def __init__(self, in_channels, out_channels, hidden_channels, params):
super().__init__()
self.prenet = Conv1dBN(in_channels, hidden_channels, 1, 1)
self.rel_pos_transformer = RelativePositionTransformer(
in_channels, out_channels, hidden_channels, **params
)
def __init__(self, in_channels, out_channels, hidden_channels, params):
super().__init__()
self.prenet = ResidualConv1dBNBlock(in_channels,
hidden_channels,
hidden_channels,
kernel_size=5,
num_res_blocks=3,
num_conv_blocks=1,
dilations=[1, 1, 1])
self.rel_pos_transformer = RelativePositionTransformer(
hidden_channels, out_channels, hidden_channels, **params)
def __init__(
self,
n_vocab: int,
out_channels: int,
hidden_channels: int,
hidden_channels_ffn: int,
num_heads: int,
num_layers: int,
kernel_size: int,
dropout_p: float,
language_emb_dim: int = None,
):
"""Text Encoder for VITS model.
Args:
n_vocab (int): Number of characters for the embedding layer.
out_channels (int): Number of channels for the output.
hidden_channels (int): Number of channels for the hidden layers.
hidden_channels_ffn (int): Number of channels for the convolutional layers.
num_heads (int): Number of attention heads for the Transformer layers.
num_layers (int): Number of Transformer layers.
kernel_size (int): Kernel size for the FFN layers in Transformer network.
dropout_p (float): Dropout rate for the Transformer layers.
"""
super().__init__()
self.out_channels = out_channels
self.hidden_channels = hidden_channels
self.emb = nn.Embedding(n_vocab, hidden_channels)
nn.init.normal_(self.emb.weight, 0.0, hidden_channels**-0.5)
if language_emb_dim:
hidden_channels += language_emb_dim
self.encoder = RelativePositionTransformer(
in_channels=hidden_channels,
out_channels=hidden_channels,
hidden_channels=hidden_channels,
hidden_channels_ffn=hidden_channels_ffn,
num_heads=num_heads,
num_layers=num_layers,
kernel_size=kernel_size,
dropout_p=dropout_p,
layer_norm_type="2",
rel_attn_window_size=4,
)
self.proj = nn.Conv1d(hidden_channels, out_channels * 2, 1)
def __init__(
self,
num_chars,
out_channels,
hidden_channels,
hidden_channels_dp,
encoder_type,
encoder_params,
dropout_p_dp=0.1,
mean_only=False,
use_prenet=True,
c_in_channels=0,
):
super().__init__()
# class arguments
self.num_chars = num_chars
self.out_channels = out_channels
self.hidden_channels = hidden_channels
self.hidden_channels_dp = hidden_channels_dp
self.dropout_p_dp = dropout_p_dp
self.mean_only = mean_only
self.use_prenet = use_prenet
self.c_in_channels = c_in_channels
self.encoder_type = encoder_type
# embedding layer
self.emb = nn.Embedding(num_chars, hidden_channels)
nn.init.normal_(self.emb.weight, 0.0, hidden_channels ** -0.5)
# init encoder module
if encoder_type.lower() == "rel_pos_transformer":
if use_prenet:
self.prenet = ResidualConv1dLayerNormBlock(
hidden_channels, hidden_channels, hidden_channels, kernel_size=5, num_layers=3, dropout_p=0.5
)
self.encoder = RelativePositionTransformer(
hidden_channels, hidden_channels, hidden_channels, **encoder_params
)
elif encoder_type.lower() == "gated_conv":
self.encoder = GatedConvBlock(hidden_channels, **encoder_params)
elif encoder_type.lower() == "residual_conv_bn":
if use_prenet:
self.prenet = nn.Sequential(nn.Conv1d(hidden_channels, hidden_channels, 1), nn.ReLU())
self.encoder = ResidualConv1dBNBlock(hidden_channels, hidden_channels, hidden_channels, **encoder_params)
self.postnet = nn.Sequential(
nn.Conv1d(self.hidden_channels, self.hidden_channels, 1), nn.BatchNorm1d(self.hidden_channels)
)
elif encoder_type.lower() == "time_depth_separable":
if use_prenet:
self.prenet = ResidualConv1dLayerNormBlock(
hidden_channels, hidden_channels, hidden_channels, kernel_size=5, num_layers=3, dropout_p=0.5
)
self.encoder = TimeDepthSeparableConvBlock(
hidden_channels, hidden_channels, hidden_channels, **encoder_params
)
else:
raise ValueError(" [!] Unkown encoder type.")
# final projection layers
self.proj_m = nn.Conv1d(hidden_channels, out_channels, 1)
if not mean_only:
self.proj_s = nn.Conv1d(hidden_channels, out_channels, 1)
# duration predictor
self.duration_predictor = DurationPredictor(
hidden_channels + c_in_channels, hidden_channels_dp, 3, dropout_p_dp
)
