testadmin
/
ParakeetEricRoss
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Merge pull request #69 from lfchener/develop 

fix the behavior of dropout in eval of tacotron2
This commit is contained in:
Li Fuchen 2020-12-28 16:30:14 +08:00 committed by GitHub
parent 9d06ec2d91 80bf04b710
commit c1de6a1e49
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 22 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

32
parakeet/models/tacotron2.py
View File

@ -71,8 +71,10 @@ class DecoderPreNet(nn.Layer):
"""
x = F.dropout(F.relu(self.linear1(x)), self.dropout_rate)
output = F.dropout(F.relu(self.linear2(x)), self.dropout_rate)
x = F.dropout(
F.relu(self.linear1(x)), self.dropout_rate, training=True)
output = F.dropout(
F.relu(self.linear2(x)), self.dropout_rate, training=True)
return output
@ -161,9 +163,13 @@ class DecoderPostNet(nn.Layer):
for i in range(len(self.conv_batchnorms) - 1):
input = F.dropout(
F.tanh(self.conv_batchnorms[i](input), self.dropout))
output = F.dropout(self.conv_batchnorms[self.num_layers - 1](input),
self.dropout)
F.tanh(self.conv_batchnorms[i](input)),
self.dropout,
training=self.training)
output = F.dropout(
self.conv_batchnorms[self.num_layers - 1](input),
self.dropout,
training=self.training)
return output
@ -228,8 +234,10 @@ class Tacotron2Encoder(nn.Layer):
"""
for conv_batchnorm in self.conv_batchnorms:
x = F.dropout(F.relu(conv_batchnorm(x)),
self.p_dropout) #(B, T, C)
x = F.dropout(
F.relu(conv_batchnorm(x)),
self.p_dropout,
training=self.training)
output, _ = self.lstm(inputs=x, sequence_length=input_lens)
return output
@ -350,8 +358,10 @@ class Tacotron2Decoder(nn.Layer):
# The first lstm layer
_, (self.attention_hidden, self.attention_cell) = self.attention_rnn(
cell_input, (self.attention_hidden, self.attention_cell))
self.attention_hidden = F.dropout(self.attention_hidden,
self.p_attention_dropout)
self.attention_hidden = F.dropout(
self.attention_hidden,
self.p_attention_dropout,
training=self.training)
# Loaction sensitive attention
attention_weights_cat = paddle.stack(
@ -367,7 +377,9 @@ class Tacotron2Decoder(nn.Layer):
_, (self.decoder_hidden, self.decoder_cell) = self.decoder_rnn(
decoder_input, (self.decoder_hidden, self.decoder_cell))
self.decoder_hidden = F.dropout(
self.decoder_hidden, p=self.p_decoder_dropout)
self.decoder_hidden,
p=self.p_decoder_dropout,
training=self.training)
# decode output one step
decoder_hidden_attention_context = paddle.concat(