File size: 58,182 Bytes
7962681 67a854c 7962681 67a854c 7962681 67a854c 7962681 67a854c 7962681 67a854c 7962681 67a854c 7962681 67a854c 7962681 17d1cb0 7962681 17d1cb0 7962681 17d1cb0 7962681 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 |
# coding=utf-8
# Copyright 2022 Meta and The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
""" PyTorch ESM model."""
import math
from dataclasses import dataclass
from typing import List, Optional, Tuple, Union
import torch
import torch.utils.checkpoint
from torch import nn
from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss, SiLU
from transformers.file_utils import (
add_code_sample_docstrings,
add_start_docstrings,
add_start_docstrings_to_model_forward,
)
from transformers.modeling_outputs import (
BaseModelOutputWithPastAndCrossAttentions,
BaseModelOutputWithPoolingAndCrossAttentions,
MaskedLMOutput,
SequenceClassifierOutput,
TokenClassifierOutput,
)
from transformers.modeling_utils import (
PreTrainedModel,
find_pruneable_heads_and_indices,
prune_linear_layer,
)
from transformers.utils import logging
from .segment_nt_config import SegmentNTConfig
logger = logging.get_logger(__name__)
_CHECKPOINT_FOR_DOC = "facebook/esm2_t6_8M_UR50D"
_CONFIG_FOR_DOC = "SegmentNTConfig"
ESM_PRETRAINED_MODEL_ARCHIVE_LIST = [
"facebook/esm2_t6_8M_UR50D",
"facebook/esm2_t12_35M_UR50D",
# This is not a complete list of all ESM models!
# See all ESM models at https://huggingface.co/models?filter=esm
]
def rotate_half(x):
x1, x2 = x.chunk(2, dim=-1)
return torch.cat((-x2, x1), dim=-1)
def apply_rotary_pos_emb(x, cos, sin):
cos = cos[:, :, : x.shape[-2], :]
sin = sin[:, :, : x.shape[-2], :]
return (x * cos) + (rotate_half(x) * sin)
def gelu(x):
"""
This is the gelu implementation from the original ESM repo. Using F.gelu yields subtly wrong results.
"""
return x * 0.5 * (1.0 + torch.erf(x / math.sqrt(2.0)))
def symmetrize(x):
"Make layer symmetric in final two dimensions, used for contact prediction."
return x + x.transpose(-1, -2)
def average_product_correct(x):
"Perform average product correct, used for contact prediction."
a1 = x.sum(-1, keepdims=True)
a2 = x.sum(-2, keepdims=True)
a12 = x.sum((-1, -2), keepdims=True)
avg = a1 * a2
avg.div_(a12) # in-place to reduce memory
normalized = x - avg
return normalized
@dataclass
class RotaryEmbeddingConfig:
"""
Parameters to initialize the RotaryEmbedding layer. The rescaling factor allows
to adapt the rotary embeddings to larger lengths than what was used for training.
One of this strategy is presented in the Yarn paper: https://arxiv.org/pdf/2309.00071.pdf. # noqa
Args:
"""
rescaling_factor: Optional[float]
class RotaryEmbedding(torch.nn.Module):
"""
Rotary position embeddings based on those in
[RoFormer](https://huggingface.co/docs/transformers/model_doc/roformer). Query and keys are transformed by rotation
matrices which depend on their relative positions.
"""
def __init__(self, dim: int, rotary_embedding_config: RotaryEmbeddingConfig):
super().__init__()
# Extract argument from the config
self.rescaling_factor = rotary_embedding_config.rescaling_factor
self.upper_freq = 10000
self.dim = dim
self._seq_len_cached = None
self._cos_cached = None
self._sin_cached = None
def _compute_cos_sin_tables(self, x, inv_freq, seq_dimension=2):
seq_len = x.shape[seq_dimension]
# Reset the tables if the sequence length has changed,
# or if we're on a new device (possibly due to tracing for instance)
self._seq_len_cached = seq_len
t = torch.arange(x.shape[seq_dimension], device=x.device).type_as(
inv_freq
)
freqs = torch.outer(t, inv_freq)
emb = torch.cat((freqs, freqs), dim=-1).to(x.device)
self._cos_cached = emb.cos()[None, None, :, :]
self._sin_cached = emb.sin()[None, None, :, :]
return self._cos_cached, self._sin_cached
def forward(
self, q: torch.Tensor, k: torch.Tensor
) -> Tuple[torch.Tensor, torch.Tensor]:
if self.rescaling_factor is None:
inv_freq = 1.0 / (self.upper_freq ** (torch.arange(0, self.dim, 2).float() / self.dim))
else:
updated_base = self.upper_freq * (
self.rescaling_factor ** (self.dim / (self.dim - 2))
)
inv_freq = 1.0 / (
updated_base ** (torch.arange(0, self.dim, 2).float() / self.dim)
)
self._cos_cached, self._sin_cached = self._compute_cos_sin_tables(
k, inv_freq, seq_dimension=-2,
)
return (
apply_rotary_pos_emb(q, self._cos_cached, self._sin_cached),
apply_rotary_pos_emb(k, self._cos_cached, self._sin_cached),
)
class EsmContactPredictionHead(nn.Module):
"""Performs symmetrization, apc, and computes a logistic regression on the output features"""
def __init__(
self,
in_features: int,
bias=True,
eos_idx: int = 2,
):
super().__init__()
self.in_features = in_features
self.eos_idx = eos_idx
self.regression = nn.Linear(in_features, 1, bias)
self.activation = nn.Sigmoid()
def forward(self, tokens, attentions):
# remove eos token attentions
eos_mask = tokens.ne(self.eos_idx).to(attentions)
eos_mask = eos_mask.unsqueeze(1) * eos_mask.unsqueeze(2)
attentions = attentions * eos_mask[:, None, None, :, :]
attentions = attentions[..., :-1, :-1]
# remove cls token attentions
attentions = attentions[..., 1:, 1:]
batch_size, layers, heads, seqlen, _ = attentions.size()
attentions = attentions.view(batch_size, layers * heads, seqlen, seqlen)
# features: batch x channels x tokens x tokens (symmetric)
attentions = attentions.to(
self.regression.weight.device
) # attentions always float32, may need to convert to float16
attentions = average_product_correct(symmetrize(attentions))
attentions = attentions.permute(0, 2, 3, 1)
return self.activation(self.regression(attentions).squeeze(3))
class EsmEmbeddings(nn.Module):
"""
Same as BertEmbeddings with a tiny tweak for positional embeddings indexing.
"""
def __init__(self, config):
super().__init__()
self.word_embeddings = nn.Embedding(
config.vocab_size, config.hidden_size, padding_idx=config.pad_token_id
)
if config.emb_layer_norm_before:
self.layer_norm = nn.LayerNorm(
config.hidden_size, eps=config.layer_norm_eps
)
else:
self.layer_norm = None
self.dropout = nn.Dropout(config.hidden_dropout_prob)
# position_ids (1, len position emb) is contiguous in memory and exported when serialized
self.position_embedding_type = getattr(
config, "position_embedding_type", "absolute"
)
self.register_buffer(
"position_ids",
torch.arange(config.max_position_embeddings).expand((1, -1)),
persistent=False,
)
self.padding_idx = config.pad_token_id
self.position_embeddings = nn.Embedding(
config.max_position_embeddings,
config.hidden_size,
padding_idx=self.padding_idx,
)
self.token_dropout = config.token_dropout
self.mask_token_id = config.mask_token_id
def forward(
self,
input_ids=None,
attention_mask=None,
position_ids=None,
inputs_embeds=None,
past_key_values_length=0,
):
if position_ids is None:
if input_ids is not None:
# Create the position ids from the input token ids. Any padded tokens remain padded.
position_ids = create_position_ids_from_input_ids(
input_ids, self.padding_idx, past_key_values_length
)
else:
position_ids = self.create_position_ids_from_inputs_embeds(
inputs_embeds
)
if inputs_embeds is None:
inputs_embeds = self.word_embeddings(input_ids)
# Note that if we want to support ESM-1 (not 1b!) in future then we need to support an
# embedding_scale factor here.
embeddings = inputs_embeds
# Matt: ESM has the option to handle masking in MLM in a slightly unusual way. If the token_dropout
# flag is False then it is handled in the same was as BERT/RoBERTa. If it is set to True, however,
# masked tokens are treated as if they were selected for input dropout and zeroed out.
# This "mask-dropout" is compensated for when masked tokens are not present, by scaling embeddings by
# a factor of (fraction of unmasked tokens during training) / (fraction of unmasked tokens in sample).
# This is analogous to the way that dropout layers scale down outputs during evaluation when not
# actually dropping out values (or, equivalently, scale up their un-dropped outputs in training).
if self.token_dropout:
embeddings.masked_fill_(
(input_ids == self.mask_token_id).unsqueeze(-1), 0.0
)
mask_ratio_train = (
0.15 * 0.8
) # Hardcoded as the ratio used in all ESM model training runs
src_lengths = attention_mask.sum(-1)
mask_ratio_observed = (input_ids == self.mask_token_id).sum(
-1
).float() / src_lengths
embeddings = (
embeddings
* (1 - mask_ratio_train)
/ (1 - mask_ratio_observed)[:, None, None]
).to(embeddings.dtype)
if self.position_embedding_type == "absolute":
position_embeddings = self.position_embeddings(position_ids)
embeddings += position_embeddings
if self.layer_norm is not None:
embeddings = self.layer_norm(embeddings)
if attention_mask is not None:
embeddings = (embeddings * attention_mask.unsqueeze(-1)).to(
embeddings.dtype
)
# Matt: I think this line was copied incorrectly from BERT, disabling it for now.
# embeddings = self.dropout(embeddings)
return embeddings
def create_position_ids_from_inputs_embeds(self, inputs_embeds):
"""
We are provided embeddings directly. We cannot infer which are padded so just generate sequential position ids.
Args:
inputs_embeds: torch.Tensor
Returns: torch.Tensor
"""
input_shape = inputs_embeds.size()[:-1]
sequence_length = input_shape[1]
position_ids = torch.arange(
self.padding_idx + 1,
sequence_length + self.padding_idx + 1,
dtype=torch.long,
device=inputs_embeds.device,
)
return position_ids.unsqueeze(0).expand(input_shape)
class EsmSelfAttention(nn.Module):
def __init__(self, config, position_embedding_type=None):
super().__init__()
if config.hidden_size % config.num_attention_heads != 0 and not hasattr(
config, "embedding_size"
):
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_attention_heads = config.num_attention_heads
self.attention_head_size = int(config.hidden_size / config.num_attention_heads)
self.all_head_size = self.num_attention_heads * self.attention_head_size
self.query = nn.Linear(config.hidden_size, self.all_head_size)
self.key = nn.Linear(config.hidden_size, self.all_head_size)
self.value = nn.Linear(config.hidden_size, self.all_head_size)
self.dropout = nn.Dropout(config.attention_probs_dropout_prob)
self.position_embedding_type = position_embedding_type or getattr(
config, "position_embedding_type", "absolute"
)
self.rotary_embeddings = None
if (
self.position_embedding_type == "relative_key"
or self.position_embedding_type == "relative_key_query"
):
self.max_position_embeddings = config.max_position_embeddings
self.distance_embedding = nn.Embedding(
2 * config.max_position_embeddings - 1, self.attention_head_size
)
elif self.position_embedding_type == "rotary":
# Initiliaze rotary embedding config
rescaling_factor = config.rescaling_factor
rotary_embedding_config = RotaryEmbeddingConfig(rescaling_factor=rescaling_factor)
self.rotary_embeddings = RotaryEmbedding(dim=self.attention_head_size, rotary_embedding_config=rotary_embedding_config)
self.is_decoder = config.is_decoder
def transpose_for_scores(self, x: torch.Tensor) -> torch.Tensor:
new_x_shape = x.size()[:-1] + (
self.num_attention_heads,
self.attention_head_size,
)
x = x.view(new_x_shape)
return x.permute(0, 2, 1, 3)
def forward(
self,
hidden_states: torch.Tensor,
attention_mask: Optional[torch.FloatTensor] = None,
head_mask: Optional[torch.FloatTensor] = None,
encoder_hidden_states: Optional[torch.FloatTensor] = None,
encoder_attention_mask: Optional[torch.FloatTensor] = None,
past_key_value: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
output_attentions: Optional[bool] = False,
) -> Tuple[torch.Tensor]:
mixed_query_layer = self.query(hidden_states)
# If this is instantiated as a cross-attention module, the keys
# and values come from an encoder; the attention mask needs to be
# such that the encoder's padding tokens are not attended to.
is_cross_attention = encoder_hidden_states is not None
if is_cross_attention and past_key_value is not None:
# reuse k,v, cross_attentions
key_layer = past_key_value[0]
value_layer = past_key_value[1]
attention_mask = encoder_attention_mask
elif is_cross_attention:
key_layer = self.transpose_for_scores(self.key(encoder_hidden_states))
value_layer = self.transpose_for_scores(self.value(encoder_hidden_states))
attention_mask = encoder_attention_mask
elif past_key_value is not None:
key_layer = self.transpose_for_scores(self.key(hidden_states))
value_layer = self.transpose_for_scores(self.value(hidden_states))
key_layer = torch.cat([past_key_value[0], key_layer], dim=2)
value_layer = torch.cat([past_key_value[1], value_layer], dim=2)
else:
key_layer = self.transpose_for_scores(self.key(hidden_states))
value_layer = self.transpose_for_scores(self.value(hidden_states))
query_layer = self.transpose_for_scores(mixed_query_layer)
# Matt: Our BERT model (which this code was derived from) scales attention logits down by sqrt(head_dim).
# ESM scales the query down by the same factor instead. Modulo numerical stability these are equivalent,
# but not when rotary embeddings get involved. Therefore, we scale the query here to match the original
# ESM code and fix rotary embeddings.
query_layer = query_layer * self.attention_head_size**-0.5
if self.is_decoder:
# if cross_attention save Tuple(torch.Tensor, torch.Tensor) of all cross attention key/value_states.
# Further calls to cross_attention layer can then reuse all cross-attention
# key/value_states (first "if" case)
# if uni-directional self-attention (decoder) save Tuple(torch.Tensor, torch.Tensor) of
# all previous decoder key/value_states. Further calls to uni-directional self-attention
# can concat previous decoder key/value_states to current projected key/value_states (third "elif" case)
# if encoder bi-directional self-attention `past_key_value` is always `None`
past_key_value = (key_layer, value_layer)
if self.position_embedding_type == "rotary":
query_layer, key_layer = self.rotary_embeddings(query_layer, key_layer)
# Take the dot product between "query" and "key" to get the raw attention scores.
attention_scores = torch.matmul(query_layer, key_layer.transpose(-1, -2))
if (
self.position_embedding_type == "relative_key"
or self.position_embedding_type == "relative_key_query"
):
seq_length = hidden_states.size()[1]
position_ids_l = torch.arange(
seq_length, dtype=torch.long, device=hidden_states.device
).view(-1, 1)
position_ids_r = torch.arange(
seq_length, dtype=torch.long, device=hidden_states.device
).view(1, -1)
distance = position_ids_l - position_ids_r
positional_embedding = self.distance_embedding(
distance + self.max_position_embeddings - 1
)
positional_embedding = positional_embedding.to(
dtype=query_layer.dtype
) # fp16 compatibility
if self.position_embedding_type == "relative_key":
relative_position_scores = torch.einsum(
"bhld,lrd->bhlr", query_layer, positional_embedding
)
attention_scores = attention_scores + relative_position_scores
elif self.position_embedding_type == "relative_key_query":
relative_position_scores_query = torch.einsum(
"bhld,lrd->bhlr", query_layer, positional_embedding
)
relative_position_scores_key = torch.einsum(
"bhrd,lrd->bhlr", key_layer, positional_embedding
)
attention_scores = (
attention_scores
+ relative_position_scores_query
+ relative_position_scores_key
)
if attention_mask is not None:
# Apply the attention mask is (precomputed for all layers in EsmModel forward() function)
attention_scores = attention_scores + attention_mask
# Normalize the attention scores to probabilities.
attention_probs = nn.functional.softmax(attention_scores, dim=-1)
# This is actually dropping out entire tokens to attend to, which might
# seem a bit unusual, but is taken from the original Transformer paper.
attention_probs = self.dropout(attention_probs)
# Mask heads if we want to
if head_mask is not None:
attention_probs = attention_probs * head_mask
context_layer = torch.matmul(attention_probs, value_layer)
context_layer = context_layer.permute(0, 2, 1, 3).contiguous()
new_context_layer_shape = context_layer.size()[:-2] + (self.all_head_size,)
context_layer = context_layer.view(new_context_layer_shape)
outputs = (
(context_layer, attention_probs) if output_attentions else (context_layer,)
)
if self.is_decoder:
outputs = outputs + (past_key_value,)
return outputs
class EsmSelfOutput(nn.Module):
def __init__(self, config):
super().__init__()
self.dense = nn.Linear(config.hidden_size, config.hidden_size)
self.dropout = nn.Dropout(config.hidden_dropout_prob)
def forward(self, hidden_states, input_tensor):
hidden_states = self.dense(hidden_states)
hidden_states = self.dropout(hidden_states)
hidden_states += input_tensor
return hidden_states
class EsmAttention(nn.Module):
def __init__(self, config):
super().__init__()
self.self = EsmSelfAttention(config)
self.output = EsmSelfOutput(config)
self.pruned_heads = set()
self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
def prune_heads(self, heads):
if len(heads) == 0:
return
heads, index = find_pruneable_heads_and_indices(
heads,
self.self.num_attention_heads,
self.self.attention_head_size,
self.pruned_heads,
)
# Prune linear layers
self.self.query = prune_linear_layer(self.self.query, index)
self.self.key = prune_linear_layer(self.self.key, index)
self.self.value = prune_linear_layer(self.self.value, index)
self.output.dense = prune_linear_layer(self.output.dense, index, dim=1)
# Update hyper params and store pruned heads
self.self.num_attention_heads = self.self.num_attention_heads - len(heads)
self.self.all_head_size = (
self.self.attention_head_size * self.self.num_attention_heads
)
self.pruned_heads = self.pruned_heads.union(heads)
def forward(
self,
hidden_states,
attention_mask=None,
head_mask=None,
encoder_hidden_states=None,
encoder_attention_mask=None,
past_key_value=None,
output_attentions=False,
):
hidden_states_ln = self.LayerNorm(hidden_states)
self_outputs = self.self(
hidden_states_ln,
attention_mask,
head_mask,
encoder_hidden_states,
encoder_attention_mask,
past_key_value,
output_attentions,
)
attention_output = self.output(self_outputs[0], hidden_states)
outputs = (attention_output,) + self_outputs[
1:
] # add attentions if we output them
return outputs
class EsmIntermediate(nn.Module):
def __init__(self, config):
super().__init__()
self.dense = nn.Linear(
config.hidden_size,
int(config.intermediate_size * 2),
bias=config.add_bias_fnn,
)
self.activation_fn = SiLU()
def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
hidden_states = self.dense(hidden_states)
# GLU
x1, x2 = hidden_states.split(int(hidden_states.size(-1) / 2), -1)
hidden_states = self.activation_fn(x1) * x2
return hidden_states
class EsmOutput(nn.Module):
def __init__(self, config):
super().__init__()
self.dense = nn.Linear(
config.intermediate_size, config.hidden_size, bias=config.add_bias_fnn
)
self.dropout = nn.Dropout(config.hidden_dropout_prob)
def forward(self, hidden_states, input_tensor):
hidden_states = self.dense(hidden_states)
hidden_states = self.dropout(hidden_states)
hidden_states += input_tensor
return hidden_states
class EsmLayer(nn.Module):
def __init__(self, config):
super().__init__()
self.chunk_size_feed_forward = config.chunk_size_feed_forward
self.seq_len_dim = 1
self.attention = EsmAttention(config)
self.is_decoder = config.is_decoder
self.add_cross_attention = config.add_cross_attention
if self.add_cross_attention:
if not self.is_decoder:
raise RuntimeError(
f"{self} should be used as a decoder model if cross attention is added"
)
self.crossattention = EsmAttention(config)
self.intermediate = EsmIntermediate(config)
self.output = EsmOutput(config)
self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
def forward(
self,
hidden_states,
attention_mask=None,
head_mask=None,
encoder_hidden_states=None,
encoder_attention_mask=None,
past_key_value=None,
output_attentions=False,
):
# decoder uni-directional self-attention cached key/values tuple is at positions 1,2
self_attn_past_key_value = (
past_key_value[:2] if past_key_value is not None else None
)
self_attention_outputs = self.attention(
hidden_states,
attention_mask,
head_mask,
output_attentions=output_attentions,
past_key_value=self_attn_past_key_value,
)
attention_output = self_attention_outputs[0]
# if decoder, the last output is tuple of self-attn cache
if self.is_decoder:
outputs = self_attention_outputs[1:-1]
present_key_value = self_attention_outputs[-1]
else:
outputs = self_attention_outputs[
1:
] # add self attentions if we output attention weights
cross_attn_present_key_value = None
if self.is_decoder and encoder_hidden_states is not None:
if not hasattr(self, "crossattention"):
raise AttributeError(
f"If `encoder_hidden_states` are passed, {self} has to be instantiated"
" with cross-attention layers by setting `config.add_cross_attention=True`"
)
# cross_attn cached key/values tuple is at positions 3,4 of past_key_value tuple
cross_attn_past_key_value = (
past_key_value[-2:] if past_key_value is not None else None
)
cross_attention_outputs = self.crossattention(
attention_output,
attention_mask,
head_mask,
encoder_hidden_states,
encoder_attention_mask,
cross_attn_past_key_value,
output_attentions,
)
attention_output = cross_attention_outputs[0]
outputs = (
outputs + cross_attention_outputs[1:-1]
) # add cross attentions if we output attention weights
# add cross-attn cache to positions 3,4 of present_key_value tuple
cross_attn_present_key_value = cross_attention_outputs[-1]
present_key_value = present_key_value + cross_attn_present_key_value
layer_output = self.feed_forward_chunk(attention_output)
outputs = (layer_output,) + outputs
# if decoder, return the attn key/values as the last output
if self.is_decoder:
outputs = outputs + (present_key_value,)
return outputs
def feed_forward_chunk(self, attention_output):
attention_output_ln = self.LayerNorm(attention_output)
intermediate_output = self.intermediate(attention_output_ln)
layer_output = self.output(intermediate_output, attention_output)
return layer_output
class EsmEncoder(nn.Module):
def __init__(self, config):
super().__init__()
self.config = config
self.layer = nn.ModuleList(
[EsmLayer(config) for _ in range(config.num_hidden_layers)]
)
self.emb_layer_norm_after = nn.LayerNorm(
config.hidden_size, eps=config.layer_norm_eps
)
self.gradient_checkpointing = False
def forward(
self,
hidden_states,
attention_mask=None,
head_mask=None,
encoder_hidden_states=None,
encoder_attention_mask=None,
past_key_values=None,
use_cache=None,
output_attentions=False,
output_hidden_states=False,
return_dict=True,
):
if self.gradient_checkpointing and self.training:
if use_cache:
logger.warning_once(
"`use_cache=True` is incompatible with `config.gradient_checkpointing=True`. Setting "
"`use_cache=False`..."
)
use_cache = False
all_hidden_states = () if output_hidden_states else None
all_self_attentions = () if output_attentions else None
all_cross_attentions = (
() if output_attentions and self.config.add_cross_attention else None
)
next_decoder_cache = () if use_cache else None
for i, layer_module in enumerate(self.layer):
if output_hidden_states:
all_hidden_states = all_hidden_states + (hidden_states,)
layer_head_mask = head_mask[i] if head_mask is not None else None
past_key_value = past_key_values[i] if past_key_values is not None else None
if self.gradient_checkpointing and self.training:
def create_custom_forward(module):
def custom_forward(*inputs):
return module(*inputs, past_key_value, output_attentions)
return custom_forward
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(layer_module),
hidden_states,
attention_mask,
layer_head_mask,
encoder_hidden_states,
encoder_attention_mask,
)
else:
layer_outputs = layer_module(
hidden_states,
attention_mask,
layer_head_mask,
encoder_hidden_states,
encoder_attention_mask,
past_key_value,
output_attentions,
)
hidden_states = layer_outputs[0]
if use_cache:
next_decoder_cache += (layer_outputs[-1],)
if output_attentions:
all_self_attentions = all_self_attentions + (layer_outputs[1],)
if self.config.add_cross_attention:
all_cross_attentions = all_cross_attentions + (layer_outputs[2],)
if self.emb_layer_norm_after:
hidden_states = self.emb_layer_norm_after(hidden_states)
if output_hidden_states:
all_hidden_states = all_hidden_states + (hidden_states,)
if not return_dict:
return tuple(
v
for v in [
hidden_states,
next_decoder_cache,
all_hidden_states,
all_self_attentions,
all_cross_attentions,
]
if v is not None
)
return BaseModelOutputWithPastAndCrossAttentions(
last_hidden_state=hidden_states,
past_key_values=next_decoder_cache,
hidden_states=all_hidden_states,
attentions=all_self_attentions,
cross_attentions=all_cross_attentions,
)
# Copied from transformers.models.bert.modeling_bert.BertPooler
class EsmPooler(nn.Module):
def __init__(self, config):
super().__init__()
self.dense = nn.Linear(config.hidden_size, config.hidden_size)
self.activation = nn.Tanh()
def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
# We "pool" the model by simply taking the hidden state corresponding
# to the first token.
first_token_tensor = hidden_states[:, 0]
pooled_output = self.dense(first_token_tensor)
pooled_output = self.activation(pooled_output)
return pooled_output
class EsmPreTrainedModel(PreTrainedModel):
"""
An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
models.
"""
config_class = SegmentNTConfig
base_model_prefix = "esm"
_no_split_modules = ["EsmLayer", "EsmFoldTriangularSelfAttentionBlock"]
# Copied from transformers.models.bert.modeling_bert.BertPreTrainedModel._init_weights
def _init_weights(self, module):
"""Initialize the weights"""
if isinstance(module, nn.Linear):
# Slightly different from the TF version which uses truncated_normal for initialization
# cf https://github.com/pytorch/pytorch/pull/5617
module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
if module.bias is not None:
module.bias.data.zero_()
elif isinstance(module, nn.Embedding):
module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
if module.padding_idx is not None:
module.weight.data[module.padding_idx].zero_()
elif isinstance(module, nn.LayerNorm):
module.bias.data.zero_()
module.weight.data.fill_(1.0)
ESM_START_DOCSTRING = r"""
This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the
library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
etc.)
This model is also a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass.
Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage
and behavior.
Parameters:
config ([`EsmConfig`]): Model configuration class with all the parameters of the
model. Initializing with a config file does not load the weights associated with the model, only the
configuration. Check out the [`~PreTrainedModel.from_pretrained`] method to load the model weights.
"""
ESM_INPUTS_DOCSTRING = r"""
Args:
input_ids (`torch.LongTensor` of shape `({0})`):
Indices of input sequence tokens in the vocabulary.
Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
[`PreTrainedTokenizer.__call__`] for details.
[What are input IDs?](../glossary#input-ids)
attention_mask (`torch.FloatTensor` of shape `({0})`, *optional*):
Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
- 1 for tokens that are **not masked**,
- 0 for tokens that are **masked**.
[What are attention masks?](../glossary#attention-mask)
position_ids (`torch.LongTensor` of shape `({0})`, *optional*):
Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0,
config.max_position_embeddings - 1]`.
[What are position IDs?](../glossary#position-ids)
head_mask (`torch.FloatTensor` of shape `(num_heads,)` or `(num_layers, num_heads)`, *optional*):
Mask to nullify selected heads of the self-attention modules. Mask values selected in `[0, 1]`:
- 1 indicates the head is **not masked**,
- 0 indicates the head is **masked**.
inputs_embeds (`torch.FloatTensor` of shape `({0}, hidden_size)`, *optional*):
Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This
is useful if you want more control over how to convert `input_ids` indices into associated vectors than the
model's internal embedding lookup matrix.
output_attentions (`bool`, *optional*):
Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
tensors for more detail.
output_hidden_states (`bool`, *optional*):
Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
more detail.
return_dict (`bool`, *optional*):
Whether or not to return a [`~file_utils.ModelOutput`] instead of a plain tuple.
"""
@add_start_docstrings(
"The bare ESM Model transformer outputting raw hidden-states without any specific head on top.",
ESM_START_DOCSTRING,
)
class EsmModel(EsmPreTrainedModel):
"""
The model can behave as an encoder (with only self-attention) as well as a decoder, in which case a layer of
cross-attention is added between the self-attention layers, following the architecture described in [Attention is
all you need](https://arxiv.org/abs/1706.03762) by Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit,
Llion Jones, Aidan N. Gomez, Lukasz Kaiser and Illia Polosukhin.
To behave as an decoder the model needs to be initialized with the `is_decoder` argument of the configuration set
to `True`. To be used in a Seq2Seq model, the model needs to initialized with both `is_decoder` argument and
`add_cross_attention` set to `True`; an `encoder_hidden_states` is then expected as an input to the forward pass.
"""
supports_gradient_checkpointing = False
def __init__(self, config, add_pooling_layer=True):
super().__init__(config)
self.config = config
self.embeddings = EsmEmbeddings(config)
self.encoder = EsmEncoder(config)
self.pooler = EsmPooler(config) if add_pooling_layer else None
self.contact_head = EsmContactPredictionHead(
in_features=config.num_hidden_layers * config.num_attention_heads, bias=True
)
# Initialize weights and apply final processing
self.post_init()
def _set_gradient_checkpointing(self, module, value=False):
if isinstance(module, EsmEncoder):
module.gradient_checkpointing = value
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)
@add_start_docstrings_to_model_forward(
ESM_INPUTS_DOCSTRING.format("(batch_size, sequence_length)")
)
@add_code_sample_docstrings(
checkpoint=_CHECKPOINT_FOR_DOC,
output_type=BaseModelOutputWithPoolingAndCrossAttentions,
config_class=_CONFIG_FOR_DOC,
)
def forward(
self,
input_ids: Optional[torch.Tensor] = None,
attention_mask: Optional[torch.Tensor] = None,
position_ids: Optional[torch.Tensor] = None,
head_mask: Optional[torch.Tensor] = None,
inputs_embeds: Optional[torch.Tensor] = None,
encoder_hidden_states: Optional[torch.Tensor] = None,
encoder_attention_mask: Optional[torch.Tensor] = None,
past_key_values: Optional[List[torch.FloatTensor]] = None,
use_cache: Optional[bool] = None,
output_attentions: Optional[bool] = None,
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
) -> Union[Tuple[torch.Tensor], BaseModelOutputWithPoolingAndCrossAttentions]:
r"""
encoder_hidden_states (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention if
the model is configured as a decoder.
encoder_attention_mask (`torch.FloatTensor` of shape `(batch_size, sequence_length)`, *optional*):
Mask to avoid performing attention on the padding token indices of the encoder input. This mask is used in
the cross-attention if the model is configured as a decoder. Mask values selected in `[0, 1]`:
- 1 for tokens that are **not masked**,
- 0 for tokens that are **masked**.
past_key_values (`tuple(tuple(torch.FloatTensor))` of length `config.n_layers` with each tuple having 4 tensors of shape `(batch_size, num_heads, sequence_length - 1, embed_size_per_head)`):
Contains precomputed key and value hidden states of the attention blocks. Can be used to speed up decoding.
If `past_key_values` are used, the user can optionally input only the last `decoder_input_ids` (those that
don't have their past key value states given to this model) of shape `(batch_size, 1)` instead of all
`decoder_input_ids` of shape `(batch_size, sequence_length)`.
use_cache (`bool`, *optional*):
If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see
`past_key_values`).
"""
output_attentions = (
output_attentions
if output_attentions is not None
else self.config.output_attentions
)
output_hidden_states = (
output_hidden_states
if output_hidden_states is not None
else self.config.output_hidden_states
)
return_dict = (
return_dict if return_dict is not None else self.config.use_return_dict
)
if self.config.is_decoder:
use_cache = use_cache if use_cache is not None else self.config.use_cache
else:
use_cache = False
if input_ids is not None and inputs_embeds is not None:
raise ValueError(
"You cannot specify both input_ids and inputs_embeds at the same time"
)
elif input_ids is not None:
input_shape = input_ids.size()
elif inputs_embeds is not None:
input_shape = inputs_embeds.size()[:-1]
else:
raise ValueError("You have to specify either input_ids or inputs_embeds")
batch_size, seq_length = input_shape
device = input_ids.device if input_ids is not None else inputs_embeds.device
# past_key_values_length
past_key_values_length = (
past_key_values[0][0].shape[2] if past_key_values is not None else 0
)
if attention_mask is None:
attention_mask = torch.ones(
((batch_size, seq_length + past_key_values_length)), device=device
)
# We can provide a self-attention mask of dimensions [batch_size, from_seq_length, to_seq_length]
# ourselves in which case we just need to make it broadcastable to all heads.
extended_attention_mask: torch.Tensor = self.get_extended_attention_mask(
attention_mask, input_shape
)
# If a 2D or 3D attention mask is provided for the cross-attention
# we need to make broadcastable to [batch_size, num_heads, seq_length, seq_length]
if self.config.is_decoder and encoder_hidden_states is not None:
(
encoder_batch_size,
encoder_sequence_length,
_,
) = encoder_hidden_states.size()
encoder_hidden_shape = (encoder_batch_size, encoder_sequence_length)
if encoder_attention_mask is None:
encoder_attention_mask = torch.ones(encoder_hidden_shape, device=device)
encoder_extended_attention_mask = self.invert_attention_mask(
encoder_attention_mask
)
else:
encoder_extended_attention_mask = None
# Prepare head mask if needed
# 1.0 in head_mask indicate we keep the head
# attention_probs has shape bsz x n_heads x N x N
# input head_mask has shape [num_heads] or [num_hidden_layers x num_heads]
# and head_mask is converted to shape [num_hidden_layers x batch x num_heads x seq_length x seq_length]
head_mask = self.get_head_mask(head_mask, self.config.num_hidden_layers)
embedding_output = self.embeddings(
input_ids=input_ids,
position_ids=position_ids,
attention_mask=attention_mask,
inputs_embeds=inputs_embeds,
past_key_values_length=past_key_values_length,
)
encoder_outputs = self.encoder(
embedding_output,
attention_mask=extended_attention_mask,
head_mask=head_mask,
encoder_hidden_states=encoder_hidden_states,
encoder_attention_mask=encoder_extended_attention_mask,
past_key_values=past_key_values,
use_cache=use_cache,
output_attentions=output_attentions,
output_hidden_states=output_hidden_states,
return_dict=return_dict,
)
sequence_output = encoder_outputs[0]
pooled_output = (
self.pooler(sequence_output) if self.pooler is not None else None
)
if not return_dict:
return (sequence_output, pooled_output) + encoder_outputs[1:]
return BaseModelOutputWithPoolingAndCrossAttentions(
last_hidden_state=sequence_output,
pooler_output=pooled_output,
past_key_values=encoder_outputs.past_key_values,
hidden_states=encoder_outputs.hidden_states,
attentions=encoder_outputs.attentions,
cross_attentions=encoder_outputs.cross_attentions,
)
def predict_contacts(self, tokens, attention_mask):
attns = self(
tokens,
attention_mask=attention_mask,
return_dict=True,
output_attentions=True,
).attentions
attns = torch.stack(attns, dim=1) # Matches the original model layout
# In the original model, attentions for padding tokens are completely zeroed out.
# This makes no difference most of the time because the other tokens won't attend to them,
# but it does for the contact prediction task, which takes attentions as input,
# so we have to mimic that here.
attns *= attention_mask.unsqueeze(1).unsqueeze(2).unsqueeze(3)
attns *= attention_mask.unsqueeze(1).unsqueeze(2).unsqueeze(4)
return self.contact_head(tokens, attns)
def create_position_ids_from_input_ids(
input_ids, padding_idx, past_key_values_length=0
):
"""
Replace non-padding symbols with their position numbers. Position numbers begin at padding_idx+1. Padding symbols
are ignored. This is modified from fairseq's `utils.make_positions`.
Args:
x: torch.Tensor x:
Returns: torch.Tensor
"""
# The series of casts and type-conversions here are carefully balanced to both work with ONNX export and XLA.
mask = input_ids.ne(padding_idx).int()
incremental_indices = (
torch.cumsum(mask, dim=1).type_as(mask) + past_key_values_length
) * mask
return incremental_indices.long() + padding_idx
class SegmentNT(EsmPreTrainedModel):
def __init__(self, config):
super().__init__(config)
self.num_labels = config.num_labels
self.config = config
self.num_features = len(config.features)
self.esm = EsmModel(config, add_pooling_layer=False)
embed_dim = config.hidden_size
num_layers = config.num_layers_head
self.unet = UNET1DSegmentationHead(
embed_dim=embed_dim,
num_classes=embed_dim // 2,
output_channels_list=tuple(
embed_dim * (2**i) for i in range(num_layers)
),
)
self.fc = nn.Linear(in_features=embed_dim, out_features=6 * 2 * self.num_features)
self.activation_fn = nn.SiLU()
self.init_weights()
# @add_start_docstrings_to_model_forward(
# ESM_INPUTS_DOCSTRING.format("batch_size, sequence_length")
# )
# @add_code_sample_docstrings(
# checkpoint=_CHECKPOINT_FOR_DOC,
# output_type=SequenceClassifierOutput,
# config_class=_CONFIG_FOR_DOC,
# )
def forward(
self,
input_ids: Optional[torch.LongTensor] = None,
attention_mask: Optional[torch.Tensor] = None,
position_ids: Optional[torch.LongTensor] = None,
head_mask: Optional[torch.Tensor] = None,
inputs_embeds: Optional[torch.FloatTensor] = None,
labels: Optional[torch.LongTensor] = None,
output_attentions: Optional[bool] = None,
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
) -> Union[Tuple, SequenceClassifierOutput]:
r"""
labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
`config.num_labels > 1` a classification loss is computed (Cross-Entropy).
"""
return_dict = (
return_dict if return_dict is not None else self.config.use_return_dict
)
outputs = self.esm(
input_ids,
attention_mask=attention_mask,
position_ids=position_ids,
head_mask=head_mask,
inputs_embeds=inputs_embeds,
output_attentions=output_attentions,
output_hidden_states=output_hidden_states,
return_dict=return_dict,
)
sequence_output = outputs[0]
# Remove CLS token
sequence_output = sequence_output[:,1:,:]
# Invert the channels and sequence length channel
sequence_output = torch.transpose(sequence_output, 2,1)
x = self.activation_fn(self.unet(sequence_output))
# Invert the channels and sequence length channel
x = torch.transpose(x, 2,1)
logits = self.fc(x)
# Final reshape to have logits per nucleotides, per feature
logits = torch.reshape(logits, (x.shape[0], x.shape[1] * 6, self.num_features, 2))
# Add logits to the ESM outputs
outputs["logits"] = logits
return outputs
class DownSample1D(nn.Module):
"""
1D-UNET downsampling block.
"""
def __init__(
self,
input_channels: int,
output_channels: int,
num_layers: int = 2,
):
"""
Args:
output_channels: number of output channels.
activation_fn: name of the activation function to use.
Should be one of "gelu",
"gelu-no-approx", "relu", "swish", "silu", "sin".
num_layers: number of convolution layers.
name: module name.
"""
super().__init__()
self.first_layer = [nn.Conv1d(
in_channels=input_channels,
out_channels=output_channels,
kernel_size=3,
stride=1,
dilation=1,
padding="same",
)]
self.next_layers = [
nn.Conv1d(
in_channels=output_channels,
out_channels=output_channels,
kernel_size=3,
stride=1,
dilation=1,
padding="same",
)
for _ in range(num_layers-1)
]
self.conv_layers = nn.ModuleList(self.first_layer + self.next_layers)
self.avg_pool = nn.AvgPool1d(
kernel_size=2,
stride=2,
padding=0,
)
self.activation_fn = nn.SiLU()
def forward(self, x: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:
for i, conv_layer in enumerate(self.conv_layers):
x = self.activation_fn(conv_layer(x))
hidden = x
x = self.avg_pool(hidden)
return x, hidden
class UpSample1D(nn.Module):
"""
1D-UNET upsampling block.
"""
def __init__(
self,
input_channels: int,
output_channels: int,
num_layers: int = 2,
):
"""
Args:
output_channels: number of output channels.
activation_fn: name of the activation function to use.
Should be one of "gelu",
"gelu-no-approx", "relu", "swish", "silu", "sin".
interpolation_method: Method to be used for upsampling interpolation.
Should be one of "nearest", "linear", "cubic", "lanczos3", "lanczos5".
num_layers: number of convolution layers.
name: module name.
"""
super().__init__()
self._first_layer = [nn.ConvTranspose1d(
in_channels=input_channels,
out_channels=output_channels,
kernel_size=3,
stride=1,
padding=1,
)]
self._next_layers = [
nn.ConvTranspose1d(
in_channels=output_channels,
out_channels=output_channels,
kernel_size=3,
stride=1,
padding=1,
)
for _ in range(num_layers-1)
]
self.conv_layers = nn.ModuleList(self._first_layer + self._next_layers)
self._activation_fn = nn.SiLU()
def forward(self, x: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:
for i, conv_layer in enumerate(self.conv_layers):
x = self._activation_fn(conv_layer(x))
# Different order than in Haiku because the channels are changed when going
# from Haiku to Torch.
x = nn.functional.interpolate(x, size=2 * x.shape[2], mode="nearest")
return x
class FinalConv1D(nn.Module):
"""
Final output block of the 1D-UNET.
"""
def __init__(
self,
input_channels: int,
output_channels: int,
num_layers: int = 2,
):
"""
Args:
output_channels: number of output channels.
activation_fn: name of the activation function to use.
Should be one of "gelu",
"gelu-no-approx", "relu", "swish", "silu", "sin".
num_layers: number of convolution layers.
name: module name.
"""
super().__init__()
self._first_layer = [nn.Conv1d(
in_channels=input_channels,
out_channels=output_channels,
kernel_size=3,
stride=1,
dilation=1,
padding="same",
)]
self._next_layers = [
nn.Conv1d(
in_channels=output_channels,
out_channels=output_channels,
kernel_size=3,
stride=1,
dilation=1,
padding="same",
)
for _ in range(num_layers-1)
]
self.conv_layers = nn.ModuleList(self._first_layer + self._next_layers)
self._activation_fn = nn.SiLU()
def forward(self, x: torch.Tensor) -> torch.Tensor:
for i, conv_layer in enumerate(self.conv_layers):
x = conv_layer(x)
if i < len(self.conv_layers) - 1:
x = self._activation_fn(x)
return x
class UNET1DSegmentationHead(nn.Module):
"""
1D-UNET based head to be plugged on top of a pretrained model to perform
semantic segmentation.
"""
def __init__(
self,
embed_dim: int,
num_classes: int,
output_channels_list: Tuple[int, ...] = (64, 128, 256),
num_conv_layers_per_block: int = 2,
):
"""
Args:
num_classes: number of classes to segment
output_channels_list: list of the number of output channel at each level of
the UNET
num_conv_layers_per_block: number of convolution layers per block.
"""
super().__init__()
self._num_pooling_layers = len(output_channels_list)
downsample_input_channels_list = (embed_dim, ) + output_channels_list[:-1]
output_channels_list_reversed = tuple(reversed(output_channels_list))
upsample_input_channels_list = (output_channels_list[-1],) + output_channels_list_reversed
upsample_output_channels_list = output_channels_list_reversed
self._downsample_blocks = nn.ModuleList([
DownSample1D(
input_channels= input_channels,
output_channels=output_channels,
num_layers=num_conv_layers_per_block,
)
for input_channels, output_channels in zip(downsample_input_channels_list, output_channels_list)
])
self._upsample_blocks = nn.ModuleList([
UpSample1D(
input_channels = input_channels,
output_channels=output_channels,
num_layers=num_conv_layers_per_block,
)
for input_channels, output_channels in zip(upsample_input_channels_list, upsample_output_channels_list)
])
self.final_block = FinalConv1D(
input_channels=output_channels_list[0],
output_channels=num_classes * 2,
num_layers=num_conv_layers_per_block,
)
def forward(self, x: torch.Tensor) -> torch.Tensor:
if x.shape[2] % 2**self._num_pooling_layers:
raise ValueError(
"Input length must be divisible by the 2 to the power of"
" number of poolign layers."
)
hiddens = []
for downsample_block in self._downsample_blocks:
x, hidden = downsample_block(x)
hiddens.append(hidden)
for i, (upsample_block, hidden) in enumerate(zip(self._upsample_blocks, reversed(hiddens))):
x = upsample_block(x) + hidden
x = self.final_block(x)
return x
|