def __init__(self, attention_rnn_dim, embedding_dim, attention_dim,
attention_location_n_filters, attention_location_kernel_size):
super(Attention, self).__init__()
self.query_layer = LinearNorm(attention_rnn_dim, attention_dim,
bias=False, w_init_gain='tanh')
self.memory_layer = LinearNorm(embedding_dim, attention_dim, bias=False,
w_init_gain='tanh')
self.v = LinearNorm(attention_dim, 1, bias=False)
self.location_layer = LocationLayer(attention_location_n_filters,
attention_location_kernel_size,
attention_dim)
self.score_mask_value = -float("inf")
def __init__(self, hparams):
super(Decoder, self).__init__()
self.n_mel_channels = hparams.n_mel_channels
self.n_frames_per_step = hparams.n_frames_per_step
self.encoder_embedding_dim = hparams.encoder_embedding_dim + hparams.token_embedding_size + hparams.speaker_embedding_dim
self.attention_rnn_dim = hparams.attention_rnn_dim
self.decoder_rnn_dim = hparams.decoder_rnn_dim
self.prenet_dim = hparams.prenet_dim
self.max_decoder_steps = hparams.max_decoder_steps
self.gate_threshold = hparams.gate_threshold
self.p_attention_dropout = hparams.p_attention_dropout
self.p_decoder_dropout = hparams.p_decoder_dropout
self.p_teacher_forcing = hparams.p_teacher_forcing
if hparams.prenet_f0_dim > 0:
self.prenet_f0 = ConvNorm(
hparams.prenet_f0_dim, # 输入f0s的维度。
hparams.prenet_f0_dim,
kernel_size=hparams.prenet_f0_kernel_size,
padding=max(0, int(hparams.prenet_f0_kernel_size / 2)),
bias=False,
stride=1,
dilation=1)
self.prenet = Prenet(
hparams.n_mel_channels * hparams.n_frames_per_step,
[hparams.prenet_dim, hparams.prenet_dim])
self.attention_rnn = nn.LSTMCell(
hparams.prenet_dim + hparams.prenet_f0_dim +
self.encoder_embedding_dim, hparams.attention_rnn_dim)
self.attention_layer = Attention(
hparams.attention_rnn_dim, self.encoder_embedding_dim,
hparams.attention_dim, hparams.attention_location_n_filters,
hparams.attention_location_kernel_size)
self.decoder_rnn = nn.LSTMCell(
hparams.attention_rnn_dim + self.encoder_embedding_dim,
hparams.decoder_rnn_dim, 1)
self.linear_projection = LinearNorm(
hparams.decoder_rnn_dim + self.encoder_embedding_dim,
hparams.n_mel_channels * hparams.n_frames_per_step)
self.gate_layer = LinearNorm(hparams.decoder_rnn_dim +
self.encoder_embedding_dim,
1,
bias=True,
w_init_gain='sigmoid')
def __init__(self, in_dim, sizes):
super(Prenet, self).__init__()
in_sizes = [in_dim] + sizes[:-1]
self.layers = nn.ModuleList([
LinearNorm(in_size, out_size, bias=False)
for (in_size, out_size) in zip(in_sizes, sizes)
])
def __init__(self, attention_n_filters, attention_kernel_size,
attention_dim):
super(LocationLayer, self).__init__()
padding = int((attention_kernel_size - 1) / 2)
self.location_conv = ConvNorm(2, attention_n_filters,
kernel_size=attention_kernel_size,
padding=padding, bias=False, stride=1,
dilation=1)
self.location_dense = LinearNorm(attention_n_filters, attention_dim,
bias=False, w_init_gain='tanh')