File size: 36,478 Bytes
906041e 11a6736 906041e 39804e8 906041e 39804e8 906041e c6d1efe 906041e 349a00a 906041e 45f82b9 906041e 45f82b9 106f086 906041e 0d4e4d8 906041e afdfb6c c0bb180 906041e ea77f15 906041e 11a6736 906041e 754cfc0 8810f3f 11a6736 77e26a3 ea77f15 9bd3985 d046804 9bd3985 ea77f15 d046804 ea77f15 9bd3985 d046804 ea77f15 9bd3985 11a6736 d046804 11a6736 9bd3985 11a6736 9bd3985 11a6736 9bd3985 11a6736 9bd3985 11a6736 2a3bbbd d046804 11a6736 ea77f15 e6dedf0 11a6736 e6dedf0 11a6736 75fb4e4 11a6736 9bd3985 11a6736 b1adb62 9bd3985 11a6736 6d641db 11a6736 9bd3985 11a6736 9bd3985 e22454c 9bd3985 11a6736 9bd3985 11a6736 e22454c 9bd3985 e22454c 9bd3985 11a6736 9bd3985 e22454c 9bd3985 e22454c 9bd3985 11a6736 9bd3985 acebd75 9bd3985 9469b79 9bd3985 44522e0 b1adb62 349a00a 2a3bbbd 349a00a 9bd3985 a9f561a ac1e96b 9bd3985 132d93f 9bd3985 5d16db2 77e26a3 2a3bbbd a9f561a 11a6736 45f82b9 11a6736 45f82b9 a9a3194 45f82b9 11a6736 abd1504 45f82b9 11a6736 9873fe3 45f82b9 11a6736 45f82b9 11a6736 45f82b9 11a6736 45f82b9 11a6736 45f82b9 11a6736 45f82b9 11a6736 45f82b9 11a6736 45f82b9 11a6736 45f82b9 11a6736 45f82b9 11a6736 45f82b9 11a6736 45f82b9 ad5f9fa 45f82b9 11a6736 45f82b9 11a6736 45f82b9 11a6736 45f82b9 11a6736 45f82b9 11a6736 906041e 968690f 906041e 4b26dae 906041e ea77f15 906041e 45f82b9 906041e c85ad25 906041e c85ad25 906041e 968690f 906041e 968690f 906041e 45f82b9 906041e 45f82b9 906041e 45f82b9 906041e 5df2393 906041e 968690f 906041e 10018f1 a5444d9 d046804 a5444d9 d046804 a5444d9 d046804 a5444d9 d046804 a5444d9 906041e a5444d9 906041e a5444d9 26cec45 a5444d9 26cec45 9469b79 d046804 a5444d9 d046804 a5444d9 26cec45 9469b79 a5444d9 26cec45 906041e a5444d9 906041e c0bb180 df864ac 906041e a5444d9 906041e a5444d9 906041e c0bb180 906041e e2d5188 45f82b9 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 | from typing import Callable, Optional
import math
import torch
from torch import nn
from transformers.activations import ACT2FN
from transformers.integrations import use_kernel_forward_from_hub
from transformers.modeling_flash_attention_utils import FlashAttentionKwargs
from transformers.modeling_layers import GradientCheckpointingLayer
from transformers.modeling_outputs import BaseModelOutputWithPast
from transformers.modeling_rope_utils import ROPE_INIT_FUNCTIONS, dynamic_rope_update
from transformers.modeling_utils import ALL_ATTENTION_FUNCTIONS, PreTrainedModel
from transformers.processing_utils import Unpack
from transformers.utils import TransformersKwargs, auto_docstring
from transformers.utils.generic import check_model_inputs
from .configuration_hinvec import HinvecConfig
SUPPORTED_SLIDING_BACKENDS=["flash_attention_3", "flash_attention_2"]
class HinvecMLP(nn.Module):
def __init__(self, config):
super().__init__()
self.config = config
self.hidden_size = config.hidden_size
self.intermediate_size = config.intermediate_size
self.gate_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
self.up_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
self.down_proj = nn.Linear(self.intermediate_size, self.hidden_size, bias=False)
self.act_fn = ACT2FN[config.hidden_act]
def forward(self, x):
down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))
return down_proj
def rotate_half(x):
"""Rotates half the hidden dims of the input."""
x1 = x[..., : x.shape[-1] // 2]
x2 = x[..., x.shape[-1] // 2 :]
return torch.cat((-x2, x1), dim=-1)
def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_dim=1):
"""Applies Rotary Position Embedding to the query and key tensors."""
cos = cos.unsqueeze(unsqueeze_dim)
sin = sin.unsqueeze(unsqueeze_dim)
q_embed = (q * cos) + (rotate_half(q) * sin)
k_embed = (k * cos) + (rotate_half(k) * sin)
return q_embed, k_embed
def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
"""
This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
"""
batch, num_key_value_heads, slen, head_dim = hidden_states.shape
if n_rep == 1:
return hidden_states
hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)
return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)
def eager_attention_forward(
module: nn.Module,
query: torch.Tensor,
key: torch.Tensor,
value: torch.Tensor,
attention_mask: Optional[torch.Tensor],
scaling: float,
dropout: float = 0.0,
**kwargs: Unpack[TransformersKwargs],
):
"""Standard eager attention without sliding window optimization."""
key_states = repeat_kv(key, 1)
value_states = repeat_kv(value, 1)
attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling
if attention_mask is not None:
attention_mask = attention_mask[:, None, None, :]
attn_weights = attn_weights + attention_mask
attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)
attn_weights = nn.functional.dropout(attn_weights, p=dropout, training=module.training)
attn_output = torch.matmul(attn_weights, value_states)
attn_output = attn_output.transpose(1, 2).contiguous()
return attn_output, attn_weights
class HinvecAttention(nn.Module):
def __init__(self, config, layer_idx):
super().__init__()
self.config = config
self.head_dim = getattr(config, "head_dim", config.hidden_size // config.num_attention_heads)
self.num_key_value_groups = config.num_attention_heads // config.num_key_value_heads
self.scaling = self.head_dim**-0.5
self.dropout = config.attention_dropout
self.is_causal = False
if config.hidden_size % config.num_attention_heads != 0:
raise ValueError(
f"The hidden size ({config.hidden_size}) is not a multiple of the number of attention "
f"heads ({config.num_attention_heads})"
)
self.num_heads = config.num_attention_heads
self.head_dim = int(config.hidden_size / config.num_attention_heads)
self.embed_dim = config.hidden_size
self.query = nn.Linear(config.hidden_size, self.embed_dim)
self.key = nn.Linear(config.hidden_size, self.embed_dim)
self.value = nn.Linear(config.hidden_size, self.embed_dim)
self.o_proj = nn.Linear(config.hidden_size, config.hidden_size, bias=True)
self.layer_idx = layer_idx
# Initialize sliding_window and sliding_window
self.sliding_window = None
if config.layer_types[layer_idx] == "sliding_attention":
self.sliding_window = config.sliding_window
assert self.sliding_window % 2 == 0, (
f"`sliding_window` for layer {self.layer_idx} has to be an even value. Given {self.sliding_window}"
)
assert self.sliding_window > 0, (
f"`sliding_window` for layer {self.layer_idx} has to be positive. Given {self.sliding_window}"
)
self.one_sided_attn_window_size = self.sliding_window // 2
# Store attention_dropout as instance variable
self.attention_dropout = config.attention_dropout
def forward(
self,
hidden_states,
position_embeddings: tuple[torch.Tensor, torch.Tensor],
attention_mask=None,
position_ids: Optional[torch.LongTensor] = None,
is_index_masked=None,
output_attentions=False,
**kwargs,
):
"""
[`HinvecAttention`] expects *len(hidden_states)* to be multiple of *sliding_window*. Padding to
*sliding_window* happens in [`LongformerModel.forward`] to avoid redoing the padding on each layer.
The *attention_mask* is changed in [`LongformerModel.forward`] from 0, 1, 2 to:
- -10000: no attention
- 0: local attention
"""
print(self.layer_idx)
input_shape = hidden_states.shape[:-1]
hidden_shape = (*input_shape, -1, self.head_dim)
# project hidden states
query_vectors = self.query(hidden_states).view(hidden_shape).transpose(1, 2)
key_vectors = self.key(hidden_states).view(hidden_shape).transpose(1, 2)
value_vectors = self.value(hidden_states).view(hidden_shape).transpose(1, 2)
batch_size, seq_len, embed_dim = hidden_states.size()
assert embed_dim == self.embed_dim, (
f"hidden_states should have embed_dim = {self.embed_dim}, but has {embed_dim}"
)
cos, sin = position_embeddings
query_vectors, key_vectors = apply_rotary_pos_emb(query_vectors, key_vectors, cos, sin)
if self.sliding_window and self.config._attn_implementation == "eager":
# normalize query
query_vectors /= math.sqrt(self.head_dim)
query_vectors = query_vectors.transpose(1, 2)
key_vectors = key_vectors.transpose(1, 2)
attn_scores = self._sliding_chunks_query_key_matmul(
query_vectors, key_vectors, self.one_sided_attn_window_size
)
# values to pad for attention probs
remove_from_windowed_attention_mask = (attention_mask != 0)[:, :, None, None]
# cast to fp32/fp16 then replace 1's with -inf
float_mask = remove_from_windowed_attention_mask.type_as(query_vectors).masked_fill(
remove_from_windowed_attention_mask, torch.finfo(query_vectors.dtype).min
)
# diagonal mask with zeros everywhere and -inf inplace of padding
diagonal_mask = self._sliding_chunks_query_key_matmul(
float_mask.new_ones(size=float_mask.size()), float_mask, self.one_sided_attn_window_size
)
# pad local attention probs
attn_scores += diagonal_mask
assert list(attn_scores.size()) == [
batch_size,
seq_len,
self.num_heads,
self.one_sided_attn_window_size * 2 + 1,
], (
f"local_attn_probs should be of size ({batch_size}, {seq_len}, {self.num_heads},"
f" {self.one_sided_attn_window_size * 2 + 1}), but is of size {attn_scores.size()}"
)
attn_probs = nn.functional.softmax(
attn_scores, dim=-1, dtype=torch.float32
) # use fp32 for numerical stability
# softmax sometimes inserts NaN if all positions are masked, replace them with 0
attn_probs = torch.masked_fill(attn_probs, is_index_masked[:, :, None, None], 0.0)
attn_probs = attn_probs.type_as(attn_scores)
# free memory
del attn_scores
# apply dropout
attn_probs = nn.functional.dropout(attn_probs, p=self.dropout, training=self.training)
value_vectors = value_vectors.transpose(1, 2)
# compute local attn only
attn_output = self._sliding_chunks_matmul_attn_probs_value(
attn_probs, value_vectors, self.one_sided_attn_window_size
)
assert attn_output.size() == (batch_size, seq_len, self.num_heads, self.head_dim), "Unexpected size"
attn_output = attn_output.transpose(0, 1).reshape(seq_len, batch_size, embed_dim).contiguous()
attn_weights = attn_probs.transpose(1, 2)
elif self.sliding_window:
if self.config._attn_implementation in SUPPORTED_SLIDING_BACKENDS:
attention_interface = ALL_ATTENTION_FUNCTIONS[self.config._attn_implementation]
attn_output, attn_weights = attention_interface(
self,
query_vectors,
key_vectors,
value_vectors,
(attention_mask < 0)*torch.finfo(query_vectors.dtype).min,
dropout=0.0 if not self.training else self.attention_dropout,
scaling=self.scaling,
sliding_window=self.sliding_window,
**kwargs,
)
else:
# Construct sliding window mask
seq_len = query_vectors.size(-2)
sw_mask = self._build_sliding_window_mask(seq_len, query_vectors.device, query_vectors.dtype)
print(sw_mask)
attention_interface = ALL_ATTENTION_FUNCTIONS[self.config._attn_implementation]
attn_output, attn_weights = attention_interface(
self,
query_vectors,
key_vectors,
value_vectors,
# Combine with user-provided mask by addition
(attention_mask < 0) * torch.finfo(query_vectors.dtype).min + sw_mask,
dropout=0.0 if not self.training else self.attention_dropout,
scaling=self.scaling,
is_causal=False, # IMPORTANT
**kwargs,
)
else:
# Use standard attention implementation
attention_interface: Callable = eager_attention_forward
if self.config._attn_implementation != "eager":
attention_interface = ALL_ATTENTION_FUNCTIONS[self.config._attn_implementation]
attn = (attention_mask < 0)*torch.finfo(query_vectors.dtype).min
attn = attn[:, None, :, None].expand(-1, 1, attn.size(1), attn.size(1))
attn_output, attn_weights = attention_interface(
self,
query_vectors,
key_vectors,
value_vectors,
attn,
dropout=0.0 if not self.training else self.attention_dropout,
scaling=self.scaling,
**kwargs,
)
attn_output = attn_output.reshape(*input_shape, -1).contiguous()
attn_output = self.o_proj(attn_output)
return attn_output, attn_weights
def _build_sliding_window_mask(self, seq_len, device, dtype):
"""
Creates an additive attention mask of shape (1, 1, seq_len, seq_len)
where each token attends only to a window around itself.
Allowed positions: i-w ... i+w
Others masked with -∞
"""
w = self.one_sided_attn_window_size
# Base mask: True means "mask this (disallowed)"
mask = torch.ones(seq_len, seq_len, device=device, dtype=torch.bool)
# Fill allowed window as False
arange = torch.arange(seq_len, device=device)
for i in range(seq_len):
left = max(0, i - w)
right = min(seq_len, i + w + 1)
mask[i, left:right] = False
# Convert boolean mask → additive logit mask
# True → -inf, False → 0
mask = mask.masked_fill(mask, torch.finfo(dtype).min)
# Shape to (batch, heads, Q, K) broadcastable format
return mask.view(1, 1, seq_len, seq_len)
@staticmethod
def _pad_and_transpose_last_two_dims(hidden_states_padded, padding):
"""pads rows and then flips rows and columns"""
hidden_states_padded = nn.functional.pad(
hidden_states_padded, padding
) # padding value is not important because it will be overwritten
hidden_states_padded = hidden_states_padded.view(
*hidden_states_padded.size()[:-2], hidden_states_padded.size(-1), hidden_states_padded.size(-2)
)
return hidden_states_padded
@staticmethod
def _pad_and_diagonalize(chunked_hidden_states):
"""
shift every row 1 step right, converting columns into diagonals.
Example:
```python
chunked_hidden_states: [
0.4983,
2.6918,
-0.0071,
1.0492,
-1.8348,
0.7672,
0.2986,
0.0285,
-0.7584,
0.4206,
-0.0405,
0.1599,
2.0514,
-1.1600,
0.5372,
0.2629,
]
window_overlap = num_rows = 4
```
(pad & diagonalize) => [ 0.4983, 2.6918, -0.0071, 1.0492, 0.0000, 0.0000, 0.0000
0.0000, -1.8348, 0.7672, 0.2986, 0.0285, 0.0000, 0.0000 0.0000, 0.0000, -0.7584, 0.4206,
-0.0405, 0.1599, 0.0000 0.0000, 0.0000, 0.0000, 2.0514, -1.1600, 0.5372, 0.2629 ]
"""
total_num_heads, num_chunks, window_overlap, hidden_dim = chunked_hidden_states.size()
chunked_hidden_states = nn.functional.pad(
chunked_hidden_states, (0, window_overlap + 1)
) # total_num_heads x num_chunks x window_overlap x (hidden_dim+window_overlap+1). Padding value is not important because it'll be overwritten
chunked_hidden_states = chunked_hidden_states.view(
total_num_heads, num_chunks, -1
) # total_num_heads x num_chunks x window_overlap*window_overlap+window_overlap
chunked_hidden_states = chunked_hidden_states[
:, :, :-window_overlap
] # total_num_heads x num_chunks x window_overlap*window_overlap
chunked_hidden_states = chunked_hidden_states.view(
total_num_heads, num_chunks, window_overlap, window_overlap + hidden_dim
)
chunked_hidden_states = chunked_hidden_states[:, :, :, :-1]
return chunked_hidden_states
@staticmethod
def _chunk(hidden_states, window_overlap, onnx_export: bool = False):
"""convert into overlapping chunks. Chunk size = 2w, overlap size = w"""
if not onnx_export:
# non-overlapping chunks of size = 2w
hidden_states = hidden_states.view(
hidden_states.size(0),
torch.div(hidden_states.size(1), (window_overlap * 2), rounding_mode="trunc"),
window_overlap * 2,
hidden_states.size(2),
)
# use `as_strided` to make the chunks overlap with an overlap size = window_overlap
chunk_size = list(hidden_states.size())
chunk_size[1] = chunk_size[1] * 2 - 1
chunk_stride = list(hidden_states.stride())
chunk_stride[1] = chunk_stride[1] // 2
return hidden_states.as_strided(size=chunk_size, stride=chunk_stride)
# When exporting to ONNX, use this separate logic
# have to use slow implementation since as_strided, unfold and 2d-tensor indexing aren't supported (yet) in ONNX export
# TODO replace this with
# > return hidden_states.unfold(dimension=1, size=window_overlap * 2, step=window_overlap).transpose(2, 3)
# once `unfold` is supported
# the case hidden_states.size(1) == window_overlap * 2 can also simply return hidden_states.unsqueeze(1), but that's control flow
chunk_size = [
hidden_states.size(0),
torch.div(hidden_states.size(1), window_overlap, rounding_mode="trunc") - 1,
window_overlap * 2,
hidden_states.size(2),
]
overlapping_chunks = torch.empty(chunk_size, device=hidden_states.device)
for chunk in range(chunk_size[1]):
overlapping_chunks[:, chunk, :, :] = hidden_states[
:, chunk * window_overlap : chunk * window_overlap + 2 * window_overlap, :
]
return overlapping_chunks
@staticmethod
def _mask_invalid_locations(input_tensor, affected_seq_len) -> torch.Tensor:
beginning_mask_2d = input_tensor.new_ones(affected_seq_len, affected_seq_len + 1).tril().flip(dims=[0])
beginning_mask = beginning_mask_2d[None, :, None, :]
ending_mask = beginning_mask.flip(dims=(1, 3))
beginning_input = input_tensor[:, :affected_seq_len, :, : affected_seq_len + 1]
beginning_mask = beginning_mask.expand(beginning_input.size())
input_tensor[:, :affected_seq_len, :, : affected_seq_len + 1] = torch.full_like(
beginning_input, -float("inf")
).where(beginning_mask.bool(), beginning_input)
ending_input = input_tensor[:, -affected_seq_len:, :, -(affected_seq_len + 1) :]
ending_mask = ending_mask.expand(ending_input.size())
input_tensor[:, -affected_seq_len:, :, -(affected_seq_len + 1) :] = torch.full_like(
ending_input, -float("inf")
).where(ending_mask.bool(), ending_input)
def _sliding_chunks_query_key_matmul(self, query: torch.Tensor, key: torch.Tensor, window_overlap: int):
"""
Matrix multiplication of query and key tensors using with a sliding window attention pattern. This
implementation splits the input into overlapping chunks of size 2w (e.g. 512 for pretrained Longformer) with an
overlap of size window_overlap
"""
batch_size, seq_len, num_heads, head_dim = query.size()
assert seq_len % (window_overlap * 2) == 0, (
f"Sequence length should be multiple of {window_overlap * 2}. Given {seq_len}"
)
assert query.size() == key.size()
chunks_count = torch.div(seq_len, window_overlap, rounding_mode="trunc") - 1
# group batch_size and num_heads dimensions into one, then chunk seq_len into chunks of size window_overlap * 2
query = query.reshape(batch_size * num_heads, seq_len, head_dim)
key = key.reshape(batch_size * num_heads, seq_len, head_dim)
query = self._chunk(query, window_overlap, getattr(self.config, "onnx_export", False))
key = self._chunk(key, window_overlap, getattr(self.config, "onnx_export", False))
# matrix multiplication
# bcxd: batch_size * num_heads x chunks x 2window_overlap x head_dim
# bcyd: batch_size * num_heads x chunks x 2window_overlap x head_dim
# bcxy: batch_size * num_heads x chunks x 2window_overlap x 2window_overlap
diagonal_chunked_attention_scores = torch.einsum("bcxd,bcyd->bcxy", (query, key)) # multiply
# convert diagonals into columns
diagonal_chunked_attention_scores = self._pad_and_transpose_last_two_dims(
diagonal_chunked_attention_scores, padding=(0, 0, 0, 1)
)
# allocate space for the overall attention matrix where the chunks are combined. The last dimension
# has (window_overlap * 2 + 1) columns. The first (window_overlap) columns are the window_overlap lower triangles (attention from a word to
# window_overlap previous words). The following column is attention score from each word to itself, then
# followed by window_overlap columns for the upper triangle.
diagonal_attention_scores = diagonal_chunked_attention_scores.new_zeros(
(batch_size * num_heads, chunks_count + 1, window_overlap, window_overlap * 2 + 1)
)
# copy parts from diagonal_chunked_attention_scores into the combined matrix of attentions
# - copying the main diagonal and the upper triangle
diagonal_attention_scores[:, :-1, :, window_overlap:] = diagonal_chunked_attention_scores[
:, :, :window_overlap, : window_overlap + 1
]
diagonal_attention_scores[:, -1, :, window_overlap:] = diagonal_chunked_attention_scores[
:, -1, window_overlap:, : window_overlap + 1
]
# - copying the lower triangle
diagonal_attention_scores[:, 1:, :, :window_overlap] = diagonal_chunked_attention_scores[
:, :, -(window_overlap + 1) : -1, window_overlap + 1 :
]
diagonal_attention_scores[:, 0, 1:window_overlap, 1:window_overlap] = diagonal_chunked_attention_scores[
:, 0, : window_overlap - 1, 1 - window_overlap :
]
# separate batch_size and num_heads dimensions again
diagonal_attention_scores = diagonal_attention_scores.view(
batch_size, num_heads, seq_len, 2 * window_overlap + 1
).transpose(2, 1)
self._mask_invalid_locations(diagonal_attention_scores, window_overlap)
return diagonal_attention_scores
def _sliding_chunks_matmul_attn_probs_value(
self, attn_probs: torch.Tensor, value: torch.Tensor, window_overlap: int
):
"""
Same as _sliding_chunks_query_key_matmul but for attn_probs and value tensors. Returned tensor will be of the
same shape as `attn_probs`
"""
batch_size, seq_len, num_heads, head_dim = value.size()
assert seq_len % (window_overlap * 2) == 0
assert attn_probs.size()[:3] == value.size()[:3]
assert attn_probs.size(3) == 2 * window_overlap + 1
chunks_count = torch.div(seq_len, window_overlap, rounding_mode="trunc") - 1
# group batch_size and num_heads dimensions into one, then chunk seq_len into chunks of size 2 window overlap
chunked_attn_probs = attn_probs.transpose(1, 2).reshape(
batch_size * num_heads,
torch.div(seq_len, window_overlap, rounding_mode="trunc"),
window_overlap,
2 * window_overlap + 1,
)
# group batch_size and num_heads dimensions into one
value = value.transpose(1, 2).reshape(batch_size * num_heads, seq_len, head_dim)
# pad seq_len with w at the beginning of the sequence and another window overlap at the end
padded_value = nn.functional.pad(value, (0, 0, window_overlap, window_overlap), value=-1)
# chunk padded_value into chunks of size 3 window overlap and an overlap of size window overlap
chunked_value_size = (batch_size * num_heads, chunks_count + 1, 3 * window_overlap, head_dim)
chunked_value_stride = padded_value.stride()
chunked_value_stride = (
chunked_value_stride[0],
window_overlap * chunked_value_stride[1],
chunked_value_stride[1],
chunked_value_stride[2],
)
chunked_value = padded_value.as_strided(size=chunked_value_size, stride=chunked_value_stride)
chunked_attn_probs = self._pad_and_diagonalize(chunked_attn_probs)
context = torch.einsum("bcwd,bcdh->bcwh", (chunked_attn_probs, chunked_value))
return context.view(batch_size, num_heads, seq_len, head_dim).transpose(1, 2)
@use_kernel_forward_from_hub("RMSNorm")
class HinvecRMSNorm(nn.Module):
def __init__(self, hidden_size, eps: float = 1e-6) -> None:
"""
HinvecRMSNorm is equivalent to T5LayerNorm
"""
super().__init__()
self.weight = nn.Parameter(torch.ones(hidden_size))
self.variance_epsilon = eps
def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
input_dtype = hidden_states.dtype
hidden_states = hidden_states.to(torch.float32)
variance = hidden_states.pow(2).mean(-1, keepdim=True)
hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)
return self.weight * hidden_states.to(input_dtype)
def extra_repr(self):
return f"{tuple(self.weight.shape)}, eps={self.variance_epsilon}"
class HinvecEncoderLayer(GradientCheckpointingLayer):
def __init__(self, config: HinvecConfig, layer_idx: int):
super().__init__()
self.layer_idx = layer_idx
self.hidden_size = config.hidden_size
self.self_attn = HinvecAttention(config=config, layer_idx=layer_idx)
self.mlp = HinvecMLP(config)
self.input_layernorm = HinvecRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
self.post_attention_layernorm = HinvecRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
self.attention_type = config.layer_types[layer_idx]
def forward(
self,
hidden_states: torch.Tensor,
attention_mask: Optional[torch.Tensor] = None,
position_ids: Optional[torch.LongTensor] = None,
position_embeddings: Optional[tuple[torch.Tensor, torch.Tensor]] = None,
**kwargs: Unpack[TransformersKwargs],
) -> torch.Tensor:
is_index_masked = attention_mask < 0
residual = hidden_states
hidden_states = self.input_layernorm(hidden_states)
# Self Attention
hidden_states, _ = self.self_attn(
hidden_states=hidden_states,
attention_mask=attention_mask,
position_ids=position_ids,
position_embeddings=position_embeddings,
is_index_masked=is_index_masked,
**kwargs,
)
hidden_states = residual + hidden_states
# Fully Connected
residual = hidden_states
hidden_states = self.post_attention_layernorm(hidden_states)
hidden_states = self.mlp(hidden_states)
hidden_states = residual + hidden_states
return hidden_states
@auto_docstring
class HinvecPreTrainedModel(PreTrainedModel):
config: HinvecConfig
base_model_prefix = "model"
supports_gradient_checkpointing = True
_no_split_modules = ["HinvecEncoderLayer"]
_supports_flash_attn = True
_supports_sdpa = True
_supports_flex_attn = True
_can_compile_fullgraph = True
_supports_attention_backend = True
_can_record_outputs = {
"hidden_states": HinvecEncoderLayer,
"attentions": HinvecAttention,
}
class HinvecRotaryEmbedding(nn.Module):
inv_freq: torch.Tensor
def __init__(self, config: HinvecConfig, device=None):
super().__init__()
if hasattr(config, "rope_scaling") and isinstance(config.rope_scaling, dict):
self.rope_type = config.rope_scaling.get("rope_type", config.rope_scaling.get("type"))
else:
self.rope_type = "default"
self.max_seq_len_cached = config.max_position_embeddings
self.original_max_seq_len = config.max_position_embeddings
self.config = config
self.rope_init_fn = ROPE_INIT_FUNCTIONS[self.rope_type]
inv_freq, self.attention_scaling = self.rope_init_fn(self.config, device)
self.register_buffer("inv_freq", inv_freq, persistent=False)
self.original_inv_freq = self.inv_freq
@torch.no_grad()
@dynamic_rope_update
def forward(self, x, position_ids):
inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1).to(x.device)
position_ids_expanded = position_ids[:, None, :].float()
device_type = x.device.type if isinstance(x.device.type, str) and x.device.type != "mps" else "cpu"
with torch.autocast(device_type=device_type, enabled=False):
freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).transpose(1, 2)
emb = torch.cat((freqs, freqs), dim=-1)
cos = emb.cos() * self.attention_scaling
sin = emb.sin() * self.attention_scaling
return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)
@auto_docstring
class HinvecModel(HinvecPreTrainedModel):
config_class = HinvecConfig
def __init__(self, config: HinvecConfig):
super().__init__(config)
self.padding_idx = config.pad_token_id
self.vocab_size = config.vocab_size
self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx)
self.layers = nn.ModuleList(
[HinvecEncoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
)
self.norm = HinvecRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
self.rotary_emb = HinvecRotaryEmbedding(config=config)
self.gradient_checkpointing = False
self.has_sliding_layers = "sliding_attention" in self.config.layer_types
# Initialize weights and apply final processing
self.post_init()
def get_input_embeddings(self):
return self.embeddings.word_embeddings
def set_input_embeddings(self, value):
self.embeddings.word_embeddings = value
def _prune_heads(self, heads_to_prune):
"""
Prunes heads of the model. heads_to_prune: dict of {layer_num: list of heads to prune in this layer} See base
class PreTrainedModel
"""
for layer, heads in heads_to_prune.items():
self.encoder.layer[layer].attention.prune_heads(heads)
def _pad_to_window_size(
self,
input_ids: torch.Tensor,
attention_mask: torch.Tensor,
token_type_ids: torch.Tensor,
position_ids: torch.Tensor,
inputs_embeds: torch.Tensor,
pad_token_id: int,
):
"""A helper function to pad tokens and mask to work with implementation of Longformer self-attention."""
# padding
sliding_window = (
self.config.sliding_window
if isinstance(self.config.sliding_window, int)
else max(self.config.sliding_window)
)
assert sliding_window % 2 == 0, f"`sliding_window` should be an even value. Given {sliding_window}"
input_shape = input_ids.shape if input_ids is not None else inputs_embeds.shape
batch_size, seq_len = input_shape[:2]
padding_len = (sliding_window - seq_len % sliding_window) % sliding_window
# this path should be recorded in the ONNX export, it is fine with padding_len == 0 as well
if padding_len > 0:
logger.warning_once(
f"Input ids are automatically padded to be a multiple of `config.sliding_window`: {sliding_window}"
)
if input_ids is not None:
input_ids = nn.functional.pad(input_ids, (0, padding_len), value=pad_token_id)
if position_ids is not None:
# pad with position_id = pad_token_id as in modeling_roberta.RobertaEmbeddings
position_ids = nn.functional.pad(position_ids, (0, padding_len), value=pad_token_id)
if inputs_embeds is not None:
input_ids_padding = inputs_embeds.new_full(
(batch_size, padding_len),
self.config.pad_token_id,
dtype=torch.long,
)
inputs_embeds_padding = self.embeddings(input_ids_padding)
inputs_embeds = torch.cat([inputs_embeds, inputs_embeds_padding], dim=-2)
attention_mask = nn.functional.pad(
attention_mask, (0, padding_len), value=0
) # no attention on the padding tokens
token_type_ids = nn.functional.pad(token_type_ids, (0, padding_len), value=0) # pad with token_type_id = 0
return padding_len, input_ids, attention_mask, token_type_ids, position_ids, inputs_embeds
@check_model_inputs
@auto_docstring
def forward(
self,
input_ids: Optional[torch.LongTensor] = None,
attention_mask: Optional[torch.Tensor] = None,
position_ids: Optional[torch.LongTensor] = None,
inputs_embeds: Optional[torch.FloatTensor] = None,
**kwargs: Unpack[TransformersKwargs],
) -> BaseModelOutputWithPast:
if (input_ids is None) ^ (inputs_embeds is not None):
raise ValueError("You must specify exactly one of input_ids or inputs_embeds")
if inputs_embeds is None:
inputs_embeds = self.embed_tokens(input_ids)
device = inputs_embeds.device
batch_size, seq_len = inputs_embeds.shape[:2]
if position_ids is None:
position_ids = torch.arange(seq_len, device=device).unsqueeze(0)
position_ids = position_ids.expand(batch_size, -1)
# 1. ----- CREATE ATTENTION MASK -----
if attention_mask is None:
attention_mask = torch.ones((batch_size, seq_len), device=device, dtype=torch.long)
# 2. ----- PAD TO WINDOW SIZE -----
sliding_window = (
self.config.sliding_window
if isinstance(self.config.sliding_window, int)
else max(self.config.sliding_window)
)
padding_len = (sliding_window - seq_len % sliding_window) % sliding_window
if padding_len > 0:
# Pad input_ids if available
if input_ids is not None:
input_ids = torch.nn.functional.pad(
input_ids, (0, padding_len), value=self.config.pad_token_id
)
# Pad position_ids
position_ids = torch.nn.functional.pad(
position_ids, (0, padding_len), value=self.config.pad_token_id
)
# Pad inputs_embeds
input_ids_padding = inputs_embeds.new_full(
(batch_size, padding_len),
self.config.pad_token_id,
dtype=torch.long,
)
inputs_embeds_padding = self.embed_tokens(input_ids_padding)
inputs_embeds = torch.cat([inputs_embeds, inputs_embeds_padding], dim=-2)
# Pad attention_mask with 0s (no attention on padding)
attention_mask = torch.nn.functional.pad(
attention_mask, (0, padding_len), value=0
)
# 3. ----- CREATE PADDING MASK -----
# Convert tri-level mask to continuous mask for attention
# Tokens with attention_mask == 0 get -inf, others get 0
expanded_mask = (attention_mask > 0).to(inputs_embeds.dtype)
padding_mask = (1.0 - expanded_mask) * torch.finfo(inputs_embeds.dtype).min
hidden_states = inputs_embeds
# Create position embeddings to be shared across the decoder layers
position_embeddings = self.rotary_emb(hidden_states, position_ids)
for encoder_layer in self.layers:
hidden_states = encoder_layer(
hidden_states,
attention_mask=padding_mask,
position_ids=position_ids,
position_embeddings=position_embeddings,
**kwargs,
)
hidden_states = self.norm(hidden_states)
return BaseModelOutputWithPast(
last_hidden_state=hidden_states,
)
__all__ = [
"HinvecPreTrainedModel",
"HinvecModel",
"HinvecRMSNorm"
] |