def __init__(
self,
config: "TacotronConfig",
ap: "AudioProcessor" = None,
tokenizer: "TTSTokenizer" = None,
speaker_manager: SpeakerManager = None,
):
super().__init__(config, ap, tokenizer, speaker_manager)
# pass all config fields to `self`
# for fewer code change
for key in config:
setattr(self, key, config[key])
# set speaker embedding channel size for determining `in_channels` for the connected layers.
# `init_multispeaker` needs to be called once more in training to initialize the speaker embedding layer based
# on the number of speakers infered from the dataset.
if self.use_speaker_embedding or self.use_d_vector_file:
self.init_multispeaker(config)
self.decoder_in_features += self.embedded_speaker_dim # add speaker embedding dim
if self.use_gst:
self.decoder_in_features += self.gst.gst_embedding_dim
if self.use_capacitron_vae:
self.decoder_in_features += self.capacitron_vae.capacitron_VAE_embedding_dim
# embedding layer
self.embedding = nn.Embedding(self.num_chars, 256, padding_idx=0)
self.embedding.weight.data.normal_(0, 0.3)
# base model layers
self.encoder = Encoder(self.encoder_in_features)
self.decoder = Decoder(
self.decoder_in_features,
self.decoder_output_dim,
self.r,
self.memory_size,
self.attention_type,
self.windowing,
self.attention_norm,
self.prenet_type,
self.prenet_dropout,
self.use_forward_attn,
self.transition_agent,
self.forward_attn_mask,
self.location_attn,
self.attention_heads,
self.separate_stopnet,
self.max_decoder_steps,
)
self.postnet = PostCBHG(self.decoder_output_dim)
self.last_linear = nn.Linear(self.postnet.cbhg.gru_features * 2,
self.out_channels)
# setup prenet dropout
self.decoder.prenet.dropout_at_inference = self.prenet_dropout_at_inference
# global style token layers
if self.gst and self.use_gst:
self.gst_layer = GST(
num_mel=self.decoder_output_dim,
num_heads=self.gst.gst_num_heads,
num_style_tokens=self.gst.gst_num_style_tokens,
gst_embedding_dim=self.gst.gst_embedding_dim,
)
# Capacitron layers
if self.capacitron_vae and self.use_capacitron_vae:
self.capacitron_vae_layer = CapacitronVAE(
num_mel=self.decoder_output_dim,
encoder_output_dim=self.encoder_in_features,
capacitron_VAE_embedding_dim=self.capacitron_vae.
capacitron_VAE_embedding_dim,
speaker_embedding_dim=self.embedded_speaker_dim
if self.use_speaker_embedding and
self.capacitron_vae.capacitron_use_speaker_embedding else None,
text_summary_embedding_dim=self.capacitron_vae.
capacitron_text_summary_embedding_dim
if self.capacitron_vae.capacitron_use_text_summary_embeddings
else None,
)
# backward pass decoder
if self.bidirectional_decoder:
self._init_backward_decoder()
# setup DDC
if self.double_decoder_consistency:
self.coarse_decoder = Decoder(
self.decoder_in_features,
self.decoder_output_dim,
self.ddc_r,
self.memory_size,
self.attention_type,
self.windowing,
self.attention_norm,
self.prenet_type,
self.prenet_dropout,
self.use_forward_attn,
self.transition_agent,
self.forward_attn_mask,
self.location_attn,
self.attention_heads,
self.separate_stopnet,
self.max_decoder_steps,
)
def __init__(self,
num_chars,
num_speakers,
r=5,
postnet_output_dim=1025,
decoder_output_dim=80,
attn_type='original',
attn_win=False,
attn_norm="sigmoid",
prenet_type="original",
prenet_dropout=True,
forward_attn=False,
trans_agent=False,
forward_attn_mask=False,
location_attn=True,
attn_K=5,
separate_stopnet=True,
bidirectional_decoder=False,
double_decoder_consistency=False,
ddc_r=None,
encoder_in_features=256,
decoder_in_features=256,
speaker_embedding_dim=None,
gst=False,
gst_embedding_dim=256,
gst_num_heads=4,
gst_style_tokens=10,
memory_size=5,
gst_use_speaker_embedding=False):
super(Tacotron, self).__init__(
num_chars, num_speakers, r, postnet_output_dim, decoder_output_dim,
attn_type, attn_win, attn_norm, prenet_type, prenet_dropout,
forward_attn, trans_agent, forward_attn_mask, location_attn,
attn_K, separate_stopnet, bidirectional_decoder,
double_decoder_consistency, ddc_r, encoder_in_features,
decoder_in_features, speaker_embedding_dim, gst, gst_embedding_dim,
gst_num_heads, gst_style_tokens, gst_use_speaker_embedding)
# speaker embedding layers
if self.num_speakers > 1:
if not self.embeddings_per_sample:
speaker_embedding_dim = 256
self.speaker_embedding = nn.Embedding(self.num_speakers,
speaker_embedding_dim)
self.speaker_embedding.weight.data.normal_(0, 0.3)
# speaker and gst embeddings is concat in decoder input
if self.num_speakers > 1:
self.decoder_in_features += speaker_embedding_dim # add speaker embedding dim
# embedding layer
self.embedding = nn.Embedding(num_chars, 256, padding_idx=0)
self.embedding.weight.data.normal_(0, 0.3)
# base model layers
self.encoder = Encoder(self.encoder_in_features)
self.decoder = Decoder(self.decoder_in_features, decoder_output_dim, r,
memory_size, attn_type, attn_win, attn_norm,
prenet_type, prenet_dropout, forward_attn,
trans_agent, forward_attn_mask, location_attn,
attn_K, separate_stopnet)
self.postnet = PostCBHG(decoder_output_dim)
self.last_linear = nn.Linear(self.postnet.cbhg.gru_features * 2,
postnet_output_dim)
# global style token layers
if self.gst:
self.gst_layer = GST(num_mel=80,
num_heads=gst_num_heads,
num_style_tokens=gst_style_tokens,
gst_embedding_dim=self.gst_embedding_dim,
speaker_embedding_dim=speaker_embedding_dim
if self.embeddings_per_sample
and self.gst_use_speaker_embedding else None)
# backward pass decoder
if self.bidirectional_decoder:
self._init_backward_decoder()
# setup DDC
if self.double_decoder_consistency:
self.coarse_decoder = Decoder(
self.decoder_in_features, decoder_output_dim, ddc_r,
memory_size, attn_type, attn_win, attn_norm, prenet_type,
prenet_dropout, forward_attn, trans_agent, forward_attn_mask,
location_attn, attn_K, separate_stopnet)