import math
import random
from dataclasses import dataclass, field
from itertools import chain
from typing import Dict, List, Tuple
import torch
import torchaudio
from coqpit import Coqpit
from torch import nn
from torch.cuda.amp.autocast_mode import autocast
from torch.nn import functional as F
from TTS.tts.layers.glow_tts.duration_predictor import DurationPredictor
from TTS.tts.layers.vits.discriminator import VitsDiscriminator
from TTS.tts.layers.vits.networks import PosteriorEncoder, ResidualCouplingBlocks, TextEncoder
from TTS.tts.layers.vits.stochastic_duration_predictor import StochasticDurationPredictor
from TTS.tts.models.base_tts import BaseTTS
from TTS.tts.utils.helpers import generate_path, maximum_path, rand_segments, segment, sequence_mask
from TTS.tts.utils.speakers import SpeakerManager
from TTS.tts.utils.languages import LanguageManager
from TTS.tts.utils.synthesis import synthesis
from TTS.tts.utils.visual import plot_alignment
from TTS.utils.trainer_utils import get_optimizer, get_scheduler
from TTS.vocoder.models.hifigan_generator import HifiganGenerator
from TTS.vocoder.utils.generic_utils import plot_results
@dataclass
class VitsArgs(Coqpit):
"""VITS model arguments.
Args:
num_chars (int):
Number of characters in the vocabulary. Defaults to 100.
out_channels (int):
Number of output channels. Defaults to 513.
spec_segment_size (int):
Decoder input segment size. Defaults to 32 `(32 * hoplength = waveform length)`.
hidden_channels (int):
Number of hidden channels of the model. Defaults to 192.
hidden_channels_ffn_text_encoder (int):
Number of hidden channels of the feed-forward layers of the text encoder transformer. Defaults to 256.
num_heads_text_encoder (int):
Number of attention heads of the text encoder transformer. Defaults to 2.
num_layers_text_encoder (int):
Number of transformer layers in the text encoder. Defaults to 6.
kernel_size_text_encoder (int):
Kernel size of the text encoder transformer FFN layers. Defaults to 3.
dropout_p_text_encoder (float):
Dropout rate of the text encoder. Defaults to 0.1.
dropout_p_duration_predictor (float):
Dropout rate of the duration predictor. Defaults to 0.1.
kernel_size_posterior_encoder (int):
Kernel size of the posterior encoder's WaveNet layers. Defaults to 5.
dilatation_posterior_encoder (int):
Dilation rate of the posterior encoder's WaveNet layers. Defaults to 1.
num_layers_posterior_encoder (int):
Number of posterior encoder's WaveNet layers. Defaults to 16.
kernel_size_flow (int):
Kernel size of the Residual Coupling layers of the flow network. Defaults to 5.
dilatation_flow (int):
Dilation rate of the Residual Coupling WaveNet layers of the flow network. Defaults to 1.
num_layers_flow (int):
Number of Residual Coupling WaveNet layers of the flow network. Defaults to 6.
resblock_type_decoder (str):
Type of the residual block in the decoder network. Defaults to "1".
resblock_kernel_sizes_decoder (List[int]):
Kernel sizes of the residual blocks in the decoder network. Defaults to `[3, 7, 11]`.
resblock_dilation_sizes_decoder (List[List[int]]):
Dilation sizes of the residual blocks in the decoder network. Defaults to `[[1, 3, 5], [1, 3, 5], [1, 3, 5]]`.
upsample_rates_decoder (List[int]):
Upsampling rates for each concecutive upsampling layer in the decoder network. The multiply of these
values must be equal to the kop length used for computing spectrograms. Defaults to `[8, 8, 2, 2]`.
upsample_initial_channel_decoder (int):
Number of hidden channels of the first upsampling convolution layer of the decoder network. Defaults to 512.
upsample_kernel_sizes_decoder (List[int]):
Kernel sizes for each upsampling layer of the decoder network. Defaults to `[16, 16, 4, 4]`.
use_sdp (bool):
Use Stochastic Duration Predictor. Defaults to True.
noise_scale (float):
Noise scale used for the sample noise tensor in training. Defaults to 1.0.
inference_noise_scale (float):
Noise scale used for the sample noise tensor in inference. Defaults to 0.667.
length_scale (float):
Scale factor for the predicted duration values. Smaller values result faster speech. Defaults to 1.
noise_scale_dp (float):
Noise scale used by the Stochastic Duration Predictor sample noise in training. Defaults to 1.0.
inference_noise_scale_dp (float):
Noise scale for the Stochastic Duration Predictor in inference. Defaults to 0.8.
max_inference_len (int):
Maximum inference length to limit the memory use. Defaults to None.
init_discriminator (bool):
Initialize the disciminator network if set True. Set False for inference. Defaults to True.
use_spectral_norm_disriminator (bool):
Use spectral normalization over weight norm in the discriminator. Defaults to False.
use_speaker_embedding (bool):
Enable/Disable speaker embedding for multi-speaker models. Defaults to False.
num_speakers (int):
Number of speakers for the speaker embedding layer. Defaults to 0.
speakers_file (str):
Path to the speaker mapping file for the Speaker Manager. Defaults to None.
speaker_embedding_channels (int):
Number of speaker embedding channels. Defaults to 256.
use_d_vector_file (bool):
Enable/Disable the use of d-vectors for multi-speaker training. Defaults to False.
d_vector_file (str):
Path to the file including pre-computed speaker embeddings. Defaults to None.
d_vector_dim (int):
Number of d-vector channels. Defaults to 0.
detach_dp_input (bool):
Detach duration predictor's input from the network for stopping the gradients. Defaults to True.
use_language_embedding (bool):
Enable/Disable language embedding for multilingual models. Defaults to False.
embedded_language_dim (int):
Number of language embedding channels. Defaults to 4.
num_languages (int):
Number of languages for the language embedding layer. Defaults to 0.
use_speaker_encoder_as_loss (bool):
Enable/Disable Speaker Consistency Loss (SCL). Defaults to False.
speaker_encoder_config_path (str):
Path to the file speaker encoder config file, to use for SCL. Defaults to "".
speaker_encoder_model_path (str):
Path to the file speaker encoder checkpoint file, to use for SCL. Defaults to "".
fine_tuning_mode (int):
Fine tuning only the vocoder part of the model, while the rest will be frozen. Defaults to 0.
Mode 0: Disabled;
Mode 1: uses the distribution predicted by the encoder and It's recommended for TTS;
Mode 2: uses the distribution predicted by the encoder and It's recommended for voice conversion.
"""
num_chars: int = 100
out_channels: int = 513
spec_segment_size: int = 32
hidden_channels: int = 192
hidden_channels_ffn_text_encoder: int = 768
num_heads_text_encoder: int = 2
num_layers_text_encoder: int = 6
kernel_size_text_encoder: int = 3
dropout_p_text_encoder: float = 0.1
dropout_p_duration_predictor: float = 0.5
kernel_size_posterior_encoder: int = 5
dilation_rate_posterior_encoder: int = 1
num_layers_posterior_encoder: int = 16
kernel_size_flow: int = 5
dilation_rate_flow: int = 1
num_layers_flow: int = 4
resblock_type_decoder: str = "1"
resblock_kernel_sizes_decoder: List[int] = field(default_factory=lambda: [3, 7, 11])
resblock_dilation_sizes_decoder: List[List[int]] = field(default_factory=lambda: [[1, 3, 5], [1, 3, 5], [1, 3, 5]])
upsample_rates_decoder: List[int] = field(default_factory=lambda: [8, 8, 2, 2])
upsample_initial_channel_decoder: int = 512
upsample_kernel_sizes_decoder: List[int] = field(default_factory=lambda: [16, 16, 4, 4])
use_sdp: bool = True
noise_scale: float = 1.0
inference_noise_scale: float = 0.667
length_scale: float = 1
noise_scale_dp: float = 1.0
inference_noise_scale_dp: float = 1.0
max_inference_len: int = None
init_discriminator: bool = True
use_spectral_norm_disriminator: bool = False
use_speaker_embedding: bool = False
num_speakers: int = 0
speakers_file: str = None
d_vector_file: str = None
speaker_embedding_channels: int = 256
use_d_vector_file: bool = False
d_vector_file: str = None
d_vector_dim: int = 0
detach_dp_input: bool = True
use_language_embedding: bool = False
embedded_language_dim: int = 4
num_languages: int = 0
use_speaker_encoder_as_loss: bool = False
speaker_encoder_config_path: str = ""
speaker_encoder_model_path: str = ""
fine_tuning_mode: int = 0
freeze_encoder: bool = False
freeze_DP: bool = False
freeze_PE: bool = False
freeze_flow_decoder: bool = False
freeze_waveform_decoder: bool = False
class Vits(BaseTTS):
"""VITS TTS model
Paper::
https://arxiv.org/pdf/2106.06103.pdf
Paper Abstract::
Several recent end-to-end text-to-speech (TTS) models enabling single-stage training and parallel
sampling have been proposed, but their sample quality does not match that of two-stage TTS systems.
In this work, we present a parallel endto-end TTS method that generates more natural sounding audio than
current two-stage models. Our method adopts variational inference augmented with normalizing flows and
an adversarial training process, which improves the expressive power of generative modeling. We also propose a
stochastic duration predictor to synthesize speech with diverse rhythms from input text. With the
uncertainty modeling over latent variables and the stochastic duration predictor, our method expresses the
natural one-to-many relationship in which a text input can be spoken in multiple ways
with different pitches and rhythms. A subjective human evaluation (mean opinion score, or MOS)
on the LJ Speech, a single speaker dataset, shows that our method outperforms the best publicly
available TTS systems and achieves a MOS comparable to ground truth.
Check :class:`TTS.tts.configs.vits_config.VitsConfig` for class arguments.
Examples:
>>> from TTS.tts.configs.vits_config import VitsConfig
>>> from TTS.tts.models.vits import Vits
>>> config = VitsConfig()
>>> model = Vits(config)
"""
# pylint: disable=dangerous-default-value
def __init__(self, config: Coqpit, speaker_manager: SpeakerManager = None):
super().__init__(config)
self.END2END = True
self.speaker_manager = speaker_manager
self.audio_config = config["audio"]
if config.__class__.__name__ == "VitsConfig":
# loading from VitsConfig
if "num_chars" not in config:
_, self.config, num_chars = self.get_characters(config)
config.model_args.num_chars = num_chars
else:
self.config = config
config.model_args.num_chars = config.num_chars
args = self.config.model_args
elif isinstance(config, VitsArgs):
# loading from VitsArgs
self.config = config
args = config
else:
raise ValueError("config must be either a VitsConfig or VitsArgs")
self.args = args
self.init_multispeaker(config)
self.init_multilingual(config)
self.length_scale = args.length_scale
self.noise_scale = args.noise_scale
self.inference_noise_scale = args.inference_noise_scale
self.inference_noise_scale_dp = args.inference_noise_scale_dp
self.noise_scale_dp = args.noise_scale_dp
self.max_inference_len = args.max_inference_len
self.spec_segment_size = args.spec_segment_size
self.text_encoder = TextEncoder(
args.num_chars,
args.hidden_channels,
args.hidden_channels,
args.hidden_channels_ffn_text_encoder,
args.num_heads_text_encoder,
args.num_layers_text_encoder,
args.kernel_size_text_encoder,
args.dropout_p_text_encoder,
language_emb_dim=self.embedded_language_dim
)
self.posterior_encoder = PosteriorEncoder(
args.out_channels,
args.hidden_channels,
args.hidden_channels,
kernel_size=args.kernel_size_posterior_encoder,
dilation_rate=args.dilation_rate_posterior_encoder,
num_layers=args.num_layers_posterior_encoder,
cond_channels=self.embedded_speaker_dim,
)
self.flow = ResidualCouplingBlocks(
args.hidden_channels,
args.hidden_channels,
kernel_size=args.kernel_size_flow,
dilation_rate=args.dilation_rate_flow,
num_layers=args.num_layers_flow,
cond_channels=self.embedded_speaker_dim,
)
if args.use_sdp:
self.duration_predictor = StochasticDurationPredictor(
args.hidden_channels + self.embedded_language_dim,
192,
3,
args.dropout_p_duration_predictor,
4,
cond_channels=self.embedded_speaker_dim,
language_emb_dim=self.embedded_language_dim,
)
else:
self.duration_predictor = DurationPredictor(
args.hidden_channels + self.embedded_language_dim,
256,
3,
args.dropout_p_duration_predictor,
cond_channels=self.embedded_speaker_dim,
language_emb_dim=self.embedded_language_dim,
)
self.waveform_decoder = HifiganGenerator(
args.hidden_channels,
1,
args.resblock_type_decoder,
args.resblock_dilation_sizes_decoder,
args.resblock_kernel_sizes_decoder,
args.upsample_kernel_sizes_decoder,
args.upsample_initial_channel_decoder,
args.upsample_rates_decoder,
inference_padding=0,
cond_channels=self.embedded_speaker_dim,
conv_pre_weight_norm=False,
conv_post_weight_norm=False,
conv_post_bias=False,
)
if args.init_discriminator:
self.disc = VitsDiscriminator(use_spectral_norm=args.use_spectral_norm_disriminator)
def init_multispeaker(self, config: Coqpit):
"""Initialize multi-speaker modules of a model. A model can be trained either with a speaker embedding layer
or with external `d_vectors` computed from a speaker encoder model.
Args:
config (Coqpit): Model configuration.
data (List, optional): Dataset items to infer number of speakers. Defaults to None.
"""
self.embedded_speaker_dim = 0
config = config.model_args
self.num_speakers = config.num_speakers
if config.use_speaker_embedding:
self._init_speaker_embedding(config)
if config.use_d_vector_file:
self._init_d_vector(config)
# TODO: make this a function
if config.use_speaker_encoder_as_loss:
if not config.speaker_encoder_model_path or not config.speaker_encoder_config_path:
raise RuntimeError(" [!] To use the speaker encoder loss you need to specify speaker_encoder_model_path and speaker_encoder_config_path !!")
self.speaker_manager.init_speaker_encoder(config.speaker_encoder_model_path, config.speaker_encoder_config_path)
self.speaker_encoder = self.speaker_manager.speaker_encoder.train()
for param in self.speaker_encoder.parameters():
param.requires_grad = False
print(" > External Speaker Encoder Loaded !!")
if hasattr(self.speaker_encoder, "audio_config") and self.audio_config["sample_rate"] != self.speaker_encoder.audio_config["sample_rate"]:
self.audio_transform = torchaudio.transforms.Resample(orig_freq=self.audio_config["sample_rate"], new_freq=self.speaker_encoder.audio_config["sample_rate"])
else:
self.audio_transform = None
else:
self.audio_transform = None
self.speaker_encoder = None
def _init_speaker_embedding(self, config):
# pylint: disable=attribute-defined-outside-init
if config.speakers_file is not None:
self.speaker_manager = SpeakerManager(speaker_id_file_path=config.speakers_file)
if self.num_speakers > 0:
print(" > initialization of speaker-embedding layers.")
self.embedded_speaker_dim = config.speaker_embedding_channels
self.emb_g = nn.Embedding(self.num_speakers, self.embedded_speaker_dim)
def _init_d_vector(self, config):
# pylint: disable=attribute-defined-outside-init
if hasattr(self, "emb_g"):
raise ValueError("[!] Speaker embedding layer already initialized before d_vector settings.")
self.speaker_manager = SpeakerManager(d_vectors_file_path=config.d_vector_file)
self.embedded_speaker_dim = config.d_vector_dim
if config.use_speaker_encoder_as_loss:
if not config.speaker_encoder_model_path or not config.speaker_encoder_config_path:
raise RuntimeError(" [!] To use the speaker encoder loss you need to specify speaker_encoder_model_path and speaker_encoder_config_path !!")
self.speaker_manager.init_speaker_encoder(config.speaker_encoder_model_path, config.speaker_encoder_config_path)
self.speaker_encoder = self.speaker_manager.speaker_encoder.train()
for param in self.speaker_encoder.parameters():
param.requires_grad = False
print(" > External Speaker Encoder Loaded !!")
if hasattr(self.speaker_encoder, "audio_config") and self.audio_config["sample_rate"] != self.speaker_encoder.audio_config["sample_rate"]:
self.audio_transform = torchaudio.transforms.Resample(orig_freq=self.audio_config["sample_rate"], new_freq=self.speaker_encoder.audio_config["sample_rate"])
else:
self.audio_transform = None
else:
self.audio_transform = None
self.speaker_encoder = None
def init_multilingual(self, config: Coqpit, data: List = None):
"""Initialize multilingual modules of a model.
Args:
config (Coqpit): Model configuration.
data (List, optional): Dataset items to infer number of speakers. Defaults to None.
"""
if hasattr(config, "model_args"):
config = config.model_args
# init language manager
self.language_manager = LanguageManager(config, data=data)
# init language embedding layer
if config.use_language_embedding:
if config.num_languages > 0 and self.language_manager.num_languages == 0:
self.num_languages = config.num_languages
else:
self.num_languages = self.language_manager.num_languages
self.embedded_language_dim = config.embedded_language_dim
self.emb_l = nn.Embedding(self.num_languages, self.embedded_language_dim)
torch.nn.init.xavier_uniform_(self.emb_l.weight)
else:
self.embedded_language_dim = 0
self.emb_l = None
@staticmethod
def _set_cond_input(aux_input: Dict):
"""Set the speaker conditioning input based on the multi-speaker mode."""
sid, g, lid = None, None, None
if "speaker_ids" in aux_input and aux_input["speaker_ids"] is not None:
sid = aux_input["speaker_ids"]
if sid.ndim == 0:
sid = sid.unsqueeze_(0)
if "d_vectors" in aux_input and aux_input["d_vectors"] is not None:
g = F.normalize(aux_input["d_vectors"]).unsqueeze(-1)
if g.ndim == 2:
g = g.unsqueeze_(0)
if "language_ids" in aux_input and aux_input["language_ids"] is not None:
lid = aux_input["language_ids"]
if lid.ndim == 0:
lid = lid.unsqueeze_(0)
return sid, g, lid
def get_aux_input(self, aux_input: Dict):
sid, g, lid = self._set_cond_input(aux_input)
return {"speaker_id": sid, "style_wav": None, "d_vector": g, "language_id": lid}
def get_aux_input_from_test_setences(self, sentence_info):
if hasattr(self.config, "model_args"):
config = self.config.model_args
else:
config = self.config
# extract speaker and language info
text, speaker_name, style_wav, language_name = None, None, None, None
if isinstance(sentence_info, list):
if len(sentence_info) == 1:
text = sentence_info[0]
elif len(sentence_info) == 2:
text, speaker_name = sentence_info
elif len(sentence_info) == 3:
text, speaker_name, style_wav = sentence_info
elif len(sentence_info) == 4:
text, speaker_name, style_wav, language_name = sentence_info
else:
text = sentence_info
# get speaker id/d_vector
speaker_id, d_vector, language_id = None, None, None
if hasattr(self, "speaker_manager"):
if config.use_d_vector_file:
if speaker_name is None:
d_vector = self.speaker_manager.get_random_d_vector()
else:
d_vector = self.speaker_manager.get_d_vector_by_speaker(speaker_name)
elif config.use_speaker_embedding:
if speaker_name is None:
speaker_id = self.speaker_manager.get_random_speaker_id()
else:
speaker_id = self.speaker_manager.speaker_ids[speaker_name]
# get language id
if hasattr(self, "language_manager") and config.use_language_embedding and language_name is not None:
language_id = self.language_manager.language_id_mapping[language_name]
return {"text": text, "speaker_id": speaker_id, "style_wav": style_wav, "d_vector": d_vector, "language_id": language_id}
def forward(
self,
x: torch.tensor,
x_lengths: torch.tensor,
y: torch.tensor,
y_lengths: torch.tensor,
aux_input={"d_vectors": None, "speaker_ids": None, "language_ids": None},
waveform=None,
) -> Dict:
"""Forward pass of the model.
Args:
x (torch.tensor): Batch of input character sequence IDs.
x_lengths (torch.tensor): Batch of input character sequence lengths.
y (torch.tensor): Batch of input spectrograms.
y_lengths (torch.tensor): Batch of input spectrogram lengths.
aux_input (dict, optional): Auxiliary inputs for multi-speaker training. Defaults to {"d_vectors": None, "speaker_ids": None}.
Returns:
Dict: model outputs keyed by the output name.
Shapes:
- x: :math:`[B, T_seq]`
- x_lengths: :math:`[B]`
- y: :math:`[B, C, T_spec]`
- y_lengths: :math:`[B]`
- d_vectors: :math:`[B, C, 1]`
- speaker_ids: :math:`[B]`
"""
outputs = {}
sid, g, lid = self._set_cond_input(aux_input)
# speaker embedding
if self.args.use_speaker_embedding and sid is not None and not self.use_d_vector:
g = self.emb_g(sid).unsqueeze(-1) # [b, h, 1]
# language embedding
lang_emb=None
if self.args.use_language_embedding and lid is not None:
lang_emb = self.emb_l(lid).unsqueeze(-1)
x, m_p, logs_p, x_mask = self.text_encoder(x, x_lengths, lang_emb=lang_emb)
# posterior encoder
z, m_q, logs_q, y_mask = self.posterior_encoder(y, y_lengths, g=g)
# flow layers
z_p = self.flow(z, y_mask, g=g)
# find the alignment path
attn_mask = torch.unsqueeze(x_mask, -1) * torch.unsqueeze(y_mask, 2)
with torch.no_grad():
o_scale = torch.exp(-2 * logs_p)
logp1 = torch.sum(-0.5 * math.log(2 * math.pi) - logs_p, [1]).unsqueeze(-1) # [b, t, 1]
logp2 = torch.einsum("klm, kln -> kmn", [o_scale, -0.5 * (z_p ** 2)])
logp3 = torch.einsum("klm, kln -> kmn", [m_p * o_scale, z_p])
logp4 = torch.sum(-0.5 * (m_p ** 2) * o_scale, [1]).unsqueeze(-1) # [b, t, 1]
logp = logp2 + logp3 + logp1 + logp4
attn = maximum_path(logp, attn_mask.squeeze(1)).unsqueeze(1).detach()
# duration predictor
attn_durations = attn.sum(3)
if self.args.use_sdp:
loss_duration = self.duration_predictor(
x.detach() if self.args.detach_dp_input else x,
x_mask,
attn_durations,
g=g.detach() if self.args.detach_dp_input and g is not None else g,
lang_emb=lang_emb.detach() if self.args.detach_dp_input and lang_emb is not None else lang_emb,
)
loss_duration = loss_duration / torch.sum(x_mask)
else:
attn_log_durations = torch.log(attn_durations + 1e-6) * x_mask
log_durations = self.duration_predictor(
x.detach() if self.args.detach_dp_input else x,
x_mask,
g=g.detach() if self.args.detach_dp_input and g is not None else g,
lang_emb=lang_emb.detach() if self.args.detach_dp_input and lang_emb is not None else lang_emb,
)
loss_duration = torch.sum((log_durations - attn_log_durations) ** 2, [1, 2]) / torch.sum(x_mask)
outputs["loss_duration"] = loss_duration
# expand prior
m_p = torch.einsum("klmn, kjm -> kjn", [attn, m_p])
logs_p = torch.einsum("klmn, kjm -> kjn", [attn, logs_p])
# select a random feature segment for the waveform decoder
z_slice, slice_ids = rand_segments(z, y_lengths, self.spec_segment_size)
o = self.waveform_decoder(z_slice, g=g)
wav_seg = segment(
waveform.transpose(1, 2),
slice_ids * self.config.audio.hop_length,
self.args.spec_segment_size * self.config.audio.hop_length,
)
if self.args.use_speaker_encoder_as_loss and self.speaker_encoder is not None:
# concate generated and GT waveforms
wavs_batch = torch.cat((wav_seg, o), dim=0).squeeze(1)
# resample audio to speaker encoder sample_rate
if self.audio_transform is not None:
wavs_batch = self.audio_transform(wavs_batch)
pred_embs = self.speaker_encoder.forward(wavs_batch, l2_norm=True)
# split generated and GT speaker embeddings
gt_spk_emb, syn_spk_emb = torch.chunk(pred_embs, 2, dim=0)
else:
gt_spk_emb, syn_spk_emb = None, None
outputs.update(
{
"model_outputs": o,
"alignments": attn.squeeze(1),
"z": z,
"z_p": z_p,
"m_p": m_p,
"logs_p": logs_p,
"m_q": m_q,
"logs_q": logs_q,
"waveform_seg": wav_seg,
"gt_spk_emb": gt_spk_emb,
"syn_spk_emb": syn_spk_emb
}
)
return outputs
def forward_fine_tuning(
self,
x: torch.tensor,
x_lengths: torch.tensor,
y: torch.tensor,
y_lengths: torch.tensor,
aux_input={"d_vectors": None, "speaker_ids": None, "language_ids": None},
waveform=None,
) -> Dict:
"""Forward pass of the model.
Args:
x (torch.tensor): Batch of input character sequence IDs.
x_lengths (torch.tensor): Batch of input character sequence lengths.
y (torch.tensor): Batch of input spectrograms.
y_lengths (torch.tensor): Batch of input spectrogram lengths.
aux_input (dict, optional): Auxiliary inputs for multi-speaker training. Defaults to {"d_vectors": None, "speaker_ids": None}.
Returns:
Dict: model outputs keyed by the output name.
Shapes:
- x: :math:`[B, T_seq]`
- x_lengths: :math:`[B]`
- y: :math:`[B, C, T_spec]`
- y_lengths: :math:`[B]`
- d_vectors: :math:`[B, C, 1]`
- speaker_ids: :math:`[B]`
"""
with torch.no_grad():
outputs = {}
sid, g, lid = self._set_cond_input(aux_input)
# speaker embedding
if self.args.use_speaker_embedding and sid is not None and not self.use_d_vector:
g = self.emb_g(sid).unsqueeze(-1) # [b, h, 1]
# language embedding
lang_emb=None
if self.args.use_language_embedding and lid is not None:
lang_emb = self.emb_l(lid).unsqueeze(-1)
x, m_p, logs_p, x_mask = self.text_encoder(x, x_lengths, lang_emb=lang_emb)
# posterior encoder
z, m_q, logs_q, y_mask = self.posterior_encoder(y, y_lengths, g=g)
# flow layers
z_p = self.flow(z, y_mask, g=g)
# find the alignment path
attn_mask = torch.unsqueeze(x_mask, -1) * torch.unsqueeze(y_mask, 2)
with torch.no_grad():
o_scale = torch.exp(-2 * logs_p)
logp1 = torch.sum(-0.5 * math.log(2 * math.pi) - logs_p, [1]).unsqueeze(-1) # [b, t, 1]
logp2 = torch.einsum("klm, kln -> kmn", [o_scale, -0.5 * (z_p ** 2)])
logp3 = torch.einsum("klm, kln -> kmn", [m_p * o_scale, z_p])
logp4 = torch.sum(-0.5 * (m_p ** 2) * o_scale, [1]).unsqueeze(-1) # [b, t, 1]
logp = logp2 + logp3 + logp1 + logp4
attn = maximum_path(logp, attn_mask.squeeze(1)).unsqueeze(1).detach()
# expand prior
m_p = torch.einsum("klmn, kjm -> kjn", [attn, m_p])
logs_p = torch.einsum("klmn, kjm -> kjn", [attn, logs_p])
# mode 1: like SC-GlowTTS paper; mode 2: recommended for voice conversion
if self.args.fine_tuning_mode == 1:
z_ft = m_p
elif self.args.fine_tuning_mode == 2:
z_ft = z_p
else:
raise RuntimeError(" [!] Invalid Fine Tunning Mode !")
# inverse decoder and get the output
z_f_pred = self.flow(z_ft, y_mask, g=g, reverse=True)
z_slice, slice_ids = rand_segment(z_f_pred, y_lengths, self.spec_segment_size)
o = self.waveform_decoder(z_slice, g=g)
wav_seg = segment(
waveform.transpose(1, 2),
slice_ids * self.config.audio.hop_length,
self.args.spec_segment_size * self.config.audio.hop_length,
)
if self.args.use_speaker_encoder_as_loss and self.speaker_encoder is not None:
# concate generated and GT waveforms
wavs_batch = torch.cat((wav_seg, o), dim=0).squeeze(1)
# resample audio to speaker encoder sample_rate
if self.audio_transform is not None:
wavs_batch = self.audio_transform(wavs_batch)
pred_embs = self.speaker_encoder.forward(wavs_batch, l2_norm=True)
# split generated and GT speaker embeddings
gt_spk_emb, syn_spk_emb = torch.chunk(pred_embs, 2, dim=0)
else:
gt_spk_emb, syn_spk_emb = None, None
outputs.update(
{
"model_outputs": o,
"alignments": attn.squeeze(1),
"loss_duration": 0.0,
"z": z,
"z_p": z_p,
"m_p": m_p,
"logs_p": logs_p,
"m_q": m_q,
"logs_q": logs_q,
"waveform_seg": wav_seg,
"gt_spk_emb": gt_spk_emb,
"syn_spk_emb": syn_spk_emb
}
)
return outputs
def inference(self, x, aux_input={"d_vectors": None, "speaker_ids": None, "language_ids": None}):
"""
Shapes:
- x: :math:`[B, T_seq]`
- d_vectors: :math:`[B, C, 1]`
- speaker_ids: :math:`[B]`
"""
sid, g, lid = self._set_cond_input(aux_input)
x_lengths = torch.tensor(x.shape[1:2]).to(x.device)
# speaker embedding
if self.args.use_speaker_embedding and sid is not None and not self.use_d_vector:
g = self.emb_g(sid).unsqueeze(-1)
# language embedding
lang_emb=None
if self.args.use_language_embedding and lid is not None:
lang_emb = self.emb_l(lid).unsqueeze(-1)
x, m_p, logs_p, x_mask = self.text_encoder(x, x_lengths, lang_emb=lang_emb)
if self.args.use_sdp:
logw = self.duration_predictor(x, x_mask, g=g, reverse=True, noise_scale=self.inference_noise_scale_dp, lang_emb=lang_emb)
else:
logw = self.duration_predictor(x, x_mask, g=g, lang_emb=lang_emb)
w = torch.exp(logw) * x_mask * self.length_scale
w_ceil = torch.ceil(w)
y_lengths = torch.clamp_min(torch.sum(w_ceil, [1, 2]), 1).long()
y_mask = sequence_mask(y_lengths, None).to(x_mask.dtype)
attn_mask = torch.unsqueeze(x_mask, 2) * torch.unsqueeze(y_mask, -1)
attn = generate_path(w_ceil.squeeze(1), attn_mask.squeeze(1).transpose(1, 2))
m_p = torch.matmul(attn.transpose(1, 2), m_p.transpose(1, 2)).transpose(1, 2)
logs_p = torch.matmul(attn.transpose(1, 2), logs_p.transpose(1, 2)).transpose(1, 2)
z_p = m_p + torch.randn_like(m_p) * torch.exp(logs_p) * self.inference_noise_scale
z = self.flow(z_p, y_mask, g=g, reverse=True)
o = self.waveform_decoder((z * y_mask)[:, :, : self.max_inference_len], g=g)
outputs = {"model_outputs": o, "alignments": attn.squeeze(1), "z": z, "z_p": z_p, "m_p": m_p, "logs_p": logs_p}
return outputs
def voice_conversion(self, y, y_lengths, speaker_cond_src, speaker_cond_tgt):
"""TODO: create an end-point for voice conversion"""
assert self.num_speakers > 0, "num_speakers have to be larger than 0."
# speaker embedding
if self.args.use_speaker_embedding and not self.use_d_vector:
g_src = self.emb_g(speaker_cond_src).unsqueeze(-1)
g_tgt = self.emb_g(speaker_cond_tgt).unsqueeze(-1)
elif self.args.use_speaker_embedding and self.use_d_vector:
g_src = F.normalize(speaker_cond_src).unsqueeze(-1)
g_tgt = F.normalize(speaker_cond_tgt).unsqueeze(-1)
else:
raise RuntimeError(" [!] Voice conversion is only supported on multi-speaker models.")
z, _, _, y_mask = self.posterior_encoder(y, y_lengths, g=g_src)
z_p = self.flow(z, y_mask, g=g_src)
z_hat = self.flow(z_p, y_mask, g=g_tgt, reverse=True)
o_hat = self.waveform_decoder(z_hat * y_mask, g=g_tgt)
return o_hat, y_mask, (z, z_p, z_hat)
def train_step(self, batch: dict, criterion: nn.Module, optimizer_idx: int) -> Tuple[Dict, Dict]:
"""Perform a single training step. Run the model forward pass and compute losses.
Args:
batch (Dict): Input tensors.
criterion (nn.Module): Loss layer designed for the model.
optimizer_idx (int): Index of optimizer to use. 0 for the generator and 1 for the discriminator networks.
Returns:
Tuple[Dict, Dict]: Model ouputs and computed losses.
"""
# pylint: disable=attribute-defined-outside-init
if optimizer_idx not in [0, 1]:
raise ValueError(" [!] Unexpected `optimizer_idx`.")
# generator pass
if self.args.fine_tuning_mode:
# ToDo: find better place fot it
# force eval mode
self.eval()
# restore train mode for the vocoder part
self.waveform_decoder.train()
self.disc.train()
if self.args.freeze_encoder:
for param in self.text_encoder.parameters():
param.requires_grad = False
if hasattr(self, 'emb_l'):
for param in self.emb_l.parameters():
param.requires_grad = False
if self.args.freeze_PE:
for param in self.posterior_encoder.parameters():
param.requires_grad = False
if self.args.freeze_DP:
for param in self.duration_predictor.parameters():
param.requires_grad = False
if self.args.freeze_flow_decoder:
for param in self.flow.parameters():
param.requires_grad = False
if self.args.freeze_waveform_decoder:
for param in self.waveform_decoder.parameters():
param.requires_grad = False
if optimizer_idx == 0:
text_input = batch["text_input"]
text_lengths = batch["text_lengths"]
mel_lengths = batch["mel_lengths"]
linear_input = batch["linear_input"]
d_vectors = batch["d_vectors"]
speaker_ids = batch["speaker_ids"]
language_ids = batch["language_ids"]
waveform = batch["waveform"]
# generator pass
if self.args.fine_tuning_mode:
# model forward
outputs = self.forward_fine_tuning(
text_input,
text_lengths,
linear_input.transpose(1, 2),
mel_lengths,
aux_input={"d_vectors": d_vectors, "speaker_ids": speaker_ids, "language_ids": language_ids},
waveform=waveform,
)
else:
outputs = self.forward(
text_input,
text_lengths,
linear_input.transpose(1, 2),
mel_lengths,
aux_input={"d_vectors": d_vectors, "speaker_ids": speaker_ids, "language_ids": language_ids},
waveform=waveform,
)
# cache tensors for the discriminator
self.y_disc_cache = None
self.wav_seg_disc_cache = None
self.y_disc_cache = outputs["model_outputs"]
self.wav_seg_disc_cache = outputs["waveform_seg"]
# compute discriminator scores and features
outputs["scores_disc_fake"], outputs["feats_disc_fake"], _, outputs["feats_disc_real"] = self.disc(
outputs["model_outputs"], outputs["waveform_seg"]
)
# compute losses
with autocast(enabled=False): # use float32 for the criterion
loss_dict = criterion[optimizer_idx](
waveform_hat=outputs["model_outputs"].float(),
waveform= outputs["waveform_seg"].float(),
z_p=outputs["z_p"].float(),
logs_q=outputs["logs_q"].float(),
m_p=outputs["m_p"].float(),
logs_p=outputs["logs_p"].float(),
z_len=mel_lengths,
scores_disc_fake=outputs["scores_disc_fake"],
feats_disc_fake=outputs["feats_disc_fake"],
feats_disc_real=outputs["feats_disc_real"],
loss_duration=outputs["loss_duration"],
fine_tuning_mode=self.args.fine_tuning_mode,
use_speaker_encoder_as_loss=self.args.use_speaker_encoder_as_loss,
gt_spk_emb=outputs["gt_spk_emb"],
syn_spk_emb=outputs["syn_spk_emb"]
)
# ignore duration loss if fine tuning mode is on
if not self.args.fine_tuning_mode:
# handle the duration loss
if self.args.use_sdp:
loss_dict["nll_duration"] = outputs["nll_duration"]
loss_dict["loss"] += outputs["nll_duration"]
else:
loss_dict["loss_duration"] = outputs["loss_duration"]
loss_dict["loss"] += outputs["loss_duration"]
elif optimizer_idx == 1:
# discriminator pass
outputs = {}
# compute scores and features
outputs["scores_disc_fake"], _, outputs["scores_disc_real"], _ = self.disc(
self.y_disc_cache.detach(), self.wav_seg_disc_cache
)
# compute loss
with autocast(enabled=False): # use float32 for the criterion
loss_dict = criterion[optimizer_idx](
outputs["scores_disc_real"],
outputs["scores_disc_fake"],
)
return outputs, loss_dict
def _log(self, ap, batch, outputs, name_prefix="train"): # pylint: disable=unused-argument,no-self-use
y_hat = outputs[0]["model_outputs"]
y = outputs[0]["waveform_seg"]
figures = plot_results(y_hat, y, ap, name_prefix)
sample_voice = y_hat[0].squeeze(0).detach().cpu().numpy()
audios = {f"{name_prefix}/audio": sample_voice}
alignments = outputs[0]["alignments"]
align_img = alignments[0].data.cpu().numpy().T
figures.update(
{
"alignment": plot_alignment(align_img, output_fig=False),
}
)
return figures, audios
def train_log(
self, batch: dict, outputs: dict, logger: "Logger", assets: dict, steps: int
): # pylint: disable=no-self-use
"""Create visualizations and waveform examples.
For example, here you can plot spectrograms and generate sample sample waveforms from these spectrograms to
be projected onto Tensorboard.
Args:
ap (AudioProcessor): audio processor used at training.
batch (Dict): Model inputs used at the previous training step.
outputs (Dict): Model outputs generated at the previoud training step.
Returns:
Tuple[Dict, np.ndarray]: training plots and output waveform.
"""
ap = assets["audio_processor"]
self._log(ap, batch, outputs, "train")
@torch.no_grad()
def eval_step(self, batch: dict, criterion: nn.Module, optimizer_idx: int):
return self.train_step(batch, criterion, optimizer_idx)
def eval_log(self, batch: dict, outputs: dict, logger: "Logger", assets: dict, steps: int) -> None:
ap = assets["audio_processor"]
return self._log(ap, batch, outputs, "eval")
@torch.no_grad()
def test_run(self, ap) -> Tuple[Dict, Dict]:
"""Generic test run for `tts` models used by `Trainer`.
You can override this for a different behaviour.
Returns:
Tuple[Dict, Dict]: Test figures and audios to be projected to Tensorboard.
"""
print(" | > Synthesizing test sentences.")
test_audios = {}
test_figures = {}
test_sentences = self.config.test_sentences
for idx, s_info in enumerate(test_sentences):
try:
aux_inputs = self.get_aux_input_from_test_setences(s_info)
wav, alignment, _, _ = synthesis(
self,
aux_inputs["text"],
self.config,
"cuda" in str(next(self.parameters()).device),
ap,
speaker_id=aux_inputs["speaker_id"],
d_vector=aux_inputs["d_vector"],
style_wav=aux_inputs["style_wav"],
language_id=aux_inputs["language_id"],
enable_eos_bos_chars=self.config.enable_eos_bos_chars,
use_griffin_lim=True,
do_trim_silence=False,
).values()
test_audios["{}-audio".format(idx)] = wav
test_figures["{}-alignment".format(idx)] = plot_alignment(alignment.T, output_fig=False)
except: # pylint: disable=bare-except
print(" !! Error creating Test Sentence -", idx)
return test_figures, test_audios
def get_optimizer(self) -> List:
"""Initiate and return the GAN optimizers based on the config parameters.
It returnes 2 optimizers in a list. First one is for the generator and the second one is for the discriminator.
Returns:
List: optimizers.
"""
gen_parameters = chain(
self.text_encoder.parameters(),
self.posterior_encoder.parameters(),
self.flow.parameters(),
self.duration_predictor.parameters(),
self.waveform_decoder.parameters(),
)
# add the speaker embedding layer
if hasattr(self, "emb_g") and self.args.use_speaker_embedding and not self.args.use_d_vector_file:
gen_parameters = chain(gen_parameters, self.emb_g.parameters())
# add the language embedding layer
if hasattr(self, "emb_l") and self.args.use_language_embedding:
gen_parameters = chain(gen_parameters, self.emb_l.parameters())
optimizer0 = get_optimizer(
self.config.optimizer, self.config.optimizer_params, self.config.lr_gen, parameters=gen_parameters
)
optimizer1 = get_optimizer(self.config.optimizer, self.config.optimizer_params, self.config.lr_disc, self.disc)
return [optimizer0, optimizer1]
def get_lr(self) -> List:
"""Set the initial learning rates for each optimizer.
Returns:
List: learning rates for each optimizer.
"""
return [self.config.lr_gen, self.config.lr_disc]
def get_scheduler(self, optimizer) -> List:
"""Set the schedulers for each optimizer.
Args:
optimizer (List[`torch.optim.Optimizer`]): List of optimizers.
Returns:
List: Schedulers, one for each optimizer.
"""
scheduler0 = get_scheduler(self.config.lr_scheduler_gen, self.config.lr_scheduler_gen_params, optimizer[0])
scheduler1 = get_scheduler(self.config.lr_scheduler_disc, self.config.lr_scheduler_disc_params, optimizer[1])
return [scheduler0, scheduler1]
def get_criterion(self):
"""Get criterions for each optimizer. The index in the output list matches the optimizer idx used in
`train_step()`"""
from TTS.tts.layers.losses import ( # pylint: disable=import-outside-toplevel
VitsDiscriminatorLoss,
VitsGeneratorLoss,
)
return [VitsGeneratorLoss(self.config), VitsDiscriminatorLoss(self.config)]
@staticmethod
def make_symbols(config):
"""Create a custom arrangement of symbols used by the model. The output list of symbols propagate along the
whole training and inference steps."""
_pad = config.characters["pad"]
_punctuations = config.characters["punctuations"]
_letters = config.characters["characters"]
_letters_ipa = config.characters["phonemes"]
symbols = [_pad] + list(_punctuations) + list(_letters)
if config.use_phonemes:
symbols += list(_letters_ipa)
return symbols
@staticmethod
def get_characters(config: Coqpit):
if config.characters is not None:
symbols = Vits.make_symbols(config)
else:
from TTS.tts.utils.text.symbols import ( # pylint: disable=import-outside-toplevel
parse_symbols,
phonemes,
symbols,
)
config.characters = parse_symbols()
if config.use_phonemes:
symbols = phonemes
num_chars = len(symbols) + getattr(config, "add_blank", False)
return symbols, config, num_chars
def load_checkpoint(
self, config, checkpoint_path, eval=False
): # pylint: disable=unused-argument, redefined-builtin
"""Load the model checkpoint and setup for training or inference"""
state = torch.load(checkpoint_path, map_location=torch.device("cpu"))
self.load_state_dict(state["model"])
if eval:
self.eval()
assert not self.training