From 9c63bcc8c6eb0c9a22a34ab14d4eeb5fa7d53186 Mon Sep 17 00:00:00 2001 From: RVC-Boss <129054828+RVC-Boss@users.noreply.github.com> Date: Tue, 11 Jul 2023 11:49:56 +0800 Subject: [PATCH] add rmvpe support add rmvpe support --- infer-web.py | 4 +- rmvpe.py | 344 +++++++++++++++++++++++++++++++++++++++++++ vc_infer_pipeline.py | 10 +- 3 files changed, 355 insertions(+), 3 deletions(-) create mode 100644 rmvpe.py diff --git a/infer-web.py b/infer-web.py index 7de75cc..db6ebc2 100644 --- a/infer-web.py +++ b/infer-web.py @@ -1340,7 +1340,7 @@ with gr.Blocks() as app: label=i18n( "选择音高提取算法,输入歌声可用pm提速,harvest低音好但巨慢无比,crepe效果好但吃GPU" ), - choices=["pm", "harvest", "crepe"], + choices=["pm", "harvest", "crepe", "rmvpe"], value="pm", interactive=True, ) @@ -1442,7 +1442,7 @@ with gr.Blocks() as app: label=i18n( "选择音高提取算法,输入歌声可用pm提速,harvest低音好但巨慢无比,crepe效果好但吃GPU" ), - choices=["pm", "harvest", "crepe"], + choices=["pm", "harvest", "crepe", "rmvpe"], value="pm", interactive=True, ) diff --git a/rmvpe.py b/rmvpe.py new file mode 100644 index 0000000..9936bc5 --- /dev/null +++ b/rmvpe.py @@ -0,0 +1,344 @@ +import sys,torch,numpy as np,traceback,pdb +import torch.nn as nn +from time import time as ttime +import torch.nn.functional as F + +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(), + ) + if in_channels != out_channels: + self.shortcut = nn.Conv2d(in_channels, out_channels, (1, 1)) + self.is_shortcut = True + else: + self.is_shortcut = False + + def forward(self, x): + if self.is_shortcut: + return self.conv(x) + self.shortcut(x) + else: + return self.conv(x) + 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): + concat_tensors = [] + x = self.bn(x) + for i in range(self.n_encoders): + _, x = self.layers[i](x) + concat_tensors.append(_) + 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 in range(self.n_blocks): + x = self.conv[i](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 in range(self.n_inters): + x = self.layers[i](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 in range(self.n_blocks): + x = self.conv2[i](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, concat_tensors): + for i in range(self.n_decoders): + x = self.layers[i](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): + 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 * N_MELS, N_CLASS), + nn.Dropout(0.25), + nn.Sigmoid() + ) + + def forward(self, mel): + mel = mel.transpose(-1, -2).unsqueeze(1) + x = self.cnn(self.unet(mel)).transpose(1, 2).flatten(-2) + x = self.fc(x) + return x +from librosa.filters import mel +class MelSpectrogram(torch.nn.Module): + def __init__( + self, + is_half, + n_mel_channels, + sampling_rate, + win_length, + hop_length, + n_fft=None, + mel_fmin=0, + mel_fmax=None, + clamp=1e-5 + ): + super().__init__() + n_fft = win_length if n_fft is None else n_fft + self.hann_window = {} + mel_basis = mel( + sr=sampling_rate, + n_fft=n_fft, + n_mels=n_mel_channels, + fmin=mel_fmin, + fmax=mel_fmax, + htk=True) + mel_basis = torch.from_numpy(mel_basis).float() + self.register_buffer("mel_basis", mel_basis) + self.n_fft = win_length if n_fft is None else n_fft + self.hop_length = hop_length + self.win_length = win_length + self.sampling_rate = sampling_rate + self.n_mel_channels = n_mel_channels + self.clamp = clamp + self.is_half=is_half + + def forward(self, audio, keyshift=0, speed=1, center=True): + factor = 2 ** (keyshift / 12) + n_fft_new = int(np.round(self.n_fft * factor)) + win_length_new = int(np.round(self.win_length * factor)) + hop_length_new = int(np.round(self.hop_length * speed)) + keyshift_key = str(keyshift) + '_' + str(audio.device) + if keyshift_key not in self.hann_window: + self.hann_window[keyshift_key] = torch.hann_window(win_length_new).to(audio.device) + fft = torch.stft( + audio, + n_fft=n_fft_new, + hop_length=hop_length_new, + win_length=win_length_new, + window=self.hann_window[keyshift_key], + center=center, + return_complex=True) + magnitude = torch.sqrt(fft.real.pow(2) + fft.imag.pow(2)) + if keyshift != 0: + size = self.n_fft // 2 + 1 + resize = magnitude.size(1) + if resize < size: + magnitude = F.pad(magnitude, (0, 0, 0, size - resize)) + magnitude = magnitude[:, :size, :]* self.win_length / win_length_new + mel_output = torch.matmul(self.mel_basis, magnitude) + if(self.is_half==True):mel_output=mel_output.half() + log_mel_spec = torch.log(torch.clamp(mel_output, min=self.clamp)) + return log_mel_spec + + + +class RMVPE: + def __init__(self, model_path,is_half, device=None): + self.resample_kernel = {} + model = E2E(4, 1, (2, 2)) + ckpt = torch.load(model_path,map_location="cpu") + model.load_state_dict(ckpt) + model.eval() + if(is_half==True):model=model.half() + self.model = model + self.resample_kernel = {} + self.is_half=is_half + if device is None: + device = 'cuda' if torch.cuda.is_available() else 'cpu' + self.device=device + self.mel_extractor = MelSpectrogram(is_half,128, 16000, 1024, 160, None, 30, 8000).to(device) + 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] + mel = F.pad(mel, (0, 32 * ((n_frames - 1) // 32 + 1) - n_frames), mode='reflect') + 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): + audio = torch.from_numpy(audio).float().to(self.device).unsqueeze(0) + # torch.cuda.synchronize() + # t0=ttime() + mel = self.mel_extractor(audio, center=True) + # torch.cuda.synchronize() + # t1=ttime() + hidden = self.mel2hidden(mel) + # torch.cuda.synchronize() + # t2=ttime() + hidden=hidden.squeeze(0).cpu().numpy() + 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__': +# audio, sampling_rate = sf.read("卢本伟语录~1.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 = "/bili-coeus/jupyter/jupyterhub-liujing04/vits_ch/test-RMVPE/weights/rmvpe_llc_half.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() +# print(f0.shape,t1-t0) diff --git a/vc_infer_pipeline.py b/vc_infer_pipeline.py index 8e530fc..d7312d4 100644 --- a/vc_infer_pipeline.py +++ b/vc_infer_pipeline.py @@ -1,10 +1,12 @@ -import numpy as np, parselmouth, torch, pdb +import numpy as np, parselmouth, torch, pdb,sys,os from time import time as ttime import torch.nn.functional as F import scipy.signal as signal import pyworld, os, traceback, faiss, librosa, torchcrepe from scipy import signal from functools import lru_cache +now_dir = os.getcwd() +sys.path.append(now_dir) bh, ah = signal.butter(N=5, Wn=48, btype="high", fs=16000) @@ -124,6 +126,12 @@ class VC(object): f0 = torchcrepe.filter.mean(f0, 3) f0[pd < 0.1] = 0 f0 = f0[0].cpu().numpy() + elif f0_method == "rmvpe": + if(hasattr(self,"model_rmvpe")==False): + from rmvpe import RMVPE + print("loading rmvpe model") + self.model_rmvpe = RMVPE("rmvpe.pt",is_half=self.is_half, device=self.device) + f0 = self.model_rmvpe.infer_from_audio(x, thred=0.03) f0 *= pow(2, f0_up_key / 12) # with open("test.txt","w")as f:f.write("\n".join([str(i)for i in f0.tolist()])) tf0 = self.sr // self.window # 每秒f0点数