_spec(mel2)
T_min = min(mel2.shape[2], fea_ref.shape[2])
mel2 = mel2[:, :, :T_min]
fea_ref = fea_ref[:, :, :T_min]
if T_min > 468:
mel2 = mel2[:, :, -468:]
fea_ref = fea_ref[:, :, -468:]
T_min = 468
chunk_len = 934 - T_min
Tref = 468 if model_version == "v3" else 500
Tchunk = 934 if model_version == "v3" else 1000
if T_min > Tref:
mel2 = mel2[:, :, -Tref:]
fea_ref = fea_ref[:, :, -Tref:]
T_min = Tref
chunk_len = Tchunk - T_min
mel2 = mel2.to(self.precision)
@ -1413,7 +1426,7 @@ class TTS:
pred_spec = denorm_spec(pred_spec)
with torch.no_grad():
wav_gen = self.bigvgan_model(pred_spec)
wav_gen = self.vocoder_model(pred_spec)
audio = wav_gen[0][0] # .cpu().detach().numpy()
audio_fragments = []