wassname/TTS
1
0
Fork 0
mirror of https://github.com/wassname/TTS.git synced 2026-09-09 11:16:00 +08:00
Code Issues Packages Projects Releases Wiki Activity

update separate stopnet flow to make it faster.

Browse Source
This commit is contained in:
Eren Golge
2019-05-17 16:15:43 +02:00
parent 788c8100ba
commit e62659da94
8 changed files with 110 additions and 69 deletions
Download Patch File Download Diff File Expand all files Collapse all files
models/tacotron.py
+4 -2
View File
@@ -15,7 +15,8 @@ class Tacotron(nn.Module):
padding_idx=None,
memory_size=5,
attn_win=False,
attn_norm="sigmoid"):
attn_norm="sigmoid",
separate_stopnet=True):
super(Tacotron, self).__init__()
self.r = r
self.mel_dim = mel_dim
@@ -23,7 +24,8 @@ class Tacotron(nn.Module):
self.embedding = nn.Embedding(num_chars, 256, padding_idx=padding_idx)
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)
self.decoder = Decoder(256, mel_dim, r, memory_size, attn_win,
attn_norm, separate_stopnet)
self.postnet = PostCBHG(mel_dim)
self.last_linear = nn.Sequential(
nn.Linear(self.postnet.cbhg.gru_features * 2, linear_dim),
models/tacotron2.py
+17 -4
View File
@@ -9,7 +9,17 @@ from utils.generic_utils import sequence_mask
# TODO: match function arguments with tacotron
class Tacotron2(nn.Module):
def __init__(self, num_chars, r, attn_win=False, attn_norm="softmax", prenet_type="original", prenet_dropout=True, forward_attn=False, trans_agent=False, location_attn=True):
def __init__(self,
num_chars,
r,
attn_win=False,
attn_norm="softmax",
prenet_type="original",
prenet_dropout=True,
forward_attn=False,
trans_agent=False,
location_attn=True,
separate_stopnet=True):
super(Tacotron2, self).__init__()
self.n_mel_channels = 80
self.n_frames_per_step = r
@@ -18,7 +28,10 @@ class Tacotron2(nn.Module):
val = sqrt(3.0) * std # uniform bounds for std
self.embedding.weight.data.uniform_(-val, val)
self.encoder = Encoder(512)
self.decoder = Decoder(512, self.n_mel_channels, r, attn_win, attn_norm, prenet_type, prenet_dropout, forward_attn, trans_agent, location_attn)
self.decoder = Decoder(512, self.n_mel_channels, r, attn_win,
attn_norm, prenet_type, prenet_dropout,
forward_attn, trans_agent, location_attn,
separate_stopnet)
self.postnet = Postnet(self.n_mel_channels)
def shape_outputs(self, mel_outputs, mel_outputs_postnet, alignments):
@@ -50,14 +63,14 @@ class Tacotron2(nn.Module):
mel_outputs, mel_outputs_postnet, alignments)
return mel_outputs, mel_outputs_postnet, alignments, stop_tokens
def inference_truncated(self, text):
"""
Preserve model states for continuous inference
"""
embedded_inputs = self.embedding(text).transpose(1, 2)
encoder_outputs = self.encoder.inference_truncated(embedded_inputs)
mel_outputs, stop_tokens, alignments = self.decoder.inference_truncated(encoder_outputs)
mel_outputs, stop_tokens, alignments = self.decoder.inference_truncated(
encoder_outputs)
mel_outputs_postnet = self.postnet(mel_outputs)
mel_outputs_postnet = mel_outputs + mel_outputs_postnet
mel_outputs, mel_outputs_postnet, alignments = self.shape_outputs(