File size: 12,332 Bytes
d758c99 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 |
import copy
import math
import torch
import torch.nn as nn
import torch.nn.functional as F
import entmax
# Adapted from
# https://github.com/tensorflow/tensor2tensor/blob/0b156ac533ab53f65f44966381f6e147c7371eee/tensor2tensor/layers/common_attention.py
def relative_attention_logits(query, key, relation):
# We can't reuse the same logic as tensor2tensor because we don't share relation vectors across the batch.
# In this version, relation vectors are shared across heads.
# query: [batch, heads, num queries, depth].
# key: [batch, heads, num kvs, depth].
# relation: [batch, num queries, num kvs, depth].
# qk_matmul is [batch, heads, num queries, num kvs]
qk_matmul = torch.matmul(query, key.transpose(-2, -1))
# q_t is [batch, num queries, heads, depth]
q_t = query.permute(0, 2, 1, 3)
# r_t is [batch, num queries, depth, num kvs]
r_t = relation.transpose(-2, -1)
# [batch, num queries, heads, depth]
# * [batch, num queries, depth, num kvs]
# = [batch, num queries, heads, num kvs]
# For each batch and query, we have a query vector per head.
# We take its dot product with the relation vector for each kv.
q_tr_t_matmul = torch.matmul(q_t, r_t)
# qtr_t_matmul_t is [batch, heads, num queries, num kvs]
q_tr_tmatmul_t = q_tr_t_matmul.permute(0, 2, 1, 3)
# [batch, heads, num queries, num kvs]
return (qk_matmul + q_tr_tmatmul_t) / math.sqrt(query.shape[-1])
# Sharing relation vectors across batch and heads:
# query: [batch, heads, num queries, depth].
# key: [batch, heads, num kvs, depth].
# relation: [num queries, num kvs, depth].
#
# Then take
# key reshaped
# [num queries, batch * heads, depth]
# relation.transpose(-2, -1)
# [num queries, depth, num kvs]
# and multiply them together.
#
# Without sharing relation vectors across heads:
# query: [batch, heads, num queries, depth].
# key: [batch, heads, num kvs, depth].
# relation: [batch, heads, num queries, num kvs, depth].
#
# Then take
# key.unsqueeze(3)
# [batch, heads, num queries, 1, depth]
# relation.transpose(-2, -1)
# [batch, heads, num queries, depth, num kvs]
# and multiply them together:
# [batch, heads, num queries, 1, depth]
# * [batch, heads, num queries, depth, num kvs]
# = [batch, heads, num queries, 1, num kvs]
# and squeeze
# [batch, heads, num queries, num kvs]
def relative_attention_values(weight, value, relation):
# In this version, relation vectors are shared across heads.
# weight: [batch, heads, num queries, num kvs].
# value: [batch, heads, num kvs, depth].
# relation: [batch, num queries, num kvs, depth].
# wv_matmul is [batch, heads, num queries, depth]
wv_matmul = torch.matmul(weight, value)
# w_t is [batch, num queries, heads, num kvs]
w_t = weight.permute(0, 2, 1, 3)
# [batch, num queries, heads, num kvs]
# * [batch, num queries, num kvs, depth]
# = [batch, num queries, heads, depth]
w_tr_matmul = torch.matmul(w_t, relation)
# w_tr_matmul_t is [batch, heads, num queries, depth]
w_tr_matmul_t = w_tr_matmul.permute(0, 2, 1, 3)
return wv_matmul + w_tr_matmul_t
# Adapted from The Annotated Transformer
def clones(module_fn, N):
return nn.ModuleList([module_fn() for _ in range(N)])
def attention(query, key, value, mask=None, dropout=None):
"Compute 'Scaled Dot Product Attention'"
d_k = query.size(-1)
scores = torch.matmul(query, key.transpose(-2, -1)) \
/ math.sqrt(d_k)
if mask is not None:
scores = scores.masked_fill(mask == 0, -1e9)
p_attn = F.softmax(scores, dim = -1)
if dropout is not None:
p_attn = dropout(p_attn)
# return torch.matmul(p_attn, value), scores.squeeze(1).squeeze(1)
return torch.matmul(p_attn, value), p_attn
def sparse_attention(query, key, value, alpha, mask=None, dropout=None):
"Compute 'Scaled Dot Product Attention'"
d_k = query.size(-1)
scores = torch.matmul(query, key.transpose(-2, -1)) \
/ math.sqrt(d_k)
if mask is not None:
scores = scores.masked_fill(mask == 0, -1e9)
if alpha == 2:
p_attn = entmax.sparsemax(scores, -1)
elif alpha == 1.5:
p_attn = entmax.entmax15(scores, -1)
else:
raise NotImplementedError
if dropout is not None:
p_attn = dropout(p_attn)
# return torch.matmul(p_attn, value), scores.squeeze(1).squeeze(1)
return torch.matmul(p_attn, value), p_attn
# Adapted from The Annotated Transformers
class MultiHeadedAttention(nn.Module):
def __init__(self, h, d_model, dropout=0.1):
"Take in model size and number of heads."
super(MultiHeadedAttention, self).__init__()
assert d_model % h == 0
# We assume d_v always equals d_k
self.d_k = d_model // h
self.h = h
self.linears = clones(lambda: nn.Linear(d_model, d_model), 4)
self.attn = None
self.dropout = nn.Dropout(p=dropout)
def forward(self, query, key, value, mask=None):
"Implements Figure 2"
if mask is not None:
# Same mask applied to all h heads.
mask = mask.unsqueeze(1)
nbatches = query.size(0)
# 1) Do all the linear projections in batch from d_model => h x d_k
query, key, value = \
[l(x).view(nbatches, -1, self.h, self.d_k).transpose(1, 2)
for l, x in zip(self.linears, (query, key, value))]
# 2) Apply attention on all the projected vectors in batch.
x, self.attn = attention(query, key, value, mask=mask,
dropout=self.dropout)
# 3) "Concat" using a view and apply a final linear.
x = x.transpose(1, 2).contiguous() \
.view(nbatches, -1, self.h * self.d_k)
if query.dim() == 3:
x = x.squeeze(1)
return self.linears[-1](x)
# Adapted from The Annotated Transformer
def attention_with_relations(query, key, value, relation_k, relation_v, mask=None, dropout=None):
"Compute 'Scaled Dot Product Attention'"
d_k = query.size(-1)
scores = relative_attention_logits(query, key, relation_k)
if mask is not None:
scores = scores.masked_fill(mask == 0, -1e9)
p_attn_orig = F.softmax(scores, dim = -1)
if dropout is not None:
p_attn = dropout(p_attn_orig)
return relative_attention_values(p_attn, value, relation_v), p_attn_orig
class PointerWithRelations(nn.Module):
def __init__(self, hidden_size, num_relation_kinds, dropout=0.2):
super(PointerWithRelations, self).__init__()
self.hidden_size = hidden_size
self.linears = clones(lambda: nn.Linear(hidden_size, hidden_size), 3)
self.attn = None
self.dropout = nn.Dropout(p=dropout)
self.relation_k_emb = nn.Embedding(num_relation_kinds, self.hidden_size)
self.relation_v_emb = nn.Embedding(num_relation_kinds, self.hidden_size)
def forward(self, query, key, value, relation, mask=None):
relation_k = self.relation_k_emb(relation)
relation_v = self.relation_v_emb(relation)
if mask is not None:
mask = mask.unsqueeze(0)
nbatches = query.size(0)
query, key, value = \
[l(x).view(nbatches, -1, 1, self.hidden_size).transpose(1, 2)
for l, x in zip(self.linears, (query, key, value))]
_, self.attn = attention_with_relations(
query,
key,
value,
relation_k,
relation_v,
mask=mask,
dropout=self.dropout)
return self.attn[0,0]
# Adapted from The Annotated Transformer
class MultiHeadedAttentionWithRelations(nn.Module):
def __init__(self, h, d_model, dropout=0.1):
"Take in model size and number of heads."
super(MultiHeadedAttentionWithRelations, self).__init__()
assert d_model % h == 0
# We assume d_v always equals d_k
self.d_k = d_model // h
self.h = h
self.linears = clones(lambda: nn.Linear(d_model, d_model), 4)
self.attn = None
self.dropout = nn.Dropout(p=dropout)
def forward(self, query, key, value, relation_k, relation_v, mask=None):
# query shape: [batch, num queries, d_model]
# key shape: [batch, num kv, d_model]
# value shape: [batch, num kv, d_model]
# relations_k shape: [batch, num queries, num kv, (d_model // h)]
# relations_v shape: [batch, num queries, num kv, (d_model // h)]
# mask shape: [batch, num queries, num kv]
if mask is not None:
# Same mask applied to all h heads.
# mask shape: [batch, 1, num queries, num kv]
mask = mask.unsqueeze(1)
nbatches = query.size(0)
# 1) Do all the linear projections in batch from d_model => h x d_k
query, key, value = \
[l(x).view(nbatches, -1, self.h, self.d_k).transpose(1, 2)
for l, x in zip(self.linears, (query, key, value))]
# 2) Apply attention on all the projected vectors in batch.
# x shape: [batch, heads, num queries, depth]
x, self.attn = attention_with_relations(
query,
key,
value,
relation_k,
relation_v,
mask=mask,
dropout=self.dropout)
# 3) "Concat" using a view and apply a final linear.
x = x.transpose(1, 2).contiguous() \
.view(nbatches, -1, self.h * self.d_k)
return self.linears[-1](x)
# Adapted from The Annotated Transformer
class Encoder(nn.Module):
"Core encoder is a stack of N layers"
def __init__(self, layer, layer_size, N, tie_layers=False):
super(Encoder, self).__init__()
if tie_layers:
self.layer = layer()
self.layers = [self.layer for _ in range(N)]
else:
self.layers = clones(layer, N)
self.norm = nn.LayerNorm(layer_size)
# TODO initialize using xavier
def forward(self, x, relation, mask):
"Pass the input (and mask) through each layer in turn."
for layer in self.layers:
x = layer(x, relation, mask)
return self.norm(x)
# Adapted from The Annotated Transformer
class SublayerConnection(nn.Module):
"""
A residual connection followed by a layer norm.
Note for code simplicity the norm is first as opposed to last.
"""
def __init__(self, size, dropout):
super(SublayerConnection, self).__init__()
self.norm = nn.LayerNorm(size)
self.dropout = nn.Dropout(dropout)
def forward(self, x, sublayer):
"Apply residual connection to any sublayer with the same size."
return x + self.dropout(sublayer(self.norm(x)))
# Adapted from The Annotated Transformer
class EncoderLayer(nn.Module):
"Encoder is made up of self-attn and feed forward (defined below)"
def __init__(self, size, self_attn, feed_forward, num_relation_kinds, dropout):
super(EncoderLayer, self).__init__()
self.self_attn = self_attn
self.feed_forward = feed_forward
self.sublayer = clones(lambda: SublayerConnection(size, dropout), 2)
self.size = size
self.relation_k_emb = nn.Embedding(num_relation_kinds, self.self_attn.d_k)
self.relation_v_emb = nn.Embedding(num_relation_kinds, self.self_attn.d_k)
def forward(self, x, relation, mask):
"Follow Figure 1 (left) for connections."
relation_k = self.relation_k_emb(relation)
relation_v = self.relation_v_emb(relation)
x = self.sublayer[0](x, lambda x: self.self_attn(x, x, x, relation_k, relation_v, mask))
return self.sublayer[1](x, self.feed_forward)
# Adapted from The Annotated Transformer
class PositionwiseFeedForward(nn.Module):
"Implements FFN equation."
def __init__(self, d_model, d_ff, dropout=0.1):
super(PositionwiseFeedForward, self).__init__()
self.w_1 = nn.Linear(d_model, d_ff)
self.w_2 = nn.Linear(d_ff, d_model)
self.dropout = nn.Dropout(dropout)
def forward(self, x):
return self.w_2(self.dropout(F.relu(self.w_1(x))))
|