You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
183 lines
5.9 KiB
183 lines
5.9 KiB
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
|
|
#
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
# you may not use this file except in compliance with the License.
|
|
# You may obtain a copy of the License at
|
|
#
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
# See the License for the specific language governing permissions and
|
|
# limitations under the License.
|
|
"""Generator module in JETS.
|
|
|
|
This code is based on https://github.com/imdanboy/jets.
|
|
|
|
"""
|
|
import numpy as np
|
|
import paddle
|
|
import paddle.nn.functional as F
|
|
from numba import jit
|
|
from paddle import nn
|
|
|
|
from paddlespeech.t2s.modules.masked_fill import masked_fill
|
|
|
|
|
|
class AlignmentModule(nn.Layer):
|
|
"""Alignment Learning Framework proposed for parallel TTS models in:
|
|
https://arxiv.org/abs/2108.10447
|
|
"""
|
|
|
|
def __init__(self, adim, odim):
|
|
super().__init__()
|
|
self.t_conv1 = nn.Conv1D(adim, adim, kernel_size=3, padding=1)
|
|
self.t_conv2 = nn.Conv1D(adim, adim, kernel_size=1, padding=0)
|
|
|
|
self.f_conv1 = nn.Conv1D(odim, adim, kernel_size=3, padding=1)
|
|
self.f_conv2 = nn.Conv1D(adim, adim, kernel_size=3, padding=1)
|
|
self.f_conv3 = nn.Conv1D(adim, adim, kernel_size=1, padding=0)
|
|
|
|
def forward(self, text, feats, x_masks=None):
|
|
"""
|
|
Args:
|
|
text (Tensor): Batched text embedding (B, T_text, adim)
|
|
feats (Tensor): Batched acoustic feature (B, T_feats, odim)
|
|
x_masks (Tensor): Mask tensor (B, T_text)
|
|
|
|
Returns:
|
|
Tensor: log probability of attention matrix (B, T_feats, T_text)
|
|
"""
|
|
|
|
text = text.transpose((0, 2, 1))
|
|
text = F.relu(self.t_conv1(text))
|
|
text = self.t_conv2(text)
|
|
text = text.transpose((0, 2, 1))
|
|
|
|
feats = feats.transpose((0, 2, 1))
|
|
feats = F.relu(self.f_conv1(feats))
|
|
feats = F.relu(self.f_conv2(feats))
|
|
feats = self.f_conv3(feats)
|
|
feats = feats.transpose((0, 2, 1))
|
|
|
|
dist = feats.unsqueeze(2) - text.unsqueeze(1)
|
|
dist = paddle.linalg.norm(dist, p=2, axis=3)
|
|
score = -dist
|
|
|
|
if x_masks is not None:
|
|
x_masks = x_masks.unsqueeze(-2)
|
|
score = masked_fill(score, x_masks, -np.inf)
|
|
log_p_attn = F.log_softmax(score, axis=-1)
|
|
return log_p_attn, score
|
|
|
|
|
|
@jit(nopython=True)
|
|
def _monotonic_alignment_search(log_p_attn):
|
|
# https://arxiv.org/abs/2005.11129
|
|
T_mel = log_p_attn.shape[0]
|
|
T_inp = log_p_attn.shape[1]
|
|
Q = np.full((T_inp, T_mel), fill_value=-np.inf)
|
|
|
|
log_prob = log_p_attn.transpose(1, 0) # -> (T_inp,T_mel)
|
|
# 1. Q <- init first row for all j
|
|
for j in range(T_mel):
|
|
Q[0, j] = log_prob[0, :j + 1].sum()
|
|
|
|
# 2.
|
|
for j in range(1, T_mel):
|
|
for i in range(1, min(j + 1, T_inp)):
|
|
Q[i, j] = max(Q[i - 1, j - 1], Q[i, j - 1]) + log_prob[i, j]
|
|
|
|
# 3.
|
|
A = np.full((T_mel, ), fill_value=T_inp - 1)
|
|
for j in range(T_mel - 2, -1, -1): # T_mel-2, ..., 0
|
|
# 'i' in {A[j+1]-1, A[j+1]}
|
|
i_a = A[j + 1] - 1
|
|
i_b = A[j + 1]
|
|
if i_b == 0:
|
|
argmax_i = 0
|
|
elif Q[i_a, j] >= Q[i_b, j]:
|
|
argmax_i = i_a
|
|
else:
|
|
argmax_i = i_b
|
|
A[j] = argmax_i
|
|
return A
|
|
|
|
|
|
def viterbi_decode(log_p_attn, text_lengths, feats_lengths):
|
|
"""
|
|
Args:
|
|
log_p_attn (Tensor):
|
|
Batched log probability of attention matrix (B, T_feats, T_text)
|
|
text_lengths (Tensor):
|
|
Text length tensor (B,)
|
|
feats_legnths (Tensor):
|
|
Feature length tensor (B,)
|
|
Returns:
|
|
Tensor:
|
|
Batched token duration extracted from `log_p_attn` (B,T_text)
|
|
Tensor:
|
|
binarization loss tensor ()
|
|
"""
|
|
B = log_p_attn.shape[0]
|
|
T_text = log_p_attn.shape[2]
|
|
device = log_p_attn.place
|
|
|
|
bin_loss = 0
|
|
ds = paddle.zeros((B, T_text), dtype="int32")
|
|
for b in range(B):
|
|
cur_log_p_attn = log_p_attn[b, :feats_lengths[b], :text_lengths[b]]
|
|
viterbi = _monotonic_alignment_search(cur_log_p_attn.numpy())
|
|
_ds = np.bincount(viterbi)
|
|
ds[b, :len(_ds)] = paddle.to_tensor(
|
|
_ds, place=device, dtype="int32")
|
|
|
|
t_idx = paddle.arange(feats_lengths[b])
|
|
bin_loss = bin_loss - cur_log_p_attn[t_idx, viterbi].mean()
|
|
bin_loss = bin_loss / B
|
|
return ds, bin_loss
|
|
|
|
|
|
@jit(nopython=True)
|
|
def _average_by_duration(ds, xs, text_lengths, feats_lengths):
|
|
B = ds.shape[0]
|
|
# xs_avg = np.zeros_like(ds)
|
|
xs_avg = np.zeros(shape=ds.shape, dtype=np.float32)
|
|
ds = ds.astype(np.int32)
|
|
for b in range(B):
|
|
t_text = text_lengths[b]
|
|
t_feats = feats_lengths[b]
|
|
d = ds[b, :t_text]
|
|
d_cumsum = d.cumsum()
|
|
d_cumsum = [0] + list(d_cumsum)
|
|
x = xs[b, :t_feats]
|
|
for n, (start, end) in enumerate(zip(d_cumsum[:-1], d_cumsum[1:])):
|
|
if len(x[start:end]) != 0:
|
|
xs_avg[b, n] = x[start:end].mean()
|
|
else:
|
|
xs_avg[b, n] = 0
|
|
return xs_avg
|
|
|
|
|
|
def average_by_duration(ds, xs, text_lengths, feats_lengths):
|
|
"""
|
|
Args:
|
|
ds (Tensor):
|
|
Batched token duration (B,T_text)
|
|
xs (Tensor):
|
|
Batched feature sequences to be averaged (B,T_feats)
|
|
text_lengths (Tensor):
|
|
Text length tensor (B,)
|
|
feats_lengths (Tensor):
|
|
Feature length tensor (B,)
|
|
Returns:
|
|
Tensor: Batched feature averaged according to the token duration (B, T_text)
|
|
"""
|
|
device = ds.place
|
|
args = [ds, xs, text_lengths, feats_lengths]
|
|
args = [arg.numpy() for arg in args]
|
|
xs_avg = _average_by_duration(*args)
|
|
xs_avg = paddle.to_tensor(xs_avg, place=device)
|
|
return xs_avg
|