File size: 7,622 Bytes
4f175c5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import math
import torch
from .commons import convert_pad_shape


class MultiHeadAttention(torch.nn.Module):

    def __init__(
        self,
        channels: int,
        out_channels: int,
        n_heads: int,
        p_dropout: float = 0.0,
        window_size: int = None,
        heads_share: bool = True,
        block_length: int = None,
        proximal_bias: bool = False,
        proximal_init: bool = False,
    ):
        super().__init__()
        assert (
            channels % n_heads == 0
        ), "Channels must be divisible by the number of heads."

        self.channels = channels
        self.out_channels = out_channels
        self.n_heads = n_heads
        self.k_channels = channels // n_heads
        self.window_size = window_size
        self.block_length = block_length
        self.proximal_bias = proximal_bias

        self.conv_q = torch.nn.Conv1d(channels, channels, 1)
        self.conv_k = torch.nn.Conv1d(channels, channels, 1)
        self.conv_v = torch.nn.Conv1d(channels, channels, 1)
        self.conv_o = torch.nn.Conv1d(channels, out_channels, 1)

        self.drop = torch.nn.Dropout(p_dropout)

        if window_size:
            n_heads_rel = 1 if heads_share else n_heads
            rel_stddev = self.k_channels**-0.5
            self.emb_rel_k = torch.nn.Parameter(
                torch.randn(n_heads_rel, 2 * window_size + 1, self.k_channels)
                * rel_stddev
            )
            self.emb_rel_v = torch.nn.Parameter(
                torch.randn(n_heads_rel, 2 * window_size + 1, self.k_channels)
                * rel_stddev
            )

        torch.nn.init.xavier_uniform_(self.conv_q.weight)
        torch.nn.init.xavier_uniform_(self.conv_k.weight)
        torch.nn.init.xavier_uniform_(self.conv_v.weight)
        torch.nn.init.xavier_uniform_(self.conv_o.weight)

        if proximal_init:
            with torch.no_grad():
                self.conv_k.weight.copy_(self.conv_q.weight)
                self.conv_k.bias.copy_(self.conv_q.bias)

    def forward(self, x, c, attn_mask=None):
        q, k, v = self.conv_q(x), self.conv_k(c), self.conv_v(c)

        x, self.attn = self.attention(q, k, v, mask=attn_mask)

        return self.conv_o(x)

    def attention(self, query, key, value, mask=None):
        b, d, t_s, t_t = (*key.size(), query.size(2))
        query = query.view(b, self.n_heads, self.k_channels, t_t).transpose(2, 3)
        key = key.view(b, self.n_heads, self.k_channels, t_s).transpose(2, 3)
        value = value.view(b, self.n_heads, self.k_channels, t_s).transpose(2, 3)

        scores = torch.matmul(query / math.sqrt(self.k_channels), key.transpose(-2, -1))

        if self.window_size:
            assert t_s == t_t, "Relative attention only supports self-attention."
            scores += self._compute_relative_scores(query, t_s)

        if self.proximal_bias:
            assert t_s == t_t, "Proximal bias only supports self-attention."
            scores += self._attention_bias_proximal(t_s).to(scores.device, scores.dtype)

        if mask is not None:
            scores = scores.masked_fill(mask == 0, -1e4)
            if self.block_length:
                block_mask = (
                    torch.ones_like(scores)
                    .triu(-self.block_length)
                    .tril(self.block_length)
                )
                scores = scores.masked_fill(block_mask == 0, -1e4)

        p_attn = self.drop(torch.nn.functional.softmax(scores, dim=-1))

        output = torch.matmul(p_attn, value)

        if self.window_size:
            output += self._apply_relative_values(p_attn, t_s)

        return output.transpose(2, 3).contiguous().view(b, d, t_t), p_attn

    def _compute_relative_scores(self, query, length):
        rel_emb = self._get_relative_embeddings(self.emb_rel_k, length)
        rel_logits = self._matmul_with_relative_keys(
            query / math.sqrt(self.k_channels), rel_emb
        )
        return self._relative_position_to_absolute_position(rel_logits)

    def _apply_relative_values(self, p_attn, length):
        rel_weights = self._absolute_position_to_relative_position(p_attn)
        rel_emb = self._get_relative_embeddings(self.emb_rel_v, length)
        return self._matmul_with_relative_values(rel_weights, rel_emb)

    def _matmul_with_relative_values(self, x, y):
        return torch.matmul(x, y.unsqueeze(0))

    def _matmul_with_relative_keys(self, x, y):
        return torch.matmul(x, y.unsqueeze(0).transpose(-2, -1))

    def _get_relative_embeddings(self, embeddings, length):
        pad_length = max(length - (self.window_size + 1), 0)
        start = max((self.window_size + 1) - length, 0)
        end = start + 2 * length - 1

        if pad_length > 0:
            embeddings = torch.nn.functional.pad(
                embeddings,
                convert_pad_shape([[0, 0], [pad_length, pad_length], [0, 0]]),
            )
        return embeddings[:, start:end]

    def _relative_position_to_absolute_position(self, x):
        batch, heads, length, _ = x.size()
        x = torch.nn.functional.pad(
            x, convert_pad_shape([[0, 0], [0, 0], [0, 0], [0, 1]])
        )
        x_flat = x.view(batch, heads, length * 2 * length)
        x_flat = torch.nn.functional.pad(
            x_flat, convert_pad_shape([[0, 0], [0, 0], [0, length - 1]])
        )
        return x_flat.view(batch, heads, length + 1, 2 * length - 1)[
            :, :, :length, length - 1 :
        ]

    def _absolute_position_to_relative_position(self, x):
        batch, heads, length, _ = x.size()
        x = torch.nn.functional.pad(
            x, convert_pad_shape([[0, 0], [0, 0], [0, 0], [0, length - 1]])
        )
        x_flat = x.view(batch, heads, length**2 + length * (length - 1))
        x_flat = torch.nn.functional.pad(
            x_flat, convert_pad_shape([[0, 0], [0, 0], [length, 0]])
        )
        return x_flat.view(batch, heads, length, 2 * length)[:, :, :, 1:]

    def _attention_bias_proximal(self, length):
        r = torch.arange(length, dtype=torch.float32)
        diff = r.unsqueeze(0) - r.unsqueeze(1)
        return -torch.log1p(torch.abs(diff)).unsqueeze(0).unsqueeze(0)


class FFN(torch.nn.Module):

    def __init__(
        self,
        in_channels: int,
        out_channels: int,
        filter_channels: int,
        kernel_size: int,
        p_dropout: float = 0.0,
        activation: str = None,
        causal: bool = False,
    ):
        super().__init__()
        self.padding_fn = self._causal_padding if causal else self._same_padding

        self.conv_1 = torch.nn.Conv1d(in_channels, filter_channels, kernel_size)
        self.conv_2 = torch.nn.Conv1d(filter_channels, out_channels, kernel_size)
        self.drop = torch.nn.Dropout(p_dropout)

        self.activation = activation

    def forward(self, x, x_mask):
        x = self.conv_1(self.padding_fn(x * x_mask))
        x = self._apply_activation(x)
        x = self.drop(x)
        x = self.conv_2(self.padding_fn(x * x_mask))
        return x * x_mask

    def _apply_activation(self, x):
        if self.activation == "gelu":
            return x * torch.sigmoid(1.702 * x)
        return torch.relu(x)

    def _causal_padding(self, x):
        pad_l, pad_r = self.conv_1.kernel_size[0] - 1, 0
        return torch.nn.functional.pad(
            x, convert_pad_shape([[0, 0], [0, 0], [pad_l, pad_r]])
        )

    def _same_padding(self, x):
        pad = (self.conv_1.kernel_size[0] - 1) // 2
        return torch.nn.functional.pad(
            x, convert_pad_shape([[0, 0], [0, 0], [pad, pad]])
        )