def __init__(self,
num_chars,
num_speakers,
r=5,
linear_dim=1025,
mel_dim=80,
memory_size=5,
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,
separate_stopnet=True):
super(Tacotron, self).__init__()
self.r = r
self.mel_dim = mel_dim
self.linear_dim = linear_dim
self.embedding = nn.Embedding(num_chars, 256)
self.embedding.weight.data.normal_(0, 0.3)
if num_speakers > 1:
self.speaker_embedding = nn.Embedding(num_speakers, 256)
self.speaker_embedding.weight.data.normal_(0, 0.3)
self.encoder = Encoder(256)
self.decoder = Decoder(256, mel_dim, r, memory_size, attn_win,
attn_norm, prenet_type, prenet_dropout,
forward_attn, trans_agent, forward_attn_mask,
location_attn, separate_stopnet)
self.postnet = PostCBHG(mel_dim)
self.last_linear = nn.Linear(self.postnet.cbhg.gru_features * 2,
linear_dim)
def __init__(self,
num_chars,
embedding_dim=256,
linear_dim=1025,
mel_dim=80,
r=5,
padding_idx=None,
memory_size=5,
attn_windowing=False,
forward_attention=False):
super(Tacotron, self).__init__()
self.r = r
self.mel_dim = mel_dim
self.linear_dim = linear_dim
self.embedding = nn.Embedding(num_chars,
embedding_dim,
padding_idx=padding_idx)
self.embedding.weight.data.normal_(0, 0.3)
self.encoder = Encoder(embedding_dim)
self.decoder = Decoder(256, mel_dim, r, memory_size, attn_windowing,
forward_attention)
self.postnet = PostCBHG(mel_dim)
self.last_linear = nn.Sequential(
nn.Linear(self.postnet.cbhg.gru_features * 2, linear_dim),
nn.Sigmoid())
def test_in_out():
layer = Decoder(
in_features=256,
memory_dim=80,
r=2,
memory_size=4,
attn_windowing=False,
attn_norm="sigmoid",
prenet_type='original',
prenet_dropout=True,
forward_attn=True,
trans_agent=True,
forward_attn_mask=True,
location_attn=True,
separate_stopnet=True)
dummy_input = T.rand(4, 8, 256)
dummy_memory = T.rand(4, 2, 80)
output, alignment, stop_tokens = layer(
dummy_input, dummy_memory, mask=None)
assert output.shape[0] == 4
assert output.shape[1] == 1, "size not {}".format(output.shape[1])
assert output.shape[2] == 80 * 2, "size not {}".format(output.shape[2])
assert stop_tokens.shape[0] == 4
def test_in_out(self):
layer = Decoder(in_features=256, memory_dim=80, r=2, memory_size=4, attn_windowing=False, attn_norm="sigmoid")
dummy_input = T.rand(4, 8, 256)
dummy_memory = T.rand(4, 2, 80)
output, alignment, stop_tokens = layer(dummy_input, dummy_memory, mask=None)
assert output.shape[0] == 4
assert output.shape[1] == 1, "size not {}".format(output.shape[1])
assert output.shape[2] == 80 * 2, "size not {}".format(output.shape[2])
assert stop_tokens.shape[0] == 4
class Tacotron(nn.Module):
def __init__(self,
num_chars,
r=5,
linear_dim=1025,
mel_dim=80,
memory_size=5,
attn_win=False,
attn_norm="sigmoid",
prenet_type="original",
prenet_dropout=True,
forward_attn=False,
trans_agent=False,
location_attn=True,
separate_stopnet=True):
super(Tacotron, self).__init__()
self.r = r
self.mel_dim = mel_dim
self.linear_dim = linear_dim
self.embedding = nn.Embedding(num_chars, 256)
self.embedding.weight.data.normal_(0, 0.3)
self.encoder = Encoder(256)
self.decoder = Decoder(256, mel_dim, r, memory_size, attn_win,
attn_norm, prenet_type, prenet_dropout,
forward_attn, trans_agent, location_attn,
separate_stopnet)
self.postnet = PostCBHG(mel_dim)
self.last_linear = nn.Sequential(
nn.Linear(self.postnet.cbhg.gru_features * 2, linear_dim),
nn.Sigmoid())
def forward(self, characters, text_lengths, mel_specs):
B = characters.size(0)
mask = sequence_mask(text_lengths).to(characters.device)
inputs = self.embedding(characters)
encoder_outputs = self.encoder(inputs)
mel_outputs, alignments, stop_tokens = self.decoder(
encoder_outputs, mel_specs, mask)
mel_outputs = mel_outputs.view(B, -1, self.mel_dim)
linear_outputs = self.postnet(mel_outputs)
linear_outputs = self.last_linear(linear_outputs)
return mel_outputs, linear_outputs, alignments, stop_tokens
def inference(self, characters):
B = characters.size(0)
inputs = self.embedding(characters)
encoder_outputs = self.encoder(inputs)
mel_outputs, alignments, stop_tokens = self.decoder.inference(
encoder_outputs)
mel_outputs = mel_outputs.view(B, -1, self.mel_dim)
linear_outputs = self.postnet(mel_outputs)
linear_outputs = self.last_linear(linear_outputs)
return mel_outputs, linear_outputs, alignments, stop_tokens
def test_in_out(self):
layer = Decoder(in_features=256, memory_dim=80, r=2)
dummy_input = T.rand(4, 8, 256)
dummy_memory = T.rand(4, 2, 80)
output, alignment, stop_tokens = layer(dummy_input, dummy_memory)
assert output.shape[0] == 4
assert output.shape[1] == 1, "size not {}".format(output.shape[1])
assert output.shape[2] == 80 * 2, "size not {}".format(output.shape[2])
assert stop_tokens.shape[0] == 4
assert stop_tokens.max() <= 1.0
assert stop_tokens.min() >= 0
def __init__(self,
embedding_dim=256,
linear_dim=1025,
mel_dim=80,
r=5,
padding_idx=None):
super(Tacotron, self).__init__()
self.r = r
self.mel_dim = mel_dim
self.linear_dim = linear_dim
self.embedding = nn.Embedding(len(symbols),
embedding_dim,
padding_idx=padding_idx)
print(" | > Number of characters : {}".format(len(symbols)))
self.embedding.weight.data.normal_(0, 0.3)
self.encoder = Encoder(embedding_dim)
self.decoder = Decoder(256, mel_dim, r)
self.postnet = CBHG(mel_dim, K=8, projections=[256, mel_dim])
self.last_linear = nn.Linear(mel_dim * 2, linear_dim)
def __init__(self,
embedding_dim=256,
linear_dim=1025,
mel_dim=80,
r=5,
padding_idx=None):
super(Tacotron, self).__init__()
self.r = r
self.mel_dim = mel_dim
self.linear_dim = linear_dim
self.embedding = nn.Embedding(
len(symbols), embedding_dim, padding_idx=padding_idx)
print(" | > Number of characters : {}".format(len(symbols)))
self.embedding.weight.data.normal_(0, 0.3)
self.encoder = Encoder(embedding_dim)
self.decoder = Decoder(256, mel_dim, r)
self.postnet = PostCBHG(mel_dim)
self.last_linear = nn.Sequential(
nn.Linear(self.postnet.cbhg.gru_features * 2, linear_dim),
nn.Sigmoid())
class TacotronGST(nn.Module):
def __init__(self,
num_chars,
num_speakers,
r=5,
linear_dim=1025,
mel_dim=80,
memory_size=5,
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,
separate_stopnet=True):
super(TacotronGST, self).__init__()
self.r = r
self.mel_dim = mel_dim
self.linear_dim = linear_dim
self.embedding = nn.Embedding(num_chars, 256)
self.embedding.weight.data.normal_(0, 0.3)
if num_speakers > 1:
self.speaker_embedding = nn.Embedding(num_speakers, 256)
self.speaker_embedding.weight.data.normal_(0, 0.3)
self.encoder = Encoder(256)
self.gst = GST(num_mel=80,
num_heads=4,
num_style_tokens=10,
embedding_dim=256)
self.decoder = Decoder(256, mel_dim, r, memory_size, attn_win,
attn_norm, prenet_type, prenet_dropout,
forward_attn, trans_agent, forward_attn_mask,
location_attn, separate_stopnet)
self.postnet = PostCBHG(mel_dim)
self.last_linear = nn.Sequential(
nn.Linear(self.postnet.cbhg.gru_features * 2, linear_dim),
nn.Sigmoid())
def forward(self, characters, text_lengths, mel_specs, speaker_ids=None):
B = characters.size(0)
mask = sequence_mask(text_lengths).to(characters.device)
inputs = self.embedding(characters)
encoder_outputs = self.encoder(inputs)
encoder_outputs = self._add_speaker_embedding(encoder_outputs,
speaker_ids)
gst_outputs = self.gst(mel_specs)
gst_outputs = gst_outputs.expand(-1, encoder_outputs.size(1), -1)
encoder_outputs = encoder_outputs + gst_outputs
mel_outputs, alignments, stop_tokens = self.decoder(
encoder_outputs, mel_specs, mask)
mel_outputs = mel_outputs.view(B, -1, self.mel_dim)
linear_outputs = self.postnet(mel_outputs)
linear_outputs = self.last_linear(linear_outputs)
return mel_outputs, linear_outputs, alignments, stop_tokens
def inference(self, characters, speaker_ids=None, style_mel=None):
B = characters.size(0)
inputs = self.embedding(characters)
encoder_outputs = self.encoder(inputs)
encoder_outputs = self._add_speaker_embedding(encoder_outputs,
speaker_ids)
if style_mel is not None:
gst_outputs = self.gst(style_mel)
gst_outputs = gst_outputs.expand(-1, encoder_outputs.size(1), -1)
encoder_outputs = encoder_outputs + gst_outputs
mel_outputs, alignments, stop_tokens = self.decoder.inference(
encoder_outputs)
mel_outputs = mel_outputs.view(B, -1, self.mel_dim)
linear_outputs = self.postnet(mel_outputs)
linear_outputs = self.last_linear(linear_outputs)
return mel_outputs, linear_outputs, alignments, stop_tokens
def _add_speaker_embedding(self, encoder_outputs, speaker_ids):
if hasattr(self, "speaker_embedding") and speaker_ids is not None:
speaker_embeddings = self.speaker_embedding(speaker_ids)
speaker_embeddings.unsqueeze_(1)
speaker_embeddings = speaker_embeddings.expand(
encoder_outputs.size(0), encoder_outputs.size(1), -1)
encoder_outputs = encoder_outputs + speaker_embeddings
return encoder_outputs
