File size: 12,752 Bytes
cea6632
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
# Written by Shigeki Karita, 2019
# Published under Apache 2.0  (http://www.apache.org/licenses/LICENSE-2.0)
# Adapted by Florian Lux, 2021

"""Multi-Head Attention layer definition."""

import math

import numpy
import torch
from torch import nn

from Utility.utils import make_non_pad_mask


class MultiHeadedAttention(nn.Module):
    """
    Multi-Head Attention layer.

    Args:
        n_head (int): The number of heads.
        n_feat (int): The number of features.
        dropout_rate (float): Dropout rate.
    """

    def __init__(self, n_head, n_feat, dropout_rate):
        """
        Construct an MultiHeadedAttention object.
        """
        super(MultiHeadedAttention, self).__init__()
        assert n_feat % n_head == 0
        # We assume d_v always equals d_k
        self.d_k = n_feat // n_head
        self.h = n_head
        self.linear_q = nn.Linear(n_feat, n_feat)
        self.linear_k = nn.Linear(n_feat, n_feat)
        self.linear_v = nn.Linear(n_feat, n_feat)
        self.linear_out = nn.Linear(n_feat, n_feat)
        self.attn = None
        self.dropout = nn.Dropout(p=dropout_rate)

    def forward_qkv(self, query, key, value):
        """
        Transform query, key and value.

        Args:
            query (torch.Tensor): Query tensor (#batch, time1, size).
            key (torch.Tensor): Key tensor (#batch, time2, size).
            value (torch.Tensor): Value tensor (#batch, time2, size).

        Returns:
            torch.Tensor: Transformed query tensor (#batch, n_head, time1, d_k).
            torch.Tensor: Transformed key tensor (#batch, n_head, time2, d_k).
            torch.Tensor: Transformed value tensor (#batch, n_head, time2, d_k).
        """
        n_batch = query.size(0)
        q = self.linear_q(query).view(n_batch, -1, self.h, self.d_k)
        k = self.linear_k(key).view(n_batch, -1, self.h, self.d_k)
        v = self.linear_v(value).view(n_batch, -1, self.h, self.d_k)
        q = q.transpose(1, 2)  # (batch, head, time1, d_k)
        k = k.transpose(1, 2)  # (batch, head, time2, d_k)
        v = v.transpose(1, 2)  # (batch, head, time2, d_k)

        return q, k, v

    def forward_attention(self, value, scores, mask):
        """
        Compute attention context vector.

        Args:
            value (torch.Tensor): Transformed value (#batch, n_head, time2, d_k).
            scores (torch.Tensor): Attention score (#batch, n_head, time1, time2).
            mask (torch.Tensor): Mask (#batch, 1, time2) or (#batch, time1, time2).

        Returns:
            torch.Tensor: Transformed value (#batch, time1, d_model)
                weighted by the attention score (#batch, time1, time2).
        """
        n_batch = value.size(0)
        if mask is not None:
            mask = mask.unsqueeze(1).eq(0)  # (batch, 1, *, time2)
            min_value = float(numpy.finfo(torch.tensor(0, dtype=scores.dtype).numpy().dtype).min)
            scores = scores.masked_fill(mask, min_value)
            self.attn = torch.softmax(scores, dim=-1).masked_fill(mask, 0.0)  # (batch, head, time1, time2)
        else:
            self.attn = torch.softmax(scores, dim=-1)  # (batch, head, time1, time2)

        p_attn = self.dropout(self.attn)
        x = torch.matmul(p_attn, value)  # (batch, head, time1, d_k)
        x = (x.transpose(1, 2).contiguous().view(n_batch, -1, self.h * self.d_k))  # (batch, time1, d_model)

        return self.linear_out(x)  # (batch, time1, d_model)

    def forward(self, query, key, value, mask):
        """
        Compute scaled dot product attention.

        Args:
            query (torch.Tensor): Query tensor (#batch, time1, size).
            key (torch.Tensor): Key tensor (#batch, time2, size).
            value (torch.Tensor): Value tensor (#batch, time2, size).
            mask (torch.Tensor): Mask tensor (#batch, 1, time2) or
                (#batch, time1, time2).

        Returns:
            torch.Tensor: Output tensor (#batch, time1, d_model).
        """
        q, k, v = self.forward_qkv(query, key, value)
        scores = torch.matmul(q, k.transpose(-2, -1)) / math.sqrt(self.d_k)
        return self.forward_attention(v, scores, mask)


class RelPositionMultiHeadedAttention(MultiHeadedAttention):
    """
    Multi-Head Attention layer with relative position encoding.
    Details can be found in https://github.com/espnet/espnet/pull/2816.
    Paper: https://arxiv.org/abs/1901.02860
    Args:
        n_head (int): The number of heads.
        n_feat (int): The number of features.
        dropout_rate (float): Dropout rate.
        zero_triu (bool): Whether to zero the upper triangular part of attention matrix.
    """

    def __init__(self, n_head, n_feat, dropout_rate, zero_triu=False):
        """Construct an RelPositionMultiHeadedAttention object."""
        super().__init__(n_head, n_feat, dropout_rate)
        self.zero_triu = zero_triu
        # linear transformation for positional encoding
        self.linear_pos = nn.Linear(n_feat, n_feat, bias=False)
        # these two learnable bias are used in matrix c and matrix d
        # as described in https://arxiv.org/abs/1901.02860 Section 3.3
        self.pos_bias_u = nn.Parameter(torch.Tensor(self.h, self.d_k))
        self.pos_bias_v = nn.Parameter(torch.Tensor(self.h, self.d_k))
        torch.nn.init.xavier_uniform_(self.pos_bias_u)
        torch.nn.init.xavier_uniform_(self.pos_bias_v)

    def rel_shift(self, x):
        """
        Compute relative positional encoding.
        Args:
            x (torch.Tensor): Input tensor (batch, head, time1, 2*time1-1).
            time1 means the length of query vector.
        Returns:
            torch.Tensor: Output tensor.
        """
        zero_pad = torch.zeros((*x.size()[:3], 1), device=x.device, dtype=x.dtype)
        x_padded = torch.cat([zero_pad, x], dim=-1)

        x_padded = x_padded.view(*x.size()[:2], x.size(3) + 1, x.size(2))
        x = x_padded[:, :, 1:].view_as(x)[:, :, :, : x.size(-1) // 2 + 1]  # only keep the positions from 0 to time2

        if self.zero_triu:
            ones = torch.ones((x.size(2), x.size(3)), device=x.device)
            x = x * torch.tril(ones, x.size(3) - x.size(2))[None, None, :, :]

        return x

    def forward(self, query, key, value, pos_emb, mask):
        """
        Compute 'Scaled Dot Product Attention' with rel. positional encoding.
        Args:
            query (torch.Tensor): Query tensor (#batch, time1, size).
            key (torch.Tensor): Key tensor (#batch, time2, size).
            value (torch.Tensor): Value tensor (#batch, time2, size).
            pos_emb (torch.Tensor): Positional embedding tensor
                (#batch, 2*time1-1, size).
            mask (torch.Tensor): Mask tensor (#batch, 1, time2) or
                (#batch, time1, time2).
        Returns:
            torch.Tensor: Output tensor (#batch, time1, d_model).
        """
        q, k, v = self.forward_qkv(query, key, value)
        q = q.transpose(1, 2)  # (batch, time1, head, d_k)

        n_batch_pos = pos_emb.size(0)
        p = self.linear_pos(pos_emb).view(n_batch_pos, -1, self.h, self.d_k)
        p = p.transpose(1, 2)  # (batch, head, 2*time1-1, d_k)

        # (batch, head, time1, d_k)
        q_with_bias_u = (q + self.pos_bias_u).transpose(1, 2)
        # (batch, head, time1, d_k)
        q_with_bias_v = (q + self.pos_bias_v).transpose(1, 2)

        # compute attention score
        # first compute matrix a and matrix c
        # as described in https://arxiv.org/abs/1901.02860 Section 3.3
        # (batch, head, time1, time2)
        matrix_ac = torch.matmul(q_with_bias_u, k.transpose(-2, -1))

        # compute matrix b and matrix d
        # (batch, head, time1, 2*time1-1)
        matrix_bd = torch.matmul(q_with_bias_v, p.transpose(-2, -1))
        matrix_bd = self.rel_shift(matrix_bd)

        scores = (matrix_ac + matrix_bd) / math.sqrt(self.d_k)  # (batch, head, time1, time2)

        return self.forward_attention(v, scores, mask)


class GuidedAttentionLoss(torch.nn.Module):
    """
    Guided attention loss function module.

    This module calculates the guided attention loss described
    in `Efficiently Trainable Text-to-Speech System Based
    on Deep Convolutional Networks with Guided Attention`_,
    which forces the attention to be diagonal.

    .. _`Efficiently Trainable Text-to-Speech System
        Based on Deep Convolutional Networks with Guided Attention`:
        https://arxiv.org/abs/1710.08969
    """

    def __init__(self, sigma=0.4, alpha=1.0):
        """
        Initialize guided attention loss module.

        Args:
            sigma (float, optional): Standard deviation to control
                how close attention to a diagonal.
            alpha (float, optional): Scaling coefficient (lambda).
            reset_always (bool, optional): Whether to always reset masks.
        """
        super(GuidedAttentionLoss, self).__init__()
        self.sigma = sigma
        self.alpha = alpha
        self.guided_attn_masks = None
        self.masks = None

    def _reset_masks(self):
        self.guided_attn_masks = None
        self.masks = None

    def forward(self, att_ws, ilens, olens):
        """
        Calculate forward propagation.

        Args:
            att_ws (Tensor): Batch of attention weights (B, T_max_out, T_max_in).
            ilens (LongTensor): Batch of input lenghts (B,).
            olens (LongTensor): Batch of output lenghts (B,).

        Returns:
            Tensor: Guided attention loss value.
        """
        self._reset_masks()
        self.guided_attn_masks = self._make_guided_attention_masks(ilens, olens).to(att_ws.device)
        self.masks = self._make_masks(ilens, olens).to(att_ws.device)
        losses = self.guided_attn_masks * att_ws
        loss = torch.mean(losses.masked_select(self.masks))
        self._reset_masks()
        return self.alpha * loss

    def _make_guided_attention_masks(self, ilens, olens):
        n_batches = len(ilens)
        max_ilen = max(ilens)
        max_olen = max(olens)
        guided_attn_masks = torch.zeros((n_batches, max_olen, max_ilen), device=ilens.device)
        for idx, (ilen, olen) in enumerate(zip(ilens, olens)):
            guided_attn_masks[idx, :olen, :ilen] = self._make_guided_attention_mask(ilen, olen, self.sigma)
        return guided_attn_masks

    @staticmethod
    def _make_guided_attention_mask(ilen, olen, sigma):
        """
        Make guided attention mask.
        """
        grid_x, grid_y = torch.meshgrid(torch.arange(olen, device=olen.device).float(), torch.arange(ilen, device=ilen.device).float())
        return 1.0 - torch.exp(-((grid_y / ilen - grid_x / olen) ** 2) / (2 * (sigma ** 2)))

    @staticmethod
    def _make_masks(ilens, olens):
        """
        Make masks indicating non-padded part.

        Args:
            ilens (LongTensor or List): Batch of lengths (B,).
            olens (LongTensor or List): Batch of lengths (B,).

        Returns:
            Tensor: Mask tensor indicating non-padded part.
                    dtype=torch.uint8 in PyTorch 1.2-
                    dtype=torch.bool in PyTorch 1.2+ (including 1.2)
        """
        in_masks = make_non_pad_mask(ilens, device=ilens.device)  # (B, T_in)
        out_masks = make_non_pad_mask(olens, device=olens.device)  # (B, T_out)
        return out_masks.unsqueeze(-1) & in_masks.unsqueeze(-2)  # (B, T_out, T_in)


class GuidedMultiHeadAttentionLoss(GuidedAttentionLoss):
    """
    Guided attention loss function module for multi head attention.

    Args:
        sigma (float, optional): Standard deviation to control
        how close attention to a diagonal.
        alpha (float, optional): Scaling coefficient (lambda).
        reset_always (bool, optional): Whether to always reset masks.
    """

    def forward(self, att_ws, ilens, olens):
        """
        Calculate forward propagation.

        Args:
            att_ws (Tensor):
                Batch of multi head attention weights (B, H, T_max_out, T_max_in).
            ilens (LongTensor): Batch of input lenghts (B,).
            olens (LongTensor): Batch of output lenghts (B,).

        Returns:
            Tensor: Guided attention loss value.
        """
        if self.guided_attn_masks is None:
            self.guided_attn_masks = (self._make_guided_attention_masks(ilens, olens).to(att_ws.device).unsqueeze(1))
        if self.masks is None:
            self.masks = self._make_masks(ilens, olens).to(att_ws.device).unsqueeze(1)
        losses = self.guided_attn_masks * att_ws
        loss = torch.mean(losses.masked_select(self.masks))
        if self.reset_always:
            self._reset_masks()

        return self.alpha * loss