File size: 72,873 Bytes
3d3face 32f768a 3d3face 32f768a 3d3face 32f768a 3d3face 32f768a 3d3face 32f768a 3d3face 32f768a 3d3face 32f768a 3d3face 32f768a 3d3face 32f768a 3d3face 32f768a 3d3face 32f768a 3d3face 32f768a 3d3face 32f768a 3d3face 32f768a 3d3face 32f768a 3d3face 32f768a 3d3face 32f768a 3d3face 32f768a 3d3face 32f768a 3d3face 32f768a 3d3face 32f768a 3d3face 32f768a 3d3face 32f768a 3d3face 32f768a 3d3face |
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 |
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# FinGPT"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Part 1: Preparing the Data"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"1.1 Initialize Directories:"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [],
"source": [
"import os\n",
"import shutil\n",
"\n",
"jsonl_path = \"data/dataset_new.jsonl\"\n",
"save_path = 'data/dataset_new'\n",
"\n",
"\n",
"if os.path.exists(jsonl_path):\n",
" os.remove(jsonl_path)\n",
"\n",
"if os.path.exists(save_path):\n",
" shutil.rmtree(save_path)\n",
"\n",
"directory = \"data\"\n",
"if not os.path.exists(directory):\n",
" os.makedirs(directory)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"1.2 Load and Prepare Dataset:"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"Dataset({\n",
" features: ['input', 'output', 'instruction'],\n",
" num_rows: 9543\n",
"})"
]
},
"execution_count": 3,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"from datasets import load_dataset\n",
"import datasets\n",
"\n",
"dic = {\n",
" 0:\"negative\",\n",
" 1:'positive',\n",
" 2:'neutral',\n",
"}\n",
"\n",
"tfns = load_dataset('zeroshot/twitter-financial-news-sentiment')\n",
"tfns = tfns['train']\n",
"tfns = tfns.to_pandas()\n",
"tfns['label'] = tfns['label'].apply(lambda x:dic[x])\n",
"tfns['instruction'] = 'What is the sentiment of this tweet? Please choose an answer from {negative/neutral/positive}.'\n",
"tfns.columns = ['input', 'output', 'instruction']\n",
"tfns = datasets.Dataset.from_pandas(tfns)\n",
"tfns"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"1.3 Concatenate and Shuffle Dataset"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"19086\n"
]
},
{
"data": {
"text/plain": [
"(19086, 3)"
]
},
"execution_count": 4,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"tmp_dataset = datasets.concatenate_datasets([tfns]*2)\n",
"train_dataset = tmp_dataset\n",
"print(tmp_dataset.num_rows)\n",
"\n",
"all_dataset = train_dataset.shuffle(seed = 42)\n",
"all_dataset.shape"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Part 2: Dataset Formatting and Tokenization"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"2.1 Dataset Formatting:"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "8204ff4d7ae048508ff011ff341df7b3",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"formatting..: 0%| | 0/19086 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"import json\n",
"from tqdm.notebook import tqdm\n",
"\n",
"\n",
"def format_example(example: dict) -> dict:\n",
" context = f\"Instruction: {example['instruction']}\\n\"\n",
" if example.get(\"input\"):\n",
" context += f\"Input: {example['input']}\\n\"\n",
" context += \"Answer: \"\n",
" target = example[\"output\"]\n",
" return {\"context\": context, \"target\": target}\n",
"\n",
"\n",
"data_list = []\n",
"for item in all_dataset.to_pandas().itertuples():\n",
" tmp = {}\n",
" tmp[\"instruction\"] = item.instruction\n",
" tmp[\"input\"] = item.input\n",
" tmp[\"output\"] = item.output\n",
" data_list.append(tmp)\n",
"\n",
"\n",
"# save to a jsonl file\n",
"with open(\"data/dataset_new.jsonl\", 'w') as f:\n",
" for example in tqdm(data_list, desc=\"formatting..\"):\n",
" f.write(json.dumps(format_example(example)) + '\\n')"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"2.2 Tokenization"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [],
"source": [
"from transformers import AutoTokenizer, AutoConfig\n",
"\n",
"model_name = \"THUDM/chatglm2-6b\"\n",
"jsonl_path = \"data/dataset_new.jsonl\" # updated path\n",
"save_path = 'data/dataset_new' # updated path\n",
"max_seq_length = 512\n",
"skip_overlength = True\n",
"\n",
"# The preprocess function tokenizes the prompt and target, combines them into input IDs,\n",
"# and then trims or pads the sequence to the maximum sequence length.\n",
"def preprocess(tokenizer, config, example, max_seq_length):\n",
" prompt = example[\"context\"]\n",
" target = example[\"target\"]\n",
" prompt_ids = tokenizer.encode(prompt, max_length=max_seq_length, truncation=True)\n",
" target_ids = tokenizer.encode(\n",
" target,\n",
" max_length=max_seq_length,\n",
" truncation=True,\n",
" add_special_tokens=False)\n",
" input_ids = prompt_ids + target_ids + [config.eos_token_id]\n",
" return {\"input_ids\": input_ids, \"seq_len\": len(prompt_ids)}\n",
"\n",
"# The read_jsonl function reads each line from the JSONL file, preprocesses it using the preprocess function,\n",
"# and then yields each preprocessed example.\n",
"def read_jsonl(path, max_seq_length, skip_overlength=False):\n",
" tokenizer = AutoTokenizer.from_pretrained(\n",
" model_name, trust_remote_code=True)\n",
" config = AutoConfig.from_pretrained(\n",
" model_name, trust_remote_code=True, device_map='auto')\n",
" with open(path, \"r\") as f:\n",
" for line in tqdm(f.readlines()):\n",
" example = json.loads(line)\n",
" feature = preprocess(tokenizer, config, example, max_seq_length)\n",
" if skip_overlength and len(feature[\"input_ids\"]) > max_seq_length:\n",
" continue\n",
" feature[\"input_ids\"] = feature[\"input_ids\"][:max_seq_length]\n",
" yield feature"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"2.3 Save the dataset"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {},
"outputs": [
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "ee7846a30c5d4d59be6bd5e2cf6c1870",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Saving the dataset (0/1 shards): 0%| | 0/19086 [00:00<?, ? examples/s]"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"# The script then creates a Hugging Face Dataset object from the generator and saves it to disk.\n",
"save_path = './data/dataset_new'\n",
"\n",
"dataset = datasets.Dataset.from_generator(\n",
" lambda: read_jsonl(jsonl_path, max_seq_length, skip_overlength)\n",
" )\n",
"dataset.save_to_disk(save_path)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Part 3: Setup FinGPT training parameters with LoRA on ChatGlm2–6b"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"3.1 Training Arguments Setup:"
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"W0804 14:22:55.473000 584 site-packages\\torch\\distributed\\elastic\\multiprocessing\\redirects.py:29] NOTE: Redirects are currently not supported in Windows or MacOs.\n"
]
}
],
"source": [
"from typing import List, Dict, Optional\n",
"import torch\n",
"from loguru import logger\n",
"from transformers import (\n",
" AutoModel,\n",
" AutoTokenizer,\n",
" TrainingArguments,\n",
" Trainer,\n",
" BitsAndBytesConfig\n",
")\n",
"from peft import (\n",
" TaskType,\n",
" LoraConfig,\n",
" get_peft_model,\n",
" set_peft_model_state_dict,\n",
" prepare_model_for_kbit_training,\n",
" prepare_model_for_int8_training,\n",
")\n",
"from peft.utils import TRANSFORMERS_MODELS_TO_LORA_TARGET_MODULES_MAPPING\n",
"\n",
"training_args = TrainingArguments(\n",
" output_dir='./finetuned_model', # saved model path\n",
" # max_steps=10000,\n",
" num_train_epochs = 2,\n",
" per_device_train_batch_size=4,\n",
" gradient_accumulation_steps=8,\n",
" learning_rate=1e-4,\n",
" weight_decay=0.01,\n",
" warmup_steps=10,\n",
" save_steps=50,\n",
" fp16=True,\n",
" # bf16=True,\n",
" torch_compile = False,\n",
" load_best_model_at_end = True,\n",
" evaluation_strategy=\"steps\",\n",
" remove_unused_columns=False,\n",
" logging_steps = 50,\n",
" eval_steps = 50,\n",
" logging_dir='./logs',\n",
" report_to=\"tensorboard\",\n",
" )"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"3.2 Quantization Config Setup:"
]
},
{
"cell_type": "code",
"execution_count": 9,
"metadata": {},
"outputs": [],
"source": [
"# Quantization\n",
"q_config = BitsAndBytesConfig(load_in_4bit=True,\n",
" bnb_4bit_quant_type='nf4',\n",
" bnb_4bit_use_double_quant=True,\n",
" bnb_4bit_compute_dtype=torch.float16\n",
" )"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"3.3 Model Loading & Preparation:"
]
},
{
"cell_type": "code",
"execution_count": 10,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\huggingface_hub\\file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`.\n",
" warnings.warn(\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "85d1520006a04d289fe3431da7df7e42",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Loading checkpoint shards: 0%| | 0/7 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\huggingface_hub\\file_download.py:945: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`.\n",
" warnings.warn(\n",
"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\peft\\utils\\other.py:145: FutureWarning: prepare_model_for_int8_training is deprecated and will be removed in a future version. Use prepare_model_for_kbit_training instead.\n",
" warnings.warn(\n"
]
}
],
"source": [
"# Load tokenizer & model\n",
"# need massive space\n",
"model_name = \"THUDM/chatglm2-6b\"\n",
"tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)\n",
"model = AutoModel.from_pretrained(\n",
" model_name,\n",
" quantization_config=q_config,\n",
" trust_remote_code=True,\n",
" device='cuda'\n",
" )\n",
"model = prepare_model_for_int8_training(model, use_gradient_checkpointing=True)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"3.4 LoRA Config & Setup:"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"trainable params: 1949696 || all params: 3390261248 || trainable%: 0.05750872447219737\n",
"trainable params: 1,949,696 || all params: 6,245,533,696 || trainable%: 0.031217444255383614\n"
]
}
],
"source": [
"def print_trainable_parameters(model):\n",
" \"\"\"\n",
" Prints the number of trainable parameters in the model.\n",
" \"\"\"\n",
" trainable_params = 0\n",
" all_param = 0\n",
" for _, param in model.named_parameters():\n",
" all_param += param.numel()\n",
" if param.requires_grad:\n",
" trainable_params += param.numel()\n",
" print(\n",
" f\"trainable params: {trainable_params} || all params: {all_param} || trainable%: {100 * trainable_params / all_param}\"\n",
" )\n",
"\n",
"\n",
"# LoRA\n",
"target_modules = TRANSFORMERS_MODELS_TO_LORA_TARGET_MODULES_MAPPING['chatglm']\n",
"lora_config = LoraConfig(\n",
" task_type=TaskType.CAUSAL_LM,\n",
" inference_mode=False,\n",
" r=8,\n",
" lora_alpha=32,\n",
" lora_dropout=0.1,\n",
" target_modules=target_modules,\n",
" bias='none',\n",
")\n",
"model = get_peft_model(model, lora_config)\n",
"print_trainable_parameters(model)\n",
"\n",
"resume_from_checkpoint = None\n",
"if resume_from_checkpoint is not None:\n",
" checkpoint_name = os.path.join(resume_from_checkpoint, 'pytorch_model.bin')\n",
" if not os.path.exists(checkpoint_name):\n",
" checkpoint_name = os.path.join(\n",
" resume_from_checkpoint, 'adapter_model.bin'\n",
" )\n",
" resume_from_checkpoint = False\n",
" if os.path.exists(checkpoint_name):\n",
" logger.info(f'Restarting from {checkpoint_name}')\n",
" adapters_weights = torch.load(checkpoint_name)\n",
" set_peft_model_state_dict(model, adapters_weights)\n",
" else:\n",
" logger.info(f'Checkpoint {checkpoint_name} not found')\n",
"\n",
"model.print_trainable_parameters()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Part 4: Loading Data and Training FinGPT"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"4.1 Loading Your Data:"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# load data\n",
"from datasets import load_from_disk\n",
"import datasets\n",
"import os\n",
"\n",
"dataset = datasets.load_from_disk(\"./data/dataset_new\")\n",
"dataset = dataset.train_test_split(0.2, shuffle=True, seed = 42)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"4.2 Training Configuration and Launch:"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\accelerate\\accelerator.py:449: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead.\n",
" self.scaler = torch.cuda.amp.GradScaler(**kwargs)\n",
"You are adding a <class 'transformers.integrations.TensorBoardCallback'> to the callbacks of this Trainer, but there is already one. The currentlist of callbacks is\n",
":DefaultFlowCallback\n",
"TensorBoardCallback\n",
"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\transformers\\optimization.py:411: FutureWarning: This implementation of AdamW is deprecated and will be removed in a future version. Use the PyTorch implementation torch.optim.AdamW instead, or set `no_deprecation_warning=True` to disable this warning\n",
" warnings.warn(\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "27521448ffea440ba40770ef24937509",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
" 0%| | 0/954 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`...\n",
"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\_dynamo\\eval_frame.py:1005: UserWarning: torch.utils.checkpoint: the use_reentrant parameter should be passed explicitly. In version 2.5 we will raise an exception if use_reentrant is not passed. use_reentrant=False is recommended, but if you need to preserve the current default behavior, you can pass use_reentrant=True. Refer to docs for more details on the differences between the two variants.\n",
" return fn(*args, **kwargs)\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"{'loss': 12.8503, 'learning_rate': 9.597457627118645e-05, 'epoch': 0.1}\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "afe29ea804ea4e178d58a6be2379b268",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
" 0%| | 0/478 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"{'eval_loss': 6.350033760070801, 'eval_runtime': 90.0344, 'eval_samples_per_second': 42.406, 'eval_steps_per_second': 5.309, 'epoch': 0.1}\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\_dynamo\\eval_frame.py:1005: UserWarning: torch.utils.checkpoint: the use_reentrant parameter should be passed explicitly. In version 2.5 we will raise an exception if use_reentrant is not passed. use_reentrant=False is recommended, but if you need to preserve the current default behavior, you can pass use_reentrant=True. Refer to docs for more details on the differences between the two variants.\n",
" return fn(*args, **kwargs)\n"
]
},
{
"ename": "KeyboardInterrupt",
"evalue": "",
"output_type": "error",
"traceback": [
"\u001b[31m---------------------------------------------------------------------------\u001b[39m",
"\u001b[31mKeyboardInterrupt\u001b[39m Traceback (most recent call last)",
"\u001b[36mCell\u001b[39m\u001b[36m \u001b[39m\u001b[32mIn[6]\u001b[39m\u001b[32m, line 63\u001b[39m\n\u001b[32m 54\u001b[39m writer = SummaryWriter()\n\u001b[32m 55\u001b[39m trainer = ModifiedTrainer(\n\u001b[32m 56\u001b[39m model=model,\n\u001b[32m 57\u001b[39m args=training_args, \u001b[38;5;66;03m# Trainer args\u001b[39;00m\n\u001b[32m (...)\u001b[39m\u001b[32m 61\u001b[39m callbacks=[TensorBoardCallback(writer)],\n\u001b[32m 62\u001b[39m )\n\u001b[32m---> \u001b[39m\u001b[32m63\u001b[39m trainer.train()\n\u001b[32m 64\u001b[39m writer.close()\n\u001b[32m 65\u001b[39m \u001b[38;5;66;03m# save model\u001b[39;00m\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\transformers\\trainer.py:1645\u001b[39m, in \u001b[36mTrainer.train\u001b[39m\u001b[34m(self, resume_from_checkpoint, trial, ignore_keys_for_eval, **kwargs)\u001b[39m\n\u001b[32m 1640\u001b[39m \u001b[38;5;28mself\u001b[39m.model_wrapped = \u001b[38;5;28mself\u001b[39m.model\n\u001b[32m 1642\u001b[39m inner_training_loop = find_executable_batch_size(\n\u001b[32m 1643\u001b[39m \u001b[38;5;28mself\u001b[39m._inner_training_loop, \u001b[38;5;28mself\u001b[39m._train_batch_size, args.auto_find_batch_size\n\u001b[32m 1644\u001b[39m )\n\u001b[32m-> \u001b[39m\u001b[32m1645\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m inner_training_loop(\n\u001b[32m 1646\u001b[39m args=args,\n\u001b[32m 1647\u001b[39m resume_from_checkpoint=resume_from_checkpoint,\n\u001b[32m 1648\u001b[39m trial=trial,\n\u001b[32m 1649\u001b[39m ignore_keys_for_eval=ignore_keys_for_eval,\n\u001b[32m 1650\u001b[39m )\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\transformers\\trainer.py:1938\u001b[39m, in \u001b[36mTrainer._inner_training_loop\u001b[39m\u001b[34m(self, batch_size, args, resume_from_checkpoint, trial, ignore_keys_for_eval)\u001b[39m\n\u001b[32m 1935\u001b[39m \u001b[38;5;28mself\u001b[39m.control = \u001b[38;5;28mself\u001b[39m.callback_handler.on_step_begin(args, \u001b[38;5;28mself\u001b[39m.state, \u001b[38;5;28mself\u001b[39m.control)\n\u001b[32m 1937\u001b[39m \u001b[38;5;28;01mwith\u001b[39;00m \u001b[38;5;28mself\u001b[39m.accelerator.accumulate(model):\n\u001b[32m-> \u001b[39m\u001b[32m1938\u001b[39m tr_loss_step = \u001b[38;5;28mself\u001b[39m.training_step(model, inputs)\n\u001b[32m 1940\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m (\n\u001b[32m 1941\u001b[39m args.logging_nan_inf_filter\n\u001b[32m 1942\u001b[39m \u001b[38;5;129;01mand\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m is_torch_tpu_available()\n\u001b[32m 1943\u001b[39m \u001b[38;5;129;01mand\u001b[39;00m (torch.isnan(tr_loss_step) \u001b[38;5;129;01mor\u001b[39;00m torch.isinf(tr_loss_step))\n\u001b[32m 1944\u001b[39m ):\n\u001b[32m 1945\u001b[39m \u001b[38;5;66;03m# if loss is nan or inf simply add the average of previous logged losses\u001b[39;00m\n\u001b[32m 1946\u001b[39m tr_loss += tr_loss / (\u001b[32m1\u001b[39m + \u001b[38;5;28mself\u001b[39m.state.global_step - \u001b[38;5;28mself\u001b[39m._globalstep_last_logged)\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\transformers\\trainer.py:2759\u001b[39m, in \u001b[36mTrainer.training_step\u001b[39m\u001b[34m(self, model, inputs)\u001b[39m\n\u001b[32m 2756\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m loss_mb.reduce_mean().detach().to(\u001b[38;5;28mself\u001b[39m.args.device)\n\u001b[32m 2758\u001b[39m \u001b[38;5;28;01mwith\u001b[39;00m \u001b[38;5;28mself\u001b[39m.compute_loss_context_manager():\n\u001b[32m-> \u001b[39m\u001b[32m2759\u001b[39m loss = \u001b[38;5;28mself\u001b[39m.compute_loss(model, inputs)\n\u001b[32m 2761\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28mself\u001b[39m.args.n_gpu > \u001b[32m1\u001b[39m:\n\u001b[32m 2762\u001b[39m loss = loss.mean() \u001b[38;5;66;03m# mean() to average on multi-gpu parallel training\u001b[39;00m\n",
"\u001b[36mCell\u001b[39m\u001b[36m \u001b[39m\u001b[32mIn[6]\u001b[39m\u001b[32m, line 3\u001b[39m, in \u001b[36mModifiedTrainer.compute_loss\u001b[39m\u001b[34m(self, model, inputs)\u001b[39m\n\u001b[32m 2\u001b[39m \u001b[38;5;28;01mdef\u001b[39;00m\u001b[38;5;250m \u001b[39m\u001b[34mcompute_loss\u001b[39m(\u001b[38;5;28mself\u001b[39m, model, inputs):\n\u001b[32m----> \u001b[39m\u001b[32m3\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m model(\n\u001b[32m 4\u001b[39m input_ids=inputs[\u001b[33m\"\u001b[39m\u001b[33minput_ids\u001b[39m\u001b[33m\"\u001b[39m],\n\u001b[32m 5\u001b[39m labels=inputs[\u001b[33m\"\u001b[39m\u001b[33mlabels\u001b[39m\u001b[33m\"\u001b[39m],\n\u001b[32m 6\u001b[39m ).loss\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\nn\\modules\\module.py:1773\u001b[39m, in \u001b[36mModule._wrapped_call_impl\u001b[39m\u001b[34m(self, *args, **kwargs)\u001b[39m\n\u001b[32m 1771\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m._compiled_call_impl(*args, **kwargs) \u001b[38;5;66;03m# type: ignore[misc]\u001b[39;00m\n\u001b[32m 1772\u001b[39m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[32m-> \u001b[39m\u001b[32m1773\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m._call_impl(*args, **kwargs)\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\nn\\modules\\module.py:1784\u001b[39m, in \u001b[36mModule._call_impl\u001b[39m\u001b[34m(self, *args, **kwargs)\u001b[39m\n\u001b[32m 1779\u001b[39m \u001b[38;5;66;03m# If we don't have any hooks, we want to skip the rest of the logic in\u001b[39;00m\n\u001b[32m 1780\u001b[39m \u001b[38;5;66;03m# this function, and just call forward.\u001b[39;00m\n\u001b[32m 1781\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m (\u001b[38;5;28mself\u001b[39m._backward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m._backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m._forward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m._forward_pre_hooks\n\u001b[32m 1782\u001b[39m \u001b[38;5;129;01mor\u001b[39;00m _global_backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_backward_hooks\n\u001b[32m 1783\u001b[39m \u001b[38;5;129;01mor\u001b[39;00m _global_forward_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_forward_pre_hooks):\n\u001b[32m-> \u001b[39m\u001b[32m1784\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m forward_call(*args, **kwargs)\n\u001b[32m 1786\u001b[39m result = \u001b[38;5;28;01mNone\u001b[39;00m\n\u001b[32m 1787\u001b[39m called_always_called_hooks = \u001b[38;5;28mset\u001b[39m()\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\accelerate\\utils\\operations.py:687\u001b[39m, in \u001b[36mconvert_outputs_to_fp32.<locals>.forward\u001b[39m\u001b[34m(*args, **kwargs)\u001b[39m\n\u001b[32m 686\u001b[39m \u001b[38;5;28;01mdef\u001b[39;00m\u001b[38;5;250m \u001b[39m\u001b[34mforward\u001b[39m(*args, **kwargs):\n\u001b[32m--> \u001b[39m\u001b[32m687\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m model_forward(*args, **kwargs)\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\accelerate\\utils\\operations.py:675\u001b[39m, in \u001b[36mConvertOutputsToFp32.__call__\u001b[39m\u001b[34m(self, *args, **kwargs)\u001b[39m\n\u001b[32m 674\u001b[39m \u001b[38;5;28;01mdef\u001b[39;00m\u001b[38;5;250m \u001b[39m\u001b[34m__call__\u001b[39m(\u001b[38;5;28mself\u001b[39m, *args, **kwargs):\n\u001b[32m--> \u001b[39m\u001b[32m675\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m convert_to_fp32(\u001b[38;5;28mself\u001b[39m.model_forward(*args, **kwargs))\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\amp\\autocast_mode.py:44\u001b[39m, in \u001b[36mautocast_decorator.<locals>.decorate_autocast\u001b[39m\u001b[34m(*args, **kwargs)\u001b[39m\n\u001b[32m 41\u001b[39m \u001b[38;5;129m@functools\u001b[39m.wraps(func)\n\u001b[32m 42\u001b[39m \u001b[38;5;28;01mdef\u001b[39;00m\u001b[38;5;250m \u001b[39m\u001b[34mdecorate_autocast\u001b[39m(*args, **kwargs):\n\u001b[32m 43\u001b[39m \u001b[38;5;28;01mwith\u001b[39;00m autocast_instance:\n\u001b[32m---> \u001b[39m\u001b[32m44\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m func(*args, **kwargs)\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\peft\\peft_model.py:1091\u001b[39m, in \u001b[36mPeftModelForCausalLM.forward\u001b[39m\u001b[34m(self, input_ids, attention_mask, inputs_embeds, labels, output_attentions, output_hidden_states, return_dict, task_ids, **kwargs)\u001b[39m\n\u001b[32m 1089\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m peft_config.peft_type == PeftType.POLY:\n\u001b[32m 1090\u001b[39m kwargs[\u001b[33m\"\u001b[39m\u001b[33mtask_ids\u001b[39m\u001b[33m\"\u001b[39m] = task_ids\n\u001b[32m-> \u001b[39m\u001b[32m1091\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m.base_model(\n\u001b[32m 1092\u001b[39m input_ids=input_ids,\n\u001b[32m 1093\u001b[39m attention_mask=attention_mask,\n\u001b[32m 1094\u001b[39m inputs_embeds=inputs_embeds,\n\u001b[32m 1095\u001b[39m labels=labels,\n\u001b[32m 1096\u001b[39m output_attentions=output_attentions,\n\u001b[32m 1097\u001b[39m output_hidden_states=output_hidden_states,\n\u001b[32m 1098\u001b[39m return_dict=return_dict,\n\u001b[32m 1099\u001b[39m **kwargs,\n\u001b[32m 1100\u001b[39m )\n\u001b[32m 1102\u001b[39m batch_size = _get_batch_size(input_ids, inputs_embeds)\n\u001b[32m 1103\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m attention_mask \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m:\n\u001b[32m 1104\u001b[39m \u001b[38;5;66;03m# concat prompt attention mask\u001b[39;00m\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\nn\\modules\\module.py:1773\u001b[39m, in \u001b[36mModule._wrapped_call_impl\u001b[39m\u001b[34m(self, *args, **kwargs)\u001b[39m\n\u001b[32m 1771\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m._compiled_call_impl(*args, **kwargs) \u001b[38;5;66;03m# type: ignore[misc]\u001b[39;00m\n\u001b[32m 1772\u001b[39m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[32m-> \u001b[39m\u001b[32m1773\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m._call_impl(*args, **kwargs)\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\nn\\modules\\module.py:1784\u001b[39m, in \u001b[36mModule._call_impl\u001b[39m\u001b[34m(self, *args, **kwargs)\u001b[39m\n\u001b[32m 1779\u001b[39m \u001b[38;5;66;03m# If we don't have any hooks, we want to skip the rest of the logic in\u001b[39;00m\n\u001b[32m 1780\u001b[39m \u001b[38;5;66;03m# this function, and just call forward.\u001b[39;00m\n\u001b[32m 1781\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m (\u001b[38;5;28mself\u001b[39m._backward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m._backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m._forward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m._forward_pre_hooks\n\u001b[32m 1782\u001b[39m \u001b[38;5;129;01mor\u001b[39;00m _global_backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_backward_hooks\n\u001b[32m 1783\u001b[39m \u001b[38;5;129;01mor\u001b[39;00m _global_forward_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_forward_pre_hooks):\n\u001b[32m-> \u001b[39m\u001b[32m1784\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m forward_call(*args, **kwargs)\n\u001b[32m 1786\u001b[39m result = \u001b[38;5;28;01mNone\u001b[39;00m\n\u001b[32m 1787\u001b[39m called_always_called_hooks = \u001b[38;5;28mset\u001b[39m()\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\peft\\tuners\\tuners_utils.py:160\u001b[39m, in \u001b[36mBaseTuner.forward\u001b[39m\u001b[34m(self, *args, **kwargs)\u001b[39m\n\u001b[32m 159\u001b[39m \u001b[38;5;28;01mdef\u001b[39;00m\u001b[38;5;250m \u001b[39m\u001b[34mforward\u001b[39m(\u001b[38;5;28mself\u001b[39m, *args: Any, **kwargs: Any):\n\u001b[32m--> \u001b[39m\u001b[32m160\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m.model.forward(*args, **kwargs)\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\accelerate\\hooks.py:165\u001b[39m, in \u001b[36madd_hook_to_module.<locals>.new_forward\u001b[39m\u001b[34m(module, *args, **kwargs)\u001b[39m\n\u001b[32m 163\u001b[39m output = module._old_forward(*args, **kwargs)\n\u001b[32m 164\u001b[39m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[32m--> \u001b[39m\u001b[32m165\u001b[39m output = module._old_forward(*args, **kwargs)\n\u001b[32m 166\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m module._hf_hook.post_forward(module, output)\n",
"\u001b[36mFile \u001b[39m\u001b[32m~/.cache\\huggingface\\modules\\transformers_modules\\THUDM\\chatglm2-6b\\d2e2d91789248536a747d9ce60642a336444186c\\modeling_chatglm.py:937\u001b[39m, in \u001b[36mChatGLMForConditionalGeneration.forward\u001b[39m\u001b[34m(self, input_ids, position_ids, attention_mask, past_key_values, inputs_embeds, labels, use_cache, output_attentions, output_hidden_states, return_dict, return_last_logit)\u001b[39m\n\u001b[32m 934\u001b[39m use_cache = use_cache \u001b[38;5;28;01mif\u001b[39;00m use_cache \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m \u001b[38;5;28;01melse\u001b[39;00m \u001b[38;5;28mself\u001b[39m.config.use_cache\n\u001b[32m 935\u001b[39m return_dict = return_dict \u001b[38;5;28;01mif\u001b[39;00m return_dict \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m \u001b[38;5;28;01melse\u001b[39;00m \u001b[38;5;28mself\u001b[39m.config.use_return_dict\n\u001b[32m--> \u001b[39m\u001b[32m937\u001b[39m transformer_outputs = \u001b[38;5;28mself\u001b[39m.transformer(\n\u001b[32m 938\u001b[39m input_ids=input_ids,\n\u001b[32m 939\u001b[39m position_ids=position_ids,\n\u001b[32m 940\u001b[39m attention_mask=attention_mask,\n\u001b[32m 941\u001b[39m past_key_values=past_key_values,\n\u001b[32m 942\u001b[39m inputs_embeds=inputs_embeds,\n\u001b[32m 943\u001b[39m use_cache=use_cache,\n\u001b[32m 944\u001b[39m output_hidden_states=output_hidden_states,\n\u001b[32m 945\u001b[39m return_dict=return_dict,\n\u001b[32m 946\u001b[39m )\n\u001b[32m 948\u001b[39m hidden_states = transformer_outputs[\u001b[32m0\u001b[39m]\n\u001b[32m 949\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m return_last_logit:\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\nn\\modules\\module.py:1773\u001b[39m, in \u001b[36mModule._wrapped_call_impl\u001b[39m\u001b[34m(self, *args, **kwargs)\u001b[39m\n\u001b[32m 1771\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m._compiled_call_impl(*args, **kwargs) \u001b[38;5;66;03m# type: ignore[misc]\u001b[39;00m\n\u001b[32m 1772\u001b[39m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[32m-> \u001b[39m\u001b[32m1773\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m._call_impl(*args, **kwargs)\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\nn\\modules\\module.py:1784\u001b[39m, in \u001b[36mModule._call_impl\u001b[39m\u001b[34m(self, *args, **kwargs)\u001b[39m\n\u001b[32m 1779\u001b[39m \u001b[38;5;66;03m# If we don't have any hooks, we want to skip the rest of the logic in\u001b[39;00m\n\u001b[32m 1780\u001b[39m \u001b[38;5;66;03m# this function, and just call forward.\u001b[39;00m\n\u001b[32m 1781\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m (\u001b[38;5;28mself\u001b[39m._backward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m._backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m._forward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m._forward_pre_hooks\n\u001b[32m 1782\u001b[39m \u001b[38;5;129;01mor\u001b[39;00m _global_backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_backward_hooks\n\u001b[32m 1783\u001b[39m \u001b[38;5;129;01mor\u001b[39;00m _global_forward_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_forward_pre_hooks):\n\u001b[32m-> \u001b[39m\u001b[32m1784\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m forward_call(*args, **kwargs)\n\u001b[32m 1786\u001b[39m result = \u001b[38;5;28;01mNone\u001b[39;00m\n\u001b[32m 1787\u001b[39m called_always_called_hooks = \u001b[38;5;28mset\u001b[39m()\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\accelerate\\hooks.py:165\u001b[39m, in \u001b[36madd_hook_to_module.<locals>.new_forward\u001b[39m\u001b[34m(module, *args, **kwargs)\u001b[39m\n\u001b[32m 163\u001b[39m output = module._old_forward(*args, **kwargs)\n\u001b[32m 164\u001b[39m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[32m--> \u001b[39m\u001b[32m165\u001b[39m output = module._old_forward(*args, **kwargs)\n\u001b[32m 166\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m module._hf_hook.post_forward(module, output)\n",
"\u001b[36mFile \u001b[39m\u001b[32m~/.cache\\huggingface\\modules\\transformers_modules\\THUDM\\chatglm2-6b\\d2e2d91789248536a747d9ce60642a336444186c\\modeling_chatglm.py:830\u001b[39m, in \u001b[36mChatGLMModel.forward\u001b[39m\u001b[34m(self, input_ids, position_ids, attention_mask, full_attention_mask, past_key_values, inputs_embeds, use_cache, output_hidden_states, return_dict)\u001b[39m\n\u001b[32m 827\u001b[39m rotary_pos_emb = rotary_pos_emb.transpose(\u001b[32m0\u001b[39m, \u001b[32m1\u001b[39m).contiguous()\n\u001b[32m 829\u001b[39m \u001b[38;5;66;03m# Run encoder.\u001b[39;00m\n\u001b[32m--> \u001b[39m\u001b[32m830\u001b[39m hidden_states, presents, all_hidden_states, all_self_attentions = \u001b[38;5;28mself\u001b[39m.encoder(\n\u001b[32m 831\u001b[39m inputs_embeds, full_attention_mask, rotary_pos_emb=rotary_pos_emb,\n\u001b[32m 832\u001b[39m kv_caches=past_key_values, use_cache=use_cache, output_hidden_states=output_hidden_states\n\u001b[32m 833\u001b[39m )\n\u001b[32m 835\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m return_dict:\n\u001b[32m 836\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mtuple\u001b[39m(v \u001b[38;5;28;01mfor\u001b[39;00m v \u001b[38;5;129;01min\u001b[39;00m [hidden_states, presents, all_hidden_states, all_self_attentions] \u001b[38;5;28;01mif\u001b[39;00m v \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m)\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\nn\\modules\\module.py:1773\u001b[39m, in \u001b[36mModule._wrapped_call_impl\u001b[39m\u001b[34m(self, *args, **kwargs)\u001b[39m\n\u001b[32m 1771\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m._compiled_call_impl(*args, **kwargs) \u001b[38;5;66;03m# type: ignore[misc]\u001b[39;00m\n\u001b[32m 1772\u001b[39m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[32m-> \u001b[39m\u001b[32m1773\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m._call_impl(*args, **kwargs)\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\nn\\modules\\module.py:1784\u001b[39m, in \u001b[36mModule._call_impl\u001b[39m\u001b[34m(self, *args, **kwargs)\u001b[39m\n\u001b[32m 1779\u001b[39m \u001b[38;5;66;03m# If we don't have any hooks, we want to skip the rest of the logic in\u001b[39;00m\n\u001b[32m 1780\u001b[39m \u001b[38;5;66;03m# this function, and just call forward.\u001b[39;00m\n\u001b[32m 1781\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m (\u001b[38;5;28mself\u001b[39m._backward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m._backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m._forward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m._forward_pre_hooks\n\u001b[32m 1782\u001b[39m \u001b[38;5;129;01mor\u001b[39;00m _global_backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_backward_hooks\n\u001b[32m 1783\u001b[39m \u001b[38;5;129;01mor\u001b[39;00m _global_forward_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_forward_pre_hooks):\n\u001b[32m-> \u001b[39m\u001b[32m1784\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m forward_call(*args, **kwargs)\n\u001b[32m 1786\u001b[39m result = \u001b[38;5;28;01mNone\u001b[39;00m\n\u001b[32m 1787\u001b[39m called_always_called_hooks = \u001b[38;5;28mset\u001b[39m()\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\accelerate\\hooks.py:165\u001b[39m, in \u001b[36madd_hook_to_module.<locals>.new_forward\u001b[39m\u001b[34m(module, *args, **kwargs)\u001b[39m\n\u001b[32m 163\u001b[39m output = module._old_forward(*args, **kwargs)\n\u001b[32m 164\u001b[39m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[32m--> \u001b[39m\u001b[32m165\u001b[39m output = module._old_forward(*args, **kwargs)\n\u001b[32m 166\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m module._hf_hook.post_forward(module, output)\n",
"\u001b[36mFile \u001b[39m\u001b[32m~/.cache\\huggingface\\modules\\transformers_modules\\THUDM\\chatglm2-6b\\d2e2d91789248536a747d9ce60642a336444186c\\modeling_chatglm.py:631\u001b[39m, in \u001b[36mGLMTransformer.forward\u001b[39m\u001b[34m(self, hidden_states, attention_mask, rotary_pos_emb, kv_caches, use_cache, output_hidden_states)\u001b[39m\n\u001b[32m 629\u001b[39m layer = \u001b[38;5;28mself\u001b[39m._get_layer(index)\n\u001b[32m 630\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28mself\u001b[39m.gradient_checkpointing \u001b[38;5;129;01mand\u001b[39;00m \u001b[38;5;28mself\u001b[39m.training:\n\u001b[32m--> \u001b[39m\u001b[32m631\u001b[39m layer_ret = torch.utils.checkpoint.checkpoint(\n\u001b[32m 632\u001b[39m layer,\n\u001b[32m 633\u001b[39m hidden_states,\n\u001b[32m 634\u001b[39m attention_mask,\n\u001b[32m 635\u001b[39m rotary_pos_emb,\n\u001b[32m 636\u001b[39m kv_caches[index],\n\u001b[32m 637\u001b[39m use_cache\n\u001b[32m 638\u001b[39m )\n\u001b[32m 639\u001b[39m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[32m 640\u001b[39m layer_ret = layer(\n\u001b[32m 641\u001b[39m hidden_states,\n\u001b[32m 642\u001b[39m attention_mask,\n\u001b[32m (...)\u001b[39m\u001b[32m 645\u001b[39m use_cache=use_cache\n\u001b[32m 646\u001b[39m )\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\_compile.py:53\u001b[39m, in \u001b[36m_disable_dynamo.<locals>.inner\u001b[39m\u001b[34m(*args, **kwargs)\u001b[39m\n\u001b[32m 50\u001b[39m disable_fn = torch._dynamo.disable(fn, recursive, wrapping=\u001b[38;5;28;01mFalse\u001b[39;00m)\n\u001b[32m 51\u001b[39m fn.__dynamo_disable = disable_fn \u001b[38;5;66;03m# type: ignore[attr-defined]\u001b[39;00m\n\u001b[32m---> \u001b[39m\u001b[32m53\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m disable_fn(*args, **kwargs)\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\_dynamo\\eval_frame.py:1005\u001b[39m, in \u001b[36mDisableContext.__call__.<locals>._fn\u001b[39m\u001b[34m(*args, **kwargs)\u001b[39m\n\u001b[32m 1003\u001b[39m _maybe_set_eval_frame(_callback_from_stance(\u001b[38;5;28mself\u001b[39m.callback))\n\u001b[32m 1004\u001b[39m \u001b[38;5;28;01mtry\u001b[39;00m:\n\u001b[32m-> \u001b[39m\u001b[32m1005\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m fn(*args, **kwargs)\n\u001b[32m 1006\u001b[39m \u001b[38;5;28;01mfinally\u001b[39;00m:\n\u001b[32m 1007\u001b[39m set_eval_frame(\u001b[38;5;28;01mNone\u001b[39;00m)\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\utils\\checkpoint.py:488\u001b[39m, in \u001b[36mcheckpoint\u001b[39m\u001b[34m(function, use_reentrant, context_fn, determinism_check, debug, *args, **kwargs)\u001b[39m\n\u001b[32m 483\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m context_fn \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m noop_context_fn \u001b[38;5;129;01mor\u001b[39;00m debug \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;28;01mFalse\u001b[39;00m:\n\u001b[32m 484\u001b[39m \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mValueError\u001b[39;00m(\n\u001b[32m 485\u001b[39m \u001b[33m\"\u001b[39m\u001b[33mPassing `context_fn` or `debug` is only supported when \u001b[39m\u001b[33m\"\u001b[39m\n\u001b[32m 486\u001b[39m \u001b[33m\"\u001b[39m\u001b[33muse_reentrant=False.\u001b[39m\u001b[33m\"\u001b[39m\n\u001b[32m 487\u001b[39m )\n\u001b[32m--> \u001b[39m\u001b[32m488\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m CheckpointFunction.apply(function, preserve, *args)\n\u001b[32m 489\u001b[39m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[32m 490\u001b[39m gen = _checkpoint_without_reentrant_generator(\n\u001b[32m 491\u001b[39m function, preserve, context_fn, determinism_check, debug, *args, **kwargs\n\u001b[32m 492\u001b[39m )\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\autograd\\function.py:581\u001b[39m, in \u001b[36mFunction.apply\u001b[39m\u001b[34m(cls, *args, **kwargs)\u001b[39m\n\u001b[32m 578\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m torch._C._are_functorch_transforms_active():\n\u001b[32m 579\u001b[39m \u001b[38;5;66;03m# See NOTE: [functorch vjp and autograd interaction]\u001b[39;00m\n\u001b[32m 580\u001b[39m args = _functorch.utils.unwrap_dead_wrappers(args)\n\u001b[32m--> \u001b[39m\u001b[32m581\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28msuper\u001b[39m().apply(*args, **kwargs) \u001b[38;5;66;03m# type: ignore[misc]\u001b[39;00m\n\u001b[32m 583\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m is_setup_ctx_defined:\n\u001b[32m 584\u001b[39m \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mRuntimeError\u001b[39;00m(\n\u001b[32m 585\u001b[39m \u001b[33m\"\u001b[39m\u001b[33mIn order to use an autograd.Function with functorch transforms \u001b[39m\u001b[33m\"\u001b[39m\n\u001b[32m 586\u001b[39m \u001b[33m\"\u001b[39m\u001b[33m(vmap, grad, jvp, jacrev, ...), it must override the setup_context \u001b[39m\u001b[33m\"\u001b[39m\n\u001b[32m 587\u001b[39m \u001b[33m\"\u001b[39m\u001b[33mstaticmethod. For more details, please see \u001b[39m\u001b[33m\"\u001b[39m\n\u001b[32m 588\u001b[39m \u001b[33m\"\u001b[39m\u001b[33mhttps://pytorch.org/docs/main/notes/extending.func.html\u001b[39m\u001b[33m\"\u001b[39m\n\u001b[32m 589\u001b[39m )\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\utils\\checkpoint.py:262\u001b[39m, in \u001b[36mCheckpointFunction.forward\u001b[39m\u001b[34m(ctx, run_function, preserve_rng_state, *args)\u001b[39m\n\u001b[32m 259\u001b[39m ctx.save_for_backward(*tensor_inputs)\n\u001b[32m 261\u001b[39m \u001b[38;5;28;01mwith\u001b[39;00m torch.no_grad():\n\u001b[32m--> \u001b[39m\u001b[32m262\u001b[39m outputs = run_function(*args)\n\u001b[32m 263\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m outputs\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\nn\\modules\\module.py:1773\u001b[39m, in \u001b[36mModule._wrapped_call_impl\u001b[39m\u001b[34m(self, *args, **kwargs)\u001b[39m\n\u001b[32m 1771\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m._compiled_call_impl(*args, **kwargs) \u001b[38;5;66;03m# type: ignore[misc]\u001b[39;00m\n\u001b[32m 1772\u001b[39m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[32m-> \u001b[39m\u001b[32m1773\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m._call_impl(*args, **kwargs)\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\nn\\modules\\module.py:1784\u001b[39m, in \u001b[36mModule._call_impl\u001b[39m\u001b[34m(self, *args, **kwargs)\u001b[39m\n\u001b[32m 1779\u001b[39m \u001b[38;5;66;03m# If we don't have any hooks, we want to skip the rest of the logic in\u001b[39;00m\n\u001b[32m 1780\u001b[39m \u001b[38;5;66;03m# this function, and just call forward.\u001b[39;00m\n\u001b[32m 1781\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m (\u001b[38;5;28mself\u001b[39m._backward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m._backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m._forward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m._forward_pre_hooks\n\u001b[32m 1782\u001b[39m \u001b[38;5;129;01mor\u001b[39;00m _global_backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_backward_hooks\n\u001b[32m 1783\u001b[39m \u001b[38;5;129;01mor\u001b[39;00m _global_forward_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_forward_pre_hooks):\n\u001b[32m-> \u001b[39m\u001b[32m1784\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m forward_call(*args, **kwargs)\n\u001b[32m 1786\u001b[39m result = \u001b[38;5;28;01mNone\u001b[39;00m\n\u001b[32m 1787\u001b[39m called_always_called_hooks = \u001b[38;5;28mset\u001b[39m()\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\accelerate\\hooks.py:165\u001b[39m, in \u001b[36madd_hook_to_module.<locals>.new_forward\u001b[39m\u001b[34m(module, *args, **kwargs)\u001b[39m\n\u001b[32m 163\u001b[39m output = module._old_forward(*args, **kwargs)\n\u001b[32m 164\u001b[39m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[32m--> \u001b[39m\u001b[32m165\u001b[39m output = module._old_forward(*args, **kwargs)\n\u001b[32m 166\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m module._hf_hook.post_forward(module, output)\n",
"\u001b[36mFile \u001b[39m\u001b[32m~/.cache\\huggingface\\modules\\transformers_modules\\THUDM\\chatglm2-6b\\d2e2d91789248536a747d9ce60642a336444186c\\modeling_chatglm.py:562\u001b[39m, in \u001b[36mGLMBlock.forward\u001b[39m\u001b[34m(self, hidden_states, attention_mask, rotary_pos_emb, kv_cache, use_cache)\u001b[39m\n\u001b[32m 559\u001b[39m layernorm_input = residual + layernorm_input\n\u001b[32m 561\u001b[39m \u001b[38;5;66;03m# Layer norm post the self attention.\u001b[39;00m\n\u001b[32m--> \u001b[39m\u001b[32m562\u001b[39m layernorm_output = \u001b[38;5;28mself\u001b[39m.post_attention_layernorm(layernorm_input)\n\u001b[32m 564\u001b[39m \u001b[38;5;66;03m# MLP.\u001b[39;00m\n\u001b[32m 565\u001b[39m mlp_output = \u001b[38;5;28mself\u001b[39m.mlp(layernorm_output)\n",
"\u001b[36mFile \u001b[39m\u001b[32md:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\nn\\modules\\module.py:1949\u001b[39m, in \u001b[36mModule.__getattr__\u001b[39m\u001b[34m(self, name)\u001b[39m\n\u001b[32m 1944\u001b[39m \u001b[38;5;28mself\u001b[39m._backward_pre_hooks = OrderedDict()\n\u001b[32m 1946\u001b[39m \u001b[38;5;66;03m# It is crucial that the return type is not annotated as `Any`, otherwise type checking\u001b[39;00m\n\u001b[32m 1947\u001b[39m \u001b[38;5;66;03m# on `torch.nn.Module` and all its subclasses is largely disabled as a result. See:\u001b[39;00m\n\u001b[32m 1948\u001b[39m \u001b[38;5;66;03m# https://github.com/pytorch/pytorch/pull/115074\u001b[39;00m\n\u001b[32m-> \u001b[39m\u001b[32m1949\u001b[39m \u001b[38;5;28;01mdef\u001b[39;00m\u001b[38;5;250m \u001b[39m\u001b[34m__getattr__\u001b[39m(\u001b[38;5;28mself\u001b[39m, name: \u001b[38;5;28mstr\u001b[39m) -> Union[Tensor, \u001b[33m\"\u001b[39m\u001b[33mModule\u001b[39m\u001b[33m\"\u001b[39m]:\n\u001b[32m 1950\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m \u001b[33m\"\u001b[39m\u001b[33m_parameters\u001b[39m\u001b[33m\"\u001b[39m \u001b[38;5;129;01min\u001b[39;00m \u001b[38;5;28mself\u001b[39m.\u001b[34m__dict__\u001b[39m:\n\u001b[32m 1951\u001b[39m _parameters = \u001b[38;5;28mself\u001b[39m.\u001b[34m__dict__\u001b[39m[\u001b[33m\"\u001b[39m\u001b[33m_parameters\u001b[39m\u001b[33m\"\u001b[39m]\n",
"\u001b[31mKeyboardInterrupt\u001b[39m: "
]
}
],
"source": [
"class ModifiedTrainer(Trainer):\n",
" def compute_loss(self, model, inputs):\n",
" return model(\n",
" input_ids=inputs[\"input_ids\"],\n",
" labels=inputs[\"labels\"],\n",
" ).loss\n",
"\n",
" def prediction_step(self, model: torch.nn.Module, inputs, prediction_loss_only: bool, ignore_keys = None):\n",
" with torch.no_grad():\n",
" res = model(\n",
" input_ids=inputs[\"input_ids\"].to(model.device),\n",
" labels=inputs[\"labels\"].to(model.device),\n",
" ).loss\n",
" return (res, None, None)\n",
"\n",
" def save_model(self, output_dir=None, _internal_call=False):\n",
" from transformers.trainer import TRAINING_ARGS_NAME\n",
"\n",
" os.makedirs(output_dir, exist_ok=True)\n",
" torch.save(self.args, os.path.join(output_dir, TRAINING_ARGS_NAME))\n",
" saved_params = {\n",
" k: v.to(\"cpu\") for k, v in self.model.named_parameters() if v.requires_grad\n",
" }\n",
" torch.save(saved_params, os.path.join(output_dir, \"adapter_model.bin\"))\n",
"\n",
"def data_collator(features: list) -> dict:\n",
" len_ids = [len(feature[\"input_ids\"]) for feature in features]\n",
" longest = max(len_ids)\n",
" input_ids = []\n",
" labels_list = []\n",
" for ids_l, feature in sorted(zip(len_ids, features), key=lambda x: -x[0]):\n",
" ids = feature[\"input_ids\"]\n",
" seq_len = feature[\"seq_len\"]\n",
" labels = (\n",
" [tokenizer.pad_token_id] * (seq_len - 1) + ids[(seq_len - 1) :] + [tokenizer.pad_token_id] * (longest - ids_l)\n",
" )\n",
" ids = ids + [tokenizer.pad_token_id] * (longest - ids_l)\n",
" _ids = torch.LongTensor(ids)\n",
" labels_list.append(torch.LongTensor(labels))\n",
" input_ids.append(_ids)\n",
" input_ids = torch.stack(input_ids)\n",
" labels = torch.stack(labels_list)\n",
" return {\n",
" \"input_ids\": input_ids,\n",
" \"labels\": labels,\n",
" }\n",
"\n",
"from torch.utils.tensorboard import SummaryWriter\n",
"from transformers.integrations import TensorBoardCallback\n",
"\n",
"# Train\n",
"# Took about 10 compute units\n",
"# Took 1 hour to train\n",
"writer = SummaryWriter()\n",
"trainer = ModifiedTrainer(\n",
" model=model,\n",
" args=training_args, # Trainer args\n",
" train_dataset=dataset[\"train\"], # Training set\n",
" eval_dataset=dataset[\"test\"], # Testing set\n",
" data_collator=data_collator, # Data Collator\n",
" callbacks=[TensorBoardCallback(writer)],\n",
")\n",
"trainer.train()\n",
"writer.close()\n",
"# save model\n",
"# model.save_pretrained(training_args.output_dir)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Part 5: Inference and Benchmarks using FinGPT"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### 5.1 Load the model"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Path exists.\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "f632e66d18914e13a9febb54f4e9ee42",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Loading checkpoint shards: 0%| | 0/7 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"import sys\n",
"import os\n",
"\n",
"current_working_directory = os.getcwd()\n",
"\n",
"finnlp_path = os.path.join(current_working_directory, 'FinNLP')\n",
"sys.path.append(finnlp_path)\n",
"\n",
"from transformers import AutoModel, AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig\n",
"\n",
"from peft import PeftModel\n",
"import torch\n",
"\n",
"# # Load benchmark datasets from FinNLP\n",
"# from finnlp.benchmarks.fpb import test_fpb\n",
"# from finnlp.benchmarks.fiqa import test_fiqa , add_instructions\n",
"# from finnlp.benchmarks.tfns import test_tfns\n",
"# from finnlp.benchmarks.nwgi import test_nwgi\n",
"from fpb import test_fpb\n",
"from fiqa import test_fiqa, add_instructions\n",
"from tfns import test_tfns\n",
"from nwgi import test_nwgi\n",
"\n",
"# load model from google drive\n",
"# from google.colab import drive\n",
"# drive.mount('/content/drive')\n",
"\n",
"\n",
"# Define the path you want to check\n",
"path_to_check = \"./finetuned_model_bak\"\n",
"\n",
"# Check if the specified path exists\n",
"if os.path.exists(path_to_check):\n",
" print(\"Path exists.\")\n",
"else:\n",
" print(\"Path does not exist.\")\n",
"\n",
"# ## load the chatglm2-6b base model\n",
"# peft_model = training_args.output_dir\n",
"\n",
"# tokenizer = AutoTokenizer.from_pretrained(base_model, trust_remote_code=True)\n",
"# model = AutoModel.from_pretrained(base_model, trust_remote_code=True, load_in_8bit=True, device_map=\"auto\")\n",
"\n",
"# model = PeftModel.from_pretrained(model, peft_model)\n",
"\n",
"# model = model.eval()\n",
"\n",
"# # load our finetuned model\n",
"base_model = \"THUDM/chatglm2-6b\"\n",
"peft_model = \"./finetuned_model_bak\"\n",
"\n",
"# Quantization\n",
"q_config = BitsAndBytesConfig(load_in_4bit=True,\n",
" bnb_4bit_quant_type='nf4',\n",
" bnb_4bit_use_double_quant=True,\n",
" bnb_4bit_compute_dtype=torch.float16\n",
" )\n",
"\n",
"tokenizer = AutoTokenizer.from_pretrained(base_model, trust_remote_code=True)\n",
"model = AutoModel.from_pretrained(base_model, trust_remote_code=True, quantization_config=q_config, device_map=\"cuda\")\n",
"\n",
"model = PeftModel.from_pretrained(model, peft_model)\n",
"model = model.eval()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"5.2 Run Benchmarks:"
]
},
{
"cell_type": "code",
"execution_count": 12,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"\n",
"Prompt example:\n",
"Instruction: What is the sentiment of this tweet? Please choose an answer from {negative/neutral/positive}.\n",
"Input: $ALLY - Ally Financial pulls outlook https://t.co/G9Zdi1boy5\n",
"Answer: \n",
"\n",
"\n",
"Total len: 2388. Batchsize: 8. Total steps: 299\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
" 0%| | 0/299 [00:01<?, ?it/s]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Unexpected exception formatting exception. Falling back to standard exception\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"Traceback (most recent call last):\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\IPython\\core\\interactiveshell.py\", line 3667, in run_code\n",
" exec(code_obj, self.user_global_ns, self.user_ns)\n",
" File \"C:\\Users\\23524\\AppData\\Local\\Temp\\ipykernel_584\\2487751523.py\", line 5, in <module>\n",
" res = test_tfns(model, tokenizer, batch_size = batch_size)\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\fingpt\\tfns.py\", line 62, in test_tfns\n",
" res = model.generate(**tokens, max_length=512)\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\peft\\peft_model.py\", line 1148, in generate\n",
" outputs = self.base_model.generate(*args, **kwargs)\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\utils\\_contextlib.py\", line 120, in decorate_context\n",
" return func(*args, **kwargs)\n",
" ^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\transformers\\generation\\utils.py\", line 1522, in generate\n",
" return self.greedy_search(\n",
" ^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\transformers\\generation\\utils.py\", line 2339, in greedy_search\n",
" outputs = self(\n",
" ^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\nn\\modules\\module.py\", line 1773, in _wrapped_call_impl\n",
" return self._call_impl(*args, **kwargs)\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\nn\\modules\\module.py\", line 1784, in _call_impl\n",
" return forward_call(*args, **kwargs)\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\accelerate\\hooks.py\", line 165, in new_forward\n",
" output = module._old_forward(*args, **kwargs)\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"C:\\Users\\23524/.cache\\huggingface\\modules\\transformers_modules\\THUDM\\chatglm2-6b\\d2e2d91789248536a747d9ce60642a336444186c\\modeling_chatglm.py\", line 937, in forward\n",
" transformer_outputs = self.transformer(\n",
" ^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\nn\\modules\\module.py\", line 1773, in _wrapped_call_impl\n",
" return self._call_impl(*args, **kwargs)\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\nn\\modules\\module.py\", line 1784, in _call_impl\n",
" return forward_call(*args, **kwargs)\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\accelerate\\hooks.py\", line 165, in new_forward\n",
" output = module._old_forward(*args, **kwargs)\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"C:\\Users\\23524/.cache\\huggingface\\modules\\transformers_modules\\THUDM\\chatglm2-6b\\d2e2d91789248536a747d9ce60642a336444186c\\modeling_chatglm.py\", line 807, in forward\n",
" inputs_embeds = self.embedding(input_ids)\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\nn\\modules\\module.py\", line 1773, in _wrapped_call_impl\n",
" return self._call_impl(*args, **kwargs)\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\nn\\modules\\module.py\", line 1784, in _call_impl\n",
" return forward_call(*args, **kwargs)\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\accelerate\\hooks.py\", line 165, in new_forward\n",
" output = module._old_forward(*args, **kwargs)\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"C:\\Users\\23524/.cache\\huggingface\\modules\\transformers_modules\\THUDM\\chatglm2-6b\\d2e2d91789248536a747d9ce60642a336444186c\\modeling_chatglm.py\", line 723, in forward\n",
" words_embeddings = self.word_embeddings(input_ids)\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\nn\\modules\\module.py\", line 1773, in _wrapped_call_impl\n",
" return self._call_impl(*args, **kwargs)\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\nn\\modules\\module.py\", line 1784, in _call_impl\n",
" return forward_call(*args, **kwargs)\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\accelerate\\hooks.py\", line 165, in new_forward\n",
" output = module._old_forward(*args, **kwargs)\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\nn\\modules\\sparse.py\", line 192, in forward\n",
" return F.embedding(\n",
" ^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\torch\\nn\\functional.py\", line 2546, in embedding\n",
" return torch.embedding(weight, input, padding_idx, scale_grad_by_freq, sparse)\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
"torch.AcceleratorError: CUDA error: out of memory\n",
"Search for `cudaErrorMemoryAllocation' in https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__TYPES.html for more information.\n",
"CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.\n",
"For debugging consider passing CUDA_LAUNCH_BLOCKING=1\n",
"Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.\n",
"\n",
"\n",
"During handling of the above exception, another exception occurred:\n",
"\n",
"Traceback (most recent call last):\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\IPython\\core\\interactiveshell.py\", line 2176, in showtraceback\n",
" stb = self.InteractiveTB.structured_traceback(\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\IPython\\core\\ultratb.py\", line 1182, in structured_traceback\n",
" return FormattedTB.structured_traceback(\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\IPython\\core\\ultratb.py\", line 1053, in structured_traceback\n",
" return VerboseTB.structured_traceback(\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\IPython\\core\\ultratb.py\", line 861, in structured_traceback\n",
" formatted_exceptions: list[list[str]] = self.format_exception_as_a_whole(\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\IPython\\core\\ultratb.py\", line 773, in format_exception_as_a_whole\n",
" frames.append(self.format_record(record))\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\IPython\\core\\ultratb.py\", line 651, in format_record\n",
" _format_traceback_lines(\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\IPython\\core\\tbtools.py\", line 99, in _format_traceback_lines\n",
" line = stack_line.render(pygmented=has_colors).rstrip(\"\\n\") + \"\\n\"\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\stack_data\\core.py\", line 360, in render\n",
" start_line, lines = self.frame_info._pygmented_scope_lines\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\stack_data\\utils.py\", line 145, in cached_property_wrapper\n",
" value = obj.__dict__[self.func.__name__] = self.func(obj)\n",
" ^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\stack_data\\core.py\", line 780, in _pygmented_scope_lines\n",
" lines = _pygmented_with_ranges(formatter, code, ranges)\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\stack_data\\utils.py\", line 165, in _pygmented_with_ranges\n",
" return pygments.highlight(code, lexer, formatter).splitlines()\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\pygments\\__init__.py\", line 82, in highlight\n",
" return format(lex(code, lexer), formatter, outfile)\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\pygments\\__init__.py\", line 64, in format\n",
" formatter.format(tokens, realoutfile)\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\pygments\\formatters\\terminal256.py\", line 250, in format\n",
" return Formatter.format(self, tokensource, outfile)\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\pygments\\formatter.py\", line 124, in format\n",
" return self.format_unencoded(tokensource, outfile)\n",
" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\pygments\\formatters\\terminal256.py\", line 256, in format_unencoded\n",
" for ttype, value in tokensource:\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\stack_data\\utils.py\", line 158, in get_tokens\n",
" for ttype, value in super().get_tokens(text):\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\pygments\\lexer.py\", line 270, in streamer\n",
" for _, t, v in self.get_tokens_unprocessed(text):\n",
" File \"d:\\anaconda\\envs\\fingpt-env\\Lib\\site-packages\\pygments\\lexer.py\", line 712, in get_tokens_unprocessed\n",
" m = rexmatch(text, pos)\n",
" ^^^^^^^^^^^^^^^^^^^\n",
"MemoryError\n"
]
}
],
"source": [
"batch_size = 8\n",
"\n",
"# TFNS Test Set, len 2388\n",
"# Available: 84.85 compute units\n",
"res = test_tfns(model, tokenizer, batch_size = batch_size)\n",
"# Available: 83.75 compute units\n",
"# Took about 1 compute unite to inference\n",
"\n",
"\n",
"# FPB, len 1212\n",
"# res = test_fpb(model, tokenizer, batch_size = batch_size)\n",
"\n",
"# FiQA, len 275\n",
"# res = test_fiqa(model, tokenizer, prompt_fun = add_instructions, batch_size = batch_size)\n",
"\n",
"# NWGI, len 4047\n",
"# res = test_nwgi(model, tokenizer, batch_size = batch_size)"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "fingpt-env",
"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": "3.11.13"
}
},
"nbformat": 4,
"nbformat_minor": 2
}
|