diff --git a/.gitignore b/.gitignore index 5d381cc..da62613 100644 --- a/.gitignore +++ b/.gitignore @@ -1,3 +1,6 @@ +# Unsloth +_unsloth_sentencepiece_temp/ + # ---> Python # Byte-compiled / optimized / DLL files __pycache__/ diff --git a/README.md b/README.md index 02c16bd..6c5d0dd 100644 --- a/README.md +++ b/README.md @@ -1,3 +1,23 @@ # MikuAI -Backend for Hatsune Miku discord bot, capable of text generation and voice changing \ No newline at end of file +Backend for Hatsune Miku discord bot, capable of text generation and voice changing. + +RVC code is adapted from [AICoverGen](https://github.com/SociallyIneptWeeb/AICoverGen). + +The model used for text generation is Mistral-7B with a custom LoRA trained on group chat messages, both of which are downloaded from Huggingface. The [Literally Every Hatsune Miku Voicebank](https://www.weights.gg/models/clp9qiyw50013rvv6tl2frcpv) model is used for voice synthesis, and is included with this repo. + +## Requirements + +Modern NVIDIA GPU with 8+ GB of VRAM, as well as 8+ GB of CPU RAM + +## Setup + +Install Anaconda/Miniconda3/Mamba and set up an environment with Python 3.10. + +Set up [Pytorch](https://pytorch.org/get-started/locally/) followed by [Unsloth](https://github.com/unslothai/unsloth). + +Install the `ffmpeg` and `sox` packages, you may have to get ffmpeg from `conda-forge` if it complains about missing shared libraries when you try to run it. + +Run `pip -r requirements.txt` to get the rest of the dependencies. + +Start the web server using `uvicorn api:app --host=0.0.0.0 --port=8000`. You can test the endpoints at `http://localhost:8000/docs`. diff --git a/api.py b/api.py new file mode 100644 index 0000000..5c6e65a --- /dev/null +++ b/api.py @@ -0,0 +1,52 @@ +from fastapi import FastAPI, Response, UploadFile +from fastapi.encoders import jsonable_encoder +from fastapi.responses import FileResponse +from config import TOKEN +import hmac +import model +from pydantic import BaseModel +from typing import List, Optional +import tempfile + +from rvc.main import song_cover_pipeline + +app = FastAPI() + +class Message(BaseModel): + role: str + content: str + +@app.post("/") +async def root(token: str, + messages: List[Message], + response: Response, + max_new_tokens: Optional[int] = 64, + temperature: Optional[float] = 0.9, + repetition_penalty: Optional[float] = 1.2): + if not hmac.compare_digest(token, TOKEN): + response.status_code = 401 + return {"error": "Bad token"} + + dict_in = jsonable_encoder(messages) + output = model.inference(dict_in, max_new_tokens=max_new_tokens, temperature=temperature, repetition_penalty=repetition_penalty) + return {"raw": output} + +@app.post("/rvc") +async def rvc(token: str, + file: UploadFile, + response: Response, + pitch_change_oct: Optional[int] = 1, + pitch_change_sem: Optional[int] = 0): + if not hmac.compare_digest(token, TOKEN): + response.status_code = 401 + return {"error": "Bad token"} + + with tempfile.NamedTemporaryFile() as tmp: + tmp.write(await file.read()) + ai_vocals_path = song_cover_pipeline(tmp.name, pitch_change_oct, voice_model='miku', pitch_change_sem=pitch_change_sem) + + return FileResponse(ai_vocals_path) + +@app.get("/ping") +def ping(): + return {"message": "pong"} diff --git a/config.py b/config.py new file mode 100644 index 0000000..102f41a --- /dev/null +++ b/config.py @@ -0,0 +1 @@ +TOKEN = "4BJkM2NO3Y6wFDXuHdKc" diff --git a/model.py b/model.py new file mode 100644 index 0000000..94671b2 --- /dev/null +++ b/model.py @@ -0,0 +1,38 @@ +from unsloth import FastLanguageModel +from unsloth.chat_templates import get_chat_template +from transformers import TextStreamer + +model, tokenizer = FastLanguageModel.from_pretrained( + model_name = "scoliono/groupchat_lora", + max_seq_length = 2048, + dtype = None, + load_in_4bit = True, +) +FastLanguageModel.for_inference(model) # Enable native 2x faster inference + +tokenizer = get_chat_template( + tokenizer, + chat_template = "chatml", # Supports zephyr, chatml, mistral, llama, alpaca, vicuna, vicuna_old, unsloth + map_eos_token = True, # Maps <|im_end|> to instead +) + +def inference(messages, max_new_tokens=64, temperature=0.9, repetition_penalty=1.2): + inputs = tokenizer.apply_chat_template( + messages, + tokenize = True, + add_generation_prompt = True, # Must add for generation + return_tensors = "pt", + ).to("cuda") + + #text_streamer = TextStreamer(tokenizer) + + token_ids = model.generate( + input_ids = inputs, + #streamer = text_streamer, + max_new_tokens = max_new_tokens, + use_cache = True, + temperature = temperature, + repetition_penalty = repetition_penalty + ) + + return tokenizer.batch_decode(token_ids) diff --git a/requirements.txt b/requirements.txt new file mode 100644 index 0000000..494cc4c --- /dev/null +++ b/requirements.txt @@ -0,0 +1,19 @@ +#deemix +fairseq==0.12.2 +faiss-cpu==1.7.3 +fastapi==0.110.0 +ffmpeg-python>=0.2.0 +librosa==0.9.1 +numpy==1.23.5 +onnxruntime_gpu +praat-parselmouth>=0.4.2 +#pedalboard==0.7.7 +#pydub==0.25.1 +pyworld==0.3.4 +#Requests==2.31.0 +scipy==1.11.1 +soundfile==0.12.1 +torchcrepe==0.0.20 +tqdm==4.65.0 +uvicorn==0.29.0 +sox==1.4.1 diff --git a/rvc/__init__.py b/rvc/__init__.py new file mode 100644 index 0000000..e69de29 diff --git a/rvc/download_models.py b/rvc/download_models.py new file mode 100644 index 0000000..bea953f --- /dev/null +++ b/rvc/download_models.py @@ -0,0 +1,27 @@ +from pathlib import Path +import requests + +MDX_DOWNLOAD_LINK = 'https://github.com/TRvlvr/model_repo/releases/download/all_public_uvr_models/' +RVC_DOWNLOAD_LINK = 'https://huggingface.co/lj1995/VoiceConversionWebUI/resolve/main/' + +BASE_DIR = Path(__file__).resolve().parent +mdxnet_models_dir = BASE_DIR / 'mdxnet_models' +rvc_models_dir = BASE_DIR / 'rvc_models' + + +def dl_model(link, model_name, dir_name): + with requests.get(f'{link}{model_name}') as r: + r.raise_for_status() + with open(dir_name / model_name, 'wb') as f: + for chunk in r.iter_content(chunk_size=8192): + f.write(chunk) + + +if __name__ == '__main__': + + rvc_model_names = ['hubert_base.pt', 'rmvpe.pt'] + for model in rvc_model_names: + print(f'Downloading {model}...') + dl_model(RVC_DOWNLOAD_LINK, model, rvc_models_dir) + + print('All models downloaded!') diff --git a/rvc/infer_pack/__init__.py b/rvc/infer_pack/__init__.py new file mode 100644 index 0000000..e69de29 diff --git a/rvc/infer_pack/attentions.py b/rvc/infer_pack/attentions.py new file mode 100644 index 0000000..1024e78 --- /dev/null +++ b/rvc/infer_pack/attentions.py @@ -0,0 +1,417 @@ +import copy +import math +import numpy as np +import torch +from torch import nn +from torch.nn import functional as F + +from . import commons +from . import modules +from .modules import LayerNorm + + +class Encoder(nn.Module): + def __init__( + self, + hidden_channels, + filter_channels, + n_heads, + n_layers, + kernel_size=1, + p_dropout=0.0, + window_size=10, + **kwargs + ): + super().__init__() + 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.window_size = window_size + + self.drop = nn.Dropout(p_dropout) + self.attn_layers = nn.ModuleList() + self.norm_layers_1 = nn.ModuleList() + self.ffn_layers = nn.ModuleList() + self.norm_layers_2 = nn.ModuleList() + for i in range(self.n_layers): + self.attn_layers.append( + MultiHeadAttention( + hidden_channels, + hidden_channels, + n_heads, + p_dropout=p_dropout, + window_size=window_size, + ) + ) + self.norm_layers_1.append(LayerNorm(hidden_channels)) + self.ffn_layers.append( + FFN( + hidden_channels, + hidden_channels, + filter_channels, + kernel_size, + p_dropout=p_dropout, + ) + ) + self.norm_layers_2.append(LayerNorm(hidden_channels)) + + def forward(self, x, x_mask): + attn_mask = x_mask.unsqueeze(2) * x_mask.unsqueeze(-1) + x = x * x_mask + for i in range(self.n_layers): + y = self.attn_layers[i](x, x, attn_mask) + y = self.drop(y) + x = self.norm_layers_1[i](x + y) + + y = self.ffn_layers[i](x, x_mask) + y = self.drop(y) + x = self.norm_layers_2[i](x + y) + x = x * x_mask + return x + + +class Decoder(nn.Module): + def __init__( + self, + hidden_channels, + filter_channels, + n_heads, + n_layers, + kernel_size=1, + p_dropout=0.0, + proximal_bias=False, + proximal_init=True, + **kwargs + ): + super().__init__() + 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.proximal_bias = proximal_bias + self.proximal_init = proximal_init + + self.drop = nn.Dropout(p_dropout) + self.self_attn_layers = nn.ModuleList() + self.norm_layers_0 = nn.ModuleList() + self.encdec_attn_layers = nn.ModuleList() + self.norm_layers_1 = nn.ModuleList() + self.ffn_layers = nn.ModuleList() + self.norm_layers_2 = nn.ModuleList() + for i in range(self.n_layers): + self.self_attn_layers.append( + MultiHeadAttention( + hidden_channels, + hidden_channels, + n_heads, + p_dropout=p_dropout, + proximal_bias=proximal_bias, + proximal_init=proximal_init, + ) + ) + self.norm_layers_0.append(LayerNorm(hidden_channels)) + self.encdec_attn_layers.append( + MultiHeadAttention( + hidden_channels, hidden_channels, n_heads, p_dropout=p_dropout + ) + ) + self.norm_layers_1.append(LayerNorm(hidden_channels)) + self.ffn_layers.append( + FFN( + hidden_channels, + hidden_channels, + filter_channels, + kernel_size, + p_dropout=p_dropout, + causal=True, + ) + ) + self.norm_layers_2.append(LayerNorm(hidden_channels)) + + def forward(self, x, x_mask, h, h_mask): + """ + x: decoder input + h: encoder output + """ + self_attn_mask = commons.subsequent_mask(x_mask.size(2)).to( + device=x.device, dtype=x.dtype + ) + encdec_attn_mask = h_mask.unsqueeze(2) * x_mask.unsqueeze(-1) + x = x * x_mask + for i in range(self.n_layers): + y = self.self_attn_layers[i](x, x, self_attn_mask) + y = self.drop(y) + x = self.norm_layers_0[i](x + y) + + y = self.encdec_attn_layers[i](x, h, encdec_attn_mask) + y = self.drop(y) + x = self.norm_layers_1[i](x + y) + + y = self.ffn_layers[i](x, x_mask) + y = self.drop(y) + x = self.norm_layers_2[i](x + y) + x = x * x_mask + return x + + +class MultiHeadAttention(nn.Module): + def __init__( + self, + channels, + out_channels, + n_heads, + p_dropout=0.0, + window_size=None, + heads_share=True, + block_length=None, + proximal_bias=False, + proximal_init=False, + ): + super().__init__() + assert channels % n_heads == 0 + + self.channels = channels + self.out_channels = out_channels + self.n_heads = n_heads + self.p_dropout = p_dropout + self.window_size = window_size + self.heads_share = heads_share + self.block_length = block_length + self.proximal_bias = proximal_bias + self.proximal_init = proximal_init + self.attn = None + + self.k_channels = channels // n_heads + self.conv_q = nn.Conv1d(channels, channels, 1) + self.conv_k = nn.Conv1d(channels, channels, 1) + self.conv_v = nn.Conv1d(channels, channels, 1) + self.conv_o = nn.Conv1d(channels, out_channels, 1) + self.drop = nn.Dropout(p_dropout) + + if window_size is not None: + n_heads_rel = 1 if heads_share else n_heads + rel_stddev = self.k_channels**-0.5 + self.emb_rel_k = nn.Parameter( + torch.randn(n_heads_rel, window_size * 2 + 1, self.k_channels) + * rel_stddev + ) + self.emb_rel_v = nn.Parameter( + torch.randn(n_heads_rel, window_size * 2 + 1, self.k_channels) + * rel_stddev + ) + + nn.init.xavier_uniform_(self.conv_q.weight) + nn.init.xavier_uniform_(self.conv_k.weight) + nn.init.xavier_uniform_(self.conv_v.weight) + if proximal_init: + with torch.no_grad(): + self.conv_k.weight.copy_(self.conv_q.weight) + self.conv_k.bias.copy_(self.conv_q.bias) + + def forward(self, x, c, attn_mask=None): + q = self.conv_q(x) + k = self.conv_k(c) + v = self.conv_v(c) + + x, self.attn = self.attention(q, k, v, mask=attn_mask) + + x = self.conv_o(x) + return x + + def attention(self, query, key, value, mask=None): + # reshape [b, d, t] -> [b, n_h, t, d_k] + b, d, t_s, t_t = (*key.size(), query.size(2)) + query = query.view(b, self.n_heads, self.k_channels, t_t).transpose(2, 3) + key = key.view(b, self.n_heads, self.k_channels, t_s).transpose(2, 3) + value = value.view(b, self.n_heads, self.k_channels, t_s).transpose(2, 3) + + scores = torch.matmul(query / math.sqrt(self.k_channels), key.transpose(-2, -1)) + if self.window_size is not None: + assert ( + t_s == t_t + ), "Relative attention is only available for self-attention." + key_relative_embeddings = self._get_relative_embeddings(self.emb_rel_k, t_s) + rel_logits = self._matmul_with_relative_keys( + query / math.sqrt(self.k_channels), key_relative_embeddings + ) + scores_local = self._relative_position_to_absolute_position(rel_logits) + scores = scores + scores_local + if self.proximal_bias: + assert t_s == t_t, "Proximal bias is only available for self-attention." + scores = scores + self._attention_bias_proximal(t_s).to( + device=scores.device, dtype=scores.dtype + ) + if mask is not None: + scores = scores.masked_fill(mask == 0, -1e4) + if self.block_length is not None: + assert ( + t_s == t_t + ), "Local attention is only available for self-attention." + block_mask = ( + torch.ones_like(scores) + .triu(-self.block_length) + .tril(self.block_length) + ) + scores = scores.masked_fill(block_mask == 0, -1e4) + p_attn = F.softmax(scores, dim=-1) # [b, n_h, t_t, t_s] + p_attn = self.drop(p_attn) + output = torch.matmul(p_attn, value) + if self.window_size is not None: + relative_weights = self._absolute_position_to_relative_position(p_attn) + value_relative_embeddings = self._get_relative_embeddings( + self.emb_rel_v, t_s + ) + output = output + self._matmul_with_relative_values( + relative_weights, value_relative_embeddings + ) + output = ( + output.transpose(2, 3).contiguous().view(b, d, t_t) + ) # [b, n_h, t_t, d_k] -> [b, d, t_t] + return output, p_attn + + def _matmul_with_relative_values(self, x, y): + """ + x: [b, h, l, m] + y: [h or 1, m, d] + ret: [b, h, l, d] + """ + ret = torch.matmul(x, y.unsqueeze(0)) + return ret + + def _matmul_with_relative_keys(self, x, y): + """ + x: [b, h, l, d] + y: [h or 1, m, d] + ret: [b, h, l, m] + """ + ret = torch.matmul(x, y.unsqueeze(0).transpose(-2, -1)) + return ret + + def _get_relative_embeddings(self, relative_embeddings, length): + max_relative_position = 2 * self.window_size + 1 + # Pad first before slice to avoid using cond ops. + pad_length = max(length - (self.window_size + 1), 0) + slice_start_position = max((self.window_size + 1) - length, 0) + slice_end_position = slice_start_position + 2 * length - 1 + if pad_length > 0: + padded_relative_embeddings = F.pad( + relative_embeddings, + commons.convert_pad_shape([[0, 0], [pad_length, pad_length], [0, 0]]), + ) + else: + padded_relative_embeddings = relative_embeddings + used_relative_embeddings = padded_relative_embeddings[ + :, slice_start_position:slice_end_position + ] + return used_relative_embeddings + + def _relative_position_to_absolute_position(self, x): + """ + x: [b, h, l, 2*l-1] + ret: [b, h, l, l] + """ + batch, heads, length, _ = x.size() + # Concat columns of pad to shift from relative to absolute indexing. + x = F.pad(x, commons.convert_pad_shape([[0, 0], [0, 0], [0, 0], [0, 1]])) + + # Concat extra elements so to add up to shape (len+1, 2*len-1). + x_flat = x.view([batch, heads, length * 2 * length]) + x_flat = F.pad( + x_flat, commons.convert_pad_shape([[0, 0], [0, 0], [0, length - 1]]) + ) + + # Reshape and slice out the padded elements. + x_final = x_flat.view([batch, heads, length + 1, 2 * length - 1])[ + :, :, :length, length - 1 : + ] + return x_final + + def _absolute_position_to_relative_position(self, x): + """ + x: [b, h, l, l] + ret: [b, h, l, 2*l-1] + """ + batch, heads, length, _ = x.size() + # padd along column + x = F.pad( + x, commons.convert_pad_shape([[0, 0], [0, 0], [0, 0], [0, length - 1]]) + ) + x_flat = x.view([batch, heads, length**2 + length * (length - 1)]) + # add 0's in the beginning that will skew the elements after reshape + x_flat = F.pad(x_flat, commons.convert_pad_shape([[0, 0], [0, 0], [length, 0]])) + x_final = x_flat.view([batch, heads, length, 2 * length])[:, :, :, 1:] + return x_final + + def _attention_bias_proximal(self, length): + """Bias for self-attention to encourage attention to close positions. + Args: + length: an integer scalar. + Returns: + a Tensor with shape [1, 1, length, length] + """ + r = torch.arange(length, dtype=torch.float32) + diff = torch.unsqueeze(r, 0) - torch.unsqueeze(r, 1) + return torch.unsqueeze(torch.unsqueeze(-torch.log1p(torch.abs(diff)), 0), 0) + + +class FFN(nn.Module): + def __init__( + self, + in_channels, + out_channels, + filter_channels, + kernel_size, + p_dropout=0.0, + activation=None, + causal=False, + ): + super().__init__() + self.in_channels = in_channels + self.out_channels = out_channels + self.filter_channels = filter_channels + self.kernel_size = kernel_size + self.p_dropout = p_dropout + self.activation = activation + self.causal = causal + + if causal: + self.padding = self._causal_padding + else: + self.padding = self._same_padding + + self.conv_1 = nn.Conv1d(in_channels, filter_channels, kernel_size) + self.conv_2 = nn.Conv1d(filter_channels, out_channels, kernel_size) + self.drop = nn.Dropout(p_dropout) + + def forward(self, x, x_mask): + x = self.conv_1(self.padding(x * x_mask)) + if self.activation == "gelu": + x = x * torch.sigmoid(1.702 * x) + else: + x = torch.relu(x) + x = self.drop(x) + x = self.conv_2(self.padding(x * x_mask)) + return x * x_mask + + def _causal_padding(self, x): + if self.kernel_size == 1: + return x + pad_l = self.kernel_size - 1 + pad_r = 0 + padding = [[0, 0], [0, 0], [pad_l, pad_r]] + x = F.pad(x, commons.convert_pad_shape(padding)) + return x + + def _same_padding(self, x): + if self.kernel_size == 1: + return x + pad_l = (self.kernel_size - 1) // 2 + pad_r = self.kernel_size // 2 + padding = [[0, 0], [0, 0], [pad_l, pad_r]] + x = F.pad(x, commons.convert_pad_shape(padding)) + return x diff --git a/rvc/infer_pack/commons.py b/rvc/infer_pack/commons.py new file mode 100644 index 0000000..5447098 --- /dev/null +++ b/rvc/infer_pack/commons.py @@ -0,0 +1,166 @@ +import math +import numpy as np +import torch +from torch import nn +from torch.nn import functional as F + + +def init_weights(m, mean=0.0, std=0.01): + classname = m.__class__.__name__ + if classname.find("Conv") != -1: + m.weight.data.normal_(mean, std) + + +def get_padding(kernel_size, dilation=1): + return int((kernel_size * dilation - dilation) / 2) + + +def convert_pad_shape(pad_shape): + l = pad_shape[::-1] + pad_shape = [item for sublist in l for item in sublist] + return pad_shape + + +def kl_divergence(m_p, logs_p, m_q, logs_q): + """KL(P||Q)""" + kl = (logs_q - logs_p) - 0.5 + kl += ( + 0.5 * (torch.exp(2.0 * logs_p) + ((m_p - m_q) ** 2)) * torch.exp(-2.0 * logs_q) + ) + return kl + + +def rand_gumbel(shape): + """Sample from the Gumbel distribution, protect from overflows.""" + uniform_samples = torch.rand(shape) * 0.99998 + 0.00001 + return -torch.log(-torch.log(uniform_samples)) + + +def rand_gumbel_like(x): + g = rand_gumbel(x.size()).to(dtype=x.dtype, device=x.device) + return g + + +def slice_segments(x, ids_str, segment_size=4): + ret = torch.zeros_like(x[:, :, :segment_size]) + for i in range(x.size(0)): + idx_str = ids_str[i] + idx_end = idx_str + segment_size + ret[i] = x[i, :, idx_str:idx_end] + return ret + + +def slice_segments2(x, ids_str, segment_size=4): + ret = torch.zeros_like(x[:, :segment_size]) + for i in range(x.size(0)): + idx_str = ids_str[i] + idx_end = idx_str + segment_size + ret[i] = x[i, idx_str:idx_end] + return ret + + +def rand_slice_segments(x, x_lengths=None, segment_size=4): + b, d, t = x.size() + if x_lengths is None: + x_lengths = t + ids_str_max = x_lengths - segment_size + 1 + ids_str = (torch.rand([b]).to(device=x.device) * ids_str_max).to(dtype=torch.long) + ret = slice_segments(x, ids_str, segment_size) + return ret, ids_str + + +def get_timing_signal_1d(length, channels, min_timescale=1.0, max_timescale=1.0e4): + position = torch.arange(length, dtype=torch.float) + num_timescales = channels // 2 + log_timescale_increment = math.log(float(max_timescale) / float(min_timescale)) / ( + num_timescales - 1 + ) + inv_timescales = min_timescale * torch.exp( + torch.arange(num_timescales, dtype=torch.float) * -log_timescale_increment + ) + scaled_time = position.unsqueeze(0) * inv_timescales.unsqueeze(1) + signal = torch.cat([torch.sin(scaled_time), torch.cos(scaled_time)], 0) + signal = F.pad(signal, [0, 0, 0, channels % 2]) + signal = signal.view(1, channels, length) + return signal + + +def add_timing_signal_1d(x, min_timescale=1.0, max_timescale=1.0e4): + b, channels, length = x.size() + signal = get_timing_signal_1d(length, channels, min_timescale, max_timescale) + return x + signal.to(dtype=x.dtype, device=x.device) + + +def cat_timing_signal_1d(x, min_timescale=1.0, max_timescale=1.0e4, axis=1): + b, channels, length = x.size() + signal = get_timing_signal_1d(length, channels, min_timescale, max_timescale) + return torch.cat([x, signal.to(dtype=x.dtype, device=x.device)], axis) + + +def subsequent_mask(length): + mask = torch.tril(torch.ones(length, length)).unsqueeze(0).unsqueeze(0) + return mask + + +@torch.jit.script +def fused_add_tanh_sigmoid_multiply(input_a, input_b, n_channels): + n_channels_int = n_channels[0] + in_act = input_a + input_b + t_act = torch.tanh(in_act[:, :n_channels_int, :]) + s_act = torch.sigmoid(in_act[:, n_channels_int:, :]) + acts = t_act * s_act + return acts + + +def convert_pad_shape(pad_shape): + l = pad_shape[::-1] + pad_shape = [item for sublist in l for item in sublist] + return pad_shape + + +def shift_1d(x): + x = F.pad(x, convert_pad_shape([[0, 0], [0, 0], [1, 0]]))[:, :, :-1] + return x + + +def sequence_mask(length, max_length=None): + if max_length is None: + max_length = length.max() + x = torch.arange(max_length, dtype=length.dtype, device=length.device) + return x.unsqueeze(0) < length.unsqueeze(1) + + +def generate_path(duration, mask): + """ + duration: [b, 1, t_x] + mask: [b, 1, t_y, t_x] + """ + device = duration.device + + b, _, t_y, t_x = mask.shape + cum_duration = torch.cumsum(duration, -1) + + cum_duration_flat = cum_duration.view(b * t_x) + path = sequence_mask(cum_duration_flat, t_y).to(mask.dtype) + path = path.view(b, t_x, t_y) + path = path - F.pad(path, convert_pad_shape([[0, 0], [1, 0], [0, 0]]))[:, :-1] + path = path.unsqueeze(1).transpose(2, 3) * mask + return path + + +def clip_grad_value_(parameters, clip_value, norm_type=2): + if isinstance(parameters, torch.Tensor): + parameters = [parameters] + parameters = list(filter(lambda p: p.grad is not None, parameters)) + norm_type = float(norm_type) + if clip_value is not None: + clip_value = float(clip_value) + + total_norm = 0 + for p in parameters: + param_norm = p.grad.data.norm(norm_type) + total_norm += param_norm.item() ** norm_type + if clip_value is not None: + p.grad.data.clamp_(min=-clip_value, max=clip_value) + total_norm = total_norm ** (1.0 / norm_type) + return total_norm diff --git a/rvc/infer_pack/models.py b/rvc/infer_pack/models.py new file mode 100644 index 0000000..d7d4105 --- /dev/null +++ b/rvc/infer_pack/models.py @@ -0,0 +1,1120 @@ +import math, pdb, os +from time import time as ttime +import torch +from torch import nn +from torch.nn import functional as F +from . import modules, attentions, commons +from .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 +import numpy as np + + +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 + + 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 diff --git a/rvc/infer_pack/modules.py b/rvc/infer_pack/modules.py new file mode 100644 index 0000000..6356a8c --- /dev/null +++ b/rvc/infer_pack/modules.py @@ -0,0 +1,522 @@ +import copy +import math +import numpy as np +import scipy +import torch +from torch import nn +from torch.nn import functional as F + +from torch.nn import Conv1d, ConvTranspose1d, AvgPool1d, Conv2d +from torch.nn.utils import weight_norm, remove_weight_norm + +from . import commons +from .commons import init_weights, get_padding +from .transforms import piecewise_rational_quadratic_transform + + +LRELU_SLOPE = 0.1 + + +class LayerNorm(nn.Module): + def __init__(self, channels, eps=1e-5): + super().__init__() + self.channels = channels + self.eps = eps + + self.gamma = nn.Parameter(torch.ones(channels)) + self.beta = nn.Parameter(torch.zeros(channels)) + + def forward(self, x): + x = x.transpose(1, -1) + x = F.layer_norm(x, (self.channels,), self.gamma, self.beta, self.eps) + return x.transpose(1, -1) + + +class ConvReluNorm(nn.Module): + def __init__( + self, + in_channels, + hidden_channels, + out_channels, + kernel_size, + n_layers, + p_dropout, + ): + super().__init__() + self.in_channels = in_channels + self.hidden_channels = hidden_channels + self.out_channels = out_channels + self.kernel_size = kernel_size + self.n_layers = n_layers + self.p_dropout = p_dropout + assert n_layers > 1, "Number of layers should be larger than 0." + + self.conv_layers = nn.ModuleList() + self.norm_layers = nn.ModuleList() + self.conv_layers.append( + nn.Conv1d( + in_channels, hidden_channels, kernel_size, padding=kernel_size // 2 + ) + ) + self.norm_layers.append(LayerNorm(hidden_channels)) + self.relu_drop = nn.Sequential(nn.ReLU(), nn.Dropout(p_dropout)) + for _ in range(n_layers - 1): + self.conv_layers.append( + nn.Conv1d( + hidden_channels, + hidden_channels, + kernel_size, + padding=kernel_size // 2, + ) + ) + self.norm_layers.append(LayerNorm(hidden_channels)) + self.proj = nn.Conv1d(hidden_channels, out_channels, 1) + self.proj.weight.data.zero_() + self.proj.bias.data.zero_() + + def forward(self, x, x_mask): + x_org = x + for i in range(self.n_layers): + x = self.conv_layers[i](x * x_mask) + x = self.norm_layers[i](x) + x = self.relu_drop(x) + x = x_org + self.proj(x) + return x * x_mask + + +class DDSConv(nn.Module): + """ + Dialted and Depth-Separable Convolution + """ + + def __init__(self, channels, kernel_size, n_layers, p_dropout=0.0): + super().__init__() + self.channels = channels + self.kernel_size = kernel_size + self.n_layers = n_layers + self.p_dropout = p_dropout + + self.drop = nn.Dropout(p_dropout) + self.convs_sep = nn.ModuleList() + self.convs_1x1 = nn.ModuleList() + self.norms_1 = nn.ModuleList() + self.norms_2 = nn.ModuleList() + for i in range(n_layers): + dilation = kernel_size**i + padding = (kernel_size * dilation - dilation) // 2 + self.convs_sep.append( + nn.Conv1d( + channels, + channels, + kernel_size, + groups=channels, + dilation=dilation, + padding=padding, + ) + ) + self.convs_1x1.append(nn.Conv1d(channels, channels, 1)) + self.norms_1.append(LayerNorm(channels)) + self.norms_2.append(LayerNorm(channels)) + + def forward(self, x, x_mask, g=None): + if g is not None: + x = x + g + for i in range(self.n_layers): + y = self.convs_sep[i](x * x_mask) + y = self.norms_1[i](y) + y = F.gelu(y) + y = self.convs_1x1[i](y) + y = self.norms_2[i](y) + y = F.gelu(y) + y = self.drop(y) + x = x + y + return x * x_mask + + +class WN(torch.nn.Module): + def __init__( + self, + hidden_channels, + kernel_size, + dilation_rate, + n_layers, + gin_channels=0, + p_dropout=0, + ): + super(WN, self).__init__() + assert kernel_size % 2 == 1 + 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.p_dropout = p_dropout + + self.in_layers = torch.nn.ModuleList() + self.res_skip_layers = torch.nn.ModuleList() + self.drop = nn.Dropout(p_dropout) + + if gin_channels != 0: + cond_layer = torch.nn.Conv1d( + gin_channels, 2 * hidden_channels * n_layers, 1 + ) + self.cond_layer = torch.nn.utils.weight_norm(cond_layer, name="weight") + + for i in range(n_layers): + dilation = dilation_rate**i + padding = int((kernel_size * dilation - dilation) / 2) + in_layer = torch.nn.Conv1d( + hidden_channels, + 2 * hidden_channels, + kernel_size, + dilation=dilation, + padding=padding, + ) + in_layer = torch.nn.utils.weight_norm(in_layer, name="weight") + self.in_layers.append(in_layer) + + # last one is not necessary + if i < n_layers - 1: + res_skip_channels = 2 * hidden_channels + else: + res_skip_channels = hidden_channels + + res_skip_layer = torch.nn.Conv1d(hidden_channels, res_skip_channels, 1) + res_skip_layer = torch.nn.utils.weight_norm(res_skip_layer, name="weight") + self.res_skip_layers.append(res_skip_layer) + + def forward(self, x, x_mask, g=None, **kwargs): + output = torch.zeros_like(x) + n_channels_tensor = torch.IntTensor([self.hidden_channels]) + + if g is not None: + g = self.cond_layer(g) + + for i in range(self.n_layers): + x_in = self.in_layers[i](x) + if g is not None: + cond_offset = i * 2 * self.hidden_channels + g_l = g[:, cond_offset : cond_offset + 2 * self.hidden_channels, :] + else: + g_l = torch.zeros_like(x_in) + + acts = commons.fused_add_tanh_sigmoid_multiply(x_in, g_l, n_channels_tensor) + acts = self.drop(acts) + + res_skip_acts = self.res_skip_layers[i](acts) + if i < self.n_layers - 1: + res_acts = res_skip_acts[:, : self.hidden_channels, :] + x = (x + res_acts) * x_mask + output = output + res_skip_acts[:, self.hidden_channels :, :] + else: + output = output + res_skip_acts + return output * x_mask + + def remove_weight_norm(self): + if self.gin_channels != 0: + torch.nn.utils.remove_weight_norm(self.cond_layer) + for l in self.in_layers: + torch.nn.utils.remove_weight_norm(l) + for l in self.res_skip_layers: + torch.nn.utils.remove_weight_norm(l) + + +class ResBlock1(torch.nn.Module): + def __init__(self, channels, kernel_size=3, dilation=(1, 3, 5)): + super(ResBlock1, self).__init__() + self.convs1 = nn.ModuleList( + [ + weight_norm( + Conv1d( + channels, + channels, + kernel_size, + 1, + dilation=dilation[0], + padding=get_padding(kernel_size, dilation[0]), + ) + ), + weight_norm( + Conv1d( + channels, + channels, + kernel_size, + 1, + dilation=dilation[1], + padding=get_padding(kernel_size, dilation[1]), + ) + ), + weight_norm( + Conv1d( + channels, + channels, + kernel_size, + 1, + dilation=dilation[2], + padding=get_padding(kernel_size, dilation[2]), + ) + ), + ] + ) + self.convs1.apply(init_weights) + + self.convs2 = nn.ModuleList( + [ + weight_norm( + Conv1d( + channels, + channels, + kernel_size, + 1, + dilation=1, + padding=get_padding(kernel_size, 1), + ) + ), + weight_norm( + Conv1d( + channels, + channels, + kernel_size, + 1, + dilation=1, + padding=get_padding(kernel_size, 1), + ) + ), + weight_norm( + Conv1d( + channels, + channels, + kernel_size, + 1, + dilation=1, + padding=get_padding(kernel_size, 1), + ) + ), + ] + ) + self.convs2.apply(init_weights) + + def forward(self, x, x_mask=None): + for c1, c2 in zip(self.convs1, self.convs2): + xt = F.leaky_relu(x, LRELU_SLOPE) + if x_mask is not None: + xt = xt * x_mask + xt = c1(xt) + xt = F.leaky_relu(xt, LRELU_SLOPE) + if x_mask is not None: + xt = xt * x_mask + xt = c2(xt) + x = xt + x + if x_mask is not None: + x = x * x_mask + return x + + def remove_weight_norm(self): + for l in self.convs1: + remove_weight_norm(l) + for l in self.convs2: + remove_weight_norm(l) + + +class ResBlock2(torch.nn.Module): + def __init__(self, channels, kernel_size=3, dilation=(1, 3)): + super(ResBlock2, self).__init__() + self.convs = nn.ModuleList( + [ + weight_norm( + Conv1d( + channels, + channels, + kernel_size, + 1, + dilation=dilation[0], + padding=get_padding(kernel_size, dilation[0]), + ) + ), + weight_norm( + Conv1d( + channels, + channels, + kernel_size, + 1, + dilation=dilation[1], + padding=get_padding(kernel_size, dilation[1]), + ) + ), + ] + ) + self.convs.apply(init_weights) + + def forward(self, x, x_mask=None): + for c in self.convs: + xt = F.leaky_relu(x, LRELU_SLOPE) + if x_mask is not None: + xt = xt * x_mask + xt = c(xt) + x = xt + x + if x_mask is not None: + x = x * x_mask + return x + + def remove_weight_norm(self): + for l in self.convs: + remove_weight_norm(l) + + +class Log(nn.Module): + def forward(self, x, x_mask, reverse=False, **kwargs): + if not reverse: + y = torch.log(torch.clamp_min(x, 1e-5)) * x_mask + logdet = torch.sum(-y, [1, 2]) + return y, logdet + else: + x = torch.exp(x) * x_mask + return x + + +class Flip(nn.Module): + def forward(self, x, *args, reverse=False, **kwargs): + x = torch.flip(x, [1]) + if not reverse: + logdet = torch.zeros(x.size(0)).to(dtype=x.dtype, device=x.device) + return x, logdet + else: + return x + + +class ElementwiseAffine(nn.Module): + def __init__(self, channels): + super().__init__() + self.channels = channels + self.m = nn.Parameter(torch.zeros(channels, 1)) + self.logs = nn.Parameter(torch.zeros(channels, 1)) + + def forward(self, x, x_mask, reverse=False, **kwargs): + if not reverse: + y = self.m + torch.exp(self.logs) * x + y = y * x_mask + logdet = torch.sum(self.logs * x_mask, [1, 2]) + return y, logdet + else: + x = (x - self.m) * torch.exp(-self.logs) * x_mask + return x + + +class ResidualCouplingLayer(nn.Module): + def __init__( + self, + channels, + hidden_channels, + kernel_size, + dilation_rate, + n_layers, + p_dropout=0, + gin_channels=0, + mean_only=False, + ): + assert channels % 2 == 0, "channels should be divisible by 2" + 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.half_channels = channels // 2 + self.mean_only = mean_only + + self.pre = nn.Conv1d(self.half_channels, hidden_channels, 1) + self.enc = WN( + hidden_channels, + kernel_size, + dilation_rate, + n_layers, + p_dropout=p_dropout, + gin_channels=gin_channels, + ) + self.post = nn.Conv1d(hidden_channels, self.half_channels * (2 - mean_only), 1) + self.post.weight.data.zero_() + self.post.bias.data.zero_() + + def forward(self, x, x_mask, g=None, reverse=False): + x0, x1 = torch.split(x, [self.half_channels] * 2, 1) + h = self.pre(x0) * x_mask + h = self.enc(h, x_mask, g=g) + stats = self.post(h) * x_mask + if not self.mean_only: + m, logs = torch.split(stats, [self.half_channels] * 2, 1) + else: + m = stats + logs = torch.zeros_like(m) + + if not reverse: + x1 = m + x1 * torch.exp(logs) * x_mask + x = torch.cat([x0, x1], 1) + logdet = torch.sum(logs, [1, 2]) + return x, logdet + else: + x1 = (x1 - m) * torch.exp(-logs) * x_mask + x = torch.cat([x0, x1], 1) + return x + + def remove_weight_norm(self): + self.enc.remove_weight_norm() + + +class ConvFlow(nn.Module): + def __init__( + self, + in_channels, + filter_channels, + kernel_size, + n_layers, + num_bins=10, + tail_bound=5.0, + ): + super().__init__() + self.in_channels = in_channels + self.filter_channels = filter_channels + self.kernel_size = kernel_size + self.n_layers = n_layers + self.num_bins = num_bins + self.tail_bound = tail_bound + self.half_channels = in_channels // 2 + + self.pre = nn.Conv1d(self.half_channels, filter_channels, 1) + self.convs = DDSConv(filter_channels, kernel_size, n_layers, p_dropout=0.0) + self.proj = nn.Conv1d( + filter_channels, self.half_channels * (num_bins * 3 - 1), 1 + ) + self.proj.weight.data.zero_() + self.proj.bias.data.zero_() + + def forward(self, x, x_mask, g=None, reverse=False): + x0, x1 = torch.split(x, [self.half_channels] * 2, 1) + h = self.pre(x0) + h = self.convs(h, x_mask, g=g) + h = self.proj(h) * x_mask + + b, c, t = x0.shape + h = h.reshape(b, c, -1, t).permute(0, 1, 3, 2) # [b, cx?, t] -> [b, c, t, ?] + + unnormalized_widths = h[..., : self.num_bins] / math.sqrt(self.filter_channels) + unnormalized_heights = h[..., self.num_bins : 2 * self.num_bins] / math.sqrt( + self.filter_channels + ) + unnormalized_derivatives = h[..., 2 * self.num_bins :] + + x1, logabsdet = piecewise_rational_quadratic_transform( + x1, + unnormalized_widths, + unnormalized_heights, + unnormalized_derivatives, + inverse=reverse, + tails="linear", + tail_bound=self.tail_bound, + ) + + x = torch.cat([x0, x1], 1) * x_mask + logdet = torch.sum(logabsdet * x_mask, [1, 2]) + if not reverse: + return x, logdet + else: + return x diff --git a/rvc/infer_pack/transforms.py b/rvc/infer_pack/transforms.py new file mode 100644 index 0000000..a11f799 --- /dev/null +++ b/rvc/infer_pack/transforms.py @@ -0,0 +1,209 @@ +import torch +from torch.nn import functional as F + +import numpy as np + + +DEFAULT_MIN_BIN_WIDTH = 1e-3 +DEFAULT_MIN_BIN_HEIGHT = 1e-3 +DEFAULT_MIN_DERIVATIVE = 1e-3 + + +def piecewise_rational_quadratic_transform( + inputs, + unnormalized_widths, + unnormalized_heights, + unnormalized_derivatives, + inverse=False, + tails=None, + tail_bound=1.0, + min_bin_width=DEFAULT_MIN_BIN_WIDTH, + min_bin_height=DEFAULT_MIN_BIN_HEIGHT, + min_derivative=DEFAULT_MIN_DERIVATIVE, +): + if tails is None: + spline_fn = rational_quadratic_spline + spline_kwargs = {} + else: + spline_fn = unconstrained_rational_quadratic_spline + spline_kwargs = {"tails": tails, "tail_bound": tail_bound} + + outputs, logabsdet = spline_fn( + inputs=inputs, + unnormalized_widths=unnormalized_widths, + unnormalized_heights=unnormalized_heights, + unnormalized_derivatives=unnormalized_derivatives, + inverse=inverse, + min_bin_width=min_bin_width, + min_bin_height=min_bin_height, + min_derivative=min_derivative, + **spline_kwargs + ) + return outputs, logabsdet + + +def searchsorted(bin_locations, inputs, eps=1e-6): + bin_locations[..., -1] += eps + return torch.sum(inputs[..., None] >= bin_locations, dim=-1) - 1 + + +def unconstrained_rational_quadratic_spline( + inputs, + unnormalized_widths, + unnormalized_heights, + unnormalized_derivatives, + inverse=False, + tails="linear", + tail_bound=1.0, + min_bin_width=DEFAULT_MIN_BIN_WIDTH, + min_bin_height=DEFAULT_MIN_BIN_HEIGHT, + min_derivative=DEFAULT_MIN_DERIVATIVE, +): + inside_interval_mask = (inputs >= -tail_bound) & (inputs <= tail_bound) + outside_interval_mask = ~inside_interval_mask + + outputs = torch.zeros_like(inputs) + logabsdet = torch.zeros_like(inputs) + + if tails == "linear": + unnormalized_derivatives = F.pad(unnormalized_derivatives, pad=(1, 1)) + constant = np.log(np.exp(1 - min_derivative) - 1) + unnormalized_derivatives[..., 0] = constant + unnormalized_derivatives[..., -1] = constant + + outputs[outside_interval_mask] = inputs[outside_interval_mask] + logabsdet[outside_interval_mask] = 0 + else: + raise RuntimeError("{} tails are not implemented.".format(tails)) + + ( + outputs[inside_interval_mask], + logabsdet[inside_interval_mask], + ) = rational_quadratic_spline( + inputs=inputs[inside_interval_mask], + unnormalized_widths=unnormalized_widths[inside_interval_mask, :], + unnormalized_heights=unnormalized_heights[inside_interval_mask, :], + unnormalized_derivatives=unnormalized_derivatives[inside_interval_mask, :], + inverse=inverse, + left=-tail_bound, + right=tail_bound, + bottom=-tail_bound, + top=tail_bound, + min_bin_width=min_bin_width, + min_bin_height=min_bin_height, + min_derivative=min_derivative, + ) + + return outputs, logabsdet + + +def rational_quadratic_spline( + inputs, + unnormalized_widths, + unnormalized_heights, + unnormalized_derivatives, + inverse=False, + left=0.0, + right=1.0, + bottom=0.0, + top=1.0, + min_bin_width=DEFAULT_MIN_BIN_WIDTH, + min_bin_height=DEFAULT_MIN_BIN_HEIGHT, + min_derivative=DEFAULT_MIN_DERIVATIVE, +): + if torch.min(inputs) < left or torch.max(inputs) > right: + raise ValueError("Input to a transform is not within its domain") + + num_bins = unnormalized_widths.shape[-1] + + if min_bin_width * num_bins > 1.0: + raise ValueError("Minimal bin width too large for the number of bins") + if min_bin_height * num_bins > 1.0: + raise ValueError("Minimal bin height too large for the number of bins") + + widths = F.softmax(unnormalized_widths, dim=-1) + widths = min_bin_width + (1 - min_bin_width * num_bins) * widths + cumwidths = torch.cumsum(widths, dim=-1) + cumwidths = F.pad(cumwidths, pad=(1, 0), mode="constant", value=0.0) + cumwidths = (right - left) * cumwidths + left + cumwidths[..., 0] = left + cumwidths[..., -1] = right + widths = cumwidths[..., 1:] - cumwidths[..., :-1] + + derivatives = min_derivative + F.softplus(unnormalized_derivatives) + + heights = F.softmax(unnormalized_heights, dim=-1) + heights = min_bin_height + (1 - min_bin_height * num_bins) * heights + cumheights = torch.cumsum(heights, dim=-1) + cumheights = F.pad(cumheights, pad=(1, 0), mode="constant", value=0.0) + cumheights = (top - bottom) * cumheights + bottom + cumheights[..., 0] = bottom + cumheights[..., -1] = top + heights = cumheights[..., 1:] - cumheights[..., :-1] + + if inverse: + bin_idx = searchsorted(cumheights, inputs)[..., None] + else: + bin_idx = searchsorted(cumwidths, inputs)[..., None] + + input_cumwidths = cumwidths.gather(-1, bin_idx)[..., 0] + input_bin_widths = widths.gather(-1, bin_idx)[..., 0] + + input_cumheights = cumheights.gather(-1, bin_idx)[..., 0] + delta = heights / widths + input_delta = delta.gather(-1, bin_idx)[..., 0] + + input_derivatives = derivatives.gather(-1, bin_idx)[..., 0] + input_derivatives_plus_one = derivatives[..., 1:].gather(-1, bin_idx)[..., 0] + + input_heights = heights.gather(-1, bin_idx)[..., 0] + + if inverse: + a = (inputs - input_cumheights) * ( + input_derivatives + input_derivatives_plus_one - 2 * input_delta + ) + input_heights * (input_delta - input_derivatives) + b = input_heights * input_derivatives - (inputs - input_cumheights) * ( + input_derivatives + input_derivatives_plus_one - 2 * input_delta + ) + c = -input_delta * (inputs - input_cumheights) + + discriminant = b.pow(2) - 4 * a * c + assert (discriminant >= 0).all() + + root = (2 * c) / (-b - torch.sqrt(discriminant)) + outputs = root * input_bin_widths + input_cumwidths + + theta_one_minus_theta = root * (1 - root) + denominator = input_delta + ( + (input_derivatives + input_derivatives_plus_one - 2 * input_delta) + * theta_one_minus_theta + ) + derivative_numerator = input_delta.pow(2) * ( + input_derivatives_plus_one * root.pow(2) + + 2 * input_delta * theta_one_minus_theta + + input_derivatives * (1 - root).pow(2) + ) + logabsdet = torch.log(derivative_numerator) - 2 * torch.log(denominator) + + return outputs, -logabsdet + else: + theta = (inputs - input_cumwidths) / input_bin_widths + theta_one_minus_theta = theta * (1 - theta) + + numerator = input_heights * ( + input_delta * theta.pow(2) + input_derivatives * theta_one_minus_theta + ) + denominator = input_delta + ( + (input_derivatives + input_derivatives_plus_one - 2 * input_delta) + * theta_one_minus_theta + ) + outputs = input_cumheights + numerator / denominator + + derivative_numerator = input_delta.pow(2) * ( + input_derivatives_plus_one * theta.pow(2) + + 2 * input_delta * theta_one_minus_theta + + input_derivatives * (1 - theta).pow(2) + ) + logabsdet = torch.log(derivative_numerator) - 2 * torch.log(denominator) + + return outputs, logabsdet diff --git a/rvc/main.py b/rvc/main.py new file mode 100644 index 0000000..bf64c1f --- /dev/null +++ b/rvc/main.py @@ -0,0 +1,110 @@ +import gc +import os + +import soundfile as sf +import sox +#from pedalboard import Pedalboard, Reverb, Compressor, HighpassFilter +#from pedalboard.io import AudioFile + +from .rvc import Config, load_hubert, get_vc, rvc_infer + +BASE_DIR = os.path.dirname(os.path.abspath(__file__)) + +rvc_models_dir = os.path.join(BASE_DIR, 'rvc_models') + + +def get_rvc_model(voice_model): + rvc_model_filename, rvc_index_filename = None, None + model_dir = os.path.join(rvc_models_dir, voice_model) + for file in os.listdir(model_dir): + ext = os.path.splitext(file)[1] + if ext == '.pth': + rvc_model_filename = file + if ext == '.index': + rvc_index_filename = file + + if rvc_model_filename is None: + error_msg = f'No model file exists in {model_dir}.' + raise NameError(error_msg) + + return os.path.join(model_dir, rvc_model_filename), os.path.join(model_dir, rvc_index_filename) if rvc_index_filename else '' + + +def pitch_shift(audio_path, pitch_change): + output_path = f'{os.path.splitext(audio_path)[0]}_p{pitch_change}.wav' + if not os.path.exists(output_path): + y, sr = sf.read(audio_path) + tfm = sox.Transformer() + tfm.pitch(pitch_change) + y_shifted = tfm.build_array(input_array=y, sample_rate_in=sr) + sf.write(output_path, y_shifted, sr) + + return output_path + + +def voice_change(voice_model, vocals_path, output_path, pitch_change, f0_method, index_rate, filter_radius, rms_mix_rate, protect, crepe_hop_length): + rvc_model_path, rvc_index_path = get_rvc_model(voice_model) + device = 'cuda:0' + config = Config(device, True) + hubert_model = load_hubert(device, config.is_half, os.path.join(rvc_models_dir, 'hubert_base.pt')) + cpt, version, net_g, tgt_sr, vc = get_vc(device, config.is_half, config, rvc_model_path) + + # convert main vocals + try: + rvc_infer(rvc_index_path, index_rate, vocals_path, output_path, pitch_change, f0_method, cpt, version, net_g, filter_radius, tgt_sr, rms_mix_rate, protect, crepe_hop_length, vc, hubert_model) + except Exception as err: + print(err) + finally: + del hubert_model, cpt + gc.collect() + +''' +def add_audio_effects(audio_path, reverb_rm_size, reverb_wet, reverb_dry, reverb_damping): + output_path = f'{os.path.splitext(audio_path)[0]}_mixed.wav' + + # Initialize audio effects plugins + board = Pedalboard( + [ + HighpassFilter(), + Compressor(ratio=4, threshold_db=-15), + Reverb(room_size=reverb_rm_size, dry_level=reverb_dry, wet_level=reverb_wet, damping=reverb_damping) + ] + ) + + with AudioFile(audio_path) as f: + with AudioFile(output_path, 'w', f.samplerate, f.num_channels) as o: + # Read one second of audio at a time, until the file is empty: + while f.tell() < f.frames: + chunk = f.read(int(f.samplerate)) + effected = board(chunk, f.samplerate, reset=False) + o.write(effected) + + return output_path +''' + + +def song_cover_pipeline(song_input, pitch_change_oct, voice_model='miku', + main_gain=0, backup_gain=0, inst_gain=0, index_rate=0.5, filter_radius=3, + rms_mix_rate=0.25, f0_method='rmvpe', crepe_hop_length=128, protect=0.33, pitch_change_sem=0, + reverb_rm_size=0.15, reverb_wet=0.2, reverb_dry=0.8, reverb_damping=0.7, output_format='mp3'): + try: + orig_vocals_path = song_input.strip('\"') + + pitch_change = pitch_change_oct * 12 + pitch_change_sem + ai_vocals_path = os.path.join( + os.path.dirname(song_input), + f'{os.path.splitext(os.path.basename(orig_vocals_path))[0]}_{voice_model}_p{pitch_change}_i{index_rate}_fr{filter_radius}_rms{rms_mix_rate}_pro{protect}_{f0_method}{"" if f0_method != "mangio-crepe" else f"_{crepe_hop_length}"}.wav' + ) + print(f'Output file path will be: {ai_vocals_path}') + + if not os.path.exists(ai_vocals_path): + print('[~] Converting voice using RVC...') + voice_change(voice_model, orig_vocals_path, ai_vocals_path, pitch_change, f0_method, index_rate, filter_radius, rms_mix_rate, protect, crepe_hop_length) + + #display_progress('[~] Applying audio effects to Vocals...', 0.8, is_webui, progress) + #ai_vocals_mixed_path = add_audio_effects(ai_vocals_path, reverb_rm_size, reverb_wet, reverb_dry, reverb_damping) + + return ai_vocals_path + + except Exception as e: + print(e) diff --git a/rvc/my_utils.py b/rvc/my_utils.py new file mode 100644 index 0000000..a525839 --- /dev/null +++ b/rvc/my_utils.py @@ -0,0 +1,21 @@ +import ffmpeg +import numpy as np + + +def load_audio(file, sr): + try: + # https://github.com/openai/whisper/blob/main/whisper/audio.py#L26 + # This launches a subprocess to decode audio while down-mixing and resampling as necessary. + # Requires the ffmpeg CLI and `ffmpeg-python` package to be installed. + file = ( + file.strip(" ").strip('"').strip("\n").strip('"').strip(" ") + ) # 防止小白拷路径头尾带了空格和"和回车 + out, _ = ( + ffmpeg.input(file, threads=0) + .output("-", format="f32le", acodec="pcm_f32le", ac=1, ar=sr) + .run(cmd=["ffmpeg", "-nostdin"], capture_stdout=True, capture_stderr=True) + ) + except Exception as e: + raise RuntimeError(f"Failed to load audio: {e}") + + return np.frombuffer(out, np.float32).flatten() diff --git a/rvc/rmvpe.py b/rvc/rmvpe.py new file mode 100644 index 0000000..7e83aa8 --- /dev/null +++ b/rvc/rmvpe.py @@ -0,0 +1,409 @@ +import numpy as np +import torch +import torch.nn as nn +import torch.nn.functional as F +from librosa.filters import mel + + +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 + + +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 diff --git a/rvc/rvc.ipynb b/rvc/rvc.ipynb new file mode 100644 index 0000000..01246ff --- /dev/null +++ b/rvc/rvc.ipynb @@ -0,0 +1,110 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 1, + "metadata": {}, + "outputs": [ + { + "name": "stderr", + "output_type": "stream", + "text": [ + "2024-03-31 05:59:31 | INFO | fairseq.tasks.text_to_speech | Please install tensorboardX: pip install tensorboardX\n", + "2024-03-31 05:59:32 | INFO | faiss.loader | Loading faiss with AVX2 support.\n", + "2024-03-31 05:59:32 | INFO | faiss.loader | Successfully loaded faiss with AVX2 support.\n" + ] + } + ], + "source": [ + "from main import song_cover_pipeline" + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + " % Total % Received % Xferd Average Speed Time Time Time Current\n", + " Dload Upload Total Spent Left Speed\n", + "100 90234 100 90234 0 0 1190k 0 --:--:-- --:--:-- --:--:-- 1174k\n" + ] + } + ], + "source": [ + "!curl -o '../audio/input.ogg' 'https://cdn.discordapp.com/attachments/252959763841679360/1222430418847010868/voice-message.ogg?ex=6616300b&is=6603bb0b&hm=edb5794314702eb535d16e96841587492033420c18e4e2b647947665109d172a&'" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Output file path will be: ../audio/input_miku_p12_i0.5_fr3_rms0.25_pro0.33_rmvpe.wav\n", + "[~] Converting voice using RVC...\n" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "2024-03-31 05:59:45 | INFO | fairseq.tasks.hubert_pretraining | current directory is /home/jshiffer/rvc\n", + "2024-03-31 05:59:45 | INFO | fairseq.tasks.hubert_pretraining | HubertPretrainingTask Config {'_name': 'hubert_pretraining', 'data': 'metadata', 'fine_tuning': False, 'labels': ['km'], 'label_dir': 'label', 'label_rate': 50.0, 'sample_rate': 16000, 'normalize': False, 'enable_padding': False, 'max_keep_size': None, 'max_sample_size': 250000, 'min_sample_size': 32000, 'single_target': False, 'random_crop': True, 'pad_audio': False}\n", + "2024-03-31 05:59:45 | INFO | fairseq.models.hubert.hubert | HubertModel Config: {'_name': 'hubert', 'label_rate': 50.0, 'extractor_mode': default, 'encoder_layers': 12, 'encoder_embed_dim': 768, 'encoder_ffn_embed_dim': 3072, 'encoder_attention_heads': 12, 'activation_fn': gelu, 'layer_type': transformer, 'dropout': 0.1, 'attention_dropout': 0.1, 'activation_dropout': 0.0, 'encoder_layerdrop': 0.05, 'dropout_input': 0.1, 'dropout_features': 0.1, 'final_dim': 256, 'untie_final_proj': True, 'layer_norm_first': False, 'conv_feature_layers': '[(512,10,5)] + [(512,3,2)] * 4 + [(512,2,2)] * 2', 'conv_bias': False, 'logit_temp': 0.1, 'target_glu': False, 'feature_grad_mult': 0.1, 'mask_length': 10, 'mask_prob': 0.8, 'mask_selection': static, 'mask_other': 0.0, 'no_mask_overlap': False, 'mask_min_space': 1, 'mask_channel_length': 10, 'mask_channel_prob': 0.0, 'mask_channel_selection': static, 'mask_channel_other': 0.0, 'no_mask_channel_overlap': False, 'mask_channel_min_space': 1, 'conv_pos': 128, 'conv_pos_groups': 16, 'latent_temp': [2.0, 0.5, 0.999995], 'skip_masked': False, 'skip_nomask': False, 'checkpoint_activations': False, 'required_seq_len_multiple': 2, 'depthwise_conv_kernel_size': 31, 'attn_type': '', 'pos_enc_type': 'abs', 'fp16': False}\n", + "/opt/conda/envs/rvc/lib/python3.10/site-packages/torch/nn/utils/weight_norm.py:28: UserWarning: torch.nn.utils.weight_norm is deprecated in favor of torch.nn.utils.parametrizations.weight_norm.\n", + " warnings.warn(\"torch.nn.utils.weight_norm is deprecated in favor of torch.nn.utils.parametrizations.weight_norm.\")\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "gin_channels: 256 self.spk_embed_dim: 109\n", + "\n" + ] + }, + { + "data": { + "text/plain": [ + "'../audio/input_miku_p12_i0.5_fr3_rms0.25_pro0.33_rmvpe.wav'" + ] + }, + "execution_count": 2, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "song_cover_pipeline('../audio/input.ogg', 1)" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "rvc", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.10.14" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} diff --git a/rvc/rvc.py b/rvc/rvc.py new file mode 100644 index 0000000..c645d76 --- /dev/null +++ b/rvc/rvc.py @@ -0,0 +1,151 @@ +from multiprocessing import cpu_count +from pathlib import Path + +import torch +from fairseq import checkpoint_utils +from scipy.io import wavfile + +from .infer_pack.models import ( + SynthesizerTrnMs256NSFsid, + SynthesizerTrnMs256NSFsid_nono, + SynthesizerTrnMs768NSFsid, + SynthesizerTrnMs768NSFsid_nono, +) +from .my_utils import load_audio +from .vc_infer_pipeline import VC + +BASE_DIR = Path(__file__).resolve().parent.parent + + +class Config: + def __init__(self, device, is_half): + self.device = device + self.is_half = is_half + self.n_cpu = 0 + self.gpu_name = None + self.gpu_mem = None + self.x_pad, self.x_query, self.x_center, self.x_max = self.device_config() + + def device_config(self) -> tuple: + if torch.cuda.is_available(): + i_device = int(self.device.split(":")[-1]) + self.gpu_name = torch.cuda.get_device_name(i_device) + if ( + ("16" in self.gpu_name and "V100" not in self.gpu_name.upper()) + or "P40" in self.gpu_name.upper() + or "1060" in self.gpu_name + or "1070" in self.gpu_name + or "1080" in self.gpu_name + ): + print("16 series/10 series P40 forced single precision") + self.is_half = False + for config_file in ["32k.json", "40k.json", "48k.json"]: + with open(BASE_DIR / "src" / "configs" / config_file, "r") as f: + strr = f.read().replace("true", "false") + with open(BASE_DIR / "src" / "configs" / config_file, "w") as f: + f.write(strr) + with open(BASE_DIR / "src" / "trainset_preprocess_pipeline_print.py", "r") as f: + strr = f.read().replace("3.7", "3.0") + with open(BASE_DIR / "src" / "trainset_preprocess_pipeline_print.py", "w") as f: + f.write(strr) + else: + self.gpu_name = None + self.gpu_mem = int( + torch.cuda.get_device_properties(i_device).total_memory + / 1024 + / 1024 + / 1024 + + 0.4 + ) + if self.gpu_mem <= 4: + with open(BASE_DIR / "src" / "trainset_preprocess_pipeline_print.py", "r") as f: + strr = f.read().replace("3.7", "3.0") + with open(BASE_DIR / "src" / "trainset_preprocess_pipeline_print.py", "w") as f: + f.write(strr) + elif torch.backends.mps.is_available(): + print("No supported N-card found, use MPS for inference") + self.device = "mps" + else: + print("No supported N-card found, use CPU for inference") + self.device = "cpu" + self.is_half = True + + if self.n_cpu == 0: + self.n_cpu = cpu_count() + + if self.is_half: + # 6G memory config + x_pad = 3 + x_query = 10 + x_center = 60 + x_max = 65 + else: + # 5G memory config + x_pad = 1 + x_query = 6 + x_center = 38 + x_max = 41 + + if self.gpu_mem != None and self.gpu_mem <= 4: + x_pad = 1 + x_query = 5 + x_center = 30 + x_max = 32 + + return x_pad, x_query, x_center, x_max + + +def load_hubert(device, is_half, model_path): + models, saved_cfg, task = checkpoint_utils.load_model_ensemble_and_task([model_path], suffix='', ) + hubert = models[0] + hubert = hubert.to(device) + + if is_half: + hubert = hubert.half() + else: + hubert = hubert.float() + + hubert.eval() + return hubert + + +def get_vc(device, is_half, config, model_path): + cpt = torch.load(model_path, map_location='cpu') + if "config" not in cpt or "weight" not in cpt: + raise ValueError(f'Incorrect format for {model_path}. Use a voice model trained using RVC v2 instead.') + + tgt_sr = cpt["config"][-1] + cpt["config"][-3] = cpt["weight"]["emb_g.weight"].shape[0] + if_f0 = cpt.get("f0", 1) + version = cpt.get("version", "v1") + + if version == "v1": + if if_f0 == 1: + net_g = SynthesizerTrnMs256NSFsid(*cpt["config"], is_half=is_half) + else: + net_g = SynthesizerTrnMs256NSFsid_nono(*cpt["config"]) + elif version == "v2": + if if_f0 == 1: + net_g = SynthesizerTrnMs768NSFsid(*cpt["config"], is_half=is_half) + else: + net_g = SynthesizerTrnMs768NSFsid_nono(*cpt["config"]) + + del net_g.enc_q + print(net_g.load_state_dict(cpt["weight"], strict=False)) + net_g.eval().to(device) + + if is_half: + net_g = net_g.half() + else: + net_g = net_g.float() + + vc = VC(tgt_sr, config) + return cpt, version, net_g, tgt_sr, vc + + +def rvc_infer(index_path, index_rate, input_path, output_path, pitch_change, f0_method, cpt, version, net_g, filter_radius, tgt_sr, rms_mix_rate, protect, crepe_hop_length, vc, hubert_model): + audio = load_audio(input_path, 16000) + times = [0, 0, 0] + if_f0 = cpt.get('f0', 1) + audio_opt = vc.pipeline(hubert_model, net_g, 0, audio, input_path, times, pitch_change, f0_method, index_path, index_rate, if_f0, filter_radius, tgt_sr, 0, rms_mix_rate, version, protect, crepe_hop_length) + wavfile.write(output_path, tgt_sr, audio_opt) diff --git a/rvc/rvc_models/hubert_base.pt b/rvc/rvc_models/hubert_base.pt new file mode 100644 index 0000000..7cfe58e Binary files /dev/null and b/rvc/rvc_models/hubert_base.pt differ diff --git a/rvc/rvc_models/miku/metadata.json b/rvc/rvc_models/miku/metadata.json new file mode 100644 index 0000000..9acce71 --- /dev/null +++ b/rvc/rvc_models/miku/metadata.json @@ -0,0 +1,151 @@ +{ + "title": "Literally EVERY Hatsune Miku Voicebank", + "author": { + "name": "princessprettydoll", + "discordUserId": null + }, + "md5": "cbc8c80ec9089c21e0cd676efe9c40f5", + "uploadedAt": "2023-11-22T12:22:35.045Z", + "weightsLink": "https://www.weights.gg/models/clp9qiyw50013rvv6tl2frcpv", + "id": "clp9qiyw50013rvv6tl2frcpv", + "type": "v2", + "tags": [ + "Fictional Character", + "Singer", + "Vocaloid", + "Japanese", + "English", + "Chinese" + ], + "description": "Sampled all the way from V2 to NT Including appends and her English and Chinese voicebanks", + "samples": [], + "files": [ + { + "name": "model.index", + "size": 183873059, + "md5": "73cad4d8ae6c387e7ca0875ca2bd626e" + }, + { + "name": "model.pth", + "size": 57583034, + "md5": "cbc8c80ec9089c21e0cd676efe9c40f5" + } + ], + "torchMetadata": { + "config": { + "spec_channels": 1025, + "segment_size": 32, + "inter_channels": 192, + "hidden_channels": 192, + "filter_channels": 768, + "n_heads": 2, + "n_layers": 6, + "kernel_size": 3, + "p_dropout": 0, + "resblock": "1", + "resblock_kernel_sizes": [ + 3, + 7, + 11 + ], + "resblock_dilation_sizes": [ + [ + 1, + 3, + 5 + ], + [ + 1, + 3, + 5 + ], + [ + 1, + 3, + 5 + ] + ], + "upsample_rates": [ + 12, + 10, + 2, + 2 + ], + "upsample_initial_channel": 512, + "upsample_kernel_sizes": [ + 24, + 20, + 4, + 4 + ], + "emb_channels": null, + "spk_embed_dim": 109, + "gin_channels": 256, + "sr": 48000 + }, + "f0": 1, + "version": "v2", + "extra_info": { + "config": [ + 1025, + 32, + 192, + 192, + 768, + 2, + 6, + 3, + 0, + "1", + [ + 3, + 7, + 11 + ], + [ + [ + 1, + 3, + 5 + ], + [ + 1, + 3, + 5 + ], + [ + 1, + 3, + 5 + ] + ], + [ + 12, + 10, + 2, + 2 + ], + 512, + [ + 24, + 20, + 4, + 4 + ], + 109, + 256, + 48000 + ], + "info": "300epoch", + "sr": "48k", + "f0": 1, + "version": "v2" + } + }, + "url": "https://models.weights.gg/28f532bc-9735-4465-9108-f55583a71658.zip", + "urls": [], + "originalFileList": [ + "HMVBICAE.pth", + "added_IVF1492_Flat_nprobe_1_HMVBICAE_v2.index" + ] +} \ No newline at end of file diff --git a/rvc/rvc_models/miku/model.index b/rvc/rvc_models/miku/model.index new file mode 100644 index 0000000..840957d Binary files /dev/null and b/rvc/rvc_models/miku/model.index differ diff --git a/rvc/rvc_models/miku/model.pth b/rvc/rvc_models/miku/model.pth new file mode 100644 index 0000000..5771586 Binary files /dev/null and b/rvc/rvc_models/miku/model.pth differ diff --git a/rvc/rvc_models/rmvpe.pt b/rvc/rvc_models/rmvpe.pt new file mode 100644 index 0000000..42f4ce4 Binary files /dev/null and b/rvc/rvc_models/rmvpe.pt differ diff --git a/rvc/vc_infer_pipeline.py b/rvc/vc_infer_pipeline.py new file mode 100644 index 0000000..a28a424 --- /dev/null +++ b/rvc/vc_infer_pipeline.py @@ -0,0 +1,653 @@ +from functools import lru_cache +from time import time as ttime + +import faiss +import librosa +import numpy as np +import os +import parselmouth +import pyworld +import sys +import torch +import torch.nn.functional as F +import torchcrepe +import traceback +from scipy import signal +from torch import Tensor + +BASE_DIR = os.path.dirname(os.path.abspath(__file__)) +#now_dir = os.path.join(BASE_DIR, 'src') +#sys.path.append(now_dir) + +bh, ah = signal.butter(N=5, Wn=48, btype="high", fs=16000) + +input_audio_path2wav = {} + + +@lru_cache +def cache_harvest_f0(input_audio_path, fs, f0max, f0min, frame_period): + audio = input_audio_path2wav[input_audio_path] + f0, t = pyworld.harvest( + audio, + fs=fs, + f0_ceil=f0max, + f0_floor=f0min, + frame_period=frame_period, + ) + f0 = pyworld.stonemask(audio, f0, t, fs) + return f0 + + +def change_rms(data1, sr1, data2, sr2, rate): # 1是输入音频，2是输出音频,rate是2的占比 + # print(data1.max(),data2.max()) + rms1 = librosa.feature.rms( + y=data1, frame_length=sr1 // 2 * 2, hop_length=sr1 // 2 + ) # 每半秒一个点 + rms2 = librosa.feature.rms(y=data2, frame_length=sr2 // 2 * 2, hop_length=sr2 // 2) + rms1 = torch.from_numpy(rms1) + rms1 = F.interpolate( + rms1.unsqueeze(0), size=data2.shape[0], mode="linear" + ).squeeze() + rms2 = torch.from_numpy(rms2) + rms2 = F.interpolate( + rms2.unsqueeze(0), size=data2.shape[0], mode="linear" + ).squeeze() + rms2 = torch.max(rms2, torch.zeros_like(rms2) + 1e-6) + data2 *= ( + torch.pow(rms1, torch.tensor(1 - rate)) + * torch.pow(rms2, torch.tensor(rate - 1)) + ).numpy() + return data2 + + +class VC(object): + def __init__(self, tgt_sr, config): + self.x_pad, self.x_query, self.x_center, self.x_max, self.is_half = ( + config.x_pad, + config.x_query, + config.x_center, + config.x_max, + config.is_half, + ) + self.sr = 16000 # hubert输入采样率 + self.window = 160 # 每帧点数 + self.t_pad = self.sr * self.x_pad # 每条前后pad时间 + self.t_pad_tgt = tgt_sr * self.x_pad + self.t_pad2 = self.t_pad * 2 + self.t_query = self.sr * self.x_query # 查询切点前后查询时间 + self.t_center = self.sr * self.x_center # 查询切点位置 + self.t_max = self.sr * self.x_max # 免查询时长阈值 + self.device = config.device + + # Fork Feature: Get the best torch device to use for f0 algorithms that require a torch device. Will return the type (torch.device) + def get_optimal_torch_device(self, index: int = 0) -> torch.device: + # Get cuda device + if torch.cuda.is_available(): + return torch.device( + f"cuda:{index % torch.cuda.device_count()}" + ) # Very fast + elif torch.backends.mps.is_available(): + return torch.device("mps") + # Insert an else here to grab "xla" devices if available. TO DO later. Requires the torch_xla.core.xla_model library + # Else wise return the "cpu" as a torch device, + return torch.device("cpu") + + # Fork Feature: Compute f0 with the crepe method + def get_f0_crepe_computation( + self, + x, + f0_min, + f0_max, + p_len, + hop_length=160, # 512 before. Hop length changes the speed that the voice jumps to a different dramatic pitch. Lower hop lengths means more pitch accuracy but longer inference time. + model="full", # Either use crepe-tiny "tiny" or crepe "full". Default is full + ): + x = x.astype( + np.float32 + ) # fixes the F.conv2D exception. We needed to convert double to float. + x /= np.quantile(np.abs(x), 0.999) + torch_device = self.get_optimal_torch_device() + audio = torch.from_numpy(x).to(torch_device, copy=True) + audio = torch.unsqueeze(audio, dim=0) + if audio.ndim == 2 and audio.shape[0] > 1: + audio = torch.mean(audio, dim=0, keepdim=True).detach() + audio = audio.detach() + print("Initiating prediction with a crepe_hop_length of: " + str(hop_length)) + pitch: Tensor = torchcrepe.predict( + audio, + self.sr, + hop_length, + f0_min, + f0_max, + model, + batch_size=hop_length * 2, + device=torch_device, + pad=True, + ) + p_len = p_len or x.shape[0] // hop_length + # Resize the pitch for final f0 + source = np.array(pitch.squeeze(0).cpu().float().numpy()) + source[source < 0.001] = np.nan + target = np.interp( + np.arange(0, len(source) * p_len, len(source)) / p_len, + np.arange(0, len(source)), + source, + ) + f0 = np.nan_to_num(target) + return f0 # Resized f0 + + def get_f0_official_crepe_computation( + self, + x, + f0_min, + f0_max, + model="full", + ): + # Pick a batch size that doesn't cause memory errors on your gpu + batch_size = 512 + # Compute pitch using first gpu + audio = torch.tensor(np.copy(x))[None].float() + f0, pd = torchcrepe.predict( + audio, + self.sr, + self.window, + f0_min, + f0_max, + model, + batch_size=batch_size, + device=self.device, + return_periodicity=True, + ) + pd = torchcrepe.filter.median(pd, 3) + f0 = torchcrepe.filter.mean(f0, 3) + f0[pd < 0.1] = 0 + f0 = f0[0].cpu().numpy() + return f0 + + # Fork Feature: Compute pYIN f0 method + def get_f0_pyin_computation(self, x, f0_min, f0_max): + y, sr = librosa.load("saudio/Sidney.wav", self.sr, mono=True) + f0, _, _ = librosa.pyin(y, sr=self.sr, fmin=f0_min, fmax=f0_max) + f0 = f0[1:] # Get rid of extra first frame + return f0 + + # Fork Feature: Acquire median hybrid f0 estimation calculation + def get_f0_hybrid_computation( + self, + methods_str, + input_audio_path, + x, + f0_min, + f0_max, + p_len, + filter_radius, + crepe_hop_length, + time_step, + ): + # Get various f0 methods from input to use in the computation stack + s = methods_str + s = s.split("hybrid")[1] + s = s.replace("[", "").replace("]", "") + methods = s.split("+") + f0_computation_stack = [] + + print("Calculating f0 pitch estimations for methods: %s" % str(methods)) + x = x.astype(np.float32) + x /= np.quantile(np.abs(x), 0.999) + # Get f0 calculations for all methods specified + for method in methods: + f0 = None + if method == "pm": + f0 = ( + parselmouth.Sound(x, self.sr) + .to_pitch_ac( + time_step=time_step / 1000, + voicing_threshold=0.6, + pitch_floor=f0_min, + pitch_ceiling=f0_max, + ) + .selected_array["frequency"] + ) + pad_size = (p_len - len(f0) + 1) // 2 + if pad_size > 0 or p_len - len(f0) - pad_size > 0: + f0 = np.pad( + f0, [[pad_size, p_len - len(f0) - pad_size]], mode="constant" + ) + elif method == "crepe": + f0 = self.get_f0_official_crepe_computation(x, f0_min, f0_max) + f0 = f0[1:] # Get rid of extra first frame + elif method == "crepe-tiny": + f0 = self.get_f0_official_crepe_computation(x, f0_min, f0_max, "tiny") + f0 = f0[1:] # Get rid of extra first frame + elif method == "mangio-crepe": + f0 = self.get_f0_crepe_computation( + x, f0_min, f0_max, p_len, crepe_hop_length + ) + elif method == "mangio-crepe-tiny": + f0 = self.get_f0_crepe_computation( + x, f0_min, f0_max, p_len, crepe_hop_length, "tiny" + ) + elif method == "harvest": + f0 = cache_harvest_f0(input_audio_path, self.sr, f0_max, f0_min, 10) + if filter_radius > 2: + f0 = signal.medfilt(f0, 3) + f0 = f0[1:] # Get rid of first frame. + elif method == "dio": # Potentially buggy? + f0, t = pyworld.dio( + 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 = f0[1:] + # elif method == "pyin": Not Working just yet + # f0 = self.get_f0_pyin_computation(x, f0_min, f0_max) + # Push method to the stack + f0_computation_stack.append(f0) + + for fc in f0_computation_stack: + print(len(fc)) + + print("Calculating hybrid median f0 from the stack of: %s" % str(methods)) + f0_median_hybrid = None + if len(f0_computation_stack) == 1: + f0_median_hybrid = f0_computation_stack[0] + else: + f0_median_hybrid = np.nanmedian(f0_computation_stack, axis=0) + return f0_median_hybrid + + def get_f0( + self, + input_audio_path, + x, + p_len, + f0_up_key, + f0_method, + filter_radius, + crepe_hop_length, + inp_f0=None, + ): + global input_audio_path2wav + time_step = self.window / self.sr * 1000 + 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) + if f0_method == "pm": + f0 = ( + parselmouth.Sound(x, self.sr) + .to_pitch_ac( + time_step=time_step / 1000, + voicing_threshold=0.6, + pitch_floor=f0_min, + pitch_ceiling=f0_max, + ) + .selected_array["frequency"] + ) + pad_size = (p_len - len(f0) + 1) // 2 + if pad_size > 0 or p_len - len(f0) - pad_size > 0: + f0 = np.pad( + f0, [[pad_size, p_len - len(f0) - pad_size]], mode="constant" + ) + elif f0_method == "harvest": + input_audio_path2wav[input_audio_path] = x.astype(np.double) + f0 = cache_harvest_f0(input_audio_path, self.sr, f0_max, f0_min, 10) + if filter_radius > 2: + f0 = signal.medfilt(f0, 3) + elif f0_method == "dio": # Potentially Buggy? + f0, t = pyworld.dio( + 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) + elif f0_method == "crepe": + f0 = self.get_f0_official_crepe_computation(x, f0_min, f0_max) + elif f0_method == "crepe-tiny": + f0 = self.get_f0_official_crepe_computation(x, f0_min, f0_max, "tiny") + elif f0_method == "mangio-crepe": + f0 = self.get_f0_crepe_computation( + x, f0_min, f0_max, p_len, crepe_hop_length + ) + elif f0_method == "mangio-crepe-tiny": + f0 = self.get_f0_crepe_computation( + x, f0_min, f0_max, p_len, crepe_hop_length, "tiny" + ) + elif f0_method == "rmvpe": + if hasattr(self, "model_rmvpe") == False: + from .rmvpe import RMVPE + + self.model_rmvpe = RMVPE( + os.path.join(BASE_DIR, 'rvc_models', 'rmvpe.pt'), is_half=self.is_half, device=self.device + ) + f0 = self.model_rmvpe.infer_from_audio(x, thred=0.03) + + elif "hybrid" in f0_method: + # Perform hybrid median pitch estimation + input_audio_path2wav[input_audio_path] = x.astype(np.double) + f0 = self.get_f0_hybrid_computation( + f0_method, + input_audio_path, + x, + f0_min, + f0_max, + p_len, + filter_radius, + crepe_hop_length, + time_step, + ) + + 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[self.x_pad * tf0 : self.x_pad * tf0 + len(replace_f0)].shape[0] + f0[self.x_pad * tf0 : self.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 vc( + self, + model, + net_g, + sid, + audio0, + pitch, + pitchf, + times, + index, + big_npy, + index_rate, + version, + protect, + ): # ,file_index,file_big_npy + feats = torch.from_numpy(audio0) + if self.is_half: + feats = feats.half() + else: + feats = feats.float() + if feats.dim() == 2: # double channels + feats = feats.mean(-1) + assert feats.dim() == 1, feats.dim() + feats = feats.view(1, -1) + padding_mask = torch.BoolTensor(feats.shape).to(self.device).fill_(False) + + inputs = { + "source": feats.to(self.device), + "padding_mask": padding_mask, + "output_layer": 9 if version == "v1" else 12, + } + t0 = ttime() + with torch.no_grad(): + logits = model.extract_features(**inputs) + feats = model.final_proj(logits[0]) if version == "v1" else logits[0] + if protect < 0.5 and pitch != None and pitchf != None: + feats0 = feats.clone() + if ( + isinstance(index, type(None)) == False + and isinstance(big_npy, type(None)) == False + and index_rate != 0 + ): + npy = feats[0].cpu().numpy() + if self.is_half: + npy = npy.astype("float32") + + # _, I = index.search(npy, 1) + # npy = big_npy[I.squeeze()] + + score, ix = index.search(npy, k=8) + weight = np.square(1 / score) + weight /= weight.sum(axis=1, keepdims=True) + npy = np.sum(big_npy[ix] * np.expand_dims(weight, axis=2), axis=1) + + if self.is_half: + npy = npy.astype("float16") + feats = ( + torch.from_numpy(npy).unsqueeze(0).to(self.device) * index_rate + + (1 - index_rate) * feats + ) + + feats = F.interpolate(feats.permute(0, 2, 1), scale_factor=2).permute(0, 2, 1) + if protect < 0.5 and pitch != None and pitchf != None: + feats0 = F.interpolate(feats0.permute(0, 2, 1), scale_factor=2).permute( + 0, 2, 1 + ) + t1 = ttime() + p_len = audio0.shape[0] // self.window + if feats.shape[1] < p_len: + p_len = feats.shape[1] + if pitch != None and pitchf != None: + pitch = pitch[:, :p_len] + pitchf = pitchf[:, :p_len] + + if protect < 0.5 and pitch != None and pitchf != None: + pitchff = pitchf.clone() + pitchff[pitchf > 0] = 1 + pitchff[pitchf < 1] = protect + pitchff = pitchff.unsqueeze(-1) + feats = feats * pitchff + feats0 * (1 - pitchff) + feats = feats.to(feats0.dtype) + p_len = torch.tensor([p_len], device=self.device).long() + with torch.no_grad(): + if pitch != None and pitchf != None: + audio1 = ( + (net_g.infer(feats, p_len, pitch, pitchf, sid)[0][0, 0]) + .data.cpu() + .float() + .numpy() + ) + else: + audio1 = ( + (net_g.infer(feats, p_len, sid)[0][0, 0]).data.cpu().float().numpy() + ) + del feats, p_len, padding_mask + if torch.cuda.is_available(): + torch.cuda.empty_cache() + t2 = ttime() + times[0] += t1 - t0 + times[2] += t2 - t1 + return audio1 + + def pipeline( + self, + model, + net_g, + sid, + audio, + input_audio_path, + times, + f0_up_key, + f0_method, + file_index, + # file_big_npy, + index_rate, + if_f0, + filter_radius, + tgt_sr, + resample_sr, + rms_mix_rate, + version, + protect, + crepe_hop_length, + f0_file=None, + ): + if ( + file_index != "" + # and file_big_npy != "" + # and os.path.exists(file_big_npy) == True + and os.path.exists(file_index) == True + and index_rate != 0 + ): + try: + index = faiss.read_index(file_index) + # big_npy = np.load(file_big_npy) + big_npy = index.reconstruct_n(0, index.ntotal) + except: + traceback.print_exc() + index = big_npy = None + else: + index = big_npy = None + audio = signal.filtfilt(bh, ah, audio) + audio_pad = np.pad(audio, (self.window // 2, self.window // 2), mode="reflect") + opt_ts = [] + if audio_pad.shape[0] > self.t_max: + audio_sum = np.zeros_like(audio) + for i in range(self.window): + audio_sum += audio_pad[i : i - self.window] + for t in range(self.t_center, audio.shape[0], self.t_center): + opt_ts.append( + t + - self.t_query + + np.where( + np.abs(audio_sum[t - self.t_query : t + self.t_query]) + == np.abs(audio_sum[t - self.t_query : t + self.t_query]).min() + )[0][0] + ) + s = 0 + audio_opt = [] + t = None + t1 = ttime() + audio_pad = np.pad(audio, (self.t_pad, self.t_pad), mode="reflect") + p_len = audio_pad.shape[0] // self.window + inp_f0 = None + if hasattr(f0_file, "name") == True: + try: + with open(f0_file.name, "r") as f: + lines = f.read().strip("\n").split("\n") + inp_f0 = [] + for line in lines: + inp_f0.append([float(i) for i in line.split(",")]) + inp_f0 = np.array(inp_f0, dtype="float32") + except: + traceback.print_exc() + sid = torch.tensor(sid, device=self.device).unsqueeze(0).long() + pitch, pitchf = None, None + if if_f0 == 1: + pitch, pitchf = self.get_f0( + input_audio_path, + audio_pad, + p_len, + f0_up_key, + f0_method, + filter_radius, + crepe_hop_length, + inp_f0, + ) + pitch = pitch[:p_len] + pitchf = pitchf[:p_len] + if self.device == "mps": + pitchf = pitchf.astype(np.float32) + pitch = torch.tensor(pitch, device=self.device).unsqueeze(0).long() + pitchf = torch.tensor(pitchf, device=self.device).unsqueeze(0).float() + t2 = ttime() + times[1] += t2 - t1 + for t in opt_ts: + t = t // self.window * self.window + if if_f0 == 1: + audio_opt.append( + self.vc( + model, + net_g, + sid, + audio_pad[s : t + self.t_pad2 + self.window], + pitch[:, s // self.window : (t + self.t_pad2) // self.window], + pitchf[:, s // self.window : (t + self.t_pad2) // self.window], + times, + index, + big_npy, + index_rate, + version, + protect, + )[self.t_pad_tgt : -self.t_pad_tgt] + ) + else: + audio_opt.append( + self.vc( + model, + net_g, + sid, + audio_pad[s : t + self.t_pad2 + self.window], + None, + None, + times, + index, + big_npy, + index_rate, + version, + protect, + )[self.t_pad_tgt : -self.t_pad_tgt] + ) + s = t + if if_f0 == 1: + audio_opt.append( + self.vc( + model, + net_g, + sid, + audio_pad[t:], + pitch[:, t // self.window :] if t is not None else pitch, + pitchf[:, t // self.window :] if t is not None else pitchf, + times, + index, + big_npy, + index_rate, + version, + protect, + )[self.t_pad_tgt : -self.t_pad_tgt] + ) + else: + audio_opt.append( + self.vc( + model, + net_g, + sid, + audio_pad[t:], + None, + None, + times, + index, + big_npy, + index_rate, + version, + protect, + )[self.t_pad_tgt : -self.t_pad_tgt] + ) + audio_opt = np.concatenate(audio_opt) + if rms_mix_rate != 1: + audio_opt = change_rms(audio, 16000, audio_opt, tgt_sr, rms_mix_rate) + if resample_sr >= 16000 and tgt_sr != resample_sr: + audio_opt = librosa.resample( + audio_opt, orig_sr=tgt_sr, target_sr=resample_sr + ) + audio_max = np.abs(audio_opt).max() / 0.99 + max_int16 = 32768 + if audio_max > 1: + max_int16 /= audio_max + audio_opt = (audio_opt * max_int16).astype(np.int16) + del pitch, pitchf, sid + if torch.cuda.is_available(): + torch.cuda.empty_cache() + return audio_opt