Spaces:
Running
Running
File size: 53,721 Bytes
e9e637b | 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 | {
"cells": [
{
"cell_type": "code",
"execution_count": 1,
"id": "cell-01-install",
"metadata": {},
"outputs": [
{
"ename": "",
"evalue": "",
"output_type": "error",
"traceback": [
"\u001b[1;31mThe Kernel crashed while executing code in the current cell or a previous cell. \n",
"\u001b[1;31mPlease review the code in the cell(s) to identify a possible cause of the failure. \n",
"\u001b[1;31mClick <a href='https://aka.ms/vscodeJupyterKernelCrash'>here</a> for more info. \n",
"\u001b[1;31mView Jupyter <a href='command:jupyter.viewOutput'>log</a> for further details."
]
},
{
"ename": "",
"evalue": "",
"output_type": "error",
"traceback": [
"\u001b[1;31mCanceled future for execute_request message before replies were done"
]
},
{
"ename": "",
"evalue": "",
"output_type": "error",
"traceback": [
"\u001b[1;31mCanceled future for execute_request message before replies were done. \n",
"\u001b[1;31mView Jupyter <a href='command:jupyter.viewOutput'>log</a> for further details."
]
}
],
"source": [
"# ╔══════════════════════════════════════════════════════════════╗\n",
"# ║ CELL 1 — Install packages ║\n",
"# ║ RUN THIS CELL ALONE FIRST — it will auto-restart runtime ║\n",
"# ╚══════════════════════════════════════════════════════════════╝\n",
"import subprocess, sys, os\n",
"\n",
"def pip(*pkgs):\n",
" subprocess.check_call([sys.executable, \"-m\", \"pip\", \"install\", \"-q\", *pkgs])\n",
"\n",
"# No pinned numpy/pandas — let Colab use its pre-installed compatible versions.\n",
"# Pinning old numpy/pandas causes the 'mtrand ABI mismatch' ValueError.\n",
"pip(\n",
" \"datasets>=2.18.0\",\n",
" \"transformers>=4.40.0\",\n",
" \"sentence-transformers>=2.7.0\",\n",
" \"scikit-learn>=1.4.0\",\n",
" \"tqdm>=4.66.0\",\n",
" \"accelerate>=0.26.0\",\n",
" \"evaluate\",\n",
")\n",
"\n",
"print(\"✅ Packages installed — restarting runtime now …\")\n",
"os.kill(os.getpid(), 9) # auto-restart; Colab reconnects in ~5 s"
]
},
{
"cell_type": "code",
"execution_count": 5,
"id": "22cff354",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m527.3/527.3 kB\u001b[0m \u001b[31m16.3 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
"\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m177.6/177.6 kB\u001b[0m \u001b[31m20.4 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
"\u001b[?25h\u001b[31mERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.\n",
"gcsfs 2025.3.0 requires fsspec==2025.3.0, but you have fsspec 2024.6.1 which is incompatible.\u001b[0m\u001b[31m\n",
"\u001b[0m"
]
}
],
"source": [
"!pip install -q \"datasets==2.21.0\""
]
},
{
"cell_type": "code",
"execution_count": 1,
"id": "cell-02-gpu",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n",
"CUDA available: True\n",
"Device name: Tesla T4\n"
]
}
],
"source": [
"# ╔══════════════════════════════════════════════════════════════╗\n",
"# ║ CELL 2 — Mount Drive + GPU check ║\n",
"# ║ Run AFTER the runtime has restarted ║\n",
"# ╚══════════════════════════════════════════════════════════════╝\n",
"from google.colab import drive\n",
"drive.mount(\"/content/drive\")\n",
"\n",
"import torch\n",
"print(\"CUDA available:\", torch.cuda.is_available())\n",
"print(\"Device name: \", torch.cuda.get_device_name(0) if torch.cuda.is_available() else \"CPU\")\n",
"assert torch.cuda.is_available(), \"❌ No GPU — set Runtime type to T4 GPU!\"\n",
"DEVICE = 0"
]
},
{
"cell_type": "code",
"execution_count": 2,
"id": "cell-03-config",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"✅ Config ready | output → /content/drive/MyDrive/Athernex/nli_contract_model_final\n"
]
}
],
"source": [
"# ╔══════════════════════════════════════════════════════════════╗\n",
"# ║ CELL 3 — Config ║\n",
"# ╚══════════════════════════════════════════════════════════════╝\n",
"import os\n",
"\n",
"CNLI_SIZE = 6820 # full ContractNLI train split\n",
"MNLI_SIZE = 50000 # pool; genre filter keeps ~8-10k government rows\n",
"SYNTH_SIZE = 1000 # synthetic contradiction pairs\n",
"\n",
"BASE_MODEL = \"typeform/distilbert-base-uncased-mnli\"\n",
"OUTPUT_DIR = \"/content/drive/MyDrive/Athernex/nli_contract_model_final\"\n",
"EPOCHS = 5\n",
"BATCH_SIZE = 32 # T4 handles 32 at max_length=128\n",
"LR = 2e-5\n",
"MAX_LEN = 128\n",
"\n",
"LABEL2ID = {\"entailment\": 0, \"contradiction\": 1, \"neutral\": 2}\n",
"ID2LABEL = {v: k for k, v in LABEL2ID.items()}\n",
"\n",
"os.makedirs(OUTPUT_DIR, exist_ok=True)\n",
"print(f\"✅ Config ready | output → {OUTPUT_DIR}\")"
]
},
{
"cell_type": "code",
"execution_count": 3,
"id": "cell-04-helpers",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"✅ Data helpers defined.\n"
]
}
],
"source": [
"# ╔══════════════════════════════════════════════════════════════╗\n",
"# ║ CELL 4 — Data loading helpers ║\n",
"# ╚══════════════════════════════════════════════════════════════╝\n",
"import re\n",
"import pandas as pd\n",
"from datasets import load_dataset\n",
"\n",
"def clean_clause(text: str) -> str:\n",
" text = re.sub(r'\\s+', ' ', text).strip()\n",
" text = re.sub(r'[^\\x00-\\x7F]+', '', text)\n",
" return text\n",
"\n",
"def load_contract_nli(split: str = \"train\", size: int = CNLI_SIZE):\n",
" \"\"\"Full ContractNLI — kiddothe2b/contract-nli, subset contractnli_a.\"\"\"\n",
" slice_str = f\"{split}[:{size}]\" if size else split\n",
" return load_dataset(\n",
" \"kiddothe2b/contract-nli\", \"contractnli_a\",\n",
" split=slice_str, trust_remote_code=True\n",
" )\n",
"\n",
"def process_contract_nli(dataset) -> pd.DataFrame:\n",
" \"\"\"kiddothe2b schema: 0=contradiction, 1=entailment, 2=neutral.\"\"\"\n",
" label_map = {0: \"contradiction\", 1: \"entailment\", 2: \"neutral\"}\n",
" records = []\n",
" for s in dataset:\n",
" p = clean_clause(s[\"premise\"])\n",
" h = clean_clause(s[\"hypothesis\"])\n",
" if len(p) < 20 or len(h) < 20:\n",
" continue\n",
" records.append({\"clause1\": p, \"clause2\": h,\n",
" \"label\": label_map.get(s[\"label\"], \"neutral\")})\n",
" return pd.DataFrame(records)\n",
"\n",
"def load_mnli_government(split: str = \"train\", size: int = MNLI_SIZE):\n",
" \"\"\"MultiNLI filtered to government genre.\"\"\"\n",
" if split == \"validation\":\n",
" split = \"validation_matched\"\n",
" slice_str = f\"{split}[:{size}]\" if size else split\n",
" ds = load_dataset(\"nyu-mll/multi_nli\", split=slice_str, trust_remote_code=True)\n",
" return ds.filter(lambda x: x[\"genre\"] == \"government\")\n",
"\n",
"def process_mnli_government(dataset) -> pd.DataFrame:\n",
" \"\"\"MultiNLI schema: 0=entailment, 1=neutral, 2=contradiction.\"\"\"\n",
" label_map = {0: \"entailment\", 1: \"neutral\", 2: \"contradiction\"}\n",
" records = []\n",
" for s in dataset:\n",
" if not s[\"premise\"] or not s[\"hypothesis\"]:\n",
" continue\n",
" p = clean_clause(s[\"premise\"])\n",
" h = clean_clause(s[\"hypothesis\"])\n",
" if len(p) < 20 or len(h) < 20:\n",
" continue\n",
" records.append({\"clause1\": p, \"clause2\": h,\n",
" \"label\": label_map.get(s[\"label\"], \"neutral\")})\n",
" return pd.DataFrame(records)\n",
"\n",
"NEGATION_MAP = {\n",
" \"shall\": \"shall not\", \"must\": \"must not\",\n",
" \"will\": \"will not\", \"may\": \"may not\",\n",
" \"is required to\": \"is not required to\",\n",
" \"exclusive\": \"non-exclusive\", \"limited\": \"unlimited\",\n",
" \"terminate\": \"not terminate\",\n",
"}\n",
"\n",
"def simulate_contradiction(clause: str):\n",
" for term, negated in NEGATION_MAP.items():\n",
" if term in clause.lower():\n",
" return re.sub(term, negated, clause, count=1, flags=re.IGNORECASE)\n",
" return None\n",
"\n",
"def build_synthetic_pairs(clauses: list, sample_size: int = SYNTH_SIZE) -> pd.DataFrame:\n",
" import random; random.seed(42)\n",
" sampled = random.sample(clauses, min(sample_size, len(clauses)))\n",
" records = []\n",
" for clause in sampled:\n",
" neg = simulate_contradiction(clause)\n",
" if neg:\n",
" records.append({\"clause1\": clause, \"clause2\": neg, \"label\": \"contradiction\"})\n",
" return pd.DataFrame(records)\n",
"\n",
"print(\"✅ Data helpers defined.\")"
]
},
{
"cell_type": "code",
"execution_count": 4,
"id": "cell-05-build",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"=======================================================\n",
"BUILDING FULL TRAINING DATA\n",
"=======================================================\n",
"\n",
"[1/3] ContractNLI (size=6820) ...\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"/usr/local/lib/python3.12/dist-packages/huggingface_hub/utils/_auth.py:103: UserWarning: \n",
"Error while fetching `HF_TOKEN` secret value from your vault: 'Requesting secret HF_TOKEN timed out. Secrets can only be fetched when running from the Colab UI.'.\n",
"You are not authenticated with the Hugging Face Hub in this notebook.\n",
"If the error persists, please let us know by opening an issue on GitHub (https://github.com/huggingface/huggingface_hub/issues/new).\n",
" warnings.warn(\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "3828de1843b244eab1864397b1be07ec",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Downloading data: 0%| | 0.00/796k [00:00<?, ?B/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "6d6569e0dd754662b7e7c7da5704ce01",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Downloading data: 0%| | 0.00/213k [00:00<?, ?B/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "99d706588b8f46b5bafd97aca5302e57",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Downloading data: 0%| | 0.00/114k [00:00<?, ?B/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "cb547e43501246ab9077fd18cfa7e2ca",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Generating train split: 0%| | 0/6819 [00:00<?, ? examples/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "3bcc04612f034aa6afd1f4bcfeeb0890",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Generating test split: 0%| | 0/1991 [00:00<?, ? examples/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "172e8bc3142d4905badd5e7c1b9dc649",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Generating validation split: 0%| | 0/978 [00:00<?, ? examples/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
" → 6819 pairs\n",
"label\n",
"entailment 3195\n",
"neutral 2820\n",
"contradiction 804\n",
"\n",
"[2/3] MultiNLI government (pool=50000) ...\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "f20dccd17a3648cdb76685916aa5603d",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Downloading readme: 0%| | 0.00/8.89k [00:00<?, ?B/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "a15041dfe77b4f8fbe005b02daa90f53",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Downloading data: 0%| | 0.00/214M [00:00<?, ?B/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "a3bd313424c04fa69cb074c62aa20e5b",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Downloading data: 0%| | 0.00/4.94M [00:00<?, ?B/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "c2711b551a664e6d8e2680dbf96890ac",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Downloading data: 0%| | 0.00/5.10M [00:00<?, ?B/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "acfa4ba975724a03af7232b72a6d07bf",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Generating train split: 0%| | 0/392702 [00:00<?, ? examples/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "f5d309c3902744d0921d0255f760c863",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Generating validation_matched split: 0%| | 0/9815 [00:00<?, ? examples/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "82e9096c59c24f2a83f580b310a5255a",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Generating validation_mismatched split: 0%| | 0/9832 [00:00<?, ? examples/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "97ee834b40814d619522c5852d240c66",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Filter: 0%| | 0/50000 [00:00<?, ? examples/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
" → 9937 pairs\n",
"label\n",
"contradiction 3526\n",
"entailment 3318\n",
"neutral 3093\n",
"\n",
"[3/3] Synthetic contradictions (size=1000) ...\n",
" → 842 pairs\n",
"\n",
"=======================================================\n",
"✅ Total training pairs: 17598\n",
"label\n",
"entailment 6513\n",
"neutral 5913\n",
"contradiction 5172\n",
"=======================================================\n"
]
}
],
"source": [
"# ╔══════════════════════════════════════════════════════════════╗\n",
"# ║ CELL 5 — Build full training dataset ║\n",
"# ╚══════════════════════════════════════════════════════════════╝\n",
"print(\"\\n\" + \"=\"*55)\n",
"print(\"BUILDING FULL TRAINING DATA\")\n",
"print(\"=\"*55)\n",
"\n",
"print(f\"\\n[1/3] ContractNLI (size={CNLI_SIZE}) ...\")\n",
"cnli_raw = load_contract_nli(size=CNLI_SIZE)\n",
"cnli_df = process_contract_nli(cnli_raw)\n",
"print(f\" → {len(cnli_df)} pairs\")\n",
"print(cnli_df[\"label\"].value_counts().to_string())\n",
"\n",
"print(f\"\\n[2/3] MultiNLI government (pool={MNLI_SIZE}) ...\")\n",
"mnli_raw = load_mnli_government(size=MNLI_SIZE)\n",
"mnli_df = process_mnli_government(mnli_raw)\n",
"print(f\" → {len(mnli_df)} pairs\")\n",
"print(mnli_df[\"label\"].value_counts().to_string())\n",
"\n",
"print(f\"\\n[3/3] Synthetic contradictions (size={SYNTH_SIZE}) ...\")\n",
"synth_df = build_synthetic_pairs(cnli_df[\"clause1\"].tolist(), sample_size=SYNTH_SIZE)\n",
"print(f\" → {len(synth_df)} pairs\")\n",
"\n",
"valid_labels = {\"entailment\", \"contradiction\", \"neutral\"}\n",
"full_df = (\n",
" pd.concat([cnli_df, mnli_df, synth_df], ignore_index=True)\n",
" .sample(frac=1, random_state=42)\n",
")\n",
"full_df = full_df[full_df[\"label\"].isin(valid_labels)].dropna(\n",
" subset=[\"clause1\", \"clause2\", \"label\"])\n",
"\n",
"print(f\"\\n{'='*55}\")\n",
"print(f\"✅ Total training pairs: {len(full_df)}\")\n",
"print(full_df[\"label\"].value_counts().to_string())\n",
"print(\"=\"*55)"
]
},
{
"cell_type": "code",
"execution_count": 5,
"id": "cell-06-tokenize",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"Warning: You are sending unauthenticated requests to the HF Hub. Please set a HF_TOKEN to enable higher rate limits and faster downloads.\n",
"WARNING:huggingface_hub.utils._http:Warning: You are sending unauthenticated requests to the HF Hub. Please set a HF_TOKEN to enable higher rate limits and faster downloads.\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "c4180a00e39f4b20a183ac5e50fcc535",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"config.json: 0%| | 0.00/776 [00:00<?, ?B/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "6acbaa79f2bc445a8dba7cbac5e0db63",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"tokenizer_config.json: 0%| | 0.00/258 [00:00<?, ?B/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "cf1904a867804acd804c1269a4cd0b36",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"vocab.txt: 0.00B [00:00, ?B/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "c51bd902071d4089a6e35bce0008d300",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"special_tokens_map.json: 0%| | 0.00/112 [00:00<?, ?B/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "736813af7da3439e801d423e723617c9",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Map: 0%| | 0/14958 [00:00<?, ? examples/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "58c1e30f69f84c45a120c54a3d64865e",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Map: 0%| | 0/2640 [00:00<?, ? examples/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Train: 14958 | Eval: 2640\n"
]
}
],
"source": [
"# ╔══════════════════════════════════════════════════════════════╗\n",
"# ║ CELL 6 — Tokenize & split ║\n",
"# ╚══════════════════════════════════════════════════════════════╝\n",
"from datasets import Dataset\n",
"from transformers import AutoTokenizer\n",
"\n",
"tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL)\n",
"\n",
"df_train = full_df.copy()\n",
"df_train[\"label\"] = df_train[\"label\"].map(LABEL2ID)\n",
"\n",
"hf_ds = Dataset.from_pandas(df_train[[\"clause1\", \"clause2\", \"label\"]])\n",
"splits = hf_ds.train_test_split(test_size=0.15, seed=42)\n",
"\n",
"def tokenize_fn(batch):\n",
" return tokenizer(\n",
" batch[\"clause1\"], batch[\"clause2\"],\n",
" truncation=True, padding=\"max_length\", max_length=MAX_LEN\n",
" )\n",
"\n",
"tokenized = splits.map(tokenize_fn, batched=True, batch_size=256)\n",
"# Drop raw text columns — keep only model inputs + label\n",
"tokenized = tokenized.remove_columns([\"clause1\", \"clause2\"])\n",
"tokenized.set_format(\"torch\")\n",
"\n",
"print(f\"Train: {len(tokenized['train'])} | Eval: {len(tokenized['test'])}\")"
]
},
{
"cell_type": "code",
"execution_count": 6,
"id": "cell-07-model",
"metadata": {},
"outputs": [
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "4ae908a15d7f40d1acfb2dc7bb152bdf",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"model.safetensors: 0%| | 0.00/268M [00:00<?, ?B/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "882b3a80c47e48d2bd370b812ba97b85",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Loading weights: 0%| | 0/104 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"warmup_ratio is deprecated and will be removed in v5.2. Use `warmup_steps` instead.\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"✅ Model & Trainer ready.\n"
]
}
],
"source": [
"# ╔══════════════════════════════════════════════════════════════╗\n",
"# ║ CELL 7 — Model & TrainingArguments ║\n",
"# ╚══════════════════════════════════════════════════════════════╝\n",
"from transformers import (\n",
" AutoModelForSequenceClassification,\n",
" TrainingArguments,\n",
" Trainer,\n",
" EarlyStoppingCallback,\n",
")\n",
"import numpy as np\n",
"from sklearn.metrics import f1_score, accuracy_score\n",
"\n",
"model = AutoModelForSequenceClassification.from_pretrained(\n",
" BASE_MODEL,\n",
" num_labels=3,\n",
" id2label=ID2LABEL,\n",
" label2id=LABEL2ID,\n",
" ignore_mismatched_sizes=True,\n",
")\n",
"\n",
"def compute_metrics(eval_pred):\n",
" logits, labels = eval_pred\n",
" preds = np.argmax(logits, axis=-1)\n",
" return {\n",
" \"accuracy\": accuracy_score(labels, preds),\n",
" \"f1\": f1_score(labels, preds, average=\"weighted\", zero_division=0),\n",
" }\n",
"\n",
"training_args = TrainingArguments(\n",
" output_dir=OUTPUT_DIR,\n",
" num_train_epochs=EPOCHS,\n",
" per_device_train_batch_size=BATCH_SIZE,\n",
" per_device_eval_batch_size=BATCH_SIZE,\n",
" learning_rate=LR,\n",
" weight_decay=0.01,\n",
" warmup_ratio=0.1,\n",
" lr_scheduler_type=\"cosine\",\n",
" eval_strategy=\"epoch\", # ← replaces deprecated evaluation_strategy\n",
" save_strategy=\"epoch\",\n",
" load_best_model_at_end=True,\n",
" metric_for_best_model=\"f1\",\n",
" greater_is_better=True,\n",
" logging_steps=50,\n",
" fp16=True, # T4 supports FP16 — ~2x speed boost\n",
" dataloader_num_workers=2,\n",
" report_to=\"none\",\n",
")\n",
"\n",
"trainer = Trainer(\n",
" model=model,\n",
" args=training_args,\n",
" train_dataset=tokenized[\"train\"],\n",
" eval_dataset=tokenized[\"test\"],\n",
" compute_metrics=compute_metrics,\n",
" callbacks=[EarlyStoppingCallback(early_stopping_patience=2)],\n",
")\n",
"\n",
"print(\"✅ Model & Trainer ready.\")"
]
},
{
"cell_type": "code",
"execution_count": 7,
"id": "cell-08-train",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"🚀 Starting training ...\n"
]
},
{
"data": {
"text/html": [
"\n",
" <div>\n",
" \n",
" <progress value='2340' max='2340' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
" [2340/2340 04:25, Epoch 5/5]\n",
" </div>\n",
" <table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: left;\">\n",
" <th>Epoch</th>\n",
" <th>Training Loss</th>\n",
" <th>Validation Loss</th>\n",
" <th>Accuracy</th>\n",
" <th>F1</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <td>1</td>\n",
" <td>0.331108</td>\n",
" <td>0.299753</td>\n",
" <td>0.882197</td>\n",
" <td>0.883166</td>\n",
" </tr>\n",
" <tr>\n",
" <td>2</td>\n",
" <td>0.287276</td>\n",
" <td>0.249932</td>\n",
" <td>0.909470</td>\n",
" <td>0.909536</td>\n",
" </tr>\n",
" <tr>\n",
" <td>3</td>\n",
" <td>0.193082</td>\n",
" <td>0.232398</td>\n",
" <td>0.920076</td>\n",
" <td>0.919938</td>\n",
" </tr>\n",
" <tr>\n",
" <td>4</td>\n",
" <td>0.181951</td>\n",
" <td>0.229638</td>\n",
" <td>0.923485</td>\n",
" <td>0.923234</td>\n",
" </tr>\n",
" <tr>\n",
" <td>5</td>\n",
" <td>0.148084</td>\n",
" <td>0.244799</td>\n",
" <td>0.921591</td>\n",
" <td>0.921433</td>\n",
" </tr>\n",
" </tbody>\n",
"</table><p>"
],
"text/plain": [
"<IPython.core.display.HTML object>"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "89c8b242441e433995dac74b6d960d3f",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Writing model shards: 0%| | 0/1 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "fb0fc6618ea240e3988d0db2228426e0",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Writing model shards: 0%| | 0/1 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "9607498cd8b64e098e9dd6a28d54e673",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Writing model shards: 0%| | 0/1 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "73f882613de74cb387b7c2ac3a4e4d16",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Writing model shards: 0%| | 0/1 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "3f1655b6bc364954bbbe3482ca15a110",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Writing model shards: 0%| | 0/1 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"There were missing keys in the checkpoint model loaded: ['distilbert.embeddings.LayerNorm.weight', 'distilbert.embeddings.LayerNorm.bias'].\n",
"There were unexpected keys in the checkpoint model loaded: ['distilbert.embeddings.LayerNorm.beta', 'distilbert.embeddings.LayerNorm.gamma'].\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"✅ Training complete!\n",
"{'train_runtime': 267.0102, 'train_samples_per_second': 280.102, 'train_steps_per_second': 8.764, 'total_flos': 2476853356746240.0, 'train_loss': 0.3612339029964219, 'epoch': 5.0}\n"
]
}
],
"source": [
"# ╔══════════════════════════════════════════════════════════════╗\n",
"# ║ CELL 8 — Train (~20-35 min on T4) ║\n",
"# ╚══════════════════════════════════════════════════════════════╝\n",
"print(\"\\n🚀 Starting training ...\")\n",
"train_result = trainer.train()\n",
"print(\"\\n✅ Training complete!\")\n",
"print(train_result.metrics)"
]
},
{
"cell_type": "code",
"execution_count": 8,
"id": "cell-09-save",
"metadata": {},
"outputs": [
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "08ae3f5fd0ea4ad790684abeda797454",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Writing model shards: 0%| | 0/1 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"✅ Model saved → /content/drive/MyDrive/Athernex/nli_contract_model_final\n"
]
}
],
"source": [
"# ╔══════════════════════════════════════════════════════════════╗\n",
"# ║ CELL 9 — Save model to Drive ║\n",
"# ╚══════════════════════════════════════════════════════════════╝\n",
"model.save_pretrained(OUTPUT_DIR)\n",
"tokenizer.save_pretrained(OUTPUT_DIR)\n",
"print(f\"\\n✅ Model saved → {OUTPUT_DIR}\")"
]
},
{
"cell_type": "code",
"execution_count": 9,
"id": "cell-10-eval",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"📊 Final evaluation on validation split ...\n"
]
},
{
"data": {
"text/html": [
"\n",
" <div>\n",
" \n",
" <progress value='83' max='83' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
" [83/83 00:02]\n",
" </div>\n",
" "
],
"text/plain": [
"<IPython.core.display.HTML object>"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"{'eval_loss': 0.22964176535606384, 'eval_accuracy': 0.9234848484848485, 'eval_f1': 0.9232344709378449, 'eval_runtime': 2.2543, 'eval_samples_per_second': 1171.072, 'eval_steps_per_second': 36.818, 'epoch': 5.0}\n"
]
}
],
"source": [
"# ╔══════════════════════════════════════════════════════════════╗\n",
"# ║ CELL 10 — Final eval on validation split ║\n",
"# ╚══════════════════════════════════════════════════════════════╝\n",
"print(\"\\n📊 Final evaluation on validation split ...\")\n",
"eval_results = trainer.evaluate()\n",
"print(eval_results)"
]
},
{
"cell_type": "code",
"execution_count": 10,
"id": "cell-11-heldout",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"[Held-out] Loading MultiNLI validation_matched (government) ...\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "2ee65214435a4afebff9dd1c35036c8e",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Filter: 0%| | 0/5000 [00:00<?, ? examples/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
" → 996 held-out samples\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "3c9b2f86a19f4b04a4a19a196080489b",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Loading weights: 0%| | 0/104 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"============================================================\n",
"HELD-OUT CLASSIFICATION REPORT\n",
"============================================================\n",
" precision recall f1-score support\n",
"\n",
"contradiction 0.85 0.86 0.86 335\n",
" entailment 0.90 0.84 0.87 354\n",
" neutral 0.78 0.83 0.80 307\n",
"\n",
" accuracy 0.84 996\n",
" macro avg 0.84 0.84 0.84 996\n",
" weighted avg 0.85 0.84 0.85 996\n",
"\n",
"Confusion Matrix:\n",
" Pred: entailment Pred: contradiction Pred: neutral\n",
"True: entailment 299 17 38\n",
"True: contradiction 12 288 35\n",
"True: neutral 21 32 254\n"
]
}
],
"source": [
"# ╔══════════════════════════════════════════════════════════════╗\n",
"# ║ CELL 11 — Held-out test (MultiNLI validation_matched) ║\n",
"# ╚══════════════════════════════════════════════════════════════╝\n",
"from sklearn.metrics import classification_report, confusion_matrix\n",
"from transformers import pipeline as hf_pipeline\n",
"\n",
"print(\"\\n[Held-out] Loading MultiNLI validation_matched (government) ...\")\n",
"val_raw = load_mnli_government(split=\"validation_matched\", size=5000)\n",
"val_df = process_mnli_government(val_raw)\n",
"val_df = val_df[val_df[\"label\"].isin(valid_labels)].dropna().reset_index(drop=True)\n",
"print(f\" → {len(val_df)} held-out samples\")\n",
"\n",
"# Batch pipeline — no top_k so batch_size works cleanly\n",
"batch_pipe = hf_pipeline(\n",
" \"text-classification\",\n",
" model=OUTPUT_DIR,\n",
" tokenizer=tokenizer,\n",
" device=0,\n",
" batch_size=64,\n",
" truncation=True,\n",
" max_length=MAX_LEN,\n",
")\n",
"\n",
"texts = [f\"{r.clause1} [SEP] {r.clause2}\" for r in val_df.itertuples()]\n",
"raw_preds = batch_pipe(texts)\n",
"pred_labels = [p[\"label\"] for p in raw_preds] # top-1 label per sample\n",
"\n",
"print(\"\\n\" + \"=\"*60)\n",
"print(\"HELD-OUT CLASSIFICATION REPORT\")\n",
"print(\"=\"*60)\n",
"print(classification_report(val_df[\"label\"].tolist(), pred_labels))\n",
"\n",
"cm = confusion_matrix(\n",
" val_df[\"label\"].tolist(), pred_labels,\n",
" labels=[\"entailment\", \"contradiction\", \"neutral\"]\n",
")\n",
"print(\"Confusion Matrix:\")\n",
"print(pd.DataFrame(\n",
" cm,\n",
" index=[\"True: entailment\", \"True: contradiction\", \"True: neutral\"],\n",
" columns=[\"Pred: entailment\", \"Pred: contradiction\", \"Pred: neutral\"],\n",
"))"
]
},
{
"cell_type": "code",
"execution_count": 13,
"id": "b407b6d7",
"metadata": {},
"outputs": [
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "b7abaf3296d3431daa1cf5efd7fa31d7",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Loading weights: 0%| | 0/104 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"You seem to be using the pipelines sequentially on GPU. In order to maximize efficiency please use a dataset\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"============================================================\n",
"CONTRACT CLAUSE SMOKE TEST (CHALLENGING)\n",
"============================================================\n",
"\n",
"✓ Expected: contradiction | Got: contradiction (99.81%)\n",
" C1: The agreement shall automatically renew for successive one-year terms unless ter\n",
" C2: This agreement expires on the end date and shall not renew automatically under a\n",
"\n",
"✗ Expected: contradiction | Got: neutral (41.68%)\n",
" C1: Seller warrants that all deliverables shall be free from defects for a period of\n",
" C2: Seller disclaims all warranties, express or implied, including any warranty of m\n",
"\n",
"✓ Expected: contradiction | Got: contradiction (99.94%)\n",
" C1: The Licensee is granted an exclusive, worldwide, perpetual license to use the So\n",
" C2: The license granted herein is non-exclusive, limited to the United States, and v\n",
"\n",
"✗ Expected: contradiction | Got: neutral (99.44%)\n",
" C1: All disputes arising under this agreement shall be resolved through binding arbi\n",
" C2: Either party may bring suit in any court of competent jurisdiction to resolve di\n",
"\n",
"✓ Expected: contradiction | Got: contradiction (99.10%)\n",
" C1: The contractor shall maintain professional liability insurance with coverage of \n",
" C2: The contractor is not required to carry any form of professional liability insur\n",
"\n",
"✓ Expected: entailment | Got: entailment (90.29%)\n",
" C1: Neither party shall disclose Confidential Information to any third party without\n",
" C2: Confidential Information must not be shared with outside parties unless the disc\n",
"\n",
"✓ Expected: entailment | Got: entailment (94.83%)\n",
" C1: The Company shall indemnify and hold harmless the Consultant against all claims \n",
" C2: If claims arise due to the Company's negligent acts, the Company is responsible \n",
"\n",
"✗ Expected: entailment | Got: neutral (63.14%)\n",
" C1: Force majeure events including natural disasters, war, and pandemic shall excuse\n",
" C2: Obligations under this contract are suspended during events such as pandemics, w\n",
"\n",
"✓ Expected: neutral | Got: neutral (99.97%)\n",
" C1: The governing law of this agreement shall be the laws of the State of Delaware.\n",
" C2: All invoices must be submitted within 30 days of service completion.\n",
"\n",
"✗ Expected: neutral | Got: contradiction (99.26%)\n",
" C1: Employee agrees to a 12-month non-compete restriction within a 50-mile radius.\n",
" C2: The company reserves the right to modify employee benefits at its sole discretio\n",
"\n",
"✓ Expected: neutral | Got: neutral (99.98%)\n",
" C1: The data processor shall implement AES-256 encryption for all data at rest.\n",
" C2: The data controller shall conduct annual privacy impact assessments.\n",
"\n",
"✗ Expected: contradiction | Got: neutral (99.91%)\n",
" C1: Payment of the full contract price is due upon delivery of the final deliverable\n",
" C2: Payment shall be made in four equal quarterly installments over the term of the \n",
"\n",
"✗ Expected: contradiction | Got: neutral (95.06%)\n",
" C1: Either party may terminate this agreement for convenience upon 30 days written n\n",
" C2: This agreement may only be terminated for cause, specifically material breach th\n",
"\n",
"✗ Expected: contradiction | Got: entailment (94.42%)\n",
" C1: The supplier shall deliver all goods FOB destination, with risk of loss transfer\n",
" C2: All shipments are FOB origin; risk of loss passes to the buyer when goods are te\n",
"\n",
"✓ Expected: contradiction | Got: contradiction (95.48%)\n",
" C1: Consultant retains all intellectual property rights in pre-existing materials in\n",
" C2: All work product, including any pre-existing materials used therein, shall be co\n",
"\n",
"============================================================\n",
"Smoke test accuracy: 8/15\n",
"============================================================\n",
"\n",
"🎉 All done! Model saved to Google Drive.\n"
]
}
],
"source": [
"# ╔══════════════════════════════════════════════════════════════╗\n",
"# ║ CELL 12 — Contract clause smoke test (challenging) ║\n",
"# ╚══════════════════════════════════════════════════════════════╝\n",
"from transformers import pipeline as hf_pipeline\n",
"\n",
"EXAMPLE_PAIRS = [\n",
" # --- CONTRADICTIONS: subtle conflicts ---\n",
" (\"The agreement shall automatically renew for successive one-year terms unless terminated with 90 days prior written notice.\",\n",
" \"This agreement expires on the end date and shall not renew automatically under any circumstances.\",\n",
" \"contradiction\"),\n",
"\n",
" (\"Seller warrants that all deliverables shall be free from defects for a period of 24 months from acceptance.\",\n",
" \"Seller disclaims all warranties, express or implied, including any warranty of merchantability or fitness.\",\n",
" \"contradiction\"),\n",
"\n",
" (\"The Licensee is granted an exclusive, worldwide, perpetual license to use the Software.\",\n",
" \"The license granted herein is non-exclusive, limited to the United States, and valid for 12 months only.\",\n",
" \"contradiction\"),\n",
"\n",
" (\"All disputes arising under this agreement shall be resolved through binding arbitration in New York.\",\n",
" \"Either party may bring suit in any court of competent jurisdiction to resolve disputes under this agreement.\",\n",
" \"contradiction\"),\n",
"\n",
" (\"The contractor shall maintain professional liability insurance with coverage of no less than $5,000,000.\",\n",
" \"The contractor is not required to carry any form of professional liability insurance.\",\n",
" \"contradiction\"),\n",
"\n",
" # --- ENTAILMENTS: same meaning, different wording ---\n",
" (\"Neither party shall disclose Confidential Information to any third party without prior written consent.\",\n",
" \"Confidential Information must not be shared with outside parties unless the disclosing party agrees in writing.\",\n",
" \"entailment\"),\n",
"\n",
" (\"The Company shall indemnify and hold harmless the Consultant against all claims arising from the Company's negligence.\",\n",
" \"If claims arise due to the Company's negligent acts, the Company is responsible for indemnifying the Consultant.\",\n",
" \"entailment\"),\n",
"\n",
" (\"Force majeure events including natural disasters, war, and pandemic shall excuse performance obligations.\",\n",
" \"Obligations under this contract are suspended during events such as pandemics, wars, or natural disasters beyond the parties' control.\",\n",
" \"entailment\"),\n",
"\n",
" # --- NEUTRALS: related but independent clauses ---\n",
" (\"The governing law of this agreement shall be the laws of the State of Delaware.\",\n",
" \"All invoices must be submitted within 30 days of service completion.\",\n",
" \"neutral\"),\n",
"\n",
" (\"Employee agrees to a 12-month non-compete restriction within a 50-mile radius.\",\n",
" \"The company reserves the right to modify employee benefits at its sole discretion.\",\n",
" \"neutral\"),\n",
"\n",
" (\"The data processor shall implement AES-256 encryption for all data at rest.\",\n",
" \"The data controller shall conduct annual privacy impact assessments.\",\n",
" \"neutral\"),\n",
"\n",
" # --- HARD CASES: tricky edge cases ---\n",
" (\"Payment of the full contract price is due upon delivery of the final deliverable.\",\n",
" \"Payment shall be made in four equal quarterly installments over the term of the agreement.\",\n",
" \"contradiction\"),\n",
"\n",
" (\"Either party may terminate this agreement for convenience upon 30 days written notice.\",\n",
" \"This agreement may only be terminated for cause, specifically material breach that remains uncured for 60 days.\",\n",
" \"contradiction\"),\n",
"\n",
" (\"The supplier shall deliver all goods FOB destination, with risk of loss transferring upon delivery.\",\n",
" \"All shipments are FOB origin; risk of loss passes to the buyer when goods are tendered to the carrier.\",\n",
" \"contradiction\"),\n",
"\n",
" (\"Consultant retains all intellectual property rights in pre-existing materials incorporated into the deliverables.\",\n",
" \"All work product, including any pre-existing materials used therein, shall be considered work-for-hire owned exclusively by the Client.\",\n",
" \"contradiction\"),\n",
"]\n",
"\n",
"smoke_pipe = hf_pipeline(\n",
" \"text-classification\",\n",
" model=OUTPUT_DIR,\n",
" tokenizer=tokenizer,\n",
" device=0,\n",
" top_k=None,\n",
" truncation=True,\n",
" max_length=MAX_LEN,\n",
")\n",
"\n",
"print(\"\\n\" + \"=\"*60)\n",
"print(\"CONTRACT CLAUSE SMOKE TEST (CHALLENGING)\")\n",
"print(\"=\"*60)\n",
"correct = 0\n",
"for clause1, clause2, expected in EXAMPLE_PAIRS:\n",
" result = smoke_pipe(f\"{clause1} [SEP] {clause2}\")\n",
" if result and isinstance(result[0], list):\n",
" result = result[0]\n",
" scores = {r[\"label\"]: r[\"score\"] for r in result}\n",
" got = max(scores, key=scores.get)\n",
" mark = \"✓\" if got == expected else \"✗\"\n",
" correct += (got == expected)\n",
" print(f\"\\n{mark} Expected: {expected:15s} | Got: {got:15s} ({scores[got]:.2%})\")\n",
" print(f\" C1: {clause1[:80]}\")\n",
" print(f\" C2: {clause2[:80]}\")\n",
"\n",
"print(f\"\\n{'='*60}\")\n",
"print(f\"Smoke test accuracy: {correct}/{len(EXAMPLE_PAIRS)}\")\n",
"print(\"=\"*60)\n",
"print(\"\\n🎉 All done! Model saved to Google Drive.\")\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "44317c51",
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3 (ipykernel)",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.12.13"
}
},
"nbformat": 4,
"nbformat_minor": 5
}
|