def __init__(self, num_mel=80, embedding_size=512):
super(TransformerTTS, self).__init__()
self.encoder = Encoder()
self.decoder = Decoder()
self.postnet = PostNet()
self.stop_linear = Linear(embedding_size, 1, w_init='sigmoid')
self.mel_linear = Linear(embedding_size, num_mel)
def __init__(self, use_postnet=True, n_spkers=1):
super(FastSpeech2, self).__init__()
### Speaker Embedding Table ###
self.use_spk_embed = hp.use_spk_embed
if self.use_spk_embed:
self.n_spkers = n_spkers
self.spk_embed_dim = hp.spk_embed_dim
self.spk_embed_weight_std = hp.spk_embed_weight_std
self.embed_speakers = Embedding(n_spkers, self.spk_embed_dim, padding_idx=None, std=self.spk_embed_weight_std)
self.use_emo_embed = hp.use_emo_embed
if self.use_emo_embed:
self.n_emotes = n_emotes
self.emo_embed_dim = hp.emo_embed_dim
self.emo_embed_weight_std = hp.emo_embed_weight_std
self.embed_emotions = Embedding(n_emotes, self.emo_embed_dim, padding_idx=None, std=self.emo_embed_weight_std)
### Encoder, Speaker Integrator, Variance Adaptor, Deocder, Postnet ###
self.encoder = Encoder()
if self.use_spk_embed:
self.speaker_integrator = SpeakerIntegrator()
self.variance_adaptor = VarianceAdaptor()
self.decoder = Decoder()
self.mel_linear = nn.Linear(hp.decoder_hidden, hp.n_mel_channels)
self.use_postnet = use_postnet
if self.use_postnet:
self.postnet = PostNet()
def __init__(self):
super(FastSpeech2, self).__init__()
self.encoder = Encoder()
self.variance_adaptor = VarianceAdaptor()
self.decoder = Decoder()
self.mel_linear = Linear(hp.decoder_hidden, hp.n_mel_channels)
self.postnet = PostNet()
def __init__(self):
super(FastSpeech, self).__init__()
self.encoder = Encoder()
self.length_regulator = LengthRegulator()
self.decoder = Decoder()
self.mel_linear = Linear(hp.decoder_output_size, hp.num_mels)
self.postnet = PostNet()
def __init__(self, use_postnet=True):
super(STYLER, self).__init__()
self.style_modeling = StyleModeling()
self.decoder = Decoder()
self.mel_linear = nn.Linear(hp.decoder_hidden, hp.n_mel_channels)
self.use_postnet = use_postnet
if self.use_postnet:
self.postnet = PostNet()
encoder_output = None
def __init__(self, use_postnet=True):
super(FastSpeech2, self).__init__()
self.encoder = Encoder()
self.variance_adaptor = TacotronDuration()
self.decoder = Decoder()
self.mel_linear = nn.Linear(hp.decoder_hidden, hp.n_mel_channels)
self.use_postnet = use_postnet
if self.use_postnet:
self.postnet = PostNet()
def __init__(self, use_postnet=True):
super(FastSpeech2, self).__init__()
# self.gst = GST()
self.encoder = Encoder()
self.variance_adaptor = VarianceAdaptor()
self.decoder = Decoder()
if hp.vocoder=='WORLD':
# self.f0_decoder= Decoder()
self.ap_linear = nn.Linear(hp.decoder_hidden, hp.n_ap_channels)
self.sp_linear = nn.Linear(hp.decoder_hidden, hp.n_sp_channels)
else:
self.mel_linear = nn.Linear(hp.decoder_hidden, hp.n_mel_channels)
self.use_postnet = use_postnet
if self.use_postnet:
self.postnet = PostNet()