def __init__(self, cfg: DictConfig):
super().__init__(cfg=cfg)
typecheck.set_typecheck_enabled(enabled=False)
cfg = self._cfg
self.vocab = AudioToCharWithDursF0Dataset.make_vocab(**cfg.train_ds.dataset.vocab)
self.embed = nn.Embedding(len(self.vocab.labels), cfg.d_char)
self.model = instantiate(cfg.model)
d_out = cfg.model.jasper[-1].filters
self.proj = nn.Conv1d(d_out, 1, kernel_size=1)
def __init__(self, cfg: DictConfig, trainer: 'Trainer' = None):
super().__init__(cfg=cfg, trainer=trainer)
typecheck.set_typecheck_enabled(enabled=False)
cfg = self._cfg
self.vocab = AudioToCharWithDursF0Dataset.make_vocab(
**cfg.train_ds.dataset.vocab)
self.embed = GaussianEmbedding(self.vocab, cfg.d_char)
self.model = instantiate(cfg.model)
d_out = cfg.model.jasper[-1].filters
self.sil_proj = nn.Conv1d(d_out, 1, kernel_size=1)
self.body_proj = nn.Conv1d(d_out, 1, kernel_size=1)
self.f0_mean, self.f0_std = cfg.f0_mean, cfg.f0_std
def __init__(self, cfg: DictConfig):
super().__init__(cfg=cfg)
typecheck.set_typecheck_enabled(enabled=False)
cfg = self._cfg
self.vocab = AudioToCharWithDursF0Dataset.make_vocab(**cfg.train_ds.dataset.vocab)
self.preprocessor = instantiate(cfg.preprocessor)
self.embed = GaussianEmbedding(self.vocab, cfg.d_char)
self.norm_f0 = MaskedInstanceNorm1d(1)
self.res_f0 = StyleResidual(cfg.d_char, 1, kernel_size=3)
self.model = instantiate(cfg.model)
d_out = cfg.model.jasper[-1].filters
self.proj = nn.Conv1d(d_out, cfg.n_mels, kernel_size=1)
def __init__(self, cfg: DictConfig, trainer: 'Trainer' = None):
super().__init__(cfg=cfg, trainer=trainer)
typecheck.set_typecheck_enabled(enabled=False)
cfg = self._cfg
self.vocab = AudioToCharWithDursF0Dataset.make_vocab(**cfg.train_ds.dataset.vocab)
self.embed = nn.Embedding(len(self.vocab.labels), cfg.d_char)
self.preprocessor = instantiate(cfg.preprocessor)
self.alignment_encoder = instantiate(cfg.alignment_encoder)
self.forward_sum_loss = ForwardSumLoss()
self.bin_loss = BinLoss()
self.bin_start_ratio = cfg.bin_start_ratio
self.add_bin_loss = False
