-
Notifications
You must be signed in to change notification settings - Fork 0
/
model.py
234 lines (194 loc) · 8.89 KB
/
model.py
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
# 定义bert上的模型结构
class ConvolutionLayer(nn.Module):
def __init__(self, input_size, channels, dilation, dropout=0.1):
super(ConvolutionLayer, self).__init__()
self.base = nn.Sequential(
nn.Dropout2d(dropout),
nn.Conv2d(input_size, channels, kernel_size=1),
nn.GELU(),
)
self.convs = nn.ModuleList(
[nn.Conv2d(channels, channels, kernel_size=3, groups=channels, dilation=d, padding=d) for d in dilation])
def forward(self, x):
x = x.permute(0, 3, 1, 2).contiguous()
x = self.base(x)
outputs = []
for conv in self.convs:
x = conv(x)
x = F.gelu(x)
outputs.append(x)
outputs = torch.cat(outputs, dim=1)
outputs = outputs.permute(0, 2, 3, 1).contiguous()
return outputs
# Triformer模型结构
class Triformer(nn.Module):
def __init__(self, d_model, d_char, d_lex, d_syn, n_heads, n_layers, dropout=0.1):
super().__init__()
self.d_model = d_model
self.n_heads = n_heads
self.n_layers = n_layers
# 将输入嵌入映射到模型维度
self.char_embed_proj = nn.Linear(d_char, d_model)
self.lex_embed_proj = nn.Linear(d_lex, d_model)
self.syn_embed_proj = nn.Linear(d_syn, d_model)
# 定义相对位置编码
self.pos_encoder = PositionalEncoding(d_model, dropout)
# 定义编码器层
encoder_layer = nn.TransformerEncoderLayer(d_model, n_heads, dim_feedforward=2048, dropout=dropout)
self.transformer_encoder = nn.TransformerEncoder(encoder_layer, n_layers)
# 多头注意力层
self.self_attn = nn.MultiheadAttention(d_model, n_heads, dropout=dropout)
self.cross_attn_char_lex = nn.MultiheadAttention(d_model, n_heads, dropout=dropout)
self.cross_attn_char_syn = nn.MultiheadAttention(d_model, n_heads, dropout=dropout)
self.cross_attn_lex_syn = nn.MultiheadAttention(d_model, n_heads, dropout=dropout)
# 前馈网络
self.ffn = nn.Sequential(
nn.Linear(d_model, 2048),
nn.ReLU(),
nn.Dropout(dropout),
nn.Linear(2048, d_model),
nn.Dropout(dropout)
)
def forward(self, char_embeddings, lex_embeddings, syn_embeddings):
# 映射输入嵌入到模型维度
char_embeddings = self.char_embed_proj(char_embeddings)
lex_embeddings = self.lex_embed_proj(lex_embeddings)
syn_embeddings = self.syn_embed_proj(syn_embeddings)
# 添加位置编码
char_embeddings = self.pos_encoder(char_embeddings)
lex_embeddings = self.pos_encoder(lex_embeddings)
syn_embeddings = self.pos_encoder(syn_embeddings)
# 自注意力
char_output = self.transformer_encoder(char_embeddings)
lex_output = self.transformer_encoder(lex_embeddings)
syn_output = self.transformer_encoder(syn_embeddings)
# 交叉注意力
char_lex_output, _ = self.cross_attn_char_lex(char_output, lex_output, lex_output)
char_syn_output, _ = self.cross_attn_char_syn(char_output, syn_output, syn_output)
lex_syn_output, _ = self.cross_attn_lex_syn(lex_output, syn_output, syn_output)
# 融合嵌入
fused_output = char_lex_output + char_syn_output + lex_syn_output
fused_output = self.ffn(fused_output)
return fused_output
# 位置编码
class PositionalEncoding(nn.Module):
def __init__(self, d_model, dropout=0.1, max_len=5000):
super().__init__()
self.dropout = nn.Dropout(p=dropout)
pe = torch.zeros(max_len, d_model)
position = torch.arange(0, max_len, dtype=torch.float).unsqueeze(1)
div_term = torch.exp(torch.arange(0, d_model, 2).float() * (-math.log(10000.0) / d_model))
pe[:, 0::2] = torch.sin(position * div_term)
pe[:, 1::2] = torch.cos(position * div_term)
pe = pe.unsqueeze(0)
self.register_buffer('pe', pe)
def forward(self, x):
x = x + self.pe[:, :x.size(1), :]
return self.dropout(x)
# 仿射变化
class Biaffine(nn.Module):
def __init__(self, n_in, n_out=1, bias_x=True, bias_y=True):
super(Biaffine, self).__init__()
self.n_in = n_in
self.n_out = n_out
self.bias_x = bias_x
self.bias_y = bias_y
weight = torch.zeros((n_out, n_in + int(bias_x), n_in + int(bias_y)))
nn.init.xavier_normal_(weight)
self.weight = nn.Parameter(weight, requires_grad=True)
def extra_repr(self):
s = f"n_in={self.n_in}, n_out={self.n_out}"
if self.bias_x:
s += f", bias_x={self.bias_x}"
if self.bias_y:
s += f", bias_y={self.bias_y}"
return s
def forward(self, x, y):
if self.bias_x:
x = torch.cat((x, torch.ones_like(x[..., :1])), -1)
if self.bias_y:
y = torch.cat((y, torch.ones_like(y[..., :1])), -1)
# [batch_size, n_out, seq_len, seq_len]
s = torch.einsum('bxi,oij,byj->boxy', x, self.weight, y)
# remove dim 1 if n_out == 1
s = s.permute(0, 2, 3, 1)
return s
# MLP
class MLP(nn.Module):
def __init__(self, n_in, n_out, dropout=0):
super().__init__()
self.linear = nn.Linear(n_in, n_out)
self.activation = nn.GELU()
self.dropout = nn.Dropout(dropout)
def forward(self, x):
x = self.dropout(x)
x = self.linear(x)
x = self.activation(x)
return x
# 注意力融合模块
class AttentionFusion(nn.Module):
def __init__(self, input_dim, window_size, beta=0.5):
super(AttentionFusion, self).__init__()
self.input_dim = input_dim
self.window_size = window_size
self.beta = beta
self.local_attention = LocalAttention(input_dim, window_size)
self.global_attention = GlobalAttention(input_dim)
def forward(self, inputs):
batch_size, seq_len, _ = inputs.size()
local_attn = self.local_attention(inputs)
global_attn = self.global_attention(inputs)
fused_attn = self.beta * local_attn + (1 - self.beta) * global_attn
outputs = torch.bmm(fused_attn.view(batch_size, 1, seq_len * seq_len), inputs.view(batch_size, seq_len * seq_len, self.input_dim))
return outputs
class LocalAttention(nn.Module):
def __init__(self, input_dim, window_size):
super(LocalAttention, self).__init__()
self.input_dim = input_dim
self.window_size = window_size
self.query = nn.Linear(input_dim, input_dim)
self.key = nn.Linear(input_dim, input_dim)
def forward(self, inputs):
batch_size, seq_len, _ = inputs.size()
queries = self.query(inputs).view(batch_size, seq_len, self.input_dim)
keys = self.key(inputs).view(batch_size, seq_len, self.input_dim)
attention_scores = torch.bmm(queries, keys.transpose(1, 2))
padded_mask = torch.ones(batch_size, seq_len, seq_len)
padding_window = self.window_size // 2
padded_mask[:, :, :padding_window] = 0
padded_mask[:, :, -padding_window:] = 0
for i in range(padding_window, seq_len - padding_window):
padded_mask[:, i, i - padding_window:i + padding_window + 1] = 1
attention_scores = attention_scores.masked_fill(padded_mask == 0, -1e9)
attention_weights = F.softmax(attention_scores, dim=-1)
return attention_weights
class GlobalAttention(nn.Module):
def __init__(self, input_dim):
super(GlobalAttention, self).__init__()
self.input_dim = input_dim
self.query = nn.Linear(input_dim, input_dim)
self.key = nn.Linear(input_dim, input_dim)
def forward(self, inputs):
batch_size, seq_len, _ = inputs.size()
queries = self.query(inputs).view(batch_size, seq_len, self.input_dim)
keys = self.key(inputs).view(batch_size, seq_len, self.input_dim)
attention_scores = torch.bmm(queries, keys.transpose(1, 2))
attention_weights = F.softmax(attention_scores, dim=-1)
return attention_weights
# 协同预测器
class CoPredictor(nn.Module):
def __init__(self, cls_num, hid_size, biaffine_size, channels, ffnn_hid_size, dropout=0):
super().__init__()
self.mlp1 = MLP(n_in=hid_size, n_out=biaffine_size, dropout=dropout)
self.mlp2 = MLP(n_in=hid_size, n_out=biaffine_size, dropout=dropout)
self.biaffine = Biaffine(n_in=biaffine_size, n_out=cls_num, bias_x=True, bias_y=True)
self.mlp_rel = MLP(channels, ffnn_hid_size, dropout=dropout)
self.linear = nn.Linear(ffnn_hid_size, cls_num)
self.dropout = nn.Dropout(dropout)
def forward(self, x, y, z):
h = self.dropout(self.mlp1(x))
t = self.dropout(self.mlp2(y))
o1 = self.biaffine(h, t)
z = self.dropout(self.mlp_rel(z))
o2 = self.linear(z)
return o1 + o2