wassname/TTS
1
0
Fork 0
mirror of https://github.com/wassname/TTS.git synced 2026-09-11 12:00:24 +08:00
Code Issues Packages Projects Releases Wiki Activity

Fix Pylint issues

Browse Source
This commit is contained in:
Reuben Morais
2019-07-19 09:08:51 +02:00
parent 509292d56a
commit 11e7895329
35 changed files with 270 additions and 316 deletions
Download Patch File Download Diff File Expand all files Collapse all files
layers/common_layers.py
+4 -3
View File
@@ -1,7 +1,6 @@
from math import sqrt
import torch
from torch.autograd import Variable
from torch import nn
from torch.autograd import Variable
from torch.nn import functional as F
@@ -107,6 +106,8 @@ class LocationLayer(nn.Module):
class Attention(nn.Module):
# Pylint gets confused by PyTorch conventions here
#pylint: disable=attribute-defined-outside-init
def __init__(self, attention_rnn_dim, embedding_dim, attention_dim,
location_attention, attention_location_n_filters,
attention_location_kernel_size, windowing, norm, forward_attn,
@@ -262,4 +263,4 @@ class Attention(nn.Module):
context = torch.bmm(alignment.unsqueeze(1), inputs)
context = context.squeeze(1)
self.attention_weights = alignment
return context
return context
layers/custom_layers.py
+2 -2
View File
@@ -1,6 +1,6 @@
# coding: utf-8
import torch
from torch import nn
# import torch
# from torch import nn
# class StopProjection(nn.Module):
# r""" Simple projection layer to predict the "stop token"
layers/gst_layers.py
+4 -3
View File
@@ -77,10 +77,11 @@ class ReferenceEncoder(nn.Module):
return out.squeeze(0)
def calculate_post_conv_height(self, height, kernel_size, stride, pad,
@staticmethod
def calculate_post_conv_height(height, kernel_size, stride, pad,
n_convs):
"""Height of spec after n convolutions with fixed kernel/stride/pad."""
for i in range(n_convs):
for _ in range(n_convs):
height = (height - kernel_size + 2 * pad) // stride + 1
return height
@@ -165,4 +166,4 @@ class MultiHeadAttention(nn.Module):
torch.split(out, 1, dim=0),
dim=3).squeeze(0) # [N, T_q, num_units]
return out
return out
layers/losses.py
+9 -17
View File
@@ -1,17 +1,13 @@
import torch
from torch.nn import functional
from torch import nn
from torch.nn import functional
from utils.generic_utils import sequence_mask
class L1LossMasked(nn.Module):
def __init__(self):
super(L1LossMasked, self).__init__()
def forward(self, input, target, length):
def forward(self, x, target, length):
"""
Args:
input: A Variable containing a FloatTensor of size
x: A Variable containing a FloatTensor of size
(batch, max_len, dim) which contains the
unnormalized probability for each class.
target: A Variable containing a LongTensor of size
@@ -26,21 +22,18 @@ class L1LossMasked(nn.Module):
target.requires_grad = False
mask = sequence_mask(
sequence_length=length, max_len=target.size(1)).unsqueeze(2).float()
mask = mask.expand_as(input)
mask = mask.expand_as(x)
loss = functional.l1_loss(
input * mask, target * mask, reduction="sum")
x * mask, target * mask, reduction="sum")
loss = loss / mask.sum()
return loss
class MSELossMasked(nn.Module):
def __init__(self):
super(MSELossMasked, self).__init__()
def forward(self, input, target, length):
def forward(self, x, target, length):
"""
Args:
input: A Variable containing a FloatTensor of size
x: A Variable containing a FloatTensor of size
(batch, max_len, dim) which contains the
unnormalized probability for each class.
target: A Variable containing a LongTensor of size
@@ -55,9 +48,8 @@ class MSELossMasked(nn.Module):
target.requires_grad = False
mask = sequence_mask(
sequence_length=length, max_len=target.size(1)).unsqueeze(2).float()
mask = mask.expand_as(input)
mask = mask.expand_as(x)
loss = functional.mse_loss(
input * mask, target * mask, reduction="sum")
x * mask, target * mask, reduction="sum")
loss = loss / mask.sum()
return loss
layers/tacotron.py
+14 -12
View File
@@ -177,7 +177,7 @@ class CBHG(nn.Module):
# (B, in_features, T_in)
if x.size(-1) == self.in_features:
x = x.transpose(1, 2)
T = x.size(-1)
# T = x.size(-1)
# (B, hid_features*K, T_in)
# Concat conv1d bank outputs
outs = []
@@ -261,7 +261,7 @@ class PostCBHG(nn.Module):
class Decoder(nn.Module):
r"""Decoder module.
"""Decoder module.
Args:
in_features (int): input vector (encoder output) sample size.
@@ -270,6 +270,8 @@ class Decoder(nn.Module):
memory_size (int): size of the past window. if <= 0 memory_size = r
TODO: arguments
"""
# Pylint gets confused by PyTorch conventions here
#pylint: disable=attribute-defined-outside-init
def __init__(self, in_features, memory_dim, r, memory_size, attn_windowing,
attn_norm, prenet_type, prenet_dropout, forward_attn,
@@ -290,16 +292,16 @@ class Decoder(nn.Module):
# processed_inputs, processed_memory -> |Attention| -> Attention, attention, RNN_State
self.attention_rnn = nn.GRUCell(in_features + 128, 256)
self.attention_layer = Attention(attention_rnn_dim=256,
embedding_dim=in_features,
attention_dim=128,
location_attention=location_attn,
attention_location_n_filters=32,
attention_location_kernel_size=31,
windowing=attn_windowing,
norm=attn_norm,
forward_attn=forward_attn,
trans_agent=trans_agent,
forward_attn_mask=forward_attn_mask)
embedding_dim=in_features,
attention_dim=128,
location_attention=location_attn,
attention_location_n_filters=32,
attention_location_kernel_size=31,
windowing=attn_windowing,
norm=attn_norm,
forward_attn=forward_attn,
trans_agent=trans_agent,
forward_attn_mask=forward_attn_mask)
# (processed_memory | attention context) -> |Linear| -> decoder_RNN_input
self.project_to_decoder_in = nn.Linear(256 + in_features, 256)
# decoder_RNN_input -> |RNN| -> RNN_state
layers/tacotron2.py
+10 -9
View File
@@ -1,9 +1,8 @@
from math import sqrt
import torch
from torch.autograd import Variable
from torch import nn
from torch.nn import functional as F
from .common_layers import Attention, Prenet, Linear, LinearBN
from .common_layers import Attention, Prenet, Linear
class ConvBNBlock(nn.Module):
@@ -33,7 +32,7 @@ class Postnet(nn.Module):
self.convolutions = nn.ModuleList()
self.convolutions.append(
ConvBNBlock(mel_dim, 512, kernel_size=5, nonlinear='tanh'))
for i in range(1, num_convs - 1):
for _ in range(1, num_convs - 1):
self.convolutions.append(
ConvBNBlock(512, 512, kernel_size=5, nonlinear='tanh'))
self.convolutions.append(
@@ -95,6 +94,8 @@ class Encoder(nn.Module):
# adapted from https://github.com/NVIDIA/tacotron2/
class Decoder(nn.Module):
# Pylint gets confused by PyTorch conventions here
#pylint: disable=attribute-defined-outside-init
def __init__(self, in_features, inputs_dim, r, attn_win, attn_norm,
prenet_type, prenet_dropout, forward_attn, trans_agent,
forward_attn_mask, location_attn, separate_stopnet):
@@ -118,15 +119,15 @@ class Decoder(nn.Module):
self.attention_rnn = nn.LSTMCell(self.prenet_dim + in_features,
self.attention_rnn_dim)
self.attention_layer = Attention(attention_rnn_dim=self.attention_rnn_dim,
self.attention_layer = Attention(attention_rnn_dim=self.attention_rnn_dim,
embedding_dim=in_features,
attention_dim=128,
location_attention=location_attn,
attention_dim=128,
location_attention=location_attn,
attention_location_n_filters=32,
attention_location_kernel_size=31,
windowing=attn_win,
norm=attn_norm,
forward_attn=forward_attn,
norm=attn_norm,
forward_attn=forward_attn,
trans_agent=trans_agent,
forward_attn_mask=forward_attn_mask)
@@ -156,7 +157,7 @@ class Decoder(nn.Module):
def _init_states(self, inputs, mask, keep_states=False):
B = inputs.size(0)
T = inputs.size(1)
# T = inputs.size(1)
if not keep_states:
self.attention_hidden = self.attention_rnn_init(