File size: 32,083 Bytes
c882c3e | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 | from transformers import PreTrainedModel
from configuration_LightGTS import LightGTSConfig
from ts_generation_mixin import TSGenerationMixin
import torch
from torch import nn
from torch import Tensor
from typing import Callable, Optional
import math
import torch.nn.functional as F
import numpy as np
class LightGTSPreTrainedModel(PreTrainedModel):
config_class = LightGTSConfig
base_model_prefix = "model"
supports_gradient_checkpointing = True
_no_split_modules = ["TSTEncoderLayer"]
_skip_keys_device_placement = "past_key_values"
_supports_flash_attn_2 = True
_supports_sdpa = False
_supports_cache_class = True
def _init_weights(self, module):
std = self.config.initializer_range
if isinstance(module, torch.nn.Linear):
module.weight.data.normal_(mean=0.0, std=std)
if module.bias is not None:
module.bias.data.zero_()
elif isinstance(module, torch.nn.Embedding):
module.weight.data.normal_(mean=0.0, std=std)
if module.padding_idx is not None:
module.weight.data[module.padding_idx].zero_()
class LightGTSForPrediction(LightGTSPreTrainedModel, TSGenerationMixin):
def __init__(self, config: LightGTSConfig):
super().__init__(config)
self.config = config
self.model = LightGTS(c_in=config.c_in,
target_dim=config.target_dim,
patch_len=config.patch_len,
stride=config.stride,
num_patch=config.num_patch,
e_layers=config.e_layers,
d_layers=config.d_layers,
n_heads=config.n_heads,
d_model=config.d_model,
shared_embedding=True,
d_ff=config.d_ff,
dropout=config.dropout,
attn_dropout=config.attn_dropout,
head_dropout=config.head_dropout,
act='relu',
head_type=config.head_type,
res_attention=False,
learn_pe=False
)
def forward(self, input, labels=None, patch_len=None, stride=None, target_dim=None):
self.config.patch_len = patch_len
self.config.stride = stride
self.config.target_dim = target_dim
#切patch
batch_size,seq_len,n_vars = input.shape
num_patch = (max(seq_len, self.config.patch_len)-self.config.patch_len) // self.config.stride + 1
self.config.num_patch = num_patch
outputs = input.view(batch_size, num_patch, self.config.patch_len, n_vars)
outputs = outputs.transpose(2, 3)
outputs = self.model(outputs, target_dim=self.config.target_dim, patch_len=self.config.patch_len, stride=self.config.stride)
loss = None
if labels is not None:
if outputs.shape != labels.shape:
outputs = outputs.view(labels.shape)
loss = self.loss_fn(outputs, labels)
return {"prediction": outputs, "loss": loss}
class LightGTSForFinetune(LightGTSPreTrainedModel, TSGenerationMixin):
def __init__(self, config: LightGTSConfig):
super().__init__(config)
self.config = config
self.model = LightGTS(c_in=config.c_in,
target_dim=config.target_dim,
patch_len=config.patch_len,
stride=config.stride,
num_patch=config.num_patch,
e_layers=config.e_layers,
d_layers=config.d_layers,
n_heads=config.n_heads,
d_model=config.d_model,
shared_embedding=True,
d_ff=config.d_ff,
dropout=config.dropout,
attn_dropout=config.attn_dropout,
head_dropout=config.head_dropout,
act='relu',
head_type=config.head_type,
res_attention=False,
learn_pe=False
)
def forward(self, input, labels=None, patch_len=None, stride=None, target_dim=None):
if patch_len is not None:
self.config.patch_len = patch_len
if stride is not None:
self.config.stride = stride
if target_dim is not None:
self.config.target_dim = target_dim
#切patch
batch_size,seq_len,n_vars = input.shape
num_patch = (max(seq_len, self.config.patch_len)-self.config.patch_len) // self.config.stride + 1
self.config.num_patch = num_patch
outputs = input.view(batch_size, num_patch, self.config.patch_len, n_vars)
outputs = outputs.transpose(2, 3)
outputs = self.model(outputs, target_dim=self.config.target_dim, patch_len=self.config.patch_len, stride=self.config.stride)
loss = None
if labels is not None:
if outputs.shape != labels.shape:
outputs = outputs.view(labels.shape)
loss = self.loss_fn(outputs, labels)
return {"prediction": outputs, "loss": loss}
class LightGTS(nn.Module):
"""
Output dimension:
[bs x target_dim x nvars] for prediction
[bs x target_dim] for regression
[bs x target_dim] for classification
[bs x num_patch x n_vars x patch_len] for pretrain
"""
def __init__(self, c_in:int, target_dim:int, patch_len:int, stride:int, num_patch:int, mask_mode:str = 'patch',mask_nums:int = 3,
e_layers:int=3, d_layers:int=3, d_model=128, n_heads=16, shared_embedding=True, d_ff:int=256,
norm:str='BatchNorm', attn_dropout:float=0.4, dropout:float=0., act:str="gelu",
res_attention:bool=True, pre_norm:bool=False, store_attn:bool=False,
pe:str='sincos', learn_pe:bool=False, head_dropout = 0,
head_type = "prediction", individual = False,
y_range:Optional[tuple]=None, verbose:bool=False, **kwargs):
super().__init__()
assert head_type in ['pretrain', 'prediction', 'regression', 'classification'], 'head type should be either pretrain, prediction, or regression'
# Basic
self.num_patch = num_patch
self.target_dim=target_dim
self.out_patch_num = math.ceil(target_dim / patch_len)
self.target_patch_len = 48
# Embedding
self.embedding = nn.Linear(self.target_patch_len, d_model)
# self.decoder_embedding = nn.Parameter(torch.randn(1, 1,1, d_model),requires_grad=True)
self.cls_embedding = nn.Parameter(torch.randn(1, 1, 1, d_model),requires_grad=True)
# self.sep_embedding = nn.Parameter(torch.randn(1, 1, 1, d_model),requires_grad=True)
# Position Embedding
# self.pos = positional_encoding(pe, learn_pe, 1 + num_patch + self.out_patch_num, d_model)
# self.drop_out = nn.Dropout(dropout)
# Encoder
self.encoder = TSTEncoder(d_model, n_heads, d_ff=d_ff, norm=norm, attn_dropout=attn_dropout, dropout=dropout,
pre_norm=pre_norm, activation=act, res_attention=res_attention, n_layers=e_layers,
store_attn=store_attn)
# Decoder
self.decoder = Decoder(d_layers, patch_len=patch_len, d_model=d_model, n_heads=n_heads, d_ff=d_ff,attn_dropout= attn_dropout, dropout=dropout)
# Head
self.n_vars = c_in
self.head_type = head_type
self.mask_mode = mask_mode
self.mask_nums = mask_nums
self.d_model = d_model
self.patch_len = patch_len
if head_type == "pretrain":
self.head = PretrainHead(d_model, patch_len, head_dropout) # custom head passed as a partial func with all its kwargs
elif head_type == "prediction":
self.head = decoder_PredictHead(d_model, self.patch_len, self.target_patch_len, head_dropout)
def get_dynamic_weights(self, n_preds, decay_rate=0.5):
"""
Generate dynamic weights for the replicated tokens using an exponential decay scheme.
Args:
- n_preds (int): Number of predictions to generate weights for.
- decay_rate (float): The base of the exponential decay. Lower values decay faster (default: 0.9).
Returns:
- torch.Tensor: A tensor of weights with exponential decay.
"""
# Exponential decay weights
weights = decay_rate ** torch.arange(n_preds)
return weights
def decoder_predict(self, bs, n_vars, dec_cross):
"""
dec_cross: tensor [bs x n_vars x num_patch x d_model]
"""
# dec_in = self.decoder_embedding.expand(bs, self.n_vars, self.out_patch_num, -1)
# dec_in = self.embedding(self.decoder_len).expand(bs, -1, -1, -1)
# dec_in = self.decoder_embedding.expand(bs, n_vars, self.out_patch_num, -1)
# dec_in = dec_cross.mean(2).unsqueeze(2).expand(-1,-1,self.out_patch_num,-1)
dec_in = dec_cross[:,:,-1,:].unsqueeze(2).expand(-1,-1,self.out_patch_num,-1)
weights = self.get_dynamic_weights(self.out_patch_num).to(dec_in.device)
dec_in = dec_in * weights.unsqueeze(0).unsqueeze(0).unsqueeze(-1)
# dec_in = torch.cat((dec_in, self.sep_tokens), dim=2)
# dec_in = dec_cross[:,:,-self.out_patch_num:,:]
# dec_in = torch.ones([bs, n_vars, self.out_patch_num, self.d_model]).to(dec_cross.device)
# dec_in = dec_in + self.pos[-self.out_patch_num:,:]
decoder_output = self.decoder(dec_in, dec_cross)
decoder_output = decoder_output.transpose(2,3)
return decoder_output
def forward(self, z, target_dim=None, patch_len=None, stride=None):
"""
z: tensor [bs x num_patch x n_vars x patch_len]
"""
if target_dim is not None:
self.target_dim = target_dim
if patch_len is not None:
self.patch_len = patch_len
if stride is not None:
self.stride = stride
self.out_patch_num = math.ceil(self.target_dim / self.patch_len)
bs, num_patch, n_vars, patch_len = z.shape
# tokenizer
cls_tokens = self.cls_embedding.expand(bs, n_vars, -1, -1)
embedding = nn.Linear(patch_len, self.d_model, bias=False)
embedding.weight.data = resample_patchemb(old=self.embedding.weight.data, new_patch_len=self.patch_len)
z = embedding(z).permute(0,2,1,3) # [bs x n_vars x num_patch x d_model]
z = torch.cat((cls_tokens, z), dim=2) # [bs x n_vars x (1 + num_patch) x d_model]
# z = self.drop_out(z + self.pos[:1 + self.num_patch, :])
# encoder
z = torch.reshape(z, (-1, 1 + num_patch, self.d_model)) # [bs*n_vars x num_patch x d_model]
z = self.encoder(z)
z = torch.reshape(z, (-1, n_vars, 1 + num_patch, self.d_model)) # [bs, n_vars x num_patch x d_model]
# decoder
z = self.decoder_predict(bs, n_vars, z[:,:,:,:])
# predict
z = self.head(z[:,:,:,:], self.patch_len)
z = z[:,:self.target_dim, :]
# z: [bs x target_dim x nvars] for prediction
# [bs x target_dim] for regression
# [bs x target_dim] for classification
# [bs x num_patch x n_vars x patch_len] for pretrain
return z
class TSTEncoder(nn.Module):
def __init__(self, d_model, n_heads, d_ff=None,
norm='BatchNorm', attn_dropout=0., dropout=0., activation='gelu',
res_attention=False, n_layers=1, pre_norm=False, store_attn=False):
super().__init__()
self.layers = nn.ModuleList([TSTEncoderLayer(d_model, n_heads=n_heads, d_ff=d_ff, norm=norm,
attn_dropout=attn_dropout, dropout=dropout,
activation=activation, res_attention=res_attention,
pre_norm=pre_norm, store_attn=store_attn) for i in range(n_layers)])
self.res_attention = res_attention
def forward(self, src:Tensor):
"""
src: tensor [bs x q_len x d_model]
"""
output = src
scores = None
if self.res_attention:
for mod in self.layers: output, scores = mod(output, prev=scores)
return output
else:
for mod in self.layers: output = mod(output)
return output
class TSTEncoderLayer(nn.Module):
def __init__(self, d_model, n_heads, d_ff=256, store_attn=False,
norm='LayerNorm', attn_dropout=0, dropout=0., bias=True,
activation="gelu", res_attention=False, pre_norm=False):
super().__init__()
assert not d_model%n_heads, f"d_model ({d_model}) must be divisible by n_heads ({n_heads})"
d_k = d_model // n_heads
d_v = d_model // n_heads
# Multi-Head attention
self.res_attention = res_attention
self.self_attn = MultiheadAttention(d_model, n_heads, d_k, d_v, attn_dropout=attn_dropout, proj_dropout=dropout, res_attention=res_attention)
# Add & Norm
self.dropout_attn = nn.Dropout(dropout)
if "batch" in norm.lower():
self.norm_attn = nn.Sequential(Transpose(1,2), nn.BatchNorm1d(d_model), Transpose(1,2))
else:
self.norm_attn = nn.LayerNorm(d_model)
# Position-wise Feed-Forward
self.ff = nn.Sequential(nn.Linear(d_model, d_ff, bias=bias),
get_activation_fn(activation),
nn.Dropout(dropout),
nn.Linear(d_ff, d_model, bias=bias))
# Add & Norm
self.dropout_ffn = nn.Dropout(dropout)
if "batch" in norm.lower():
self.norm_ffn = nn.Sequential(Transpose(1,2), nn.BatchNorm1d(d_model), Transpose(1,2))
else:
self.norm_ffn = nn.LayerNorm(d_model)
self.pre_norm = pre_norm
self.store_attn = store_attn
# # se block
# self.SE = SE_Block(inchannel=7)
def forward(self, src:Tensor, prev:Optional[Tensor]=None):
"""
src: tensor [bs x q_len x d_model]
"""
# Multi-Head attention sublayer
if self.pre_norm:
src = self.norm_attn(src)
## Multi-Head attention
if self.res_attention:
src2, attn, scores = self.self_attn(src, src, src, prev)
else:
# attention_mask = causal_attention_mask(src.shape[1]).to(src.device)
# src2, attn = self.self_attn(src, src, src, attn_mask=attention_mask)
src2, attn = self.self_attn(src, src, src)
if self.store_attn:
self.attn = attn
# total, num_patch, d_model = src2.size()
# bs = int(total/7)
# src2 = self.SE(src2.reshape(bs, 7, num_patch, -1)).reshape(total, num_patch, -1)
## Add & Norm
src = src + self.dropout_attn(src2) # Add: residual connection with residual dropout
if not self.pre_norm:
src = self.norm_attn(src)
# Feed-forward sublayer
if self.pre_norm:
src = self.norm_ffn(src)
## Position-wise Feed-Forward
src2 = self.ff(src)
## Add & Norm
src = src + self.dropout_ffn(src2) # Add: residual connection with residual dropout
if not self.pre_norm:
src = self.norm_ffn(src)
if self.res_attention:
return src, scores
else:
return src
class Decoder(nn.Module):
def __init__(self, d_layers, patch_len, d_model, n_heads, d_ff=None, attn_dropout=0.2, dropout=0.1):
super(Decoder, self).__init__()
self.decoder_layers = nn.ModuleList()
for i in range(d_layers):
self.decoder_layers.append(DecoderLayer(patch_len, d_model, n_heads, d_ff, attn_dropout, dropout))
def forward(self, x, cross):
output = x
for layer in self.decoder_layers:
output = layer(output, cross)
return output
class DecoderLayer(nn.Module):
def __init__(self, patch_len, d_model, n_heads, d_ff=None, attn_dropout = 0.2, dropout=0.5, norm="BatchNorm"):
super(DecoderLayer, self).__init__()
self.self_attention = MultiheadAttention(d_model, n_heads, res_attention=False, attn_dropout=attn_dropout)
self.cross_attention = MultiheadAttention(d_model, n_heads, attn_dropout=attn_dropout, rope_type=True)
# self.pos_embed = nn.Conv1d(d_model, d_model, kernel_size=3, padding=1, groups=d_model)
if 'batch' in norm.lower():
self.norm1 = nn.Sequential(Transpose(1,2), nn.BatchNorm1d(d_model), Transpose(1,2))
self.norm2 = nn.Sequential(Transpose(1,2), nn.BatchNorm1d(d_model), Transpose(1,2))
self.norm3 = nn.Sequential(Transpose(1,2), nn.BatchNorm1d(d_model), Transpose(1,2))
else:
self.norm1 = nn.LayerNorm(d_model)
self.norm2 = nn.LayerNorm(d_model)
self.norm3 = nn.LayerNorm(d_model)
self.dropout = nn.Dropout(dropout)
self.MLP1 = CMlp(in_features = d_model, hidden_features = d_ff, out_features = d_model, drop=dropout)
def forward(self, x, cross):
batch, n_vars, num_patch, d_model = x.shape
x = x.reshape(batch*n_vars, num_patch, d_model)
# x = x.permute(0,2,1)
# x = x + self.pos_embed(x)
# x = x.permute(0,2,1)
cross = cross.reshape(batch*n_vars, -1, d_model)
attention_mask = causal_attention_mask(num_patch).to(x.device)
x_attn , _= self.self_attention(x, attn_mask=attention_mask)
x_attn = self.norm1(x_attn) + x
x_cross , _ = self.cross_attention(x_attn, cross, cross)
x_cross = self.dropout(self.norm2(x_cross)) + x_attn
x_ff = self.MLP1(x_cross)
x_ff = self.norm3(x_ff) + x_cross
x_ff = x_ff.reshape(batch, n_vars, num_patch, d_model)
return x_ff
def causal_attention_mask(seq_length):
"""
创建一个因果注意力掩码。掩码中的每个位置 (i, j)
表示在计算第i个位置的attention时, 第j个位置是否可以被看见。
如果j <= i, 这个位置被设为1(可见), 否则设为0(不可见)。
Args:
seq_length (int): 序列的长度
Returns:
torch.Tensor: 因果注意力掩码,大小为 (seq_length, seq_length)
"""
mask = torch.triu(torch.ones(seq_length, seq_length) * float('-inf'), diagonal=1)
return mask
class CMlp(nn.Module):
def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.):
super().__init__()
out_features = out_features or in_features
hidden_features = hidden_features or in_features
self.fc1 = nn.Conv1d(in_features, hidden_features, 1)
self.act = act_layer()
self.fc2 = nn.Conv1d(hidden_features, out_features, 1)
self.drop = nn.Dropout(drop)
def forward(self, x):
x = x.permute(0,2,1)
x = self.fc1(x)
x = self.act(x)
x = self.drop(x)
x = self.fc2(x)
x = self.drop(x)
x = x.permute(0,2,1)
return x
class Transpose(nn.Module):
def __init__(self, *dims, contiguous=False):
super().__init__()
self.dims, self.contiguous = dims, contiguous
def forward(self, x):
if self.contiguous: return x.transpose(*self.dims).contiguous()
else: return x.transpose(*self.dims)
class MultiheadAttention(nn.Module):
def __init__(self, d_model, n_heads, d_k=None, d_v=None, res_attention=False, attn_dropout=0., proj_dropout=0., qkv_bias=True, lsa=False, rope_type=False):
"""Multi Head Attention Layer
Input shape:
Q: [batch_size (bs) x max_q_len x d_model]
K, V: [batch_size (bs) x q_len x d_model]
mask: [q_len x q_len]
"""
super().__init__()
d_k = d_model // n_heads if d_k is None else d_k
d_v = d_model // n_heads if d_v is None else d_v
self.n_heads, self.d_k, self.d_v = n_heads, d_k, d_v
self.W_Q = nn.Linear(d_model, d_k * n_heads, bias=qkv_bias)
self.W_K = nn.Linear(d_model, d_k * n_heads, bias=qkv_bias)
self.W_V = nn.Linear(d_model, d_v * n_heads, bias=qkv_bias)
# Scaled Dot-Product Attention (multiple heads)
self.res_attention = res_attention
self.sdp_attn = ScaledDotProductAttention(d_model, n_heads, attn_dropout=attn_dropout, res_attention=self.res_attention, lsa=lsa, rope_type=rope_type)
# Poject output
self.to_out = nn.Sequential(nn.Linear(n_heads * d_v, d_model), nn.Dropout(proj_dropout))
def forward(self, Q:Tensor, K:Optional[Tensor]=None, V:Optional[Tensor]=None, prev:Optional[Tensor]=None,
key_padding_mask:Optional[Tensor]=None, attn_mask:Optional[Tensor]=None):
bs = Q.size(0)
if K is None: K = Q
if V is None: V = Q
# Linear (+ split in multiple heads)
q_s = self.W_Q(Q).view(bs, -1, self.n_heads, self.d_k).transpose(1,2) # q_s : [bs x n_heads x max_q_len x d_k]
k_s = self.W_K(K).view(bs, -1, self.n_heads, self.d_k).permute(0,2,3,1) # k_s : [bs x n_heads x d_k x q_len] - transpose(1,2) + transpose(2,3)
v_s = self.W_V(V).view(bs, -1, self.n_heads, self.d_v).transpose(1,2) # v_s : [bs x n_heads x q_len x d_v]
# Apply Scaled Dot-Product Attention (multiple heads)
if self.res_attention:
output, attn_weights, attn_scores = self.sdp_attn(q_s, k_s, v_s, prev=prev, key_padding_mask=key_padding_mask, attn_mask=attn_mask)
else:
output, attn_weights = self.sdp_attn(q_s, k_s, v_s, key_padding_mask=key_padding_mask, attn_mask=attn_mask)
# output: [bs x n_heads x q_len x d_v], attn: [bs x n_heads x q_len x q_len], scores: [bs x n_heads x max_q_len x q_len]
# back to the original inputs dimensions
output = output.transpose(1, 2).contiguous().view(bs, -1, self.n_heads * self.d_v) # output: [bs x q_len x n_heads * d_v]
output = self.to_out(output)
if self.res_attention: return output, attn_weights, attn_scores
else: return output, attn_weights
class ScaledDotProductAttention(nn.Module):
r"""Scaled Dot-Product Attention module (Attention is all you need by Vaswani et al., 2017) with optional residual attention from previous layer
(Realformer: Transformer likes residual attention by He et al, 2020) and locality self sttention (Vision Transformer for Small-Size Datasets
by Lee et al, 2021)"""
def __init__(self, d_model, n_heads, attn_dropout=0., res_attention=False, lsa=False, rope_type=False):
super().__init__()
self.attn_dropout = nn.Dropout(attn_dropout)
self.res_attention = res_attention
head_dim = d_model // n_heads
self.scale = nn.Parameter(torch.tensor(head_dim ** -0.5), requires_grad=lsa)
self.lsa = lsa
self.rope_type = rope_type
def forward(self, q:Tensor, k:Tensor, v:Tensor, prev:Optional[Tensor]=None, key_padding_mask:Optional[Tensor]=None, attn_mask:Optional[Tensor]=None):
'''
Input shape:
q : [bs x n_heads x max_q_len x d_k]
k : [bs x n_heads x d_k x seq_len]
v : [bs x n_heads x seq_len x d_v]
prev : [bs x n_heads x q_len x seq_len]
key_padding_mask: [bs x seq_len]
attn_mask : [1 x seq_len x seq_len]
Output shape:
output: [bs x n_heads x q_len x d_v]
attn : [bs x n_heads x q_len x seq_len]
scores : [bs x n_heads x q_len x seq_len]
'''
# using RoPE
if self.rope_type:
q, k = RoPE_decoder(q, k.permute(0,1,3,2))
else:
q, k = RoPE(q, k.permute(0,1,3,2))
k = k.permute(0,1,3,2)
# Scaled MatMul (q, k) - similarity scores for all pairs of positions in an input sequence
attn_scores = torch.matmul(q, k) * self.scale # attn_scores : [bs x n_heads x max_q_len x q_len]
# Add pre-softmax attention scores from the previous layer (optional)
if prev is not None: attn_scores = attn_scores + prev
# Attention mask (optional)
if attn_mask is not None: # attn_mask with shape [q_len x seq_len] - only used when q_len == seq_len
if attn_mask.dtype == torch.bool:
attn_scores.masked_fill_(attn_mask, -np.inf)
else:
attn_scores += attn_mask
# Key padding mask (optional)
if key_padding_mask is not None: # mask with shape [bs x q_len] (only when max_w_len == q_len)
attn_scores.masked_fill_(key_padding_mask.unsqueeze(1).unsqueeze(2), -np.inf)
# normalize the attention weights
attn_weights = F.softmax(attn_scores, dim=-1) # attn_weights : [bs x n_heads x max_q_len x q_len]
attn_weights = self.attn_dropout(attn_weights)
# compute the new values given the attention weights
output = torch.matmul(attn_weights, v) # output: [bs x n_heads x max_q_len x d_v]
if self.res_attention: return output, attn_weights, attn_scores
else: return output, attn_weights
def RoPE(q, k):
# q,k: (bs, head, max_len, output_dim)
batch_size = q.shape[0]
nums_head = q.shape[1]
max_len = q.shape[2]
output_dim = q.shape[-1]
# (bs, head, max_len, output_dim)
pos_emb = sinusoidal_position_embedding(batch_size, nums_head, max_len, output_dim, q.device, factor=1)
# cos_pos,sin_pos: (bs, head, max_len, output_dim)
# 看rope公式可知,相邻cos,sin之间是相同的,所以复制一遍。如(1,2,3)变成(1,1,2,2,3,3)
cos_pos = pos_emb[..., 1::2].repeat_interleave(2, dim=-1) # 将奇数列信息抽取出来也就是cos 拿出来并复制
sin_pos = pos_emb[..., ::2].repeat_interleave(2, dim=-1) # 将偶数列信息抽取出来也就是sin 拿出来并复制
# q,k: (bs, head, max_len, output_dim)
q2 = torch.stack([-q[..., 1::2], q[..., ::2]], dim=-1)
q2 = q2.reshape(q.shape) # reshape后就是正负交替了
# 更新qw, *对应位置相乘
q = q * cos_pos + q2 * sin_pos
k2 = torch.stack([-k[..., 1::2], k[..., ::2]], dim=-1)
k2 = k2.reshape(k.shape)
# 更新kw, *对应位置相乘
k = k * cos_pos + k2 * sin_pos
return q, k
def RoPE_decoder(q, k):
# q,k: (bs, head, max_len, output_dim)
batch_size = q.shape[0]
nums_head = q.shape[1]
q_max_len = q.shape[2]
k_max_len = k.shape[2]
output_dim = q.shape[-1]
# (bs, head, max_len, output_dim)
pos_emb = sinusoidal_position_embedding(batch_size, nums_head, k_max_len + q_max_len, output_dim, q.device, factor=1)
# cos_pos,sin_pos: (bs, head, max_len, output_dim)
# 看rope公式可知,相邻cos,sin之间是相同的,所以复制一遍。如(1,2,3)变成(1,1,2,2,3,3)
cos_pos = pos_emb[..., 1::2].repeat_interleave(2, dim=-1) # 将奇数列信息抽取出来也就是cos 拿出来并复制
sin_pos = pos_emb[..., ::2].repeat_interleave(2, dim=-1) # 将偶数列信息抽取出来也就是sin 拿出来并复制
# q,k: (bs, head, max_len, output_dim)
q2 = torch.stack([-q[..., 1::2], q[..., ::2]], dim=-1)
q2 = q2.reshape(q.shape) # reshape后就是正负交替了
# 更新qw, *对应位置相乘
q = q * cos_pos[:,:,-q_max_len:,:] + q2 * sin_pos[:,:,-q_max_len:,:]
k2 = torch.stack([-k[..., 1::2], k[..., ::2]], dim=-1)
k2 = k2.reshape(k.shape)
# 更新kw, *对应位置相乘
k = k * cos_pos[:,:,:k_max_len,:] + k2 * sin_pos[:,:,:k_max_len,:]
return q, k
def sinusoidal_position_embedding(batch_size, nums_head, max_len, output_dim, device, factor=1.0):
# (max_len * factor, 1)
position = torch.arange(0, max_len * factor, 1 / factor, dtype=torch.float).unsqueeze(-1)
# (output_dim//2)
ids = torch.arange(0, output_dim // 2, dtype=torch.float) # i 范围是 [0, d/2]
theta = torch.pow(10000, -2 * ids / output_dim)
# (max_len * factor, output_dim//2)
embeddings = position * theta
# (max_len * factor, output_dim//2, 2)
embeddings = torch.stack([torch.sin(embeddings), torch.cos(embeddings)], dim=-1)
# (bs, head, max_len * factor, output_dim//2, 2)
embeddings = embeddings.repeat((batch_size, nums_head, *([1] * len(embeddings.shape))))
# (bs, head, max_len * factor, output_dim)
embeddings = torch.reshape(embeddings, (batch_size, nums_head, -1, output_dim))
embeddings = embeddings.to(device)
# 如果 factor > 1, 使用插值位置来生成更细粒度的嵌入
if factor > 1.0:
interpolation_indices = torch.linspace(0, embeddings.shape[2] - 1, max_len).long()
embeddings = embeddings[:, :, interpolation_indices, :]
return embeddings
class PretrainHead(nn.Module):
def __init__(self, d_model, patch_len, dropout):
super().__init__()
self.dropout = nn.Dropout(dropout)
self.linear = nn.Linear(d_model, patch_len)
def forward(self, x):
"""
x: tensor [bs x nvars x d_model x num_patch]
output: tensor [bs x nvars x num_patch x patch_len]
"""
x = x.transpose(2,3) # [bs x nvars x num_patch x d_model]
x = self.linear( self.dropout(x) ) # [bs x nvars x num_patch x patch_len]
x = x.permute(0,2,1,3) # [bs x num_patch x nvars x patch_len]
return x
class decoder_PredictHead(nn.Module):
def __init__(self, d_model, patch_len, target_patch_len, dropout):
super().__init__()
self.dropout = nn.Dropout(dropout)
self.linear = nn.Linear(d_model, target_patch_len)
self.d_model = d_model
def forward(self, x, patch_len):
"""
x: tensor [bs x nvars x d_model x num_patch]
output: tensor [bs x nvars x num_patch x patch_len]
"""
Linear = nn.Linear(self.d_model, patch_len, bias=False)
Linear.weight.data = resample_patchemb(old=self.linear.weight.data.T, new_patch_len=patch_len).T
x = x.transpose(2,3) # [bs x nvars x num_patch x d_model]
x = Linear( self.dropout(x) ) # [bs x nvars x num_patch x patch_len]
x = x.permute(0,2,3,1) # [bs x num_patch x x patch_len x nvars]
return x.reshape(x.shape[0],-1,x.shape[3])
def resample_patchemb(old: torch.Tensor, new_patch_len: int):
assert old.dim() == 2, "输入张量应为2D (d_model, patch_size)"
if old.size(1) == new_patch_len:
return old
old = old.T
old_shape = old.size(0)
factor = new_patch_len/old_shape
# 定义辅助函数:批量resize
def resize(x_tensor, new_shape):
return F.interpolate(x_tensor.unsqueeze(0), size=new_shape, mode='linear').squeeze(0)
# 构造缩放矩阵
basis_vectors = torch.eye(old_shape, dtype=torch.float32, device=old.device)
resize_mat = resize(basis_vectors, new_patch_len).T
# 计算伪逆
resize_mat_pinv = torch.linalg.pinv(resize_mat.T)
# 直接矩阵操作完成重采样
resampled_kernels = resize_mat_pinv @ old * math.sqrt(factor)
return resampled_kernels.T
def get_activation_fn(activation):
if callable(activation): return activation()
elif activation.lower() == "relu": return nn.ReLU()
elif activation.lower() == "gelu": return nn.GELU()
raise ValueError(f'{activation} is not available. You can use "relu", "gelu", or a callable') |