mirror of
https://github.com/fumiama/Retrieval-based-Voice-Conversion-WebUI.git
synced 2026-06-05 01:10:22 +08:00
469 lines
16 KiB
Python
469 lines
16 KiB
Python
from io import BytesIO
|
||
import os
|
||
from typing import List, Optional, Tuple, Union
|
||
import numpy as np
|
||
import torch
|
||
|
||
from infer.lib import jit
|
||
|
||
try:
|
||
# Fix "Torch not compiled with CUDA enabled"
|
||
import intel_extension_for_pytorch as ipex # pylint: disable=import-error, unused-import
|
||
|
||
if torch.xpu.is_available():
|
||
from infer.modules.ipex import ipex_init
|
||
|
||
ipex_init()
|
||
except Exception: # pylint: disable=broad-exception-caught
|
||
pass
|
||
import torch.nn as nn
|
||
import torch.nn.functional as F
|
||
from librosa.util import normalize, pad_center, tiny
|
||
from scipy.signal import get_window
|
||
|
||
import logging
|
||
|
||
logger = logging.getLogger(__name__)
|
||
|
||
from rvc.f0.mel import MelSpectrogram
|
||
|
||
from time import time as ttime
|
||
|
||
|
||
class BiGRU(nn.Module):
|
||
def __init__(self, input_features, hidden_features, num_layers):
|
||
super(BiGRU, self).__init__()
|
||
self.gru = nn.GRU(
|
||
input_features,
|
||
hidden_features,
|
||
num_layers=num_layers,
|
||
batch_first=True,
|
||
bidirectional=True,
|
||
)
|
||
|
||
def forward(self, x):
|
||
return self.gru(x)[0]
|
||
|
||
|
||
class ConvBlockRes(nn.Module):
|
||
def __init__(self, in_channels, out_channels, momentum=0.01):
|
||
super(ConvBlockRes, self).__init__()
|
||
self.conv = nn.Sequential(
|
||
nn.Conv2d(
|
||
in_channels=in_channels,
|
||
out_channels=out_channels,
|
||
kernel_size=(3, 3),
|
||
stride=(1, 1),
|
||
padding=(1, 1),
|
||
bias=False,
|
||
),
|
||
nn.BatchNorm2d(out_channels, momentum=momentum),
|
||
nn.ReLU(),
|
||
nn.Conv2d(
|
||
in_channels=out_channels,
|
||
out_channels=out_channels,
|
||
kernel_size=(3, 3),
|
||
stride=(1, 1),
|
||
padding=(1, 1),
|
||
bias=False,
|
||
),
|
||
nn.BatchNorm2d(out_channels, momentum=momentum),
|
||
nn.ReLU(),
|
||
)
|
||
# self.shortcut:Optional[nn.Module] = None
|
||
if in_channels != out_channels:
|
||
self.shortcut = nn.Conv2d(in_channels, out_channels, (1, 1))
|
||
|
||
def forward(self, x: torch.Tensor):
|
||
if not hasattr(self, "shortcut"):
|
||
return self.conv(x) + x
|
||
else:
|
||
return self.conv(x) + self.shortcut(x)
|
||
|
||
|
||
class Encoder(nn.Module):
|
||
def __init__(
|
||
self,
|
||
in_channels,
|
||
in_size,
|
||
n_encoders,
|
||
kernel_size,
|
||
n_blocks,
|
||
out_channels=16,
|
||
momentum=0.01,
|
||
):
|
||
super(Encoder, self).__init__()
|
||
self.n_encoders = n_encoders
|
||
self.bn = nn.BatchNorm2d(in_channels, momentum=momentum)
|
||
self.layers = nn.ModuleList()
|
||
self.latent_channels = []
|
||
for i in range(self.n_encoders):
|
||
self.layers.append(
|
||
ResEncoderBlock(
|
||
in_channels, out_channels, kernel_size, n_blocks, momentum=momentum
|
||
)
|
||
)
|
||
self.latent_channels.append([out_channels, in_size])
|
||
in_channels = out_channels
|
||
out_channels *= 2
|
||
in_size //= 2
|
||
self.out_size = in_size
|
||
self.out_channel = out_channels
|
||
|
||
def forward(self, x: torch.Tensor):
|
||
concat_tensors: List[torch.Tensor] = []
|
||
x = self.bn(x)
|
||
for i, layer in enumerate(self.layers):
|
||
t, x = layer(x)
|
||
concat_tensors.append(t)
|
||
return x, concat_tensors
|
||
|
||
|
||
class ResEncoderBlock(nn.Module):
|
||
def __init__(
|
||
self, in_channels, out_channels, kernel_size, n_blocks=1, momentum=0.01
|
||
):
|
||
super(ResEncoderBlock, self).__init__()
|
||
self.n_blocks = n_blocks
|
||
self.conv = nn.ModuleList()
|
||
self.conv.append(ConvBlockRes(in_channels, out_channels, momentum))
|
||
for i in range(n_blocks - 1):
|
||
self.conv.append(ConvBlockRes(out_channels, out_channels, momentum))
|
||
self.kernel_size = kernel_size
|
||
if self.kernel_size is not None:
|
||
self.pool = nn.AvgPool2d(kernel_size=kernel_size)
|
||
|
||
def forward(self, x):
|
||
for i, conv in enumerate(self.conv):
|
||
x = conv(x)
|
||
if self.kernel_size is not None:
|
||
return x, self.pool(x)
|
||
else:
|
||
return x
|
||
|
||
|
||
class Intermediate(nn.Module): #
|
||
def __init__(self, in_channels, out_channels, n_inters, n_blocks, momentum=0.01):
|
||
super(Intermediate, self).__init__()
|
||
self.n_inters = n_inters
|
||
self.layers = nn.ModuleList()
|
||
self.layers.append(
|
||
ResEncoderBlock(in_channels, out_channels, None, n_blocks, momentum)
|
||
)
|
||
for i in range(self.n_inters - 1):
|
||
self.layers.append(
|
||
ResEncoderBlock(out_channels, out_channels, None, n_blocks, momentum)
|
||
)
|
||
|
||
def forward(self, x):
|
||
for i, layer in enumerate(self.layers):
|
||
x = layer(x)
|
||
return x
|
||
|
||
|
||
class ResDecoderBlock(nn.Module):
|
||
def __init__(self, in_channels, out_channels, stride, n_blocks=1, momentum=0.01):
|
||
super(ResDecoderBlock, self).__init__()
|
||
out_padding = (0, 1) if stride == (1, 2) else (1, 1)
|
||
self.n_blocks = n_blocks
|
||
self.conv1 = nn.Sequential(
|
||
nn.ConvTranspose2d(
|
||
in_channels=in_channels,
|
||
out_channels=out_channels,
|
||
kernel_size=(3, 3),
|
||
stride=stride,
|
||
padding=(1, 1),
|
||
output_padding=out_padding,
|
||
bias=False,
|
||
),
|
||
nn.BatchNorm2d(out_channels, momentum=momentum),
|
||
nn.ReLU(),
|
||
)
|
||
self.conv2 = nn.ModuleList()
|
||
self.conv2.append(ConvBlockRes(out_channels * 2, out_channels, momentum))
|
||
for i in range(n_blocks - 1):
|
||
self.conv2.append(ConvBlockRes(out_channels, out_channels, momentum))
|
||
|
||
def forward(self, x, concat_tensor):
|
||
x = self.conv1(x)
|
||
x = torch.cat((x, concat_tensor), dim=1)
|
||
for i, conv2 in enumerate(self.conv2):
|
||
x = conv2(x)
|
||
return x
|
||
|
||
|
||
class Decoder(nn.Module):
|
||
def __init__(self, in_channels, n_decoders, stride, n_blocks, momentum=0.01):
|
||
super(Decoder, self).__init__()
|
||
self.layers = nn.ModuleList()
|
||
self.n_decoders = n_decoders
|
||
for i in range(self.n_decoders):
|
||
out_channels = in_channels // 2
|
||
self.layers.append(
|
||
ResDecoderBlock(in_channels, out_channels, stride, n_blocks, momentum)
|
||
)
|
||
in_channels = out_channels
|
||
|
||
def forward(self, x: torch.Tensor, concat_tensors: List[torch.Tensor]):
|
||
for i, layer in enumerate(self.layers):
|
||
x = layer(x, concat_tensors[-1 - i])
|
||
return x
|
||
|
||
|
||
class DeepUnet(nn.Module):
|
||
def __init__(
|
||
self,
|
||
kernel_size,
|
||
n_blocks,
|
||
en_de_layers=5,
|
||
inter_layers=4,
|
||
in_channels=1,
|
||
en_out_channels=16,
|
||
):
|
||
super(DeepUnet, self).__init__()
|
||
self.encoder = Encoder(
|
||
in_channels, 128, en_de_layers, kernel_size, n_blocks, en_out_channels
|
||
)
|
||
self.intermediate = Intermediate(
|
||
self.encoder.out_channel // 2,
|
||
self.encoder.out_channel,
|
||
inter_layers,
|
||
n_blocks,
|
||
)
|
||
self.decoder = Decoder(
|
||
self.encoder.out_channel, en_de_layers, kernel_size, n_blocks
|
||
)
|
||
|
||
def forward(self, x: torch.Tensor) -> torch.Tensor:
|
||
x, concat_tensors = self.encoder(x)
|
||
x = self.intermediate(x)
|
||
x = self.decoder(x, concat_tensors)
|
||
return x
|
||
|
||
|
||
class E2E(nn.Module):
|
||
def __init__(
|
||
self,
|
||
n_blocks,
|
||
n_gru,
|
||
kernel_size,
|
||
en_de_layers=5,
|
||
inter_layers=4,
|
||
in_channels=1,
|
||
en_out_channels=16,
|
||
):
|
||
super(E2E, self).__init__()
|
||
self.unet = DeepUnet(
|
||
kernel_size,
|
||
n_blocks,
|
||
en_de_layers,
|
||
inter_layers,
|
||
in_channels,
|
||
en_out_channels,
|
||
)
|
||
self.cnn = nn.Conv2d(en_out_channels, 3, (3, 3), padding=(1, 1))
|
||
if n_gru:
|
||
self.fc = nn.Sequential(
|
||
BiGRU(3 * 128, 256, n_gru),
|
||
nn.Linear(512, 360),
|
||
nn.Dropout(0.25),
|
||
nn.Sigmoid(),
|
||
)
|
||
else:
|
||
self.fc = nn.Sequential(
|
||
nn.Linear(3 * nn.N_MELS, nn.N_CLASS), nn.Dropout(0.25), nn.Sigmoid()
|
||
)
|
||
|
||
def forward(self, mel):
|
||
# print(mel.shape)
|
||
mel = mel.transpose(-1, -2).unsqueeze(1)
|
||
x = self.cnn(self.unet(mel)).transpose(1, 2).flatten(-2)
|
||
x = self.fc(x)
|
||
# print(x.shape)
|
||
return x
|
||
|
||
|
||
class RMVPE:
|
||
def __init__(self, model_path: str, is_half, device=None, use_jit=False):
|
||
self.resample_kernel = {}
|
||
self.resample_kernel = {}
|
||
self.is_half = is_half
|
||
if device is None:
|
||
device = "cuda:0" if torch.cuda.is_available() else "cpu"
|
||
self.device = device
|
||
self.mel_extractor = MelSpectrogram(
|
||
is_half=is_half,
|
||
n_mel_channels=128,
|
||
sampling_rate=16000,
|
||
win_length=1024,
|
||
hop_length=160,
|
||
mel_fmin=30,
|
||
mel_fmax=8000,
|
||
device=device,
|
||
).to(device)
|
||
if "privateuseone" in str(device):
|
||
import onnxruntime as ort
|
||
|
||
ort_session = ort.InferenceSession(
|
||
"%s/rmvpe.onnx" % os.environ["rmvpe_root"],
|
||
providers=["DmlExecutionProvider"],
|
||
)
|
||
self.model = ort_session
|
||
else:
|
||
if str(self.device) == "cuda":
|
||
self.device = torch.device("cuda:0")
|
||
|
||
def get_jit_model():
|
||
jit_model_path = model_path.rstrip(".pth")
|
||
jit_model_path += ".half.jit" if is_half else ".jit"
|
||
ckpt = None
|
||
if os.path.exists(jit_model_path):
|
||
ckpt = jit.load(jit_model_path)
|
||
model_device = ckpt["device"]
|
||
if model_device != str(self.device):
|
||
del ckpt
|
||
ckpt = None
|
||
|
||
if ckpt is None:
|
||
ckpt = jit.rmvpe_jit_export(
|
||
model_path=model_path,
|
||
mode="script",
|
||
inputs_path=None,
|
||
save_path=jit_model_path,
|
||
device=device,
|
||
is_half=is_half,
|
||
)
|
||
|
||
model = torch.jit.load(BytesIO(ckpt["model"]), map_location=device)
|
||
return model
|
||
|
||
def get_default_model():
|
||
model = E2E(4, 1, (2, 2))
|
||
ckpt = torch.load(model_path, map_location="cpu")
|
||
model.load_state_dict(ckpt)
|
||
model.eval()
|
||
if is_half:
|
||
model = model.half()
|
||
else:
|
||
model = model.float()
|
||
return model
|
||
|
||
if use_jit:
|
||
if is_half and "cpu" in str(self.device):
|
||
logger.warning(
|
||
"Use default rmvpe model. \
|
||
Jit is not supported on the CPU for half floating point"
|
||
)
|
||
self.model = get_default_model()
|
||
else:
|
||
self.model = get_jit_model()
|
||
else:
|
||
self.model = get_default_model()
|
||
|
||
self.model = self.model.to(device)
|
||
cents_mapping = 20 * np.arange(360) + 1997.3794084376191
|
||
self.cents_mapping = np.pad(cents_mapping, (4, 4)) # 368
|
||
|
||
def mel2hidden(self, mel):
|
||
with torch.no_grad():
|
||
n_frames = mel.shape[-1]
|
||
n_pad = 32 * ((n_frames - 1) // 32 + 1) - n_frames
|
||
if n_pad > 0:
|
||
mel = F.pad(mel, (0, n_pad), mode="constant")
|
||
if "privateuseone" in str(self.device):
|
||
onnx_input_name = self.model.get_inputs()[0].name
|
||
onnx_outputs_names = self.model.get_outputs()[0].name
|
||
hidden = self.model.run(
|
||
[onnx_outputs_names],
|
||
input_feed={onnx_input_name: mel.cpu().numpy()},
|
||
)[0]
|
||
else:
|
||
mel = mel.half() if self.is_half else mel.float()
|
||
hidden = self.model(mel)
|
||
return hidden[:, :n_frames]
|
||
|
||
def decode(self, hidden, thred=0.03):
|
||
cents_pred = self.to_local_average_cents(hidden, thred=thred)
|
||
f0 = 10 * (2 ** (cents_pred / 1200))
|
||
f0[f0 == 10] = 0
|
||
# f0 = np.array([10 * (2 ** (cent_pred / 1200)) if cent_pred else 0 for cent_pred in cents_pred])
|
||
return f0
|
||
|
||
def infer_from_audio(self, audio, thred=0.03):
|
||
# torch.cuda.synchronize()
|
||
# t0 = ttime()
|
||
if not torch.is_tensor(audio):
|
||
audio = torch.from_numpy(audio)
|
||
mel = self.mel_extractor(
|
||
audio.float().to(self.device).unsqueeze(0), center=True
|
||
)
|
||
# print(123123123,mel.device.type)
|
||
# torch.cuda.synchronize()
|
||
# t1 = ttime()
|
||
hidden = self.mel2hidden(mel)
|
||
# torch.cuda.synchronize()
|
||
# t2 = ttime()
|
||
# print(234234,hidden.device.type)
|
||
if "privateuseone" not in str(self.device):
|
||
hidden = hidden.squeeze(0).cpu().numpy()
|
||
else:
|
||
hidden = hidden[0]
|
||
if self.is_half == True:
|
||
hidden = hidden.astype("float32")
|
||
|
||
f0 = self.decode(hidden, thred=thred)
|
||
# torch.cuda.synchronize()
|
||
# t3 = ttime()
|
||
# print("hmvpe:%s\t%s\t%s\t%s"%(t1-t0,t2-t1,t3-t2,t3-t0))
|
||
return f0
|
||
|
||
def to_local_average_cents(self, salience, thred=0.05):
|
||
# t0 = ttime()
|
||
center = np.argmax(salience, axis=1) # 帧长#index
|
||
salience = np.pad(salience, ((0, 0), (4, 4))) # 帧长,368
|
||
# t1 = ttime()
|
||
center += 4
|
||
todo_salience = []
|
||
todo_cents_mapping = []
|
||
starts = center - 4
|
||
ends = center + 5
|
||
for idx in range(salience.shape[0]):
|
||
todo_salience.append(salience[:, starts[idx] : ends[idx]][idx])
|
||
todo_cents_mapping.append(self.cents_mapping[starts[idx] : ends[idx]])
|
||
# t2 = ttime()
|
||
todo_salience = np.array(todo_salience) # 帧长，9
|
||
todo_cents_mapping = np.array(todo_cents_mapping) # 帧长，9
|
||
product_sum = np.sum(todo_salience * todo_cents_mapping, 1)
|
||
weight_sum = np.sum(todo_salience, 1) # 帧长
|
||
devided = product_sum / weight_sum # 帧长
|
||
# t3 = ttime()
|
||
maxx = np.max(salience, axis=1) # 帧长
|
||
devided[maxx <= thred] = 0
|
||
# t4 = ttime()
|
||
# print("decode:%s\t%s\t%s\t%s" % (t1 - t0, t2 - t1, t3 - t2, t4 - t3))
|
||
return devided
|
||
|
||
|
||
if __name__ == "__main__":
|
||
import librosa
|
||
import soundfile as sf
|
||
|
||
audio, sampling_rate = sf.read(r"C:\Users\liujing04\Desktop\Z\冬之花clip1.wav")
|
||
if len(audio.shape) > 1:
|
||
audio = librosa.to_mono(audio.transpose(1, 0))
|
||
audio_bak = audio.copy()
|
||
if sampling_rate != 16000:
|
||
audio = librosa.resample(audio, orig_sr=sampling_rate, target_sr=16000)
|
||
model_path = r"D:\BaiduNetdiskDownload\RVC-beta-v2-0727AMD_realtime\rmvpe.pt"
|
||
thred = 0.03 # 0.01
|
||
device = "cuda" if torch.cuda.is_available() else "cpu"
|
||
rmvpe = RMVPE(model_path, is_half=False, device=device)
|
||
t0 = ttime()
|
||
f0 = rmvpe.infer_from_audio(audio, thred=thred)
|
||
# f0 = rmvpe.infer_from_audio(audio, thred=thred)
|
||
# f0 = rmvpe.infer_from_audio(audio, thred=thred)
|
||
# f0 = rmvpe.infer_from_audio(audio, thred=thred)
|
||
# f0 = rmvpe.infer_from_audio(audio, thred=thred)
|
||
t1 = ttime()
|
||
logger.info("%s %.2f", f0.shape, t1 - t0)
|