Spaces:
Sleeping
Sleeping
File size: 62,116 Bytes
1ce5e18 |
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 |
# coding=utf-8
# Copyright 2022 The Salesforce Team Authors and The HuggingFace 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 BLIP model."""
import warnings
from dataclasses import dataclass
from typing import Any, Optional, Tuple, Union
import torch
import torch.utils.checkpoint
from torch import nn
from torch.nn.functional import normalize
from ...activations import ACT2FN
from ...modeling_outputs import BaseModelOutput, BaseModelOutputWithPooling
from ...modeling_utils import PreTrainedModel
from ...utils import (
ModelOutput,
add_start_docstrings,
add_start_docstrings_to_model_forward,
logging,
replace_return_docstrings,
)
from .configuration_blip import BlipConfig, BlipTextConfig, BlipVisionConfig
from .modeling_blip_text import BlipTextLMHeadModel, BlipTextModel
logger = logging.get_logger(__name__)
_CHECKPOINT_FOR_DOC = "Salesforce/blip-vqa-base"
BLIP_PRETRAINED_MODEL_ARCHIVE_LIST = [
"Salesforce/blip-vqa-base",
"Salesforce/blip-vqa-capfilt-large",
"Salesforce/blip-image-captioning-base",
"Salesforce/blip-image-captioning-large",
"Salesforce/blip-itm-base-coco",
"Salesforce/blip-itm-large-coco",
"Salesforce/blip-itm-base-flickr",
"Salesforce/blip-itm-large-flickr",
# See all BLIP models at https://huggingface.co/models?filter=blip
]
# Copied from transformers.models.clip.modeling_clip.contrastive_loss
def contrastive_loss(logits: torch.Tensor) -> torch.Tensor:
return nn.functional.cross_entropy(logits, torch.arange(len(logits), device=logits.device))
# Copied from transformers.models.clip.modeling_clip.clip_loss with clip->blip
def blip_loss(similarity: torch.Tensor) -> torch.Tensor:
caption_loss = contrastive_loss(similarity)
image_loss = contrastive_loss(similarity.t())
return (caption_loss + image_loss) / 2.0
@dataclass
class BlipForConditionalGenerationModelOutput(ModelOutput):
"""
Adapted from the base class for vision model's outputs that also contains image embeddings of the pooling of the
last hidden states. This class also adds the loss term from the text decoder.
Args:
loss (`torch.FloatTensor`, *optional*, returned when `labels` is provided, `torch.FloatTensor` of shape `(1,)`):
Languge modeling loss from the text decoder.
logits (`torch.FloatTensor` of shape `(batch_size, sequence_length, config.vocab_size)`, *optional*):
Prediction scores of the language modeling head of the text decoder model.
image_embeds (`torch.FloatTensor` of shape `(batch_size, output_dim)`, *optional*):
The image embeddings obtained after applying the Vision Transformer model to the input image.
last_hidden_state (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
Sequence of hidden-states at the output of the last layer of the model.
hidden_states (`tuple(torch.FloatTensor)`, *optional*, returned when `output_hidden_states=True`):
Tuple of `torch.FloatTensor` (one for the output of the embeddings, if the model has an embedding layer, +
one for the output of each layer) of shape `(batch_size, sequence_length, hidden_size)`.
Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.
attentions (`tuple(torch.FloatTensor)`, *optional*, returned when `output_attentions=True` is passed):
Tuple of `torch.FloatTensor` (one for each layer) of shape `(batch_size, num_heads, sequence_length,
sequence_length)`.
Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
heads.
"""
loss: Optional[Tuple[torch.FloatTensor]] = None
logits: Optional[Tuple[torch.FloatTensor]] = None
image_embeds: Optional[torch.FloatTensor] = None
last_hidden_state: torch.FloatTensor = None
hidden_states: Optional[Tuple[torch.FloatTensor]] = None
attentions: Optional[Tuple[torch.FloatTensor]] = None
@property
def decoder_logits(self):
warnings.warn(
"`decoder_logits` attribute is deprecated and will be removed in version 5 of Transformers."
" Please use the `logits` attribute to retrieve the final output instead.",
FutureWarning,
)
return self.logits
@dataclass
class BlipTextVisionModelOutput(ModelOutput):
"""
Adapted from the base class for vision model's outputs that also contains image embeddings of the pooling of the
last hidden states. This class also adds the loss term from the text decoder.
Args:
loss (`torch.FloatTensor` of shape `(1,)`, *optional*, returned when `labels` is provided):
Languge modeling loss from the text decoder.
image_embeds (`torch.FloatTensor` of shape `(batch_size, output_dim)` *optional* returned when model is initialized with `with_projection=True`):
The image embeddings obtained by applying the projection layer to the pooler_output.
last_hidden_state (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`):
Sequence of hidden-states at the output of the last layer of the model.
hidden_states (`tuple(torch.FloatTensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`):
Tuple of `torch.FloatTensor` (one for the output of the embeddings, if the model has an embedding layer, +
one for the output of each layer) of shape `(batch_size, sequence_length, hidden_size)`.
Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.
attentions (`tuple(torch.FloatTensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`):
Tuple of `torch.FloatTensor` (one for each layer) of shape `(batch_size, num_heads, sequence_length,
sequence_length)`.
Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
heads.
"""
loss: Optional[torch.FloatTensor] = None
image_embeds: Optional[torch.FloatTensor] = None
last_hidden_state: torch.FloatTensor = None
hidden_states: Optional[Tuple[torch.FloatTensor]] = None
attentions: Optional[Tuple[torch.FloatTensor]] = None
@dataclass
class BlipImageTextMatchingModelOutput(ModelOutput):
"""
Adapted from the base class for vision model's outputs that also contains image embeddings of the pooling of the
last hidden states. This class also adds the loss term from the text decoder as well as the image-text similarity
scores.
Args:
itm_score (`torch.FloatTensor`):
The image-text similarity scores.
loss (`torch.FloatTensor` of shape `(1,)`, *optional*, returned when `labels` is provided):
Languge modeling loss from the text decoder.
image_embeds (`torch.FloatTensor` of shape `(batch_size, output_dim)` *optional* returned when model is initialized with `with_projection=True`):
The image embeddings obtained by applying the projection layer to the pooler_output.
last_hidden_state (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`):
Sequence of hidden-states at the output of the last layer of the model.
hidden_states (`tuple(torch.FloatTensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`):
Tuple of `torch.FloatTensor` (one for the output of the embeddings, if the model has an embedding layer, +
one for the output of each layer) of shape `(batch_size, sequence_length, hidden_size)`.
Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.
vision_pooler_output (`torch.FloatTensor` of shape `(batch_size, hidden_size)`, *optional*):
Last layer hidden-state of the vision of the vision-only branch of the model.
attentions (`tuple(torch.FloatTensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`):
Tuple of `torch.FloatTensor` (one for each layer) of shape `(batch_size, num_heads, sequence_length,
sequence_length)`.
Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
heads.
question_embeds (`torch.FloatTensor`):
The question embeddings obtained by the text projection layer.
"""
itm_score: Optional[torch.FloatTensor] = None
loss: Optional[torch.FloatTensor] = None
image_embeds: Optional[torch.FloatTensor] = None
last_hidden_state: torch.FloatTensor = None
hidden_states: Optional[Tuple[torch.FloatTensor]] = None
vision_pooler_output: Optional[torch.FloatTensor] = None
attentions: Optional[Tuple[torch.FloatTensor]] = None
question_embeds: Optional[Tuple[torch.FloatTensor]] = None
@dataclass
class BlipOutput(ModelOutput):
"""
Args:
loss (`torch.FloatTensor` of shape `(1,)`, *optional*, returned when `return_loss` is `True`):
Contrastive loss for image-text similarity.
logits_per_image:(`torch.FloatTensor` of shape `(image_batch_size, text_batch_size)`):
The scaled dot product scores between `image_embeds` and `text_embeds`. This represents the image-text
similarity scores.
logits_per_text:(`torch.FloatTensor` of shape `(text_batch_size, image_batch_size)`):
The scaled dot product scores between `text_embeds` and `image_embeds`. This represents the text-image
similarity scores.
text_embeds(`torch.FloatTensor` of shape `(batch_size, output_dim`):
The text embeddings obtained by applying the projection layer to the pooled output of [`BlipTextModel`].
image_embeds(`torch.FloatTensor` of shape `(batch_size, output_dim`):
The image embeddings obtained by applying the projection layer to the pooled output of [`BlipVisionModel`].
text_model_output(`BaseModelOutputWithPooling`):
The output of the [`BlipTextModel`].
vision_model_output(`BaseModelOutputWithPooling`):
The output of the [`BlipVisionModel`].
"""
loss: Optional[torch.FloatTensor] = None
logits_per_image: torch.FloatTensor = None
logits_per_text: torch.FloatTensor = None
text_embeds: torch.FloatTensor = None
image_embeds: torch.FloatTensor = None
text_model_output: BaseModelOutputWithPooling = None
vision_model_output: BaseModelOutputWithPooling = None
def to_tuple(self) -> Tuple[Any]:
return tuple(
self[k] if k not in ["text_model_output", "vision_model_output"] else getattr(self, k).to_tuple()
for k in self.keys()
)
class BlipVisionEmbeddings(nn.Module):
def __init__(self, config: BlipVisionConfig):
super().__init__()
self.config = config
self.embed_dim = config.hidden_size
self.image_size = config.image_size
self.patch_size = config.patch_size
self.class_embedding = nn.Parameter(torch.randn(1, 1, self.embed_dim))
self.patch_embedding = nn.Conv2d(
in_channels=3, out_channels=self.embed_dim, kernel_size=self.patch_size, stride=self.patch_size
)
self.num_patches = (self.image_size // self.patch_size) ** 2
self.num_positions = self.num_patches + 1
self.position_embedding = nn.Parameter(torch.randn(1, self.num_positions, self.embed_dim))
def forward(self, pixel_values: torch.FloatTensor) -> torch.Tensor:
batch_size = pixel_values.shape[0]
target_dtype = self.patch_embedding.weight.dtype
patch_embeds = self.patch_embedding(pixel_values.to(dtype=target_dtype)) # shape = [*, width, grid, grid]
patch_embeds = patch_embeds.flatten(2).transpose(1, 2)
class_embeds = self.class_embedding.expand(batch_size, 1, -1).to(target_dtype)
embeddings = torch.cat([class_embeds, patch_embeds], dim=1)
embeddings = embeddings + self.position_embedding[:, : embeddings.size(1), :].to(target_dtype)
return embeddings
# Copied from transformers.models.clip.modeling_clip.CLIPTextEmbeddings with CLIP->Blip
class BlipTextEmbeddings(nn.Module):
def __init__(self, config: BlipTextConfig):
super().__init__()
embed_dim = config.hidden_size
self.token_embedding = nn.Embedding(config.vocab_size, embed_dim)
self.position_embedding = nn.Embedding(config.max_position_embeddings, embed_dim)
# position_ids (1, len position emb) is contiguous in memory and exported when serialized
self.register_buffer(
"position_ids", torch.arange(config.max_position_embeddings).expand((1, -1)), persistent=False
)
def forward(
self,
input_ids: Optional[torch.LongTensor] = None,
position_ids: Optional[torch.LongTensor] = None,
inputs_embeds: Optional[torch.FloatTensor] = None,
) -> torch.Tensor:
seq_length = input_ids.shape[-1] if input_ids is not None else inputs_embeds.shape[-2]
if position_ids is None:
position_ids = self.position_ids[:, :seq_length]
if inputs_embeds is None:
inputs_embeds = self.token_embedding(input_ids)
position_embeddings = self.position_embedding(position_ids)
embeddings = inputs_embeds + position_embeddings
return embeddings
class BlipAttention(nn.Module):
"""Multi-headed attention from 'Attention Is All You Need' paper"""
def __init__(self, config):
super().__init__()
self.config = config
self.embed_dim = config.hidden_size
self.num_heads = config.num_attention_heads
self.head_dim = self.embed_dim // self.num_heads
if self.head_dim * self.num_heads != self.embed_dim:
raise ValueError(
f"embed_dim must be divisible by num_heads (got `embed_dim`: {self.embed_dim} and `num_heads`:"
f" {self.num_heads})."
)
self.scale = self.head_dim**-0.5
self.dropout = nn.Dropout(config.attention_dropout)
self.qkv = nn.Linear(self.embed_dim, 3 * self.embed_dim)
self.projection = nn.Linear(self.embed_dim, self.embed_dim)
def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int):
return tensor.view(bsz, seq_len, self.num_heads, self.head_dim).transpose(1, 2).contiguous()
def forward(
self,
hidden_states: torch.Tensor,
head_mask: Optional[torch.Tensor] = None,
output_attentions: Optional[bool] = False,
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
"""Input shape: Batch x Time x Channel"""
bsz, tgt_len, embed_dim = hidden_states.size()
mixed_qkv = (
self.qkv(hidden_states)
.reshape(bsz, tgt_len, 3, self.num_heads, embed_dim // self.num_heads)
.permute(2, 0, 3, 1, 4)
)
query_states, key_states, value_states = mixed_qkv[0], mixed_qkv[1], mixed_qkv[2]
# Take the dot product between "query" and "key" to get the raw attention scores.
attention_scores = torch.matmul(query_states, key_states.transpose(-1, -2))
attention_scores = attention_scores * self.scale
# 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_states).permute(0, 2, 1, 3)
new_context_layer_shape = context_layer.size()[:-2] + (self.embed_dim,)
context_layer = context_layer.reshape(new_context_layer_shape)
output = self.projection(context_layer)
outputs = (output, attention_probs) if output_attentions else (output, None)
return outputs
# Copied from transformers.models.clip.modeling_clip.CLIPMLP with CLIP->Blip
class BlipMLP(nn.Module):
def __init__(self, config):
super().__init__()
self.config = config
self.activation_fn = ACT2FN[config.hidden_act]
self.fc1 = nn.Linear(config.hidden_size, config.intermediate_size)
self.fc2 = nn.Linear(config.intermediate_size, config.hidden_size)
def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
hidden_states = self.fc1(hidden_states)
hidden_states = self.activation_fn(hidden_states)
hidden_states = self.fc2(hidden_states)
return hidden_states
class BlipEncoderLayer(nn.Module):
def __init__(self, config: BlipConfig):
super().__init__()
self.embed_dim = config.hidden_size
self.self_attn = BlipAttention(config)
self.layer_norm1 = nn.LayerNorm(self.embed_dim, eps=config.layer_norm_eps)
self.mlp = BlipMLP(config)
self.layer_norm2 = nn.LayerNorm(self.embed_dim, eps=config.layer_norm_eps)
def forward(
self,
hidden_states: torch.Tensor,
attention_mask: torch.Tensor,
output_attentions: Optional[bool] = False,
) -> Tuple[torch.FloatTensor]:
"""
Args:
hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
attention_mask (`torch.FloatTensor`): attention mask of size
`(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
`(config.encoder_attention_heads,)`.
output_attentions (`bool`, *optional*):
Whether or not to return the attentions tensors of all attention layers. See `attentions` under
returned tensors for more detail.
"""
residual = hidden_states
hidden_states = self.layer_norm1(hidden_states)
hidden_states, attn_weights = self.self_attn(
hidden_states=hidden_states,
head_mask=attention_mask,
output_attentions=output_attentions,
)
hidden_states = hidden_states + residual
residual = hidden_states
hidden_states = self.layer_norm2(hidden_states)
hidden_states = self.mlp(hidden_states)
hidden_states = hidden_states + residual
outputs = (hidden_states,)
if output_attentions:
outputs += (attn_weights,)
return outputs
class BlipPreTrainedModel(PreTrainedModel):
"""
An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
models.
"""
config_class = BlipConfig
base_model_prefix = "blip"
supports_gradient_checkpointing = True
def _init_weights(self, module):
"""Initialize the weights"""
factor = self.config.initializer_range
if isinstance(module, nn.Conv2d) or isinstance(module, nn.Embedding) or isinstance(module, nn.Linear):
module.weight.data.normal_(mean=0.0, std=factor)
if hasattr(module, "bias") and module.bias is not None:
module.bias.data.zero_()
if isinstance(module, BlipVisionEmbeddings):
if hasattr(self.config, "vision_config"):
factor = self.config.vision_config.initializer_range
nn.init.trunc_normal_(
module.position_embedding,
mean=0.0,
std=factor,
)
nn.init.trunc_normal_(
module.class_embedding,
mean=0.0,
std=factor,
)
elif isinstance(module, nn.LayerNorm):
module.bias.data.zero_()
module.weight.data.fill_(1.0)
elif isinstance(module, nn.Linear) and module.bias is not None:
module.bias.data.zero_()
def _set_gradient_checkpointing(self, module, value=False):
if isinstance(module, BlipEncoder):
module.gradient_checkpointing = value
BLIP_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 ([`BlipConfig`]): 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.
"""
BLIP_TEXT_INPUTS_DOCSTRING = r"""
Args:
input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
it.
Indices can be obtained using [`AutoProcessor`]. See [`BlipProcessor.__call__`] for details.
[What are input IDs?](../glossary#input-ids)
attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *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 `(batch_size, sequence_length)`, *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)
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 [`~utils.ModelOutput`] instead of a plain tuple.
"""
BLIP_VISION_INPUTS_DOCSTRING = r"""
Args:
pixel_values (`torch.FloatTensor` of shape `(batch_size, num_channels, height, width)`):
Pixel values. Padding will be ignored by default should you provide it. Pixel values can be obtained using
[`BlipImageProcessor`]. See [`BlipImageProcessor.__call__`] for details.
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 [`~utils.ModelOutput`] instead of a plain tuple.
"""
BLIP_INPUTS_DOCSTRING = r"""
Args:
input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
it.
Indices can be obtained using [`AutoProcessor`]. See [`BlipProcessor.__call__`] for details.
[What are input IDs?](../glossary#input-ids)
attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *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 `(batch_size, sequence_length)`, *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)
pixel_values (`torch.FloatTensor` of shape `(batch_size, num_channels, height, width)`):
Pixel values. Padding will be ignored by default should you provide it. Pixel values can be obtained using
[`BlipImageProcessor`]. See [`BlipImageProcessor.__call__`] for details.
return_loss (`bool`, *optional*):
Whether or not to return the contrastive loss.
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 [`~utils.ModelOutput`] instead of a plain tuple.
"""
class BlipEncoder(nn.Module):
"""
Transformer encoder consisting of `config.num_hidden_layers` self attention layers. Each layer is a
[`BlipEncoderLayer`].
Args:
config (`BlipConfig`):
The corresponding vision configuration for the `BlipEncoder`.
"""
def __init__(self, config: BlipConfig):
super().__init__()
self.config = config
self.layers = nn.ModuleList([BlipEncoderLayer(config) for _ in range(config.num_hidden_layers)])
self.gradient_checkpointing = False
def forward(
self,
inputs_embeds,
attention_mask: Optional[torch.Tensor] = None,
output_attentions: Optional[bool] = None,
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
) -> Union[Tuple, BaseModelOutput]:
r"""
Args:
inputs_embeds (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`):
Embedded representation of the inputs. Should be float, not int tokens.
attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *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)
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 [`~utils.ModelOutput`] instead of a plain tuple.
"""
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
encoder_states = () if output_hidden_states else None
all_attentions = () if output_attentions else None
hidden_states = inputs_embeds
for idx, encoder_layer in enumerate(self.layers):
if output_hidden_states:
encoder_states = encoder_states + (hidden_states,)
if self.gradient_checkpointing and self.training:
def create_custom_forward(module):
def custom_forward(*inputs):
return module(*inputs, output_attentions)
return custom_forward
layer_outputs = torch.utils.checkpoint.checkpoint(
create_custom_forward(encoder_layer),
hidden_states,
attention_mask,
)
else:
layer_outputs = encoder_layer(
hidden_states,
attention_mask,
output_attentions=output_attentions,
)
hidden_states = layer_outputs[0]
if output_attentions:
all_attentions = all_attentions + (layer_outputs[1],)
if output_hidden_states:
encoder_states = encoder_states + (hidden_states,)
if not return_dict:
return tuple(v for v in [hidden_states, encoder_states, all_attentions] if v is not None)
return BaseModelOutput(
last_hidden_state=hidden_states, hidden_states=encoder_states, attentions=all_attentions
)
class BlipVisionModel(BlipPreTrainedModel):
main_input_name = "pixel_values"
config_class = BlipVisionConfig
def __init__(self, config: BlipVisionConfig):
super().__init__(config)
self.config = config
embed_dim = config.hidden_size
self.embeddings = BlipVisionEmbeddings(config)
self.encoder = BlipEncoder(config)
self.post_layernorm = nn.LayerNorm(embed_dim, eps=config.layer_norm_eps)
self.post_init()
@add_start_docstrings_to_model_forward(BLIP_VISION_INPUTS_DOCSTRING)
@replace_return_docstrings(output_type=BaseModelOutputWithPooling, config_class=BlipVisionConfig)
def forward(
self,
pixel_values: Optional[torch.FloatTensor] = None,
output_attentions: Optional[bool] = None,
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
) -> Union[Tuple, BaseModelOutputWithPooling]:
r"""
Returns:
"""
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 pixel_values is None:
raise ValueError("You have to specify pixel_values")
hidden_states = self.embeddings(pixel_values)
encoder_outputs = self.encoder(
inputs_embeds=hidden_states,
output_attentions=output_attentions,
output_hidden_states=output_hidden_states,
return_dict=return_dict,
)
last_hidden_state = encoder_outputs[0]
last_hidden_state = self.post_layernorm(last_hidden_state)
pooled_output = last_hidden_state[:, 0, :]
pooled_output = self.post_layernorm(pooled_output)
if not return_dict:
return (last_hidden_state, pooled_output) + encoder_outputs[1:]
return BaseModelOutputWithPooling(
last_hidden_state=last_hidden_state,
pooler_output=pooled_output,
hidden_states=encoder_outputs.hidden_states,
attentions=encoder_outputs.attentions,
)
def get_input_embeddings(self):
return self.embeddings
@add_start_docstrings(BLIP_START_DOCSTRING)
class BlipModel(BlipPreTrainedModel):
config_class = BlipConfig
def __init__(self, config: BlipConfig):
super().__init__(config)
if not isinstance(config.text_config, BlipTextConfig):
raise ValueError(
"config.text_config is expected to be of type BlipTextConfig but is of type"
f" {type(config.text_config)}."
)
if not isinstance(config.vision_config, BlipVisionConfig):
raise ValueError(
"config.vision_config is expected to be of type BlipVisionConfig but is of type"
f" {type(config.vision_config)}."
)
text_config = config.text_config
vision_config = config.vision_config
self.projection_dim = config.projection_dim
self.text_embed_dim = text_config.hidden_size
self.vision_embed_dim = vision_config.hidden_size
self.text_model = BlipTextModel(text_config)
self.vision_model = BlipVisionModel(vision_config)
self.visual_projection = nn.Linear(self.vision_embed_dim, self.projection_dim, bias=False)
self.text_projection = nn.Linear(self.text_embed_dim, self.projection_dim, bias=False)
self.logit_scale = nn.Parameter(torch.tensor(self.config.logit_scale_init_value))
# Initialize weights and apply final processing
self.post_init()
@add_start_docstrings_to_model_forward(BLIP_TEXT_INPUTS_DOCSTRING)
def get_text_features(
self,
input_ids: Optional[torch.Tensor] = None,
attention_mask: Optional[torch.Tensor] = None,
position_ids: Optional[torch.Tensor] = None,
return_dict: Optional[bool] = None,
) -> torch.FloatTensor:
r"""
Returns:
text_features (`torch.FloatTensor` of shape `(batch_size, output_dim`): The text embeddings obtained by
applying the projection layer to the pooled output of [`BlipTextModel`].
Examples:
```python
>>> from transformers import AutoProcessor, BlipModel
>>> model = BlipModel.from_pretrained("Salesforce/blip-image-captioning-base")
>>> processor = AutoProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
>>> inputs = processor(text=["a photo of a cat", "a photo of a dog"], padding=True, return_tensors="pt")
>>> text_features = model.get_text_features(**inputs)
```"""
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
text_outputs = self.text_model(
input_ids=input_ids,
attention_mask=attention_mask,
position_ids=position_ids,
return_dict=return_dict,
)
pooled_output = text_outputs[1]
text_features = self.text_projection(pooled_output)
return text_features
@add_start_docstrings_to_model_forward(BLIP_VISION_INPUTS_DOCSTRING)
def get_image_features(
self,
pixel_values: Optional[torch.FloatTensor] = None,
return_dict: Optional[bool] = None,
) -> torch.FloatTensor:
r"""
Returns:
image_features (`torch.FloatTensor` of shape `(batch_size, output_dim`): The image embeddings obtained by
applying the projection layer to the pooled output of [`BlipVisionModel`].
Examples:
```python
>>> from PIL import Image
>>> import requests
>>> from transformers import AutoProcessor, BlipModel
>>> model = BlipModel.from_pretrained("Salesforce/blip-image-captioning-base")
>>> processor = AutoProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
>>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
>>> image = Image.open(requests.get(url, stream=True).raw)
>>> inputs = processor(images=image, return_tensors="pt")
>>> image_features = model.get_image_features(**inputs)
```"""
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
vision_outputs = self.vision_model(pixel_values=pixel_values, return_dict=return_dict)
pooled_output = vision_outputs[1] # pooled_output
image_features = self.visual_projection(pooled_output)
return image_features
@add_start_docstrings_to_model_forward(BLIP_INPUTS_DOCSTRING)
@replace_return_docstrings(output_type=BlipOutput, config_class=BlipConfig)
def forward(
self,
input_ids: Optional[torch.LongTensor] = None,
pixel_values: Optional[torch.FloatTensor] = None,
attention_mask: Optional[torch.Tensor] = None,
position_ids: Optional[torch.LongTensor] = None,
return_loss: Optional[bool] = None,
output_attentions: Optional[bool] = None,
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
) -> Union[Tuple, BlipOutput]:
r"""
Returns:
Examples:
```python
>>> from PIL import Image
>>> import requests
>>> from transformers import AutoProcessor, BlipModel
>>> model = BlipModel.from_pretrained("Salesforce/blip-image-captioning-base")
>>> processor = AutoProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
>>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
>>> image = Image.open(requests.get(url, stream=True).raw)
>>> inputs = processor(
... text=["a photo of a cat", "a photo of a dog"], images=image, return_tensors="pt", padding=True
... )
>>> outputs = model(**inputs)
>>> logits_per_image = outputs.logits_per_image # this is the image-text similarity score
>>> probs = logits_per_image.softmax(dim=1) # we can take the softmax to get the label probabilities
```"""
# Use BLIP model's config for some fields (if specified) instead of those of vision & text components.
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
vision_outputs = self.vision_model(
pixel_values=pixel_values,
output_attentions=output_attentions,
output_hidden_states=output_hidden_states,
return_dict=return_dict,
)
text_outputs = self.text_model(
input_ids=input_ids,
attention_mask=attention_mask,
position_ids=position_ids,
output_attentions=output_attentions,
output_hidden_states=output_hidden_states,
return_dict=return_dict,
)
image_embeds = vision_outputs[1]
image_embeds = self.visual_projection(image_embeds)
text_embeds = text_outputs[1]
text_embeds = self.text_projection(text_embeds)
# normalized features
image_embeds = image_embeds / image_embeds.norm(p=2, dim=-1, keepdim=True)
text_embeds = text_embeds / text_embeds.norm(p=2, dim=-1, keepdim=True)
# cosine similarity as logits
logit_scale = self.logit_scale.exp()
logits_per_text = torch.matmul(text_embeds, image_embeds.t()) * logit_scale
logits_per_image = logits_per_text.t()
loss = None
if return_loss:
loss = blip_loss(logits_per_text)
if not return_dict:
output = (logits_per_image, logits_per_text, text_embeds, image_embeds, text_outputs, vision_outputs)
return ((loss,) + output) if loss is not None else output
return BlipOutput(
loss=loss,
logits_per_image=logits_per_image,
logits_per_text=logits_per_text,
text_embeds=text_embeds,
image_embeds=image_embeds,
text_model_output=text_outputs,
vision_model_output=vision_outputs,
)
@add_start_docstrings(
"""
BLIP Model for image captioning. The model consists of a vision encoder and a text decoder. One can optionally pass
`input_ids` to the model, which serve as a text prompt, to make the text decoder continue the prompt. Otherwise,
the decoder starts generating text from the [BOS] (beginning-of-sequence) token. will start generating the caption
from the text input. If no text input is provided, the decoder will start with the [BOS] token only.
""",
BLIP_START_DOCSTRING,
)
class BlipForConditionalGeneration(BlipPreTrainedModel):
config_class = BlipConfig
_tied_weights_keys = ["text_decoder.cls.predictions.decoder.bias"]
main_input_name = "pixel_values"
def __init__(self, config: BlipConfig):
super().__init__(config)
self.vision_model = BlipVisionModel(config.vision_config)
self.text_decoder = BlipTextLMHeadModel(config.text_config)
self.decoder_input_ids = config.text_config.bos_token_id
self.decoder_pad_token_id = config.text_config.pad_token_id
# Initialize weights and apply final processing
self.post_init()
def get_input_embeddings(self) -> nn.Module:
return self.vision_model.embeddings.patch_embedding
@add_start_docstrings_to_model_forward(BLIP_VISION_INPUTS_DOCSTRING)
@replace_return_docstrings(output_type=BlipForConditionalGenerationModelOutput, config_class=BlipVisionConfig)
def forward(
self,
pixel_values: torch.FloatTensor,
input_ids: Optional[torch.LongTensor] = None,
attention_mask: Optional[torch.LongTensor] = None,
output_attentions: Optional[bool] = None,
output_hidden_states: Optional[bool] = None,
labels: Optional[torch.LongTensor] = None,
return_dict: Optional[bool] = None,
) -> Union[Tuple, BlipForConditionalGenerationModelOutput]:
r"""
Returns:
Examples:
```python
>>> from PIL import Image
>>> import requests
>>> from transformers import AutoProcessor, BlipForConditionalGeneration
>>> processor = AutoProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
>>> model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base")
>>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
>>> image = Image.open(requests.get(url, stream=True).raw)
>>> text = "A picture of"
>>> inputs = processor(images=image, text=text, return_tensors="pt")
>>> outputs = model(**inputs)
```"""
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
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
)
vision_outputs = self.vision_model(
pixel_values=pixel_values,
output_attentions=output_attentions,
output_hidden_states=output_hidden_states,
return_dict=return_dict,
)
image_embeds = vision_outputs[0]
outputs = self.text_decoder(
input_ids=input_ids,
attention_mask=attention_mask,
encoder_hidden_states=image_embeds,
labels=labels,
return_dict=return_dict,
reduction="mean",
)
if not return_dict:
outputs = (outputs[0], outputs[1], image_embeds, vision_outputs[0]) + vision_outputs[2:]
return tuple(output for output in outputs if output is not None)
return BlipForConditionalGenerationModelOutput(
loss=outputs.loss,
logits=outputs.logits,
image_embeds=image_embeds,
last_hidden_state=vision_outputs.last_hidden_state,
hidden_states=vision_outputs.hidden_states,
attentions=vision_outputs.attentions,
)
@torch.no_grad()
def generate(
self,
pixel_values: torch.FloatTensor,
input_ids: Optional[torch.LongTensor] = None,
attention_mask: Optional[torch.LongTensor] = None,
**generate_kwargs,
) -> torch.LongTensor:
r"""
Overrides *generate* function to be able to use the model as a conditional generator
Parameters:
pixel_values (*torch.FloatTensor* of shape *(batch_size, num_channels, image_height, image_width)*:
Input image to be processed
input_ids (*torch.LongTensor* of shape *(batch_size, sequence_length)*, *optional*):
The sequence used as a prompt for the generation.
attention_mask (*torch.LongTensor* of shape *(batch_size, sequence_length)*, *optional*):
Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
Examples:
```python
>>> from PIL import Image
>>> import requests
>>> from transformers import AutoProcessor, BlipForConditionalGeneration
>>> model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base")
>>> processor = AutoProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
>>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
>>> image = Image.open(requests.get(url, stream=True).raw)
>>> inputs = processor(images=image, return_tensors="pt")
>>> outputs = model.generate(**inputs)
>>> print(processor.decode(outputs[0], skip_special_tokens=True))
two cats sleeping on a couch
```
"""
batch_size = pixel_values.shape[0]
vision_outputs = self.vision_model(pixel_values=pixel_values)
image_embeds = vision_outputs[0]
image_attention_mask = torch.ones(image_embeds.size()[:-1], dtype=torch.long).to(image_embeds.device)
if isinstance(input_ids, list):
input_ids = torch.LongTensor(input_ids)
elif input_ids is None:
input_ids = (
torch.LongTensor([[self.decoder_input_ids, self.config.text_config.eos_token_id]])
.repeat(batch_size, 1)
.to(image_embeds.device)
)
input_ids[:, 0] = self.config.text_config.bos_token_id
attention_mask = attention_mask[:, :-1] if attention_mask is not None else None
outputs = self.text_decoder.generate(
input_ids=input_ids[:, :-1],
eos_token_id=self.config.text_config.sep_token_id,
pad_token_id=self.config.text_config.pad_token_id,
attention_mask=attention_mask,
encoder_hidden_states=image_embeds,
encoder_attention_mask=image_attention_mask,
**generate_kwargs,
)
return outputs
@add_start_docstrings(
"""
BLIP Model for visual question answering. The model consists of a vision encoder, a text encoder as well as a text
decoder. The vision encoder will encode the input image, the text encoder will encode the input question together
with the encoding of the image, and the text decoder will output the answer to the question.
""",
BLIP_START_DOCSTRING,
)
class BlipForQuestionAnswering(BlipPreTrainedModel):
config_class = BlipConfig
_tied_weights_keys = ["text_decoder.cls.predictions.decoder.bias"]
def __init__(self, config: BlipConfig):
super().__init__(config)
self.vision_model = BlipVisionModel(config.vision_config)
self.text_encoder = BlipTextModel(config.text_config, add_pooling_layer=False)
self.text_decoder = BlipTextLMHeadModel(config.text_config)
self.decoder_pad_token_id = config.text_config.pad_token_id
self.decoder_start_token_id = config.text_config.bos_token_id
# Initialize weights and apply final processing
self.post_init()
def get_input_embeddings(self) -> nn.Module:
return self.vision_model.embeddings.patch_embedding
@add_start_docstrings_to_model_forward(BLIP_VISION_INPUTS_DOCSTRING)
@replace_return_docstrings(output_type=BlipTextVisionModelOutput, config_class=BlipVisionConfig)
def forward(
self,
input_ids: torch.LongTensor,
pixel_values: torch.FloatTensor,
decoder_input_ids: Optional[torch.LongTensor] = None,
decoder_attention_mask: Optional[torch.LongTensor] = None,
attention_mask: Optional[torch.LongTensor] = None,
output_attentions: Optional[bool] = None,
output_hidden_states: Optional[bool] = None,
labels: Optional[torch.LongTensor] = None,
return_dict: Optional[bool] = None,
) -> Union[Tuple, BlipTextVisionModelOutput]:
r"""
Returns:
Examples:
```python
>>> from PIL import Image
>>> import requests
>>> from transformers import AutoProcessor, BlipForQuestionAnswering
>>> model = BlipForQuestionAnswering.from_pretrained("Salesforce/blip-vqa-base")
>>> processor = AutoProcessor.from_pretrained("Salesforce/blip-vqa-base")
>>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
>>> image = Image.open(requests.get(url, stream=True).raw)
>>> # training
>>> text = "How many cats are in the picture?"
>>> label = "2"
>>> inputs = processor(images=image, text=text, return_tensors="pt")
>>> labels = processor(text=label, return_tensors="pt").input_ids
>>> inputs["labels"] = labels
>>> outputs = model(**inputs)
>>> loss = outputs.loss
>>> loss.backward()
>>> # inference
>>> text = "How many cats are in the picture?"
>>> inputs = processor(images=image, text=text, return_tensors="pt")
>>> outputs = model.generate(**inputs)
>>> print(processor.decode(outputs[0], skip_special_tokens=True))
2
```"""
if labels is None and decoder_input_ids is None:
raise ValueError(
"Either `decoder_input_ids` or `labels` should be passed when calling `forward` with"
" `BlipForQuestionAnswering`. if you are training the model make sure that `labels` is passed, if you"
" are using the model for inference make sure that `decoder_input_ids` is passed or call `generate`"
)
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
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
)
vision_outputs = self.vision_model(
pixel_values=pixel_values,
output_attentions=output_attentions,
output_hidden_states=output_hidden_states,
return_dict=return_dict,
)
image_embeds = vision_outputs[0]
image_attention_mask = torch.ones(image_embeds.size()[:-1], dtype=torch.long)
question_embeds = self.text_encoder(
input_ids=input_ids,
attention_mask=attention_mask,
encoder_hidden_states=image_embeds,
encoder_attention_mask=image_attention_mask,
return_dict=return_dict,
)
if labels is not None and decoder_input_ids is None:
# labels are already shifted right, see: https://github.com/huggingface/transformers/pull/23153
decoder_input_ids = labels
question_embeds = question_embeds[0] if not return_dict else question_embeds.last_hidden_state
answer_output = self.text_decoder(
input_ids=decoder_input_ids,
attention_mask=decoder_attention_mask,
encoder_hidden_states=question_embeds,
encoder_attention_mask=attention_mask,
labels=labels,
return_dict=return_dict,
reduction="mean",
)
if labels is not None:
decoder_loss = answer_output.loss.mean() if return_dict else answer_output[0].mean()
else:
decoder_loss = None
if not return_dict:
outputs = (decoder_loss, image_embeds, vision_outputs[0]) + vision_outputs[2:]
return tuple(output for output in outputs if output is not None)
return BlipTextVisionModelOutput(
loss=decoder_loss,
image_embeds=image_embeds,
last_hidden_state=vision_outputs.last_hidden_state,
hidden_states=vision_outputs.hidden_states,
attentions=vision_outputs.attentions,
)
@torch.no_grad()
def generate(
self,
input_ids: torch.LongTensor,
pixel_values: torch.FloatTensor,
attention_mask: Optional[torch.LongTensor] = None,
**generate_kwargs,
) -> torch.LongTensor:
r"""
Overrides *generate* function to be able to use the model as a conditional generator
Parameters:
input_ids (*torch.LongTensor* of shape *(batch_size, sequence_length)*):
The sequence used as a prompt for the generation.
pixel_values (*torch.FloatTensor* of shape *(batch_size, num_channels, image_height, image_width)*:
Input image to be processed
attention_mask (*torch.LongTensor* of shape *(batch_size, sequence_length)*, *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 MASKED tokens.
**generate_kwargs:
Additional arguments passed to the *generate* function of the decoder
Examples:
```python
>>> from PIL import Image
>>> import requests
>>> from transformers import AutoProcessor, BlipForQuestionAnswering
>>> model = BlipForQuestionAnswering.from_pretrained("Salesforce/blip-vqa-base")
>>> processor = AutoProcessor.from_pretrained("Salesforce/blip-vqa-base")
>>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
>>> image = Image.open(requests.get(url, stream=True).raw)
>>> text = "How many cats are in the picture?"
>>> inputs = processor(images=image, text=text, return_tensors="pt")
>>> outputs = model.generate(**inputs)
>>> print(processor.decode(outputs[0], skip_special_tokens=True))
2
```
"""
vision_outputs = self.vision_model(pixel_values=pixel_values)
image_embeds = vision_outputs[0]
image_attention_mask = torch.ones(image_embeds.size()[:-1], dtype=torch.long).to(image_embeds.device)
if isinstance(input_ids, list):
input_ids = torch.LongTensor(input_ids)
question_outputs = self.text_encoder(
input_ids=input_ids,
attention_mask=attention_mask,
encoder_hidden_states=image_embeds,
encoder_attention_mask=image_attention_mask,
return_dict=False,
)
question_embeds = question_outputs[0]
question_attention_mask = torch.ones(question_embeds.size()[:-1], dtype=torch.long).to(question_embeds.device)
bos_ids = torch.full(
(question_embeds.size(0), 1), fill_value=self.decoder_start_token_id, device=question_embeds.device
)
outputs = self.text_decoder.generate(
input_ids=bos_ids,
eos_token_id=self.config.text_config.sep_token_id,
pad_token_id=self.config.text_config.pad_token_id,
encoder_hidden_states=question_embeds,
encoder_attention_mask=question_attention_mask,
**generate_kwargs,
)
return outputs
@add_start_docstrings(
"""
BLIP Model with a vision and text projector, and a classification head on top. The model is used in the context of
image-text retrieval. Given an image and a text, the model returns the probability of the text being relevant to
the image.
""",
BLIP_START_DOCSTRING,
)
class BlipForImageTextRetrieval(BlipPreTrainedModel):
config_class = BlipConfig
def __init__(self, config: BlipConfig):
super().__init__(config)
self.vision_model = BlipVisionModel(config.vision_config)
self.text_encoder = BlipTextModel(config.text_config, add_pooling_layer=False)
# vision projection layer
self.vision_proj = nn.Linear(config.vision_config.hidden_size, config.image_text_hidden_size)
# text projection layer
self.text_proj = nn.Linear(config.text_config.hidden_size, config.image_text_hidden_size)
# image text matching head
self.itm_head = nn.Linear(config.text_config.hidden_size, 2)
self.decoder_pad_token_id = (
config.text_config.pad_token_id
if not hasattr(config, "decoder_pad_token_id")
else config.decoder_pad_token_id
)
self.decoder_start_token_id = (
config.text_config.bos_token_id
if not hasattr(config, "decoder_start_token_id")
else config.decoder_start_token_id
)
# Initialize weights and apply final processing
self.post_init()
def get_input_embeddings(self) -> nn.Module:
return self.vision_model.embeddings.patch_embedding
@add_start_docstrings_to_model_forward(BLIP_VISION_INPUTS_DOCSTRING)
@replace_return_docstrings(output_type=BlipTextVisionModelOutput, config_class=BlipVisionConfig)
def forward(
self,
input_ids: torch.LongTensor,
pixel_values: torch.FloatTensor,
use_itm_head: Optional[bool] = True,
attention_mask: Optional[torch.LongTensor] = None,
output_attentions: Optional[bool] = None,
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
) -> Union[Tuple, BlipTextVisionModelOutput]:
r"""
Returns:
Examples:
```python
>>> from PIL import Image
>>> import requests
>>> from transformers import AutoProcessor, BlipForImageTextRetrieval
>>> model = BlipForImageTextRetrieval.from_pretrained("Salesforce/blip-itm-base-coco")
>>> processor = AutoProcessor.from_pretrained("Salesforce/blip-itm-base-coco")
>>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
>>> image = Image.open(requests.get(url, stream=True).raw)
>>> text = "an image of a cat"
>>> inputs = processor(images=image, text=text, return_tensors="pt")
>>> outputs = model(**inputs)
```
"""
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
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
)
vision_outputs = self.vision_model(
pixel_values=pixel_values,
output_attentions=output_attentions,
output_hidden_states=output_hidden_states,
return_dict=return_dict,
)
image_embeds = vision_outputs[0]
image_atts = torch.ones(image_embeds.size()[:-1], dtype=torch.long)
if use_itm_head:
question_embeds = self.text_encoder(
input_ids=input_ids,
attention_mask=attention_mask,
encoder_hidden_states=image_embeds,
encoder_attention_mask=image_atts,
return_dict=return_dict,
)
question_embeds = question_embeds[0] if not return_dict else question_embeds.last_hidden_state
output = self.itm_head(question_embeds[:, 0, :])
else:
question_embeds = self.text_encoder(
input_ids=input_ids,
attention_mask=attention_mask,
return_dict=return_dict,
)
question_embeds = question_embeds[0] if not return_dict else question_embeds.last_hidden_state
image_feat = normalize(self.vision_proj(image_embeds[:, 0, :]), dim=-1)
text_feat = normalize(self.text_proj(question_embeds[:, 0, :]), dim=-1)
output = image_feat @ text_feat.t()
if not return_dict:
outputs = (output, vision_outputs[0]) + vision_outputs[2:] + (question_embeds,)
return tuple(output for output in outputs if output is not None)
return BlipImageTextMatchingModelOutput(
itm_score=output,
last_hidden_state=vision_outputs.last_hidden_state,
hidden_states=vision_outputs.hidden_states,
attentions=vision_outputs.attentions,
question_embeds=question_embeds,
)
|