File size: 58,339 Bytes
44eca84 1028385 16db48a 1028385 44eca84 70d7fbb 44eca84 f22aece 1028385 19fa3ec a78db43 1028385 a78db43 1028385 a407d29 1028385 19fa3ec a78db43 19fa3ec abc0b46 551a5a7 abc0b46 a78db43 44eca84 38a8e51 a78db43 70d7fbb 44eca84 70d7fbb 44eca84 16db48a 540a0c2 16db48a a78db43 16db48a 540a0c2 16db48a 540a0c2 16db48a a78db43 16db48a a78db43 16db48a a78db43 16db48a 540a0c2 16db48a 540a0c2 16db48a a78db43 16db48a a78db43 16db48a a78db43 16db48a 70d7fbb 16db48a 04ba1c9 540a0c2 04ba1c9 540a0c2 04ba1c9 70d7fbb 540a0c2 70d7fbb 540a0c2 70d7fbb 540a0c2 70d7fbb 540a0c2 70d7fbb 540a0c2 70d7fbb 540a0c2 04ba1c9 16db48a 44eca84 19fa3ec 1028385 fa58912 811a69b fa58912 1028385 8833bbc 19fa3ec 04ba1c9 19fa3ec e754fde 16db48a 19fa3ec 8833bbc fa58912 19fa3ec 1028385 04ba1c9 1028385 e754fde 19fa3ec 8833bbc 1028385 19fa3ec 8833bbc abc0b46 8833bbc 44eca84 551a5a7 fa58912 8833bbc fa58912 8833bbc fa58912 8833bbc eabd9f8 8833bbc fa58912 8833bbc 551a5a7 8833bbc 38a8e51 04ba1c9 8833bbc e754fde 8833bbc 19fa3ec 8833bbc 1028385 44eca84 33824cf 1028385 44eca84 fa58912 1028385 8833bbc fa58912 8833bbc fa58912 8833bbc 551a5a7 44eca84 38a8e51 04ba1c9 44eca84 e754fde 38a8e51 44eca84 1028385 19fa3ec 9fb003c 8833bbc 9fb003c 1028385 811a69b 9fb003c 551a5a7 1028385 8833bbc 04ba1c9 1028385 e754fde 1028385 44eca84 551a5a7 19fa3ec 8ad9fee 44eca84 8ad9fee 44eca84 551a5a7 44eca84 38a8e51 04ba1c9 19fa3ec e754fde 19fa3ec 2b7496e 16db48a 2b7496e 16db48a 2b7496e 16db48a 2b7496e 16db48a 2b7496e 16db48a 2b7496e 16db48a 2b7496e 16db48a 2b7496e 16db48a 2b7496e 19fa3ec 16db48a 19fa3ec 70d7fbb 2b7496e 44eca84 |
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 |
{
"nbformat": 4,
"nbformat_minor": 0,
"metadata": {
"colab": {
"provenance": []
},
"kernelspec": {
"name": "python3",
"display_name": "Python 3"
},
"language_info": {
"name": "python"
}
},
"cells": [
{
"cell_type": "markdown",
"source": [
"This Notebook is a Stable-diffusion tool which allows you to find similiar tokens from the SD 1.5 vocab.json that you can use for text-to-image generation. Try this Free online SD 1.5 generator with the results: https://perchance.org/fusion-ai-image-generator\n",
"\n",
"Scroll to the bottom of the notebook to see the guide for how this works."
],
"metadata": {
"id": "L7JTcbOdBPfh"
}
},
{
"cell_type": "code",
"source": [
"# @title ✳️ Load/initialize values\n",
"# Load the tokens into the colab\n",
"!git clone https://huggingface.co/datasets/codeShare/sd_tokens\n",
"import torch\n",
"from torch import linalg as LA\n",
"device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
"%cd /content/sd_tokens\n",
"token = torch.load('sd15_tensors.pt', map_location=device, weights_only=True)\n",
"#-----#\n",
"\n",
"#Import the vocab.json\n",
"import json\n",
"import pandas as pd\n",
"with open('vocab.json', 'r') as f:\n",
" data = json.load(f)\n",
"\n",
"_df = pd.DataFrame({'count': data})['count']\n",
"\n",
"vocab = {\n",
" value: key for key, value in _df.items()\n",
"}\n",
"#-----#\n",
"\n",
"# Define functions/constants\n",
"NUM_TOKENS = 49407\n",
"\n",
"def absolute_value(x):\n",
" return max(x, -x)\n",
"\n",
"\n",
"def token_similarity(A, B):\n",
"\n",
" #Vector length#\n",
" _A = LA.vector_norm(A, ord=2)\n",
" _B = LA.vector_norm(B, ord=2)\n",
"\n",
" #----#\n",
" result = torch.dot(A,B)/(_A*_B)\n",
" #similarity_pcnt = absolute_value(result.item()*100)\n",
" similarity_pcnt = result.item()*100\n",
" similarity_pcnt_aprox = round(similarity_pcnt, 3)\n",
" result = f'{similarity_pcnt_aprox} %'\n",
" return result\n",
"\n",
"\n",
"def similarity(id_A , id_B):\n",
" #Tensors\n",
" A = token[id_A]\n",
" B = token[id_B]\n",
" return token_similarity(A, B)\n",
"#----#\n",
"\n",
"#print(vocab[8922]) #the vocab item for ID 8922\n",
"#print(token[8922].shape) #dimension of the token\n",
"\n",
"mix_with = \"\"\n",
"mix_method = \"None\"\n",
"\n",
"#-------------#\n",
"# UNUSED\n",
"\n",
"# Get the 10 lowest values from a tensor as a string\n",
"def get_valleys (A):\n",
" sorted, indices = torch.sort(A,dim=0 , descending=False)\n",
" result = \"{\"\n",
" for index in range(10):\n",
" id = indices[index].item()\n",
" result = result + f\"{id}\"\n",
" if(index<9):\n",
" result = result + \",\"\n",
" result = result + \"}\"\n",
" return result\n",
"\n",
"# Get the 10 highest values from a tensor as a string\n",
"def get_peaks (A):\n",
" sorted, indices = torch.sort(A,dim=0 , descending=True)\n",
" result = \"{\"\n",
" for index in range(10):\n",
" id = indices[index].item()\n",
" result = result + f\"{id}\"\n",
" if(index<9):\n",
" result = result + \",\"\n",
" result = result + \"}\"\n",
" return result"
],
"metadata": {
"id": "Ch9puvwKH1s3",
"collapsed": true,
"cellView": "form"
},
"execution_count": null,
"outputs": []
},
{
"cell_type": "code",
"source": [
"# @title ⚡ Get similiar tokens\n",
"from transformers import AutoTokenizer\n",
"tokenizer = AutoTokenizer.from_pretrained(\"openai/clip-vit-large-patch14\", clean_up_tokenization_spaces = False)\n",
"\n",
"# @markdown Write name of token to match against\n",
"prompt= \"banana\" # @param {type:'string',\"placeholder\":\"leave empty for random value token\"}\n",
"\n",
"tokenizer_output = tokenizer(text = prompt)\n",
"input_ids = tokenizer_output['input_ids']\n",
"print(input_ids)\n",
"\n",
"\n",
"#The prompt will be enclosed with the <|start-of-text|> and <|end-of-text|> tokens, which is why output will be [49406, ... , 49407].\n",
"\n",
"#You can leave the 'prompt' field empty to get a random value tensor. Since the tensor is random value, it will not correspond to any tensor in the vocab.json list , and this it will have no ID.\n",
"\n",
"id_A = input_ids[1]\n",
"A = token[id_A]\n",
"_A = LA.vector_norm(A, ord=2)\n",
"\n",
"#if no imput exists we just randomize the entire thing\n",
"if (prompt == \"\"):\n",
" id_A = -1\n",
" print(\"Tokenized prompt tensor A is a random valued tensor with no ID\")\n",
" R = torch.rand(768)\n",
" _R = LA.vector_norm(R, ord=2)\n",
" A = R*(_A/_R)\n",
" name_A = 'random_A'\n",
"\n",
"# @markdown (optional) Mix the token with something else\n",
"mix_with = \"\" # @param {\"type\":\"string\",\"placeholder\":\"leave empty for random value token\"}\n",
"mix_method = \"None\" # @param [\"None\" , \"Average\", \"Subtract\"] {allow-input: true}\n",
"w = 0.5 # @param {type:\"slider\", min:0, max:1, step:0.01}\n",
"\n",
"# @markdown Limit char size of included token\n",
"min_char_size = 3 # @param {type:\"slider\", min:0, max: 50, step:1}\n",
"char_range = 5 # @param {type:\"slider\", min:0, max: 50, step:1}\n",
"\n",
"tokenizer_output = tokenizer(text = mix_with)\n",
"input_ids = tokenizer_output['input_ids']\n",
"id_C = input_ids[1]\n",
"C = token[id_C]\n",
"_C = LA.vector_norm(C, ord=2)\n",
"\n",
"#if no imput exists we just randomize the entire thing\n",
"if (mix_with == \"\"):\n",
" id_C = -1\n",
" print(\"Tokenized prompt 'mix_with' tensor C is a random valued tensor with no ID\")\n",
" R = torch.rand(768)\n",
" _R = LA.vector_norm(R, ord=2)\n",
" C = R*(_C/_R)\n",
" name_C = 'random_C'\n",
"\n",
"name_A = \"A of random type\"\n",
"if (id_A>-1):\n",
" name_A = vocab[id_A]\n",
"\n",
"name_C = \"token C of random type\"\n",
"if (id_C>-1):\n",
" name_C = vocab[id_C]\n",
"\n",
"# Peaks feature\n",
"#peaks_A = get_valleys(A)\n",
"#peaks_C = get_valleys(C)\n",
"#print(f\"The elementwise top 10 highest values for A is at indices {peaks_A}\")\n",
"#print(\"-------\")\n",
"#print(f\"The elementwise top 10 highest values for C is at indices {peaks_C}\")\n",
"#print(\"-------\")\n",
"#//------//\n",
"\n",
"print(f\"The similarity between A '{name_A}' and C '{name_C}' is {token_similarity(A, C)}\")\n",
"\n",
"if (mix_method == \"None\"):\n",
" print(\"No operation\")\n",
"\n",
"if (mix_method == \"Average\"):\n",
" A = w*A + (1-w)*C\n",
" _A = LA.vector_norm(A, ord=2)\n",
" print(f\"Tokenized prompt tensor A '{name_A}' token has been recalculated as A = w*A + (1-w)*C , where C is '{name_C}' token , for w = {w} \")\n",
"\n",
"if (mix_method == \"Subtract\"):\n",
" tmp = w*A - (1-w)*C\n",
" _tmp = LA.vector_norm(tmp, ord=2)\n",
" A = (_A/_tmp)*tmp\n",
" #//---//\n",
" _A = LA.vector_norm(A, ord=2)\n",
" print(f\"Tokenized prompt tensor A '{name_A}' token has been recalculated as A = _A*norm(w*A - (1-w)*C) , where C is '{name_C}' token , for w = {w} \")\n",
"\n",
"#OPTIONAL : Add/subtract + normalize above result with another token. Leave field empty to get a random value tensor\n",
"\n",
"dots = torch.zeros(NUM_TOKENS)\n",
"for index in range(NUM_TOKENS):\n",
" id_B = index\n",
" B = token[id_B]\n",
" _B = LA.vector_norm(B, ord=2)\n",
" result = torch.dot(A,B)/(_A*_B)\n",
" #result = absolute_value(result.item())\n",
" result = result.item()\n",
" dots[index] = result\n",
"\n",
"\n",
"sorted, indices = torch.sort(dots,dim=0 , descending=True)\n",
"#----#\n",
"if (mix_method == \"Average\"):\n",
" print(f'Calculated all cosine-similarities between the average of token {name_A} and {name_C} with Id_A = {id_A} and mixed Id_C = {id_C} as a 1x{sorted.shape[0]} tensor')\n",
"if (mix_method == \"Subtract\"):\n",
" print(f'Calculated all cosine-similarities between the subtract of token {name_A} and {name_C} with Id_A = {id_A} and mixed Id_C = {id_C} as a 1x{sorted.shape[0]} tensor')\n",
"if (mix_method == \"None\"):\n",
" print(f'Calculated all cosine-similarities between the token {name_A} with Id_A = {id_A} with the the rest of the {NUM_TOKENS} tokens as a 1x{sorted.shape[0]} tensor')\n",
"\n",
"#Produce a list id IDs that are most similiar to the prompt ID at positiion 1 based on above result\n",
"\n",
"# @markdown Set print options\n",
"list_size = 100 # @param {type:'number'}\n",
"print_ID = False # @param {type:\"boolean\"}\n",
"print_Similarity = True # @param {type:\"boolean\"}\n",
"print_Name = True # @param {type:\"boolean\"}\n",
"print_Divider = True # @param {type:\"boolean\"}\n",
"\n",
"\n",
"if (print_Divider):\n",
" print('//---//')\n",
"\n",
"print('')\n",
"print('Here is the result : ')\n",
"print('')\n",
"\n",
"for index in range(list_size):\n",
" id = indices[index].item()\n",
" if (print_Name):\n",
" print(f'{vocab[id]}') # vocab item\n",
" if (print_ID):\n",
" print(f'ID = {id}') # IDs\n",
" if (print_Similarity):\n",
" print(f'similiarity = {round(sorted[index].item()*100,2)} %')\n",
" if (print_Divider):\n",
" print('--------')\n",
"\n",
"#Print the sorted list from above result"
],
"metadata": {
"id": "iWeFnT1gAx6A",
"cellView": "form"
},
"execution_count": null,
"outputs": []
},
{
"cell_type": "code",
"source": [
"# @title 💫 Compare Text encodings\n",
"\n",
"prompt_A = \"banana\" # @param {\"type\":\"string\",\"placeholder\":\"Write a prompt\"}\n",
"prompt_B = \"\" # @param {\"type\":\"string\",\"placeholder\":\"Write a prompt\"}\n",
"use_token_padding = True # @param {type:\"boolean\"}\n",
"\n",
"from transformers import CLIPProcessor, CLIPModel\n",
"\n",
"processor = CLIPProcessor.from_pretrained(\"openai/clip-vit-large-patch14\" , clean_up_tokenization_spaces = True)\n",
"\n",
"model = CLIPModel.from_pretrained(\"openai/clip-vit-large-patch14\")\n",
"\n",
"ids_A = processor.tokenizer(text=prompt_A, padding=use_token_padding, return_tensors=\"pt\")\n",
"text_encoding_A = model.get_text_features(**ids_A)\n",
"\n",
"\n",
"ids_B = processor.tokenizer(text=prompt_B, padding=use_token_padding, return_tensors=\"pt\")\n",
"text_encoding_B = model.get_text_features(**ids_B)\n",
"\n",
"similarity_str = 'The similarity between the text_encoding for A:\"' + prompt_A + '\" and B: \"' + prompt_B +'\" is ' + token_similarity(text_encoding_A[0] , text_encoding_B[0])\n",
"\n",
"\n",
"print(similarity_str)\n",
"#outputs = model(**inputs)\n",
"#logits_per_image = outputs.logits_per_image # this is the image-text similarity score\n",
"#probs = logits_per_image.softmax(dim=1) # we can take the softmax to get the label probabilities"
],
"metadata": {
"id": "QQOjh5BvnG8M",
"collapsed": true,
"cellView": "form"
},
"execution_count": null,
"outputs": []
},
{
"cell_type": "code",
"source": [
"# @title 🪐🖼️ -> 📝 Image to prompt : Add single token to existing prompt to match image\n",
"from google.colab import files\n",
"def getLocalFiles():\n",
" _files = files.upload()\n",
" if len(_files) >0:\n",
" for k,v in _files.items():\n",
" open(k,'wb').write(v)\n",
"\n",
"#Get image\n",
"# You can use \"http://images.cocodataset.org/val2017/000000039769.jpg\" for testing\n",
"url = \"http://images.cocodataset.org/val2017/000000039769.jpg\" # @param {\"type\":\"string\",\"placeholder\":\"leave empty for local upload\"}\n",
"from PIL import Image\n",
"import requests\n",
"if url == \"\":\n",
" image_A = getLocalFiles()\n",
"else:\n",
" image_A = Image.open(requests.get(url, stream=True).raw)\n",
"\n",
"\n",
"# Get image features\n",
"from transformers import CLIPProcessor, CLIPModel\n",
"processor = CLIPProcessor.from_pretrained(\"openai/clip-vit-large-patch14\" , clean_up_tokenization_spaces = True)\n",
"model = CLIPModel.from_pretrained(\"openai/clip-vit-large-patch14\")\n",
"inputs = processor(images=image_A, return_tensors=\"pt\")\n",
"image_features = model.get_image_features(**inputs)\n",
"text_encoding_A = image_features\n",
"A = text_encoding_A[0]\n",
"_A = LA.vector_norm(A, ord=2)\n",
"prompt_A = \"the image\"\n",
"name_A = prompt_A\n",
"#-----#\n",
"\n",
"# @markdown Set conditions for the output\n",
"must_start_with = \"\" # @param {\"type\":\"string\",\"placeholder\":\"write a text\"}\n",
"must_contain = \"\" # @param {\"type\":\"string\",\"placeholder\":\"write a text\"}\n",
"must_end_with = \"\" # @param {\"type\":\"string\",\"placeholder\":\"write a text\"}\n",
"token_B = must_contain\n",
"\n",
"# @markdown Limit the search\n",
"use_token_padding = True # @param {type:\"boolean\"}\n",
"start_search_at_ID = 12500 # @param {type:\"slider\", min:0, max: 49407, step:100}\n",
"search_range = 500 # @param {type:\"slider\", min:0, max: 2000, step:100}\n",
"restrictions = 'Suffix only' # @param [\"None\", \"Suffix only\", \"Prefix only\"]\n",
"\n",
"# @markdown Limit char size of included token\n",
"min_char_size = 3 # @param {type:\"slider\", min:0, max: 50, step:1}\n",
"char_range = 5 # @param {type:\"slider\", min:0, max: 50, step:1}\n",
"\n",
"#Tokenize input B\n",
"from transformers import AutoTokenizer\n",
"tokenizer = AutoTokenizer.from_pretrained(\"openai/clip-vit-large-patch14\", clean_up_tokenization_spaces = False)\n",
"tokenizer_output = tokenizer(text = token_B)\n",
"input_ids = tokenizer_output['input_ids']\n",
"#-----#\n",
"name_B = must_contain\n",
"#-----#\n",
"\n",
"START = start_search_at_ID\n",
"RANGE = min(search_range , 49407 - start_search_at_ID)\n",
"\n",
"dots = torch.zeros(RANGE)\n",
"is_BC = torch.zeros(RANGE)\n",
"for index in range(RANGE):\n",
" id_C = START + index\n",
" C = token[id_C]\n",
" _C = LA.vector_norm(C, ord=2)\n",
" name_C = vocab[id_C]\n",
"\n",
" # Decide if we should process prefix/suffix tokens\n",
" if name_C.find('</w>')<=-1:\n",
" if restrictions != \"Prefix only\":\n",
" continue\n",
" else:\n",
" if restrictions == \"Prefix only\":\n",
" continue\n",
" #-----#\n",
"\n",
" # Decide if char-size is within range\n",
" if len(name_C) < min_char_size:\n",
" continue\n",
" if len(name_C) > min_char_size + char_range:\n",
" continue\n",
" #-----#\n",
"\n",
" name_CB = must_start_with + name_C + name_B + must_end_with\n",
" if restrictions == \"Prefix only\":\n",
" name_CB = must_start_with + name_C + '-' + name_B + must_end_with\n",
" #-----#\n",
" ids_CB = processor.tokenizer(text=name_CB, padding=use_token_padding, return_tensors=\"pt\")\n",
" text_encoding_CB = model.get_text_features(**ids_CB)\n",
" CB = text_encoding_CB[0]\n",
" _CB = LA.vector_norm(CB, ord=2)\n",
" sim_CB = torch.dot(A,CB)/(_A*_CB)\n",
" #-----#\n",
" if restrictions == \"Prefix only\":\n",
" result = sim_CB\n",
" result = result.item()\n",
" dots[index] = result\n",
" continue\n",
" #-----#\n",
" name_BC = must_start_with + name_B + name_C + must_end_with\n",
" ids_BC = processor.tokenizer(text=name_BC, padding=use_token_padding, return_tensors=\"pt\")\n",
" text_encoding_BC = model.get_text_features(**ids_BC)\n",
" BC = text_encoding_BC[0]\n",
" _BC = LA.vector_norm(BC, ord=2)\n",
" sim_BC = torch.dot(A,BC)/(_A*_BC)\n",
" #-----#\n",
"\n",
" result = sim_CB\n",
" if(sim_BC > sim_CB):\n",
" is_BC[index] = 1\n",
" result = sim_BC\n",
"\n",
" #result = absolute_value(result.item())\n",
" result = result.item()\n",
" dots[index] = result\n",
"#----#\n",
"\n",
"sorted, indices = torch.sort(dots,dim=0 , descending=True)\n",
"\n",
"# @markdown Print options\n",
"list_size = 100 # @param {type:'number'}\n",
"print_ID = False # @param {type:\"boolean\"}\n",
"print_Similarity = True # @param {type:\"boolean\"}\n",
"print_Name = True # @param {type:\"boolean\"}\n",
"print_Divider = True # @param {type:\"boolean\"}\n",
"\n",
"\n",
"if (print_Divider):\n",
" print('//---//')\n",
"\n",
"print('')\n",
"print(f'These token pairings within the range ID = {START} to ID = {START + RANGE} most closely match the text_encoding for {prompt_A} : ')\n",
"print('')\n",
"\n",
"for index in range(min(list_size,RANGE)):\n",
" id = START + indices[index].item()\n",
" if (print_Name):\n",
" if(is_BC[index]>0):\n",
" print(must_start_with + name_B + vocab[id] + must_end_with)\n",
" else:\n",
" if restrictions == \"Prefix only\":\n",
" print(must_start_with + vocab[id] + '-' + name_B + must_end_with)\n",
" else:\n",
" print(must_start_with + vocab[id] + name_B + must_end_with)\n",
" if (print_ID):\n",
" print(f'ID = {id}') # IDs\n",
" if (print_Similarity):\n",
" print(f'similiarity = {round(sorted[index].item()*100,2)} %')\n",
" if (print_Divider):\n",
" print('--------')\n",
"\n",
"\n",
"\n",
"\n",
"\n"
],
"metadata": {
"collapsed": true,
"cellView": "form",
"id": "fi0jRruI0-tu",
"outputId": "6d7e8c39-a117-4b35-acfe-2a128c65aeb7",
"colab": {
"base_uri": "https://localhost:8080/"
}
},
"execution_count": 9,
"outputs": [
{
"output_type": "stream",
"name": "stdout",
"text": [
"//---//\n",
"\n",
"These token pairings within the range ID = 12500 to ID = 13000 most closely match the text_encoding for the prompt \"the image\" : \n",
"\n",
"sits</w>yellow\n",
"similiarity = 23.02 %\n",
"--------\n",
"neys</w>yellow\n",
"similiarity = 19.74 %\n",
"--------\n",
"cody</w>yellow\n",
"similiarity = 18.61 %\n",
"--------\n",
"wns</w>yellow\n",
"similiarity = 18.43 %\n",
"--------\n",
"java</w>yellow\n",
"similiarity = 18.15 %\n",
"--------\n",
"jj</w>yellow\n",
"similiarity = 18.03 %\n",
"--------\n",
"eno</w>yellow\n",
"similiarity = 17.87 %\n",
"--------\n",
"cled</w>yellow\n",
"similiarity = 17.85 %\n",
"--------\n",
"nom</w>yellow\n",
"similiarity = 17.75 %\n",
"--------\n",
"dads</w>yellow\n",
"similiarity = 17.5 %\n",
"--------\n",
"mil</w>yellow\n",
"similiarity = 17.47 %\n",
"--------\n",
"whom</w>yellow\n",
"similiarity = 17.37 %\n",
"--------\n",
"itv</w>yellow\n",
"similiarity = 17.34 %\n",
"--------\n",
"vibe</w>yellow\n",
"similiarity = 17.2 %\n",
"--------\n",
"noir</w>yellow\n",
"similiarity = 17.14 %\n",
"--------\n",
"yellowarel</w>\n",
"similiarity = 17.1 %\n",
"--------\n",
"#âĢ¦</w>yellow\n",
"similiarity = 17.04 %\n",
"--------\n",
"maya</w>yellow\n",
"similiarity = 17.03 %\n",
"--------\n",
"yellowbam</w>\n",
"similiarity = 17.01 %\n",
"--------\n",
"erts</w>yellow\n",
"similiarity = 17.01 %\n",
"--------\n",
"xc</w>yellow\n",
"similiarity = 16.98 %\n",
"--------\n",
"mob</w>yellow\n",
"similiarity = 16.89 %\n",
"--------\n",
"dees</w>yellow\n",
"similiarity = 16.87 %\n",
"--------\n",
"icc</w>yellow\n",
"similiarity = 16.75 %\n",
"--------\n",
"aly</w>yellow\n",
"similiarity = 16.63 %\n",
"--------\n",
"lis</w>yellow\n",
"similiarity = 16.63 %\n",
"--------\n",
"yellowturf</w>\n",
"similiarity = 16.62 %\n",
"--------\n",
"yellowbaba</w>\n",
"similiarity = 16.58 %\n",
"--------\n",
":*</w>yellow\n",
"similiarity = 16.42 %\n",
"--------\n",
"inho</w>yellow\n",
"similiarity = 16.39 %\n",
"--------\n",
"yellowhes</w>\n",
"similiarity = 16.37 %\n",
"--------\n",
"nity</w>yellow\n",
"similiarity = 16.3 %\n",
"--------\n",
"lust</w>yellow\n",
"similiarity = 16.3 %\n",
"--------\n",
"ikh</w>yellow\n",
"similiarity = 16.26 %\n",
"--------\n",
"nyt</w>yellow\n",
"similiarity = 16.24 %\n",
"--------\n",
"(+</w>yellow\n",
"similiarity = 16.11 %\n",
"--------\n",
"foto</w>yellow\n",
"similiarity = 16.11 %\n",
"--------\n",
"stl</w>yellow\n",
"similiarity = 16.06 %\n",
"--------\n",
"mick</w>yellow\n",
"similiarity = 16.06 %\n",
"--------\n",
"...@</w>yellow\n",
"similiarity = 16.05 %\n",
"--------\n",
"ugh</w>yellow\n",
"similiarity = 16.05 %\n",
"--------\n",
"gro</w>yellow\n",
"similiarity = 16.01 %\n",
"--------\n",
"wski</w>yellow\n",
"similiarity = 16.01 %\n",
"--------\n",
"ðŁĴ«</w>yellow\n",
"similiarity = 15.74 %\n",
"--------\n",
"deen</w>yellow\n",
"similiarity = 15.73 %\n",
"--------\n",
"assy</w>yellow\n",
"similiarity = 15.72 %\n",
"--------\n",
"mtv</w>yellow\n",
"similiarity = 15.72 %\n",
"--------\n",
"yellowðŁĺ»</w>\n",
"similiarity = 15.72 %\n",
"--------\n",
"yellowfrm</w>\n",
"similiarity = 15.65 %\n",
"--------\n",
"moss</w>yellow\n",
"similiarity = 15.64 %\n",
"--------\n",
"bart</w>yellow\n",
"similiarity = 15.61 %\n",
"--------\n",
"tw</w>yellow\n",
"similiarity = 15.51 %\n",
"--------\n",
"yellowplug</w>\n",
"similiarity = 15.46 %\n",
"--------\n",
"jen</w>yellow\n",
"similiarity = 15.45 %\n",
"--------\n",
"pst</w>yellow\n",
"similiarity = 15.43 %\n",
"--------\n",
"omfg</w>yellow\n",
"similiarity = 15.43 %\n",
"--------\n",
"dine</w>yellow\n",
"similiarity = 15.38 %\n",
"--------\n",
"vern</w>yellow\n",
"similiarity = 15.33 %\n",
"--------\n",
"reno</w>yellow\n",
"similiarity = 15.25 %\n",
"--------\n",
"yellow´</w>\n",
"similiarity = 15.14 %\n",
"--------\n",
"omic</w>yellow\n",
"similiarity = 15.14 %\n",
"--------\n",
"łï¸ı</w>yellow\n",
"similiarity = 15.11 %\n",
"--------\n",
"yellowgis</w>\n",
"similiarity = 15.06 %\n",
"--------\n",
"aunt</w>yellow\n",
"similiarity = 15.0 %\n",
"--------\n",
"joan</w>yellow\n",
"similiarity = 14.96 %\n",
"--------\n",
"anas</w>yellow\n",
"similiarity = 14.92 %\n",
"--------\n",
"ðŁĴĵ</w>yellow\n",
"similiarity = 14.9 %\n",
"--------\n",
"chad</w>yellow\n",
"similiarity = 14.89 %\n",
"--------\n",
"yellowsake</w>\n",
"similiarity = 14.88 %\n",
"--------\n",
"gues</w>yellow\n",
"similiarity = 14.84 %\n",
"--------\n",
"gian</w>yellow\n",
"similiarity = 14.84 %\n",
"--------\n",
"asi</w>yellow\n",
"similiarity = 14.83 %\n",
"--------\n",
"yellowoven</w>\n",
"similiarity = 14.82 %\n",
"--------\n",
"jury</w>yellow\n",
"similiarity = 14.79 %\n",
"--------\n",
"blvd</w>yellow\n",
"similiarity = 14.75 %\n",
"--------\n",
"omez</w>yellow\n",
"similiarity = 14.72 %\n",
"--------\n",
"yellowyang</w>\n",
"similiarity = 14.7 %\n",
"--------\n",
"gu</w>yellow\n",
"similiarity = 14.48 %\n",
"--------\n",
"yellowova</w>\n",
"similiarity = 14.45 %\n",
"--------\n",
"yellowinez</w>\n",
"similiarity = 14.44 %\n",
"--------\n",
"pei</w>yellow\n",
"similiarity = 14.44 %\n",
"--------\n",
"ãĢIJ</w>yellow\n",
"similiarity = 14.43 %\n",
"--------\n",
"ãĢij</w>yellow\n",
"similiarity = 14.43 %\n",
"--------\n",
"ðŁĮŀ</w>yellow\n",
"similiarity = 14.36 %\n",
"--------\n",
"ðŁĺĿ</w>yellow\n",
"similiarity = 14.27 %\n",
"--------\n",
"troy</w>yellow\n",
"similiarity = 14.16 %\n",
"--------\n",
"pale</w>yellow\n",
"similiarity = 14.14 %\n",
"--------\n",
"boi</w>yellow\n",
"similiarity = 14.11 %\n",
"--------\n",
"nn</w>yellow\n",
"similiarity = 14.08 %\n",
"--------\n",
"âı°</w>yellow\n",
"similiarity = 14.01 %\n",
"--------\n",
"ooth</w>yellow\n",
"similiarity = 13.93 %\n",
"--------\n",
"pied</w>yellow\n",
"similiarity = 13.9 %\n",
"--------\n",
"bola</w>yellow\n",
"similiarity = 13.79 %\n",
"--------\n",
"âŀ¡</w>yellow\n",
"similiarity = 13.77 %\n",
"--------\n",
"rena</w>yellow\n",
"similiarity = 13.75 %\n",
"--------\n",
"dley</w>yellow\n",
"similiarity = 13.73 %\n",
"--------\n",
"evan</w>yellow\n",
"similiarity = 13.67 %\n",
"--------\n",
"pony</w>yellow\n",
"similiarity = 13.63 %\n",
"--------\n",
"rene</w>yellow\n",
"similiarity = 13.62 %\n",
"--------\n",
"mock</w>yellow\n",
"similiarity = 13.57 %\n",
"--------\n"
]
}
]
},
{
"cell_type": "code",
"source": [
"# @title 🪐📝 Prompt to prompt : Add single token to existing prompt to match another prompt\n",
"# @markdown Write a text to match against...\n",
"prompt_A = \"photo of a banana\" # @param {\"type\":\"string\",\"placeholder\":\"Write a prompt\"}\n",
"\n",
"# @markdown Set conditions for the output\n",
"must_start_with = \"\" # @param {\"type\":\"string\",\"placeholder\":\"write a text\"}\n",
"must_contain = \"yellow\" # @param {\"type\":\"string\",\"placeholder\":\"write a text\"}\n",
"must_end_with = \"\" # @param {\"type\":\"string\",\"placeholder\":\"write a text\"}\n",
"token_B = must_contain\n",
"\n",
"# @markdown Limit the search\n",
"use_token_padding = True # @param {type:\"boolean\"}\n",
"start_search_at_ID = 12500 # @param {type:\"slider\", min:0, max: 49407, step:100}\n",
"search_range = 500 # @param {type:\"slider\", min:0, max: 2000, step:100}\n",
"restrictions = 'Suffix only' # @param [\"None\", \"Suffix only\", \"Prefix only\"]\n",
"\n",
"# @markdown Limit char size of included token\n",
"min_char_size = 3 # @param {type:\"slider\", min:0, max: 50, step:1}\n",
"char_range = 5 # @param {type:\"slider\", min:0, max: 50, step:1}\n",
"\n",
"#Tokenize input B\n",
"from transformers import AutoTokenizer\n",
"tokenizer = AutoTokenizer.from_pretrained(\"openai/clip-vit-large-patch14\", clean_up_tokenization_spaces = False)\n",
"tokenizer_output = tokenizer(text = token_B)\n",
"input_ids = tokenizer_output['input_ids']\n",
"#-----#\n",
"name_B = must_contain\n",
"#-----#\n",
"\n",
"from transformers import CLIPProcessor, CLIPModel\n",
"processor = CLIPProcessor.from_pretrained(\"openai/clip-vit-large-patch14\" , clean_up_tokenization_spaces = True)\n",
"model = CLIPModel.from_pretrained(\"openai/clip-vit-large-patch14\")\n",
"#-------#\n",
"ids_A = processor.tokenizer(text=prompt_A, padding=use_token_padding, return_tensors=\"pt\")\n",
"text_encoding_A = model.get_text_features(**ids_A)\n",
"A = text_encoding_A[0]\n",
"_A = LA.vector_norm(A, ord=2)\n",
"name_A = prompt_A\n",
"print(f'a text_encoding was created for the prompt \"{prompt_A}\" ')\n",
"print('')\n",
"#----#\n",
"\n",
"START = start_search_at_ID\n",
"RANGE = min(search_range , 49407 - start_search_at_ID)\n",
"\n",
"dots = torch.zeros(RANGE)\n",
"is_BC = torch.zeros(RANGE)\n",
"for index in range(RANGE):\n",
" id_C = START + index\n",
" C = token[id_C]\n",
" _C = LA.vector_norm(C, ord=2)\n",
" name_C = vocab[id_C]\n",
"\n",
" # Decide if we should process prefix/suffix tokens\n",
" if name_C.find('</w>')<=-1:\n",
" if restrictions != \"Prefix only\":\n",
" continue\n",
" else:\n",
" if restrictions == \"Prefix only\":\n",
" continue\n",
" #-----#\n",
"\n",
" # Decide if char-size is within range\n",
" if len(name_C) < min_char_size:\n",
" continue\n",
" if len(name_C) > min_char_size + char_range:\n",
" continue\n",
" #-----#\n",
"\n",
" name_CB = must_start_with + name_C + name_B + must_end_with\n",
" if restrictions == \"Prefix only\":\n",
" name_CB = must_start_with + name_C + '-' + name_B + must_end_with\n",
" #-----#\n",
" ids_CB = processor.tokenizer(text=name_CB, padding=use_token_padding, return_tensors=\"pt\")\n",
" text_encoding_CB = model.get_text_features(**ids_CB)\n",
" CB = text_encoding_CB[0]\n",
" _CB = LA.vector_norm(CB, ord=2)\n",
" sim_CB = torch.dot(A,CB)/(_A*_CB)\n",
" #-----#\n",
" if restrictions == \"Prefix only\":\n",
" result = sim_CB\n",
" result = result.item()\n",
" dots[index] = result\n",
" continue\n",
" #-----#\n",
" name_BC = must_start_with + name_B + name_C + must_end_with\n",
" ids_BC = processor.tokenizer(text=name_BC, padding=use_token_padding, return_tensors=\"pt\")\n",
" text_encoding_BC = model.get_text_features(**ids_BC)\n",
" BC = text_encoding_BC[0]\n",
" _BC = LA.vector_norm(BC, ord=2)\n",
" sim_BC = torch.dot(A,BC)/(_A*_BC)\n",
" #-----#\n",
"\n",
" result = sim_CB\n",
" if(sim_BC > sim_CB):\n",
" is_BC[index] = 1\n",
" result = sim_BC\n",
"\n",
" #result = absolute_value(result.item())\n",
" result = result.item()\n",
" dots[index] = result\n",
"#----#\n",
"\n",
"sorted, indices = torch.sort(dots,dim=0 , descending=True)\n",
"\n",
"# @markdown Print options\n",
"list_size = 100 # @param {type:'number'}\n",
"print_ID = False # @param {type:\"boolean\"}\n",
"print_Similarity = True # @param {type:\"boolean\"}\n",
"print_Name = True # @param {type:\"boolean\"}\n",
"print_Divider = True # @param {type:\"boolean\"}\n",
"\n",
"\n",
"if (print_Divider):\n",
" print('//---//')\n",
"\n",
"print('')\n",
"print(f'These token pairings within the range ID = {START} to ID = {START + RANGE} most closely match the text_encoding for the prompt \"{prompt_A}\" : ')\n",
"print('')\n",
"\n",
"for index in range(min(list_size,RANGE)):\n",
" id = START + indices[index].item()\n",
" if (print_Name):\n",
" if(is_BC[index]>0):\n",
" print(must_start_with + name_B + vocab[id] + must_end_with)\n",
" else:\n",
" if restrictions == \"Prefix only\":\n",
" print(must_start_with + vocab[id] + '-' + name_B + must_end_with)\n",
" else:\n",
" print(must_start_with + vocab[id] + name_B + must_end_with)\n",
" if (print_ID):\n",
" print(f'ID = {id}') # IDs\n",
" if (print_Similarity):\n",
" print(f'similiarity = {round(sorted[index].item()*100,2)} %')\n",
" if (print_Divider):\n",
" print('--------')"
],
"metadata": {
"cellView": "form",
"id": "uDtcm-l8UCJk"
},
"execution_count": null,
"outputs": []
},
{
"cell_type": "markdown",
"source": [
"# ↓ Sub modules (use these to build your own projects) ↓"
],
"metadata": {
"id": "_d8WtPgtAymM"
}
},
{
"cell_type": "code",
"source": [
"# @title 📝 -> 🆔 Tokenize prompt into IDs\n",
"from transformers import AutoTokenizer\n",
"tokenizer = AutoTokenizer.from_pretrained(\"openai/clip-vit-large-patch14\", clean_up_tokenization_spaces = False)\n",
"\n",
"prompt= \"banana\" # @param {type:'string'}\n",
"\n",
"tokenizer_output = tokenizer(text = prompt)\n",
"input_ids = tokenizer_output['input_ids']\n",
"print(input_ids)\n",
"\n",
"\n",
"#The prompt will be enclosed with the <|start-of-text|> and <|end-of-text|> tokens, which is why output will be [49406, ... , 49407].\n",
"\n",
"#You can leave the 'prompt' field empty to get a random value tensor. Since the tensor is random value, it will not correspond to any tensor in the vocab.json list , and this it will have no ID."
],
"metadata": {
"id": "RPdkYzT2_X85",
"cellView": "form"
},
"execution_count": null,
"outputs": []
},
{
"cell_type": "code",
"source": [
"# @title 🆔->🥢 Take the ID at index 1 from above result and get its corresponding tensor value\n",
"\n",
"id_A = input_ids[1]\n",
"A = token[id_A]\n",
"_A = LA.vector_norm(A, ord=2)\n",
"\n",
"#if no imput exists we just randomize the entire thing\n",
"if (prompt == \"\"):\n",
" id_A = -1\n",
" print(\"Tokenized prompt tensor A is a random valued tensor with no ID\")\n",
" R = torch.rand(768)\n",
" _R = LA.vector_norm(R, ord=2)\n",
" A = R*(_A/_R)\n",
"\n",
"#Save a copy of the tensor A\n",
"id_P = id_A\n",
"P = A\n",
"_P = LA.vector_norm(A, ord=2)\n"
],
"metadata": {
"id": "YqdiF8DIz9Wu",
"cellView": "form"
},
"execution_count": null,
"outputs": []
},
{
"cell_type": "code",
"source": [
"# @title 🥢 -> 🥢🔀 Take the ID at index 1 from above result and modify it (optional)\n",
"mix_with = \"\" # @param {type:'string'}\n",
"mix_method = \"None\" # @param [\"None\" , \"Average\", \"Subtract\"] {allow-input: true}\n",
"w = 0.5 # @param {type:\"slider\", min:0, max:1, step:0.01}\n",
"\n",
"#------#\n",
"#If set to TRUE , this will use the output of this cell , tensor A, as the input of this cell the 2nd time we run it. Use this feature to mix many tokens into A\n",
"re_iterate_tensor_A = True # @param {\"type\":\"boolean\"}\n",
"if (re_iterate_tensor_A == False) :\n",
" #prevent re-iterating A by reading from stored copy\n",
" id_A = id_P\n",
" A = P\n",
" _A = _P\n",
"#----#\n",
"\n",
"tokenizer_output = tokenizer(text = mix_with)\n",
"input_ids = tokenizer_output['input_ids']\n",
"id_C = input_ids[1]\n",
"C = token[id_C]\n",
"_C = LA.vector_norm(C, ord=2)\n",
"\n",
"#if no imput exists we just randomize the entire thing\n",
"if (mix_with == \"\"):\n",
" id_C = -1\n",
" print(\"Tokenized prompt 'mix_with' tensor C is a random valued tensor with no ID\")\n",
" R = torch.rand(768)\n",
" _R = LA.vector_norm(R, ord=2)\n",
" C = R*(_C/_R)\n",
"\n",
"if (mix_method == \"None\"):\n",
" print(\"No operation\")\n",
"\n",
"if (mix_method == \"Average\"):\n",
" A = w*A + (1-w)*C\n",
" _A = LA.vector_norm(A, ord=2)\n",
" print(\"Tokenized prompt tensor A has been recalculated as A = w*A + (1-w)*C , where C is the tokenized prompt 'mix_with' tensor C\")\n",
"\n",
"if (mix_method == \"Subtract\"):\n",
" tmp = (A/_A) - (C/_C)\n",
" _tmp = LA.vector_norm(tmp, ord=2)\n",
" A = tmp*((w*_A + (1-w)*_C)/_tmp)\n",
" _A = LA.vector_norm(A, ord=2)\n",
" print(\"Tokenized prompt tensor A has been recalculated as A = (w*_A + (1-w)*_C) * norm(w*A - (1-w)*C) , where C is the tokenized prompt 'mix_with' tensor C\")\n",
"\n",
"#OPTIONAL : Add/subtract + normalize above result with another token. Leave field empty to get a random value tensor"
],
"metadata": {
"id": "oXbNSRSKPgRr",
"collapsed": true,
"cellView": "form"
},
"execution_count": null,
"outputs": []
},
{
"cell_type": "code",
"source": [
"\n",
"# @title 🥢->🧾🥢 Find Similiar Tokens to ID at index 1 from above result\n",
"dots = torch.zeros(NUM_TOKENS)\n",
"for index in range(NUM_TOKENS):\n",
" id_B = index\n",
" B = token[id_B]\n",
" _B = LA.vector_norm(B, ord=2)\n",
" result = torch.dot(A,B)/(_A*_B)\n",
" #result = absolute_value(result.item())\n",
" result = result.item()\n",
" dots[index] = result\n",
"\n",
"name_A = \"A of random type\"\n",
"if (id_A>-1):\n",
" name_A = vocab[id_A]\n",
"\n",
"name_C = \"token C of random type\"\n",
"if (id_C>-1):\n",
" name_C = vocab[id_C]\n",
"\n",
"\n",
"sorted, indices = torch.sort(dots,dim=0 , descending=True)\n",
"#----#\n",
"if (mix_method == \"Average\"):\n",
" print(f'Calculated all cosine-similarities between the average of token {name_A} and {name_C} with Id_A = {id_A} and mixed Id_C = {id_C} as a 1x{sorted.shape[0]} tensor')\n",
"if (mix_method == \"Subtract\"):\n",
" print(f'Calculated all cosine-similarities between the subtract of token {name_A} and {name_C} with Id_A = {id_A} and mixed Id_C = {id_C} as a 1x{sorted.shape[0]} tensor')\n",
"if (mix_method == \"None\"):\n",
" print(f'Calculated all cosine-similarities between the token {name_A} with Id_A = {id_A} with the the rest of the {NUM_TOKENS} tokens as a 1x{sorted.shape[0]} tensor')\n",
"\n",
"#Produce a list id IDs that are most similiar to the prompt ID at positiion 1 based on above result"
],
"metadata": {
"id": "juxsvco9B0iV",
"collapsed": true,
"cellView": "form"
},
"execution_count": null,
"outputs": []
},
{
"cell_type": "markdown",
"source": [],
"metadata": {
"id": "cYYu5C5C6MHH"
}
},
{
"cell_type": "code",
"source": [
"# @title 🥢🧾 -> 🖨️ Print Result from the 'Similiar Tokens' list from above result\n",
"list_size = 100 # @param {type:'number'}\n",
"print_ID = False # @param {type:\"boolean\"}\n",
"print_Similarity = True # @param {type:\"boolean\"}\n",
"print_Name = True # @param {type:\"boolean\"}\n",
"print_Divider = True # @param {type:\"boolean\"}\n",
"\n",
"for index in range(list_size):\n",
" id = indices[index].item()\n",
" if (print_Name):\n",
" print(f'{vocab[id]}') # vocab item\n",
" if (print_ID):\n",
" print(f'ID = {id}') # IDs\n",
" if (print_Similarity):\n",
" print(f'similiarity = {round(sorted[index].item()*100,2)} %') # % value\n",
" if (print_Divider):\n",
" print('--------')\n",
"\n",
"#Print the sorted list from above result"
],
"metadata": {
"id": "YIEmLAzbHeuo",
"collapsed": true,
"cellView": "form"
},
"execution_count": null,
"outputs": []
},
{
"cell_type": "code",
"source": [
"\n",
"# @title 🆔 Get similarity % of two token IDs\n",
"id_for_token_A = 4567 # @param {type:'number'}\n",
"id_for_token_B = 4343 # @param {type:'number'}\n",
"\n",
"similarity_str = 'The similarity between tokens A and B is ' + similarity(id_for_token_A , id_for_token_B)\n",
"\n",
"print(similarity_str)\n",
"\n",
"#Valid ID ranges for id_for_token_A / id_for_token_B are between 0 and 49407"
],
"metadata": {
"id": "MwmOdC9cNZty",
"collapsed": true,
"cellView": "form"
},
"execution_count": null,
"outputs": []
},
{
"cell_type": "markdown",
"source": [
"\n",
"\n",
"# How does this notebook work?\n",
"\n",
"Similiar vectors = similiar output in the SD 1.5 / SDXL / FLUX model\n",
"\n",
"CLIP converts the prompt text to vectors (“tensors”) , with float32 values usually ranging from -1 to 1.\n",
"\n",
"Dimensions are \\[ 1x768 ] tensors for SD 1.5 , and a \\[ 1x768 , 1x1024 ] tensor for SDXL and FLUX.\n",
"\n",
"The SD models and FLUX converts these vectors to an image.\n",
"\n",
"This notebook takes an input string , tokenizes it and matches the first token against the 49407 token vectors in the vocab.json : [https://huggingface.co/black-forest-labs/FLUX.1-dev/tree/main/tokenizer](https://www.google.com/url?q=https%3A%2F%2Fhuggingface.co%2Fblack-forest-labs%2FFLUX.1-dev%2Ftree%2Fmain%2Ftokenizer)\n",
"\n",
"It finds the “most similiar tokens” in the list. Similarity is the theta angle between the token vectors.\n",
"\n",
"<div>\n",
"<img src=\"https://huggingface.co/datasets/codeShare/sd_tokens/resolve/main/cosine.jpeg\" width=\"300\"/>\n",
"</div>\n",
"\n",
"The angle is calculated using cosine similarity , where 1 = 100% similarity (parallell vectors) , and 0 = 0% similarity (perpendicular vectors).\n",
"\n",
"Negative similarity is also possible.\n",
"\n",
"# How can I use it?\n",
"\n",
"If you are bored of prompting “girl” and want something similiar you can run this notebook and use the “chick” token at 21.88% similarity , for example\n",
"\n",
"You can also run a mixed search , like “cute+girl”/2 , where for example “kpop” has a 16.71% similarity\n",
"\n",
"There are some strange tokens further down the list you go. Example: tokens similiar to the token \"pewdiepie</w>\" (yes this is an actual token that exists in CLIP)\n",
"\n",
"<div>\n",
"<img src=\"https://lemmy.world/pictrs/image/a1cd284e-3341-4284-9949-5f8b58d3bd1f.jpeg\" width=\"300\"/>\n",
"</div>\n",
"\n",
"Each of these correspond to a unique 1x768 token vector.\n",
"\n",
"The higher the ID value , the less often the token appeared in the CLIP training data.\n",
"\n",
"To reiterate; this is the CLIP model training data , not the SD-model training data.\n",
"\n",
"So for certain models , tokens with high ID can give very consistent results , if the SD model is trained to handle them.\n",
"\n",
"Example of this can be anime models , where japanese artist names can affect the output greatly. \n",
"\n",
"Tokens with high ID will often give the \"fun\" output when used in very short prompts.\n",
"\n",
"# What about token vector length?\n",
"\n",
"If you are wondering about token magnitude,\n",
"Prompt weights like (banana:1.2) will scale the magnitude of the corresponding 1x768 tensor(s) by 1.2 . So thats how prompt token magnitude works.\n",
"\n",
"Source: [https://huggingface.co/docs/diffusers/main/en/using-diffusers/weighted\\_prompts](https://www.google.com/url?q=https%3A%2F%2Fhuggingface.co%2Fdocs%2Fdiffusers%2Fmain%2Fen%2Fusing-diffusers%2Fweighted_prompts)\\*\n",
"\n",
"So TLDR; vector direction = “what to generate” , vector magnitude = “prompt weights”\n",
"\n",
"# How prompting works (technical summary)\n",
"\n",
" 1. There is no correct way to prompt.\n",
"\n",
"2. Stable diffusion reads your prompt left to right, one token at a time, finding association _from_ the previous token _to_ the current token _and to_ the image generated thus far (Cross Attention Rule)\n",
"\n",
"3. Stable Diffusion is an optimization problem that seeks to maximize similarity to prompt and minimize similarity to negatives (Optimization Rule)\n",
"\n",
"Reference material (covers entire SD , so not good source material really, but the info is there) : https://youtu.be/sFztPP9qPRc?si=ge2Ty7wnpPGmB0gi\n",
"\n",
"# The SD pipeline\n",
"\n",
"For every step (20 in total by default) for SD1.5 :\n",
"\n",
"1. Prompt text => (tokenizer)\n",
"2. => Nx768 token vectors =>(CLIP model) =>\n",
"3. 1x768 encoding => ( the SD model / Unet ) =>\n",
"4. => _Desired_ image per Rule 3 => ( sampler)\n",
"5. => Paint a section of the image => (image)\n",
"\n",
"# Disclaimer /Trivia\n",
"\n",
"This notebook should be seen as a \"dictionary search tool\" for the vocab.json , which is the same for SD1.5 , SDXL and FLUX. Feel free to verify this by checking the 'tokenizer' folder under each model.\n",
"\n",
"vocab.json in the FLUX model , for example (1 of 2 copies) : https://huggingface.co/black-forest-labs/FLUX.1-dev/tree/main/tokenizer\n",
"\n",
"I'm using Clip-vit-large-patch14 , which is used in SD 1.5 , and is one among the two tokenizers for SDXL and FLUX : https://huggingface.co/openai/clip-vit-large-patch14/blob/main/README.md\n",
"\n",
"This set of tokens has dimension 1x768. \n",
"\n",
"SDXL and FLUX uses an additional set of tokens of dimension 1x1024.\n",
"\n",
"These are not included in this notebook. Feel free to include them yourselves (I would appreciate that).\n",
"\n",
"To do so, you will have to download a FLUX and/or SDXL model\n",
"\n",
", and copy the 49407x1024 tensor list that is stored within the model and then save it as a .pt file.\n",
"\n",
"//---//\n",
"\n",
"I am aware it is actually the 1x768 text_encoding being processed into an image for the SD models + FLUX.\n",
"\n",
"As such , I've included text_encoding comparison at the bottom of the Notebook.\n",
"\n",
"I am also aware thar SDXL and FLUX uses additional encodings , which are not included in this notebook.\n",
"\n",
"* Clip-vit-bigG for SDXL: https://huggingface.co/laion/CLIP-ViT-bigG-14-laion2B-39B-b160k/blob/main/README.md\n",
"\n",
"* And the T5 text encoder for FLUX. I have 0% understanding of FLUX T5 text_encoder.\n",
"\n",
"//---//\n",
"\n",
"If you want them , feel free to include them yourself and share the results (cuz I probably won't) :)!\n",
"\n",
"That being said , being an encoding , I reckon the CLIP Nx768 => 1x768 should be \"linear\" (or whatever one might call it)\n",
"\n",
"So exchange a few tokens in the Nx768 for something similiar , and the resulting 1x768 ought to be kinda similar to 1x768 we had earlier. Hopefully.\n",
"\n",
"I feel its important to mention this , in case some wonder why the token-token similarity don't match the text-encoding to text-encoding similarity.\n",
"\n",
"# Note regarding CLIP text encoding vs. token\n",
"\n",
"*To make this disclaimer clear; Token-to-token similarity is not the same as text_encoding similarity.*\n",
"\n",
"I have to say this , since it will otherwise get (even more) confusing , as both the individual tokens , and the text_encoding have dimensions 1x768.\n",
"\n",
"They are separate things. Separate results. etc.\n",
"\n",
"As such , you will not get anything useful if you start comparing similarity between a token , and a text-encoding. So don't do that :)!\n",
"\n",
"# What about the CLIP image encoding?\n",
"\n",
"The CLIP model can also do an image_encoding of an image, where the output will be a 1x768 tensor. These _can_ be compared with the text_encoding.\n",
"\n",
"Comparing CLIP image_encoding with the CLIP text_encoding for a bunch of random prompts until you find the \"highest similarity\" , is a method used in the CLIP interrogator : https://huggingface.co/spaces/pharmapsychotic/CLIP-Interrogator\n",
"\n",
"List of random prompts for CLIP interrogator can be found here, for reference : https://github.com/pharmapsychotic/clip-interrogator/tree/main/clip_interrogator/data\n",
"\n",
"The CLIP image_encoding is not included in this Notebook.\n",
"\n",
"If you spot errors / ideas for improvememts; feel free to fix the code in your own notebook and post the results.\n",
"\n",
"I'd appreciate that over people saying \"your math is wrong you n00b!\" with no constructive feedback.\n",
"\n",
"//---//\n",
"\n",
"Regarding output\n",
"\n",
"# What are the </w> symbols?\n",
"\n",
"The whitespace symbol indicate if the tokenized item ends with whitespace ( the suffix \"banana</w>\" => \"banana \" ) or not (the prefix \"post\" in \"post-apocalyptic \")\n",
"\n",
"For ease of reference , I call them prefix-tokens and suffix-tokens.\n",
"\n",
"Sidenote:\n",
"\n",
"Prefix tokens have the unique property in that they \"mutate\" suffix tokens\n",
"\n",
"Example: \"photo of a #prefix#-banana\"\n",
"\n",
"where #prefix# is a randomly selected prefix-token from the vocab.json\n",
"\n",
"The hyphen \"-\" exists to guarantee the tokenized text splits into the written #prefix# and #suffix# token respectively. The \"-\" hypen symbol can be replaced by any other special character of your choosing.\n",
"\n",
" Capital letters work too , e.g \"photo of a #prefix#Abanana\" since the capital letters A-Z are only listed once in the entire vocab.json.\n",
"\n",
"You can also choose to omit any separator and just rawdog it with the prompt \"photo of a #prefix#banana\" , however know that this may , on occasion , be tokenized as completely different tokens of lower ID:s.\n",
"\n",
"Curiously , common NSFW terms found online have in the CLIP model have been purposefully fragmented into separate #prefix# and #suffix# counterparts in the vocab.json. Likely for PR-reasons.\n",
"\n",
"You can verify the results using this online tokenizer: https://sd-tokenizer.rocker.boo/\n",
"\n",
"<div>\n",
"<img src=\"https://lemmy.world/pictrs/image/43467d75-7406-4a13-93ca-cdc469f944fc.jpeg\" width=\"300\"/>\n",
"<img src=\"https://lemmy.world/pictrs/image/c0411565-0cb3-47b1-a788-b368924d6f17.jpeg\" width=\"300\"/>\n",
"<img src=\"https://lemmy.world/pictrs/image/c27c6550-a88b-4543-9bd7-067dff016be2.jpeg\" width=\"300\"/>\n",
"</div>\n",
"\n",
"# What is that gibberish tokens that show up?\n",
"\n",
"The gibberish tokens like \"ðŁĺħ\\</w>\" are actually emojis!\n",
"\n",
"Try writing some emojis in this online tokenizer to see the results: https://sd-tokenizer.rocker.boo/\n",
"\n",
"It is a bit borked as it can't process capital letters properly.\n",
"\n",
"Also note that this is not reversible.\n",
"\n",
"If tokenization \"😅\" => ðŁĺħ</w>\n",
"\n",
"Then you can't prompt \"ðŁĺħ\" and expect to get the same result as the tokenized original emoji , \"😅\".\n",
"\n",
"SD 1.5 models actually have training for Emojis.\n",
"\n",
"But you have to set CLIP skip to 1 for this to work is intended.\n",
"\n",
"For example, this is the result from \"photo of a 🧔🏻♂️\"\n",
"\n",
"\n",
"<div>\n",
"<img src=\"https://lemmy.world/pictrs/image/e2b51aea-6960-4ad0-867e-8ce85f2bd51e.jpeg\" width=\"300\"/>\n",
"</div>\n",
"\n",
"A tutorial on stuff you can do with the vocab.list concluded.\n",
"\n",
"Anyways, have fun with the notebook.\n",
"\n",
"There might be some updates in the future with features not mentioned here.\n",
"\n",
"//---//\n",
"\n",
"https://codeandlife.com/2023/01/26/mastering-the-huggingface-clip-model-how-to-extract-embeddings-and-calculate-similarity-for-text-and-images/"
],
"metadata": {
"id": "njeJx_nSSA8H"
}
}
]
} |