From 211e13b80a4bf13d683f199d75a2f88dd50b7444 Mon Sep 17 00:00:00 2001 From: "Zhang, Di" Date: Sun, 9 Jul 2023 18:07:02 +0800 Subject: [PATCH] Add directML support to RVC for AMD & Intel GPU supported (#707) --- environment_dml.yaml | 186 ++++++ guidml.py | 710 +++++++++++++++++++++ lib/infer_pack/models_dml.py | 1124 ++++++++++++++++++++++++++++++++++ 3 files changed, 2020 insertions(+) create mode 100644 environment_dml.yaml create mode 100644 guidml.py create mode 100644 lib/infer_pack/models_dml.py diff --git a/environment_dml.yaml b/environment_dml.yaml new file mode 100644 index 0000000..0fb3f22 --- /dev/null +++ b/environment_dml.yaml @@ -0,0 +1,186 @@ +name: pydml +channels: + - pytorch + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/ + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/ + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/ + - defaults + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/fastai/ + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/ + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/bioconda/ +dependencies: + - abseil-cpp=20211102.0=hd77b12b_0 + - absl-py=1.3.0=py310haa95532_0 + - aiohttp=3.8.3=py310h2bbff1b_0 + - aiosignal=1.2.0=pyhd3eb1b0_0 + - async-timeout=4.0.2=py310haa95532_0 + - attrs=22.1.0=py310haa95532_0 + - blas=1.0=mkl + - blinker=1.4=py310haa95532_0 + - bottleneck=1.3.5=py310h9128911_0 + - brotli=1.0.9=h2bbff1b_7 + - brotli-bin=1.0.9=h2bbff1b_7 + - brotlipy=0.7.0=py310h2bbff1b_1002 + - bzip2=1.0.8=he774522_0 + - c-ares=1.19.0=h2bbff1b_0 + - ca-certificates=2023.05.30=haa95532_0 + - cachetools=4.2.2=pyhd3eb1b0_0 + - certifi=2023.5.7=py310haa95532_0 + - cffi=1.15.1=py310h2bbff1b_3 + - charset-normalizer=2.0.4=pyhd3eb1b0_0 + - click=8.0.4=py310haa95532_0 + - colorama=0.4.6=py310haa95532_0 + - contourpy=1.0.5=py310h59b6b97_0 + - cryptography=39.0.1=py310h21b164f_0 + - cycler=0.11.0=pyhd3eb1b0_0 + - fonttools=4.25.0=pyhd3eb1b0_0 + - freetype=2.12.1=ha860e81_0 + - frozenlist=1.3.3=py310h2bbff1b_0 + - giflib=5.2.1=h8cc25b3_3 + - glib=2.69.1=h5dc1a3c_2 + - google-auth=2.6.0=pyhd3eb1b0_0 + - google-auth-oauthlib=0.4.4=pyhd3eb1b0_0 + - grpc-cpp=1.48.2=hf108199_0 + - grpcio=1.48.2=py310hf108199_0 + - gst-plugins-base=1.18.5=h9e645db_0 + - gstreamer=1.18.5=hd78058f_0 + - icu=58.2=ha925a31_3 + - idna=3.4=py310haa95532_0 + - intel-openmp=2023.1.0=h59b6b97_46319 + - jpeg=9e=h2bbff1b_1 + - kiwisolver=1.4.4=py310hd77b12b_0 + - krb5=1.19.4=h5b6d351_0 + - lerc=3.0=hd77b12b_0 + - libbrotlicommon=1.0.9=h2bbff1b_7 + - libbrotlidec=1.0.9=h2bbff1b_7 + - libbrotlienc=1.0.9=h2bbff1b_7 + - libclang=14.0.6=default_hb5a9fac_1 + - libclang13=14.0.6=default_h8e68704_1 + - libdeflate=1.17=h2bbff1b_0 + - libffi=3.4.4=hd77b12b_0 + - libiconv=1.16=h2bbff1b_2 + - libogg=1.3.5=h2bbff1b_1 + - libpng=1.6.39=h8cc25b3_0 + - libprotobuf=3.20.3=h23ce68f_0 + - libtiff=4.5.0=h6c2663c_2 + - libuv=1.44.2=h2bbff1b_0 + - libvorbis=1.3.7=he774522_0 + - libwebp=1.2.4=hbc33d0d_1 + - libwebp-base=1.2.4=h2bbff1b_1 + - libxml2=2.10.3=h0ad7f3c_0 + - libxslt=1.1.37=h2bbff1b_0 + - lz4-c=1.9.4=h2bbff1b_0 + - markdown=3.4.1=py310haa95532_0 + - markupsafe=2.1.1=py310h2bbff1b_0 + - matplotlib=3.7.1=py310haa95532_1 + - matplotlib-base=3.7.1=py310h4ed8f06_1 + - mkl=2023.1.0=h8bd8f75_46356 + - mkl-service=2.4.0=py310h2bbff1b_1 + - mkl_fft=1.3.6=py310h4ed8f06_1 + - mkl_random=1.2.2=py310h4ed8f06_1 + - multidict=6.0.2=py310h2bbff1b_0 + - munkres=1.1.4=py_0 + - numexpr=2.8.4=py310h2cd9be0_1 + - numpy=1.24.3=py310h055cbcc_1 + - numpy-base=1.24.3=py310h65a83cf_1 + - oauthlib=3.2.2=py310haa95532_0 + - openssl=1.1.1t=h2bbff1b_0 + - packaging=23.0=py310haa95532_0 + - pandas=1.5.3=py310h4ed8f06_0 + - pcre=8.45=hd77b12b_0 + - pillow=9.4.0=py310hd77b12b_0 + - pip=23.0.1=py310haa95532_0 + - ply=3.11=py310haa95532_0 + - protobuf=3.20.3=py310hd77b12b_0 + - pyasn1=0.4.8=pyhd3eb1b0_0 + - pyasn1-modules=0.2.8=py_0 + - pycparser=2.21=pyhd3eb1b0_0 + - pyjwt=2.4.0=py310haa95532_0 + - pyopenssl=23.0.0=py310haa95532_0 + - pyparsing=3.0.9=py310haa95532_0 + - pyqt=5.15.7=py310hd77b12b_0 + - pyqt5-sip=12.11.0=py310hd77b12b_0 + - pysocks=1.7.1=py310haa95532_0 + - python=3.10.11=h966fe2a_2 + - python-dateutil=2.8.2=pyhd3eb1b0_0 + - pytorch-mutex=1.0=cpu + - pytz=2022.7=py310haa95532_0 + - pyyaml=6.0=py310h2bbff1b_1 + - qt-main=5.15.2=he8e5bd7_8 + - qt-webengine=5.15.9=hb9a9bb5_5 + - qtwebkit=5.212=h2bbfb41_5 + - re2=2022.04.01=hd77b12b_0 + - requests=2.29.0=py310haa95532_0 + - requests-oauthlib=1.3.0=py_0 + - rsa=4.7.2=pyhd3eb1b0_1 + - setuptools=67.8.0=py310haa95532_0 + - sip=6.6.2=py310hd77b12b_0 + - six=1.16.0=pyhd3eb1b0_1 + - sqlite=3.41.2=h2bbff1b_0 + - tbb=2021.8.0=h59b6b97_0 + - tensorboard=2.10.0=py310haa95532_0 + - tensorboard-data-server=0.6.1=py310haa95532_0 + - tensorboard-plugin-wit=1.8.1=py310haa95532_0 + - tk=8.6.12=h2bbff1b_0 + - toml=0.10.2=pyhd3eb1b0_0 + - tornado=6.2=py310h2bbff1b_0 + - tqdm=4.65.0=py310h9909e9c_0 + - typing_extensions=4.5.0=py310haa95532_0 + - tzdata=2023c=h04d1e81_0 + - urllib3=1.26.16=py310haa95532_0 + - vc=14.2=h21ff451_1 + - vs2015_runtime=14.27.29016=h5e58377_2 + - werkzeug=2.2.3=py310haa95532_0 + - wheel=0.38.4=py310haa95532_0 + - win_inet_pton=1.1.0=py310haa95532_0 + - xz=5.4.2=h8cc25b3_0 + - yaml=0.2.5=he774522_0 + - yarl=1.8.1=py310h2bbff1b_0 + - zlib=1.2.13=h8cc25b3_0 + - zstd=1.5.5=hd43e919_0 + - pip: + - antlr4-python3-runtime==4.8 + - appdirs==1.4.4 + - audioread==3.0.0 + - bitarray==2.7.4 + - cython==0.29.35 + - decorator==5.1.1 + - fairseq==0.12.2 + - faiss-cpu==1.7.4 + - filelock==3.12.0 + - hydra-core==1.0.7 + - jinja2==3.1.2 + - joblib==1.2.0 + - lazy-loader==0.2 + - librosa==0.10.0.post2 + - llvmlite==0.40.0 + - lxml==4.9.2 + - mpmath==1.3.0 + - msgpack==1.0.5 + - networkx==3.1 + - noisereduce==2.0.1 + - numba==0.57.0 + - omegaconf==2.0.6 + - opencv-python==4.7.0.72 + - pooch==1.6.0 + - portalocker==2.7.0 + - pysimplegui==4.60.5 + - pywin32==306 + - pyworld==0.3.3 + - regex==2023.5.5 + - sacrebleu==2.3.1 + - scikit-learn==1.2.2 + - scipy==1.10.1 + - sounddevice==0.4.6 + - soundfile==0.12.1 + - soxr==0.3.5 + - sympy==1.12 + - tabulate==0.9.0 + - threadpoolctl==3.1.0 + - torch==2.0.0 + - torch-directml==0.2.0.dev230426 + - torchaudio==2.0.1 + - torchvision==0.15.1 + - wget==3.2 +prefix: D:\ProgramData\anaconda3_\envs\pydml diff --git a/guidml.py b/guidml.py new file mode 100644 index 0000000..aadf22d --- /dev/null +++ b/guidml.py @@ -0,0 +1,710 @@ +""" +0416后的更新： + 引入config中half + 重建npy而不用填写 + v2支持 + 无f0模型支持 + 修复 + + int16： + 增加无索引支持 + f0算法改harvest(怎么看就只有这个会影响CPU占用)，但是不这么改效果不好 +""" +import os, sys, traceback, re + +import json + +now_dir = os.getcwd() +sys.path.append(now_dir) +from config import Config + +Config = Config() + +import torch_directml +import PySimpleGUI as sg +import sounddevice as sd +import noisereduce as nr +import numpy as np +from fairseq import checkpoint_utils +import librosa, torch, pyworld, faiss, time, threading +import torch.nn.functional as F +import torchaudio.transforms as tat +import scipy.signal as signal + + +# import matplotlib.pyplot as plt +from lib.infer_pack.models import ( + SynthesizerTrnMs256NSFsid, + SynthesizerTrnMs256NSFsid_nono, + SynthesizerTrnMs768NSFsid, + SynthesizerTrnMs768NSFsid_nono, +) +from i18n import I18nAuto + +i18n = I18nAuto() +device = torch_directml.device(torch_directml.default_device()) +current_dir = os.getcwd() + + +class RVC: + def __init__( + self, key, hubert_path, pth_path, index_path, npy_path, index_rate + ) -> None: + """ + 初始化 + """ + try: + self.f0_up_key = key + self.time_step = 160 / 16000 * 1000 + self.f0_min = 50 + self.f0_max = 1100 + self.f0_mel_min = 1127 * np.log(1 + self.f0_min / 700) + self.f0_mel_max = 1127 * np.log(1 + self.f0_max / 700) + self.sr = 16000 + self.window = 160 + if index_rate != 0: + self.index = faiss.read_index(index_path) + # self.big_npy = np.load(npy_path) + self.big_npy = self.index.reconstruct_n(0, self.index.ntotal) + print("index search enabled") + self.index_rate = index_rate + model_path = hubert_path + print("load model(s) from {}".format(model_path)) + models, saved_cfg, task = checkpoint_utils.load_model_ensemble_and_task( + [model_path], + suffix="", + ) + self.model = models[0] + self.model = self.model.to(device) + if Config.is_half: + self.model = self.model.half() + else: + self.model = self.model.float() + self.model.eval() + cpt = torch.load(pth_path, map_location="cpu") + self.tgt_sr = cpt["config"][-1] + cpt["config"][-3] = cpt["weight"]["emb_g.weight"].shape[0] # n_spk + self.if_f0 = cpt.get("f0", 1) + self.version = cpt.get("version", "v1") + if self.version == "v1": + if self.if_f0 == 1: + self.net_g = SynthesizerTrnMs256NSFsid( + *cpt["config"], is_half=Config.is_half + ) + else: + self.net_g = SynthesizerTrnMs256NSFsid_nono(*cpt["config"]) + elif self.version == "v2": + if self.if_f0 == 1: + self.net_g = SynthesizerTrnMs768NSFsid( + *cpt["config"], is_half=Config.is_half + ) + else: + self.net_g = SynthesizerTrnMs768NSFsid_nono(*cpt["config"]) + del self.net_g.enc_q + print(self.net_g.load_state_dict(cpt["weight"], strict=False)) + self.net_g.eval().to(device) + if Config.is_half: + self.net_g = self.net_g.half() + else: + self.net_g = self.net_g.float() + except: + print(traceback.format_exc()) + + def get_f0(self, x, f0_up_key, inp_f0=None): + x_pad = 1 + f0_min = 50 + f0_max = 1100 + f0_mel_min = 1127 * np.log(1 + f0_min / 700) + f0_mel_max = 1127 * np.log(1 + f0_max / 700) + f0, t = pyworld.harvest( + x.astype(np.double), + fs=self.sr, + f0_ceil=f0_max, + f0_floor=f0_min, + frame_period=10, + ) + f0 = pyworld.stonemask(x.astype(np.double), f0, t, self.sr) + f0 = signal.medfilt(f0, 3) + 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 + if inp_f0 is not None: + delta_t = np.round( + (inp_f0[:, 0].max() - inp_f0[:, 0].min()) * tf0 + 1 + ).astype("int16") + replace_f0 = np.interp( + list(range(delta_t)), inp_f0[:, 0] * 100, inp_f0[:, 1] + ) + shape = f0[x_pad * tf0 : x_pad * tf0 + len(replace_f0)].shape[0] + f0[x_pad * tf0 : x_pad * tf0 + len(replace_f0)] = replace_f0[:shape] + # with open("test_opt.txt","w")as f:f.write("\n".join([str(i)for i in f0.tolist()])) + f0bak = f0.copy() + f0_mel = 1127 * np.log(1 + f0 / 700) + f0_mel[f0_mel > 0] = (f0_mel[f0_mel > 0] - f0_mel_min) * 254 / ( + f0_mel_max - f0_mel_min + ) + 1 + f0_mel[f0_mel <= 1] = 1 + f0_mel[f0_mel > 255] = 255 + f0_coarse = np.rint(f0_mel).astype(np.int) + return f0_coarse, f0bak # 1-0 + + def infer(self, feats: torch.Tensor) -> np.ndarray: + """ + 推理函数 + """ + audio = feats.clone().cpu().numpy() + assert feats.dim() == 1, feats.dim() + feats = feats.view(1, -1) + padding_mask = torch.BoolTensor(feats.shape).fill_(False) + if Config.is_half: + feats = feats.half() + else: + feats = feats.float() + inputs = { + "source": feats.to(device), + "padding_mask": padding_mask.to(device), + "output_layer": 9 if self.version == "v1" else 12, + } + torch.cuda.synchronize() + with torch.no_grad(): + logits = self.model.extract_features(**inputs) + feats = ( + self.model.final_proj(logits[0]) if self.version == "v1" else logits[0] + ) + + ####索引优化 + try: + if ( + hasattr(self, "index") + and hasattr(self, "big_npy") + and self.index_rate != 0 + ): + npy = feats[0].cpu().numpy().astype("float32") + score, ix = self.index.search(npy, k=8) + weight = np.square(1 / score) + weight /= weight.sum(axis=1, keepdims=True) + npy = np.sum(self.big_npy[ix] * np.expand_dims(weight, axis=2), axis=1) + if Config.is_half: + npy = npy.astype("float16") + feats = ( + torch.from_numpy(npy).unsqueeze(0).to(device) * self.index_rate + + (1 - self.index_rate) * feats + ) + else: + print("index search FAIL or disabled") + except: + traceback.print_exc() + print("index search FAIL") + feats = F.interpolate(feats.permute(0, 2, 1), scale_factor=2).permute(0, 2, 1) + torch.cuda.synchronize() + print(feats.shape) + if self.if_f0 == 1: + pitch, pitchf = self.get_f0(audio, self.f0_up_key) + p_len = min(feats.shape[1], 13000, pitch.shape[0]) # 太大了爆显存 + else: + pitch, pitchf = None, None + p_len = min(feats.shape[1], 13000) # 太大了爆显存 + torch.cuda.synchronize() + # print(feats.shape,pitch.shape) + feats = feats[:, :p_len, :] + if self.if_f0 == 1: + pitch = pitch[:p_len] + pitchf = pitchf[:p_len] + pitch = torch.LongTensor(pitch).unsqueeze(0).to(device) + pitchf = torch.FloatTensor(pitchf).unsqueeze(0).to(device) + p_len = torch.LongTensor([p_len]).to(device) + ii = 0 # sid + sid = torch.LongTensor([ii]).to(device) + with torch.no_grad(): + if self.if_f0 == 1: + infered_audio = ( + self.net_g.infer(feats, p_len, pitch, pitchf, sid)[0][0, 0] + .data.cpu() + .float() + ) + else: + infered_audio = ( + self.net_g.infer(feats, p_len, sid)[0][0, 0].data.cpu().float() + ) + torch.cuda.synchronize() + return infered_audio + + +class GUIConfig: + def __init__(self) -> None: + self.hubert_path: str = "" + self.pth_path: str = "" + self.index_path: str = "" + self.npy_path: str = "" + self.pitch: int = 12 + self.samplerate: int = 44100 + self.block_time: float = 1.0 # s + self.buffer_num: int = 1 + self.threhold: int = -30 + self.crossfade_time: float = 0.08 + self.extra_time: float = 0.04 + self.I_noise_reduce = False + self.O_noise_reduce = False + self.index_rate = 0.3 + + +class GUI: + def __init__(self) -> None: + self.config = GUIConfig() + self.flag_vc = False + + self.launcher() + + def load(self): + ( + input_devices, + output_devices, + input_devices_indices, + output_devices_indices, + ) = self.get_devices() + try: + with open("values1.json", "r") as j: + data = json.load(j) + except: + with open("values1.json", "w") as j: + data = { + "pth_path": "", + "index_path": "", + "sg_input_device": input_devices[ + input_devices_indices.index(sd.default.device[0]) + ], + "sg_output_device": output_devices[ + output_devices_indices.index(sd.default.device[1]) + ], + "threhold": "-45", + "pitch": "0", + "index_rate": "0", + "block_time": "1", + "crossfade_length": "0.04", + "extra_time": "1", + } + return data + + def launcher(self): + data = self.load() + sg.theme("LightBlue3") + input_devices, output_devices, _, _ = self.get_devices() + layout = [ + [ + sg.Frame( + title=i18n("加载模型"), + layout=[ + [ + sg.Input( + default_text="hubert_base.pt", + key="hubert_path", + disabled=True, + ), + sg.FileBrowse( + i18n("Hubert模型"), + initial_folder=os.path.join(os.getcwd()), + file_types=(("pt files", "*.pt"),), + ), + ], + [ + sg.Input( + default_text=data.get("pth_path", ""), + key="pth_path", + ), + sg.FileBrowse( + i18n("选择.pth文件"), + initial_folder=os.path.join(os.getcwd(), "weights"), + file_types=(("weight files", "*.pth"),), + ), + ], + [ + sg.Input( + default_text=data.get("index_path", ""), + key="index_path", + ), + sg.FileBrowse( + i18n("选择.index文件"), + initial_folder=os.path.join(os.getcwd(), "logs"), + file_types=(("index files", "*.index"),), + ), + ], + [ + sg.Input( + default_text="你不需要填写这个You don't need write this.", + key="npy_path", + disabled=True, + ), + sg.FileBrowse( + i18n("选择.npy文件"), + initial_folder=os.path.join(os.getcwd(), "logs"), + file_types=(("feature files", "*.npy"),), + ), + ], + ], + ) + ], + [ + sg.Frame( + layout=[ + [ + sg.Text(i18n("输入设备")), + sg.Combo( + input_devices, + key="sg_input_device", + default_value=data.get("sg_input_device", ""), + ), + ], + [ + sg.Text(i18n("输出设备")), + sg.Combo( + output_devices, + key="sg_output_device", + default_value=data.get("sg_output_device", ""), + ), + ], + ], + title=i18n("音频设备(请使用同种类驱动)"), + ) + ], + [ + sg.Frame( + layout=[ + [ + sg.Text(i18n("响应阈值")), + sg.Slider( + range=(-60, 0), + key="threhold", + resolution=1, + orientation="h", + default_value=data.get("threhold", ""), + ), + ], + [ + sg.Text(i18n("音调设置")), + sg.Slider( + range=(-24, 24), + key="pitch", + resolution=1, + orientation="h", + default_value=data.get("pitch", ""), + ), + ], + [ + sg.Text(i18n("Index Rate")), + sg.Slider( + range=(0.0, 1.0), + key="index_rate", + resolution=0.01, + orientation="h", + default_value=data.get("index_rate", ""), + ), + ], + ], + title=i18n("常规设置"), + ), + sg.Frame( + layout=[ + [ + sg.Text(i18n("采样长度")), + sg.Slider( + range=(0.1, 3.0), + key="block_time", + resolution=0.1, + orientation="h", + default_value=data.get("block_time", ""), + ), + ], + [ + sg.Text(i18n("淡入淡出长度")), + sg.Slider( + range=(0.01, 0.15), + key="crossfade_length", + resolution=0.01, + orientation="h", + default_value=data.get("crossfade_length", ""), + ), + ], + [ + sg.Text(i18n("额外推理时长")), + sg.Slider( + range=(0.05, 3.00), + key="extra_time", + resolution=0.01, + orientation="h", + default_value=data.get("extra_time", ""), + ), + ], + [ + sg.Checkbox(i18n("输入降噪"), key="I_noise_reduce"), + sg.Checkbox(i18n("输出降噪"), key="O_noise_reduce"), + ], + ], + title=i18n("性能设置"), + ), + ], + [ + sg.Button(i18n("开始音频转换"), key="start_vc"), + sg.Button(i18n("停止音频转换"), key="stop_vc"), + sg.Text(i18n("推理时间(ms):")), + sg.Text("0", key="infer_time"), + ], + ] + self.window = sg.Window("RVC - GUI", layout=layout) + self.event_handler() + + def event_handler(self): + while True: + event, values = self.window.read() + if event == sg.WINDOW_CLOSED: + self.flag_vc = False + exit() + if event == "start_vc" and self.flag_vc == False: + if self.set_values(values) == True: + print("using_cuda:" + str(torch.cuda.is_available())) + self.start_vc() + settings = { + "pth_path": values["pth_path"], + "index_path": values["index_path"], + "sg_input_device": values["sg_input_device"], + "sg_output_device": values["sg_output_device"], + "threhold": values["threhold"], + "pitch": values["pitch"], + "index_rate": values["index_rate"], + "block_time": values["block_time"], + "crossfade_length": values["crossfade_length"], + "extra_time": values["extra_time"], + } + with open("values1.json", "w") as j: + json.dump(settings, j) + if event == "stop_vc" and self.flag_vc == True: + self.flag_vc = False + + def set_values(self, values): + if len(values["pth_path"].strip()) == 0: + sg.popup(i18n("请选择pth文件")) + return False + if len(values["index_path"].strip()) == 0: + sg.popup(i18n("请选择index文件")) + return False + pattern = re.compile("[^\x00-\x7F]+") + if pattern.findall(values["hubert_path"]): + sg.popup(i18n("hubert模型路径不可包含中文")) + return False + if pattern.findall(values["pth_path"]): + sg.popup(i18n("pth文件路径不可包含中文")) + return False + if pattern.findall(values["index_path"]): + sg.popup(i18n("index文件路径不可包含中文")) + return False + self.set_devices(values["sg_input_device"], values["sg_output_device"]) + self.config.hubert_path = os.path.join(current_dir, "hubert_base.pt") + self.config.pth_path = values["pth_path"] + self.config.index_path = values["index_path"] + self.config.npy_path = values["npy_path"] + self.config.threhold = values["threhold"] + self.config.pitch = values["pitch"] + self.config.block_time = values["block_time"] + self.config.crossfade_time = values["crossfade_length"] + self.config.extra_time = values["extra_time"] + self.config.I_noise_reduce = values["I_noise_reduce"] + self.config.O_noise_reduce = values["O_noise_reduce"] + self.config.index_rate = values["index_rate"] + return True + + def start_vc(self): + torch.cuda.empty_cache() + self.flag_vc = True + self.block_frame = int(self.config.block_time * self.config.samplerate) + self.crossfade_frame = int(self.config.crossfade_time * self.config.samplerate) + self.sola_search_frame = int(0.012 * self.config.samplerate) + self.delay_frame = int(0.01 * self.config.samplerate) # 往前预留0.02s + self.extra_frame = int(self.config.extra_time * self.config.samplerate) + self.rvc = None + self.rvc = RVC( + self.config.pitch, + self.config.hubert_path, + self.config.pth_path, + self.config.index_path, + self.config.npy_path, + self.config.index_rate, + ) + self.input_wav: np.ndarray = np.zeros( + self.extra_frame + + self.crossfade_frame + + self.sola_search_frame + + self.block_frame, + dtype="float32", + ) + self.output_wav: torch.Tensor = torch.zeros( + self.block_frame, device=device, dtype=torch.float32 + ) + self.sola_buffer: torch.Tensor = torch.zeros( + self.crossfade_frame, device=device, dtype=torch.float32 + ) + self.fade_in_window: torch.Tensor = torch.linspace( + 0.0, 1.0, steps=self.crossfade_frame, device=device, dtype=torch.float32 + ) + self.fade_out_window: torch.Tensor = 1 - self.fade_in_window + self.resampler1 = tat.Resample( + orig_freq=self.config.samplerate, new_freq=16000, dtype=torch.float32 + ) + self.resampler2 = tat.Resample( + orig_freq=self.rvc.tgt_sr, + new_freq=self.config.samplerate, + dtype=torch.float32, + ) + thread_vc = threading.Thread(target=self.soundinput) + thread_vc.start() + + def soundinput(self): + """ + 接受音频输入 + """ + with sd.Stream( + channels=2, + callback=self.audio_callback, + blocksize=self.block_frame, + samplerate=self.config.samplerate, + dtype="float32", + ): + while self.flag_vc: + time.sleep(self.config.block_time) + print("Audio block passed.") + print("ENDing VC") + + def audio_callback( + self, indata: np.ndarray, outdata: np.ndarray, frames, times, status + ): + """ + 音频处理 + """ + start_time = time.perf_counter() + indata = librosa.to_mono(indata.T) + if self.config.I_noise_reduce: + indata[:] = nr.reduce_noise(y=indata, sr=self.config.samplerate) + + """noise gate""" + frame_length = 2048 + hop_length = 1024 + rms = librosa.feature.rms( + y=indata, frame_length=frame_length, hop_length=hop_length + ) + db_threhold = librosa.amplitude_to_db(rms, ref=1.0)[0] < self.config.threhold + # print(rms.shape,db.shape,db) + for i in range(db_threhold.shape[0]): + if db_threhold[i]: + indata[i * hop_length : (i + 1) * hop_length] = 0 + self.input_wav[:] = np.append(self.input_wav[self.block_frame :], indata) + + # infer + print("input_wav:" + str(self.input_wav.shape)) + # print('infered_wav:'+str(infer_wav.shape)) + infer_wav: torch.Tensor = self.resampler2( + self.rvc.infer(self.resampler1(torch.from_numpy(self.input_wav))) + )[-self.crossfade_frame - self.sola_search_frame - self.block_frame :].to( + device + ) + print("infer_wav:" + str(infer_wav.shape)) + + # SOLA algorithm from https://github.com/yxlllc/DDSP-SVC + cor_nom = F.conv1d( + infer_wav[None, None, : self.crossfade_frame + self.sola_search_frame], + self.sola_buffer[None, None, :], + ) + cor_den = torch.sqrt( + F.conv1d( + infer_wav[None, None, : self.crossfade_frame + self.sola_search_frame] + ** 2, + torch.ones(1, 1, self.crossfade_frame, device=device), + ) + + 1e-8 + ) + sola_offset = torch.argmax(cor_nom[0, 0] / cor_den[0, 0]) + print("sola offset: " + str(int(sola_offset))) + + # crossfade + self.output_wav[:] = infer_wav[sola_offset : sola_offset + self.block_frame] + self.output_wav[: self.crossfade_frame] *= self.fade_in_window + self.output_wav[: self.crossfade_frame] += self.sola_buffer[:] + if sola_offset < self.sola_search_frame: + self.sola_buffer[:] = ( + infer_wav[ + -self.sola_search_frame + - self.crossfade_frame + + sola_offset : -self.sola_search_frame + + sola_offset + ] + * self.fade_out_window + ) + else: + self.sola_buffer[:] = ( + infer_wav[-self.crossfade_frame :] * self.fade_out_window + ) + + if self.config.O_noise_reduce: + outdata[:] = np.tile( + nr.reduce_noise( + y=self.output_wav[:].cpu().numpy(), sr=self.config.samplerate + ), + (2, 1), + ).T + else: + outdata[:] = self.output_wav[:].repeat(2, 1).t().cpu().numpy() + total_time = time.perf_counter() - start_time + self.window["infer_time"].update(int(total_time * 1000)) + print("infer time:" + str(total_time)) + + def get_devices(self, update: bool = True): + """获取设备列表""" + if update: + sd._terminate() + sd._initialize() + devices = sd.query_devices() + hostapis = sd.query_hostapis() + for hostapi in hostapis: + for device_idx in hostapi["devices"]: + devices[device_idx]["hostapi_name"] = hostapi["name"] + input_devices = [ + f"{d['name']} ({d['hostapi_name']})" + for d in devices + if d["max_input_channels"] > 0 + ] + output_devices = [ + f"{d['name']} ({d['hostapi_name']})" + for d in devices + if d["max_output_channels"] > 0 + ] + input_devices_indices = [ + d["index"] if "index" in d else d["name"] + for d in devices + if d["max_input_channels"] > 0 + ] + output_devices_indices = [ + d["index"] if "index" in d else d["name"] + for d in devices + if d["max_output_channels"] > 0 + ] + return ( + input_devices, + output_devices, + input_devices_indices, + output_devices_indices, + ) + + def set_devices(self, input_device, output_device): + """设置输出设备""" + ( + input_devices, + output_devices, + input_device_indices, + output_device_indices, + ) = self.get_devices() + sd.default.device[0] = input_device_indices[input_devices.index(input_device)] + sd.default.device[1] = output_device_indices[ + output_devices.index(output_device) + ] + print("input device:" + str(sd.default.device[0]) + ":" + str(input_device)) + print("output device:" + str(sd.default.device[1]) + ":" + str(output_device)) + + +gui = GUI() diff --git a/lib/infer_pack/models_dml.py b/lib/infer_pack/models_dml.py new file mode 100644 index 0000000..958d7b2 --- /dev/null +++ b/lib/infer_pack/models_dml.py @@ -0,0 +1,1124 @@ +import math, pdb, os +from time import time as ttime +import torch +from torch import nn +from torch.nn import functional as F +from lib.infer_pack import modules +from lib.infer_pack import attentions +from lib.infer_pack import commons +from lib.infer_pack.commons import init_weights, get_padding +from torch.nn import Conv1d, ConvTranspose1d, AvgPool1d, Conv2d +from torch.nn.utils import weight_norm, remove_weight_norm, spectral_norm +from lib.infer_pack.commons import init_weights +import numpy as np +from lib.infer_pack import commons + + +class TextEncoder256(nn.Module): + def __init__( + self, + out_channels, + hidden_channels, + filter_channels, + n_heads, + n_layers, + kernel_size, + p_dropout, + f0=True, + ): + super().__init__() + self.out_channels = out_channels + self.hidden_channels = hidden_channels + self.filter_channels = filter_channels + self.n_heads = n_heads + self.n_layers = n_layers + self.kernel_size = kernel_size + self.p_dropout = p_dropout + self.emb_phone = nn.Linear(256, hidden_channels) + self.lrelu = nn.LeakyReLU(0.1, inplace=True) + if f0 == True: + self.emb_pitch = nn.Embedding(256, hidden_channels) # pitch 256 + self.encoder = attentions.Encoder( + hidden_channels, filter_channels, n_heads, n_layers, kernel_size, p_dropout + ) + self.proj = nn.Conv1d(hidden_channels, out_channels * 2, 1) + + def forward(self, phone, pitch, lengths): + if pitch == None: + x = self.emb_phone(phone) + else: + x = self.emb_phone(phone) + self.emb_pitch(pitch) + x = x * math.sqrt(self.hidden_channels) # [b, t, h] + x = self.lrelu(x) + x = torch.transpose(x, 1, -1) # [b, h, t] + x_mask = torch.unsqueeze(commons.sequence_mask(lengths, x.size(2)), 1).to( + x.dtype + ) + x = self.encoder(x * x_mask, x_mask) + stats = self.proj(x) * x_mask + + m, logs = torch.split(stats, self.out_channels, dim=1) + return m, logs, x_mask + + +class TextEncoder768(nn.Module): + def __init__( + self, + out_channels, + hidden_channels, + filter_channels, + n_heads, + n_layers, + kernel_size, + p_dropout, + f0=True, + ): + super().__init__() + self.out_channels = out_channels + self.hidden_channels = hidden_channels + self.filter_channels = filter_channels + self.n_heads = n_heads + self.n_layers = n_layers + self.kernel_size = kernel_size + self.p_dropout = p_dropout + self.emb_phone = nn.Linear(768, hidden_channels) + self.lrelu = nn.LeakyReLU(0.1, inplace=True) + if f0 == True: + self.emb_pitch = nn.Embedding(256, hidden_channels) # pitch 256 + self.encoder = attentions.Encoder( + hidden_channels, filter_channels, n_heads, n_layers, kernel_size, p_dropout + ) + self.proj = nn.Conv1d(hidden_channels, out_channels * 2, 1) + + def forward(self, phone, pitch, lengths): + if pitch == None: + x = self.emb_phone(phone) + else: + x = self.emb_phone(phone) + self.emb_pitch(pitch) + x = x * math.sqrt(self.hidden_channels) # [b, t, h] + x = self.lrelu(x) + x = torch.transpose(x, 1, -1) # [b, h, t] + x_mask = torch.unsqueeze(commons.sequence_mask(lengths, x.size(2)), 1).to( + x.dtype + ) + x = self.encoder(x * x_mask, x_mask) + stats = self.proj(x) * x_mask + + m, logs = torch.split(stats, self.out_channels, dim=1) + return m, logs, x_mask + + +class ResidualCouplingBlock(nn.Module): + def __init__( + self, + channels, + hidden_channels, + kernel_size, + dilation_rate, + n_layers, + n_flows=4, + gin_channels=0, + ): + super().__init__() + self.channels = channels + self.hidden_channels = hidden_channels + self.kernel_size = kernel_size + self.dilation_rate = dilation_rate + self.n_layers = n_layers + self.n_flows = n_flows + self.gin_channels = gin_channels + + self.flows = nn.ModuleList() + for i in range(n_flows): + self.flows.append( + modules.ResidualCouplingLayer( + channels, + hidden_channels, + kernel_size, + dilation_rate, + n_layers, + gin_channels=gin_channels, + mean_only=True, + ) + ) + self.flows.append(modules.Flip()) + + def forward(self, x, x_mask, g=None, reverse=False): + if not reverse: + for flow in self.flows: + x, _ = flow(x, x_mask, g=g, reverse=reverse) + else: + for flow in reversed(self.flows): + x = flow(x, x_mask, g=g, reverse=reverse) + return x + + def remove_weight_norm(self): + for i in range(self.n_flows): + self.flows[i * 2].remove_weight_norm() + + +class PosteriorEncoder(nn.Module): + def __init__( + self, + in_channels, + out_channels, + hidden_channels, + kernel_size, + dilation_rate, + n_layers, + gin_channels=0, + ): + super().__init__() + self.in_channels = in_channels + self.out_channels = out_channels + self.hidden_channels = hidden_channels + self.kernel_size = kernel_size + self.dilation_rate = dilation_rate + self.n_layers = n_layers + self.gin_channels = gin_channels + + self.pre = nn.Conv1d(in_channels, hidden_channels, 1) + self.enc = modules.WN( + hidden_channels, + kernel_size, + dilation_rate, + n_layers, + gin_channels=gin_channels, + ) + self.proj = nn.Conv1d(hidden_channels, out_channels * 2, 1) + + def forward(self, x, x_lengths, g=None): + x_mask = torch.unsqueeze(commons.sequence_mask(x_lengths, x.size(2)), 1).to( + x.dtype + ) + x = self.pre(x) * x_mask + x = self.enc(x, x_mask, g=g) + stats = self.proj(x) * x_mask + m, logs = torch.split(stats, self.out_channels, dim=1) + z = (m + torch.randn_like(m) * torch.exp(logs)) * x_mask + return z, m, logs, x_mask + + def remove_weight_norm(self): + self.enc.remove_weight_norm() + + +class Generator(torch.nn.Module): + def __init__( + self, + initial_channel, + resblock, + resblock_kernel_sizes, + resblock_dilation_sizes, + upsample_rates, + upsample_initial_channel, + upsample_kernel_sizes, + gin_channels=0, + ): + super(Generator, self).__init__() + self.num_kernels = len(resblock_kernel_sizes) + self.num_upsamples = len(upsample_rates) + self.conv_pre = Conv1d( + initial_channel, upsample_initial_channel, 7, 1, padding=3 + ) + resblock = modules.ResBlock1 if resblock == "1" else modules.ResBlock2 + + self.ups = nn.ModuleList() + for i, (u, k) in enumerate(zip(upsample_rates, upsample_kernel_sizes)): + self.ups.append( + weight_norm( + ConvTranspose1d( + upsample_initial_channel // (2**i), + upsample_initial_channel // (2 ** (i + 1)), + k, + u, + padding=(k - u) // 2, + ) + ) + ) + + self.resblocks = nn.ModuleList() + for i in range(len(self.ups)): + ch = upsample_initial_channel // (2 ** (i + 1)) + for j, (k, d) in enumerate( + zip(resblock_kernel_sizes, resblock_dilation_sizes) + ): + self.resblocks.append(resblock(ch, k, d)) + + self.conv_post = Conv1d(ch, 1, 7, 1, padding=3, bias=False) + self.ups.apply(init_weights) + + if gin_channels != 0: + self.cond = nn.Conv1d(gin_channels, upsample_initial_channel, 1) + + def forward(self, x, g=None): + x = self.conv_pre(x) + if g is not None: + x = x + self.cond(g) + + for i in range(self.num_upsamples): + x = F.leaky_relu(x, modules.LRELU_SLOPE) + x = self.ups[i](x) + xs = None + for j in range(self.num_kernels): + if xs is None: + xs = self.resblocks[i * self.num_kernels + j](x) + else: + xs += self.resblocks[i * self.num_kernels + j](x) + x = xs / self.num_kernels + x = F.leaky_relu(x) + x = self.conv_post(x) + x = torch.tanh(x) + + return x + + def remove_weight_norm(self): + for l in self.ups: + remove_weight_norm(l) + for l in self.resblocks: + l.remove_weight_norm() + + +class SineGen(torch.nn.Module): + """Definition of sine generator + SineGen(samp_rate, harmonic_num = 0, + sine_amp = 0.1, noise_std = 0.003, + voiced_threshold = 0, + flag_for_pulse=False) + samp_rate: sampling rate in Hz + harmonic_num: number of harmonic overtones (default 0) + sine_amp: amplitude of sine-wavefrom (default 0.1) + noise_std: std of Gaussian noise (default 0.003) + voiced_thoreshold: F0 threshold for U/V classification (default 0) + flag_for_pulse: this SinGen is used inside PulseGen (default False) + Note: when flag_for_pulse is True, the first time step of a voiced + segment is always sin(np.pi) or cos(0) + """ + + def __init__( + self, + samp_rate, + harmonic_num=0, + sine_amp=0.1, + noise_std=0.003, + voiced_threshold=0, + flag_for_pulse=False, + ): + super(SineGen, self).__init__() + self.sine_amp = sine_amp + self.noise_std = noise_std + self.harmonic_num = harmonic_num + self.dim = self.harmonic_num + 1 + self.sampling_rate = samp_rate + self.voiced_threshold = voiced_threshold + + def _f02uv(self, f0): + # generate uv signal + uv = torch.ones_like(f0) + uv = uv * (f0 > self.voiced_threshold) + return uv.float() + + def forward(self, f0, upp): + """sine_tensor, uv = forward(f0) + input F0: tensor(batchsize=1, length, dim=1) + f0 for unvoiced steps should be 0 + output sine_tensor: tensor(batchsize=1, length, dim) + output uv: tensor(batchsize=1, length, 1) + """ + with torch.no_grad(): + f0 = f0[:, None].transpose(1, 2) + f0_buf = torch.zeros(f0.shape[0], f0.shape[1], self.dim, device=f0.device) + # fundamental component + f0_buf[:, :, 0] = f0[:, :, 0] + for idx in np.arange(self.harmonic_num): + f0_buf[:, :, idx + 1] = f0_buf[:, :, 0] * ( + idx + 2 + ) # idx + 2: the (idx+1)-th overtone, (idx+2)-th harmonic + rad_values = (f0_buf / self.sampling_rate) % 1 ###%1意味着n_har的乘积无法后处理优化 + rand_ini = torch.rand( + f0_buf.shape[0], f0_buf.shape[2], device=f0_buf.device + ) + rand_ini[:, 0] = 0 + rad_values[:, 0, :] = rad_values[:, 0, :] + rand_ini + tmp_over_one = torch.cumsum(rad_values, 1) # % 1 #####%1意味着后面的cumsum无法再优化 + tmp_over_one *= upp + tmp_over_one = F.interpolate( + tmp_over_one.transpose(2, 1), + scale_factor=upp, + mode="linear", + align_corners=True, + ).transpose(2, 1) + rad_values = F.interpolate( + rad_values.transpose(2, 1), scale_factor=upp, mode="nearest" + ).transpose( + 2, 1 + ) ####### + tmp_over_one %= 1 + tmp_over_one_idx = (tmp_over_one[:, 1:, :] - tmp_over_one[:, :-1, :]) < 0 + cumsum_shift = torch.zeros_like(rad_values) + cumsum_shift[:, 1:, :] = tmp_over_one_idx * -1.0 + sine_waves = torch.sin( + torch.cumsum(rad_values + cumsum_shift, dim=1) * 2 * np.pi + ) + sine_waves = sine_waves * self.sine_amp + uv = self._f02uv(f0) + uv = F.interpolate( + uv.transpose(2, 1), scale_factor=upp, mode="nearest" + ).transpose(2, 1) + noise_amp = uv * self.noise_std + (1 - uv) * self.sine_amp / 3 + noise = noise_amp * torch.randn_like(sine_waves) + sine_waves = sine_waves * uv + noise + return sine_waves, uv, noise + + +class SourceModuleHnNSF(torch.nn.Module): + """SourceModule for hn-nsf + SourceModule(sampling_rate, harmonic_num=0, sine_amp=0.1, + add_noise_std=0.003, voiced_threshod=0) + sampling_rate: sampling_rate in Hz + harmonic_num: number of harmonic above F0 (default: 0) + sine_amp: amplitude of sine source signal (default: 0.1) + add_noise_std: std of additive Gaussian noise (default: 0.003) + note that amplitude of noise in unvoiced is decided + by sine_amp + voiced_threshold: threhold to set U/V given F0 (default: 0) + Sine_source, noise_source = SourceModuleHnNSF(F0_sampled) + F0_sampled (batchsize, length, 1) + Sine_source (batchsize, length, 1) + noise_source (batchsize, length 1) + uv (batchsize, length, 1) + """ + + def __init__( + self, + sampling_rate, + harmonic_num=0, + sine_amp=0.1, + add_noise_std=0.003, + voiced_threshod=0, + is_half=True, + ): + super(SourceModuleHnNSF, self).__init__() + + self.sine_amp = sine_amp + self.noise_std = add_noise_std + self.is_half = is_half + # to produce sine waveforms + self.l_sin_gen = SineGen( + sampling_rate, harmonic_num, sine_amp, add_noise_std, voiced_threshod + ) + + # to merge source harmonics into a single excitation + self.l_linear = torch.nn.Linear(harmonic_num + 1, 1) + self.l_tanh = torch.nn.Tanh() + + def forward(self, x, upp=None): + sine_wavs, uv, _ = self.l_sin_gen(x, upp) + if self.is_half: + sine_wavs = sine_wavs.half() + sine_merge = self.l_tanh(self.l_linear(sine_wavs)) + return sine_merge, None, None # noise, uv + + +class GeneratorNSF(torch.nn.Module): + def __init__( + self, + initial_channel, + resblock, + resblock_kernel_sizes, + resblock_dilation_sizes, + upsample_rates, + upsample_initial_channel, + upsample_kernel_sizes, + gin_channels, + sr, + is_half=False, + ): + super(GeneratorNSF, self).__init__() + self.num_kernels = len(resblock_kernel_sizes) + self.num_upsamples = len(upsample_rates) + + self.f0_upsamp = torch.nn.Upsample(scale_factor=np.prod(upsample_rates)) + self.m_source = SourceModuleHnNSF( + sampling_rate=sr, harmonic_num=0, is_half=is_half + ) + self.noise_convs = nn.ModuleList() + self.conv_pre = Conv1d( + initial_channel, upsample_initial_channel, 7, 1, padding=3 + ) + resblock = modules.ResBlock1 if resblock == "1" else modules.ResBlock2 + + self.ups = nn.ModuleList() + for i, (u, k) in enumerate(zip(upsample_rates, upsample_kernel_sizes)): + c_cur = upsample_initial_channel // (2 ** (i + 1)) + self.ups.append( + weight_norm( + ConvTranspose1d( + upsample_initial_channel // (2**i), + upsample_initial_channel // (2 ** (i + 1)), + k, + u, + padding=(k - u) // 2, + ) + ) + ) + if i + 1 < len(upsample_rates): + stride_f0 = np.prod(upsample_rates[i + 1 :]) + self.noise_convs.append( + Conv1d( + 1, + c_cur, + kernel_size=stride_f0 * 2, + stride=stride_f0, + padding=stride_f0 // 2, + ) + ) + else: + self.noise_convs.append(Conv1d(1, c_cur, kernel_size=1)) + + self.resblocks = nn.ModuleList() + for i in range(len(self.ups)): + ch = upsample_initial_channel // (2 ** (i + 1)) + for j, (k, d) in enumerate( + zip(resblock_kernel_sizes, resblock_dilation_sizes) + ): + self.resblocks.append(resblock(ch, k, d)) + + self.conv_post = Conv1d(ch, 1, 7, 1, padding=3, bias=False) + self.ups.apply(init_weights) + + if gin_channels != 0: + self.cond = nn.Conv1d(gin_channels, upsample_initial_channel, 1) + + self.upp = np.prod(upsample_rates) + + def forward(self, x, f0, g=None): + har_source, noi_source, uv = self.m_source(f0, self.upp) + har_source = har_source.transpose(1, 2) + x = self.conv_pre(x) + if g is not None: + x = x + self.cond(g) + + for i in range(self.num_upsamples): + x = F.leaky_relu(x, modules.LRELU_SLOPE) + x = self.ups[i](x) + x_source = self.noise_convs[i](har_source) + x = x + x_source + xs = None + for j in range(self.num_kernels): + if xs is None: + xs = self.resblocks[i * self.num_kernels + j](x) + else: + xs += self.resblocks[i * self.num_kernels + j](x) + x = xs / self.num_kernels + x = F.leaky_relu(x) + x = self.conv_post(x) + x = torch.tanh(x) + return x + + def remove_weight_norm(self): + for l in self.ups: + remove_weight_norm(l) + for l in self.resblocks: + l.remove_weight_norm() + + +sr2sr = { + "32k": 32000, + "40k": 40000, + "48k": 48000, +} + + +class SynthesizerTrnMs256NSFsid(nn.Module): + def __init__( + self, + spec_channels, + segment_size, + inter_channels, + hidden_channels, + filter_channels, + n_heads, + n_layers, + kernel_size, + p_dropout, + resblock, + resblock_kernel_sizes, + resblock_dilation_sizes, + upsample_rates, + upsample_initial_channel, + upsample_kernel_sizes, + spk_embed_dim, + gin_channels, + sr, + **kwargs + ): + super().__init__() + if type(sr) == type("strr"): + sr = sr2sr[sr] + self.spec_channels = spec_channels + self.inter_channels = inter_channels + self.hidden_channels = hidden_channels + self.filter_channels = filter_channels + self.n_heads = n_heads + self.n_layers = n_layers + self.kernel_size = kernel_size + self.p_dropout = p_dropout + self.resblock = resblock + self.resblock_kernel_sizes = resblock_kernel_sizes + self.resblock_dilation_sizes = resblock_dilation_sizes + self.upsample_rates = upsample_rates + self.upsample_initial_channel = upsample_initial_channel + self.upsample_kernel_sizes = upsample_kernel_sizes + self.segment_size = segment_size + self.gin_channels = gin_channels + # self.hop_length = hop_length# + self.spk_embed_dim = spk_embed_dim + self.enc_p = TextEncoder256( + inter_channels, + hidden_channels, + filter_channels, + n_heads, + n_layers, + kernel_size, + p_dropout, + ) + self.dec = GeneratorNSF( + inter_channels, + resblock, + resblock_kernel_sizes, + resblock_dilation_sizes, + upsample_rates, + upsample_initial_channel, + upsample_kernel_sizes, + gin_channels=gin_channels, + sr=sr, + is_half=kwargs["is_half"], + ) + self.enc_q = PosteriorEncoder( + spec_channels, + inter_channels, + hidden_channels, + 5, + 1, + 16, + gin_channels=gin_channels, + ) + self.flow = ResidualCouplingBlock( + inter_channels, hidden_channels, 5, 1, 3, gin_channels=gin_channels + ) + self.emb_g = nn.Embedding(self.spk_embed_dim, gin_channels) + print("gin_channels:", gin_channels, "self.spk_embed_dim:", self.spk_embed_dim) + + def remove_weight_norm(self): + self.dec.remove_weight_norm() + self.flow.remove_weight_norm() + self.enc_q.remove_weight_norm() + + def forward( + self, phone, phone_lengths, pitch, pitchf, y, y_lengths, ds + ): # 这里ds是id，[bs,1] + # print(1,pitch.shape)#[bs,t] + g = self.emb_g(ds).unsqueeze(-1) # [b, 256, 1]##1是t，广播的 + m_p, logs_p, x_mask = self.enc_p(phone, pitch, phone_lengths) + z, m_q, logs_q, y_mask = self.enc_q(y, y_lengths, g=g) + z_p = self.flow(z, y_mask, g=g) + z_slice, ids_slice = commons.rand_slice_segments( + z, y_lengths, self.segment_size + ) + # print(-1,pitchf.shape,ids_slice,self.segment_size,self.hop_length,self.segment_size//self.hop_length) + pitchf = commons.slice_segments2(pitchf, ids_slice, self.segment_size) + # print(-2,pitchf.shape,z_slice.shape) + o = self.dec(z_slice, pitchf, g=g) + return o, ids_slice, x_mask, y_mask, (z, z_p, m_p, logs_p, m_q, logs_q) + + def infer(self, phone, phone_lengths, pitch, nsff0, sid, max_len=None): + g = self.emb_g(sid).unsqueeze(-1) + m_p, logs_p, x_mask = self.enc_p(phone, pitch, phone_lengths) + z_p = (m_p + torch.exp(logs_p) * torch.randn_like(m_p) * 0.66666) * x_mask + z = self.flow(z_p, x_mask, g=g, reverse=True) + o = self.dec((z * x_mask)[:, :, :max_len], nsff0, g=g) + return o, x_mask, (z, z_p, m_p, logs_p) + + +class SynthesizerTrnMs768NSFsid(nn.Module): + def __init__( + self, + spec_channels, + segment_size, + inter_channels, + hidden_channels, + filter_channels, + n_heads, + n_layers, + kernel_size, + p_dropout, + resblock, + resblock_kernel_sizes, + resblock_dilation_sizes, + upsample_rates, + upsample_initial_channel, + upsample_kernel_sizes, + spk_embed_dim, + gin_channels, + sr, + **kwargs + ): + super().__init__() + if type(sr) == type("strr"): + sr = sr2sr[sr] + self.spec_channels = spec_channels + self.inter_channels = inter_channels + self.hidden_channels = hidden_channels + self.filter_channels = filter_channels + self.n_heads = n_heads + self.n_layers = n_layers + self.kernel_size = kernel_size + self.p_dropout = p_dropout + self.resblock = resblock + self.resblock_kernel_sizes = resblock_kernel_sizes + self.resblock_dilation_sizes = resblock_dilation_sizes + self.upsample_rates = upsample_rates + self.upsample_initial_channel = upsample_initial_channel + self.upsample_kernel_sizes = upsample_kernel_sizes + self.segment_size = segment_size + self.gin_channels = gin_channels + # self.hop_length = hop_length# + self.spk_embed_dim = spk_embed_dim + self.enc_p = TextEncoder768( + inter_channels, + hidden_channels, + filter_channels, + n_heads, + n_layers, + kernel_size, + p_dropout, + ) + self.dec = GeneratorNSF( + inter_channels, + resblock, + resblock_kernel_sizes, + resblock_dilation_sizes, + upsample_rates, + upsample_initial_channel, + upsample_kernel_sizes, + gin_channels=gin_channels, + sr=sr, + is_half=kwargs["is_half"], + ) + self.enc_q = PosteriorEncoder( + spec_channels, + inter_channels, + hidden_channels, + 5, + 1, + 16, + gin_channels=gin_channels, + ) + self.flow = ResidualCouplingBlock( + inter_channels, hidden_channels, 5, 1, 3, gin_channels=gin_channels + ) + self.emb_g = nn.Embedding(self.spk_embed_dim, gin_channels) + print("gin_channels:", gin_channels, "self.spk_embed_dim:", self.spk_embed_dim) + + def remove_weight_norm(self): + self.dec.remove_weight_norm() + self.flow.remove_weight_norm() + self.enc_q.remove_weight_norm() + + def forward( + self, phone, phone_lengths, pitch, pitchf, y, y_lengths, ds + ): # 这里ds是id，[bs,1] + # print(1,pitch.shape)#[bs,t] + g = self.emb_g(ds).unsqueeze(-1) # [b, 256, 1]##1是t，广播的 + m_p, logs_p, x_mask = self.enc_p(phone, pitch, phone_lengths) + z, m_q, logs_q, y_mask = self.enc_q(y, y_lengths, g=g) + z_p = self.flow(z, y_mask, g=g) + z_slice, ids_slice = commons.rand_slice_segments( + z, y_lengths, self.segment_size + ) + # print(-1,pitchf.shape,ids_slice,self.segment_size,self.hop_length,self.segment_size//self.hop_length) + pitchf = commons.slice_segments2(pitchf, ids_slice, self.segment_size) + # print(-2,pitchf.shape,z_slice.shape) + o = self.dec(z_slice, pitchf, g=g) + return o, ids_slice, x_mask, y_mask, (z, z_p, m_p, logs_p, m_q, logs_q) + + def infer(self, phone, phone_lengths, pitch, nsff0, sid, max_len=None): + g = self.emb_g(sid).unsqueeze(-1) + m_p, logs_p, x_mask = self.enc_p(phone, pitch, phone_lengths) + z_p = (m_p + torch.exp(logs_p) * torch.randn_like(m_p) * 0.66666) * x_mask + z = self.flow(z_p, x_mask, g=g, reverse=True) + o = self.dec((z * x_mask)[:, :, :max_len], nsff0, g=g) + return o, x_mask, (z, z_p, m_p, logs_p) + + +class SynthesizerTrnMs256NSFsid_nono(nn.Module): + def __init__( + self, + spec_channels, + segment_size, + inter_channels, + hidden_channels, + filter_channels, + n_heads, + n_layers, + kernel_size, + p_dropout, + resblock, + resblock_kernel_sizes, + resblock_dilation_sizes, + upsample_rates, + upsample_initial_channel, + upsample_kernel_sizes, + spk_embed_dim, + gin_channels, + sr=None, + **kwargs + ): + super().__init__() + self.spec_channels = spec_channels + self.inter_channels = inter_channels + self.hidden_channels = hidden_channels + self.filter_channels = filter_channels + self.n_heads = n_heads + self.n_layers = n_layers + self.kernel_size = kernel_size + self.p_dropout = p_dropout + self.resblock = resblock + self.resblock_kernel_sizes = resblock_kernel_sizes + self.resblock_dilation_sizes = resblock_dilation_sizes + self.upsample_rates = upsample_rates + self.upsample_initial_channel = upsample_initial_channel + self.upsample_kernel_sizes = upsample_kernel_sizes + self.segment_size = segment_size + self.gin_channels = gin_channels + # self.hop_length = hop_length# + self.spk_embed_dim = spk_embed_dim + self.enc_p = TextEncoder256( + inter_channels, + hidden_channels, + filter_channels, + n_heads, + n_layers, + kernel_size, + p_dropout, + f0=False, + ) + self.dec = Generator( + inter_channels, + resblock, + resblock_kernel_sizes, + resblock_dilation_sizes, + upsample_rates, + upsample_initial_channel, + upsample_kernel_sizes, + gin_channels=gin_channels, + ) + self.enc_q = PosteriorEncoder( + spec_channels, + inter_channels, + hidden_channels, + 5, + 1, + 16, + gin_channels=gin_channels, + ) + self.flow = ResidualCouplingBlock( + inter_channels, hidden_channels, 5, 1, 3, gin_channels=gin_channels + ) + self.emb_g = nn.Embedding(self.spk_embed_dim, gin_channels) + print("gin_channels:", gin_channels, "self.spk_embed_dim:", self.spk_embed_dim) + + def remove_weight_norm(self): + self.dec.remove_weight_norm() + self.flow.remove_weight_norm() + self.enc_q.remove_weight_norm() + + def forward(self, phone, phone_lengths, y, y_lengths, ds): # 这里ds是id，[bs,1] + g = self.emb_g(ds).unsqueeze(-1) # [b, 256, 1]##1是t，广播的 + m_p, logs_p, x_mask = self.enc_p(phone, None, phone_lengths) + z, m_q, logs_q, y_mask = self.enc_q(y, y_lengths, g=g) + z_p = self.flow(z, y_mask, g=g) + z_slice, ids_slice = commons.rand_slice_segments( + z, y_lengths, self.segment_size + ) + o = self.dec(z_slice, g=g) + return o, ids_slice, x_mask, y_mask, (z, z_p, m_p, logs_p, m_q, logs_q) + + def infer(self, phone, phone_lengths, sid, max_len=None): + g = self.emb_g(sid).unsqueeze(-1) + m_p, logs_p, x_mask = self.enc_p(phone, None, phone_lengths) + z_p = (m_p + torch.exp(logs_p) * torch.randn_like(m_p) * 0.66666) * x_mask + z = self.flow(z_p, x_mask, g=g, reverse=True) + o = self.dec((z * x_mask)[:, :, :max_len], g=g) + return o, x_mask, (z, z_p, m_p, logs_p) + + +class SynthesizerTrnMs768NSFsid_nono(nn.Module): + def __init__( + self, + spec_channels, + segment_size, + inter_channels, + hidden_channels, + filter_channels, + n_heads, + n_layers, + kernel_size, + p_dropout, + resblock, + resblock_kernel_sizes, + resblock_dilation_sizes, + upsample_rates, + upsample_initial_channel, + upsample_kernel_sizes, + spk_embed_dim, + gin_channels, + sr=None, + **kwargs + ): + super().__init__() + self.spec_channels = spec_channels + self.inter_channels = inter_channels + self.hidden_channels = hidden_channels + self.filter_channels = filter_channels + self.n_heads = n_heads + self.n_layers = n_layers + self.kernel_size = kernel_size + self.p_dropout = p_dropout + self.resblock = resblock + self.resblock_kernel_sizes = resblock_kernel_sizes + self.resblock_dilation_sizes = resblock_dilation_sizes + self.upsample_rates = upsample_rates + self.upsample_initial_channel = upsample_initial_channel + self.upsample_kernel_sizes = upsample_kernel_sizes + self.segment_size = segment_size + self.gin_channels = gin_channels + # self.hop_length = hop_length# + self.spk_embed_dim = spk_embed_dim + self.enc_p = TextEncoder768( + inter_channels, + hidden_channels, + filter_channels, + n_heads, + n_layers, + kernel_size, + p_dropout, + f0=False, + ) + self.dec = Generator( + inter_channels, + resblock, + resblock_kernel_sizes, + resblock_dilation_sizes, + upsample_rates, + upsample_initial_channel, + upsample_kernel_sizes, + gin_channels=gin_channels, + ) + self.enc_q = PosteriorEncoder( + spec_channels, + inter_channels, + hidden_channels, + 5, + 1, + 16, + gin_channels=gin_channels, + ) + self.flow = ResidualCouplingBlock( + inter_channels, hidden_channels, 5, 1, 3, gin_channels=gin_channels + ) + self.emb_g = nn.Embedding(self.spk_embed_dim, gin_channels) + print("gin_channels:", gin_channels, "self.spk_embed_dim:", self.spk_embed_dim) + + def remove_weight_norm(self): + self.dec.remove_weight_norm() + self.flow.remove_weight_norm() + self.enc_q.remove_weight_norm() + + def forward(self, phone, phone_lengths, y, y_lengths, ds): # 这里ds是id，[bs,1] + g = self.emb_g(ds).unsqueeze(-1) # [b, 256, 1]##1是t，广播的 + m_p, logs_p, x_mask = self.enc_p(phone, None, phone_lengths) + z, m_q, logs_q, y_mask = self.enc_q(y, y_lengths, g=g) + z_p = self.flow(z, y_mask, g=g) + z_slice, ids_slice = commons.rand_slice_segments( + z, y_lengths, self.segment_size + ) + o = self.dec(z_slice, g=g) + return o, ids_slice, x_mask, y_mask, (z, z_p, m_p, logs_p, m_q, logs_q) + + def infer(self, phone, phone_lengths, sid, max_len=None): + g = self.emb_g(sid).unsqueeze(-1) + m_p, logs_p, x_mask = self.enc_p(phone, None, phone_lengths) + z_p = (m_p + torch.exp(logs_p) * torch.randn_like(m_p) * 0.66666) * x_mask + z = self.flow(z_p, x_mask, g=g, reverse=True) + o = self.dec((z * x_mask)[:, :, :max_len], g=g) + return o, x_mask, (z, z_p, m_p, logs_p) + + +class MultiPeriodDiscriminator(torch.nn.Module): + def __init__(self, use_spectral_norm=False): + super(MultiPeriodDiscriminator, self).__init__() + periods = [2, 3, 5, 7, 11, 17] + # periods = [3, 5, 7, 11, 17, 23, 37] + + discs = [DiscriminatorS(use_spectral_norm=use_spectral_norm)] + discs = discs + [ + DiscriminatorP(i, use_spectral_norm=use_spectral_norm) for i in periods + ] + self.discriminators = nn.ModuleList(discs) + + def forward(self, y, y_hat): + y_d_rs = [] # + y_d_gs = [] + fmap_rs = [] + fmap_gs = [] + for i, d in enumerate(self.discriminators): + y_d_r, fmap_r = d(y) + y_d_g, fmap_g = d(y_hat) + # for j in range(len(fmap_r)): + # print(i,j,y.shape,y_hat.shape,fmap_r[j].shape,fmap_g[j].shape) + y_d_rs.append(y_d_r) + y_d_gs.append(y_d_g) + fmap_rs.append(fmap_r) + fmap_gs.append(fmap_g) + + return y_d_rs, y_d_gs, fmap_rs, fmap_gs + + +class MultiPeriodDiscriminatorV2(torch.nn.Module): + def __init__(self, use_spectral_norm=False): + super(MultiPeriodDiscriminatorV2, self).__init__() + # periods = [2, 3, 5, 7, 11, 17] + periods = [2, 3, 5, 7, 11, 17, 23, 37] + + discs = [DiscriminatorS(use_spectral_norm=use_spectral_norm)] + discs = discs + [ + DiscriminatorP(i, use_spectral_norm=use_spectral_norm) for i in periods + ] + self.discriminators = nn.ModuleList(discs) + + def forward(self, y, y_hat): + y_d_rs = [] # + y_d_gs = [] + fmap_rs = [] + fmap_gs = [] + for i, d in enumerate(self.discriminators): + y_d_r, fmap_r = d(y) + y_d_g, fmap_g = d(y_hat) + # for j in range(len(fmap_r)): + # print(i,j,y.shape,y_hat.shape,fmap_r[j].shape,fmap_g[j].shape) + y_d_rs.append(y_d_r) + y_d_gs.append(y_d_g) + fmap_rs.append(fmap_r) + fmap_gs.append(fmap_g) + + return y_d_rs, y_d_gs, fmap_rs, fmap_gs + + +class DiscriminatorS(torch.nn.Module): + def __init__(self, use_spectral_norm=False): + super(DiscriminatorS, self).__init__() + norm_f = weight_norm if use_spectral_norm == False else spectral_norm + self.convs = nn.ModuleList( + [ + norm_f(Conv1d(1, 16, 15, 1, padding=7)), + norm_f(Conv1d(16, 64, 41, 4, groups=4, padding=20)), + norm_f(Conv1d(64, 256, 41, 4, groups=16, padding=20)), + norm_f(Conv1d(256, 1024, 41, 4, groups=64, padding=20)), + norm_f(Conv1d(1024, 1024, 41, 4, groups=256, padding=20)), + norm_f(Conv1d(1024, 1024, 5, 1, padding=2)), + ] + ) + self.conv_post = norm_f(Conv1d(1024, 1, 3, 1, padding=1)) + + def forward(self, x): + fmap = [] + + for l in self.convs: + x = l(x) + x = F.leaky_relu(x, modules.LRELU_SLOPE) + fmap.append(x) + x = self.conv_post(x) + fmap.append(x) + x = torch.flatten(x, 1, -1) + + return x, fmap + + +class DiscriminatorP(torch.nn.Module): + def __init__(self, period, kernel_size=5, stride=3, use_spectral_norm=False): + super(DiscriminatorP, self).__init__() + self.period = period + self.use_spectral_norm = use_spectral_norm + norm_f = weight_norm if use_spectral_norm == False else spectral_norm + self.convs = nn.ModuleList( + [ + norm_f( + Conv2d( + 1, + 32, + (kernel_size, 1), + (stride, 1), + padding=(get_padding(kernel_size, 1), 0), + ) + ), + norm_f( + Conv2d( + 32, + 128, + (kernel_size, 1), + (stride, 1), + padding=(get_padding(kernel_size, 1), 0), + ) + ), + norm_f( + Conv2d( + 128, + 512, + (kernel_size, 1), + (stride, 1), + padding=(get_padding(kernel_size, 1), 0), + ) + ), + norm_f( + Conv2d( + 512, + 1024, + (kernel_size, 1), + (stride, 1), + padding=(get_padding(kernel_size, 1), 0), + ) + ), + norm_f( + Conv2d( + 1024, + 1024, + (kernel_size, 1), + 1, + padding=(get_padding(kernel_size, 1), 0), + ) + ), + ] + ) + self.conv_post = norm_f(Conv2d(1024, 1, (3, 1), 1, padding=(1, 0))) + + def forward(self, x): + fmap = [] + + # 1d to 2d + b, c, t = x.shape + if t % self.period != 0: # pad first + n_pad = self.period - (t % self.period) + x = F.pad(x, (0, n_pad), "reflect") + t = t + n_pad + x = x.view(b, c, t // self.period, self.period) + + for l in self.convs: + x = l(x) + x = F.leaky_relu(x, modules.LRELU_SLOPE) + fmap.append(x) + x = self.conv_post(x) + fmap.append(x) + x = torch.flatten(x, 1, -1) + + return x, fmap