#\u7b2c6\u7ae0\/\u52a0\u8f7d\u7f16\u7801\u5de5\u5177tokenizer\nfrom transformers import AutoTokenizer\n\ntokenizer = AutoTokenizer.from_pretrained('..\/models\/rbt3')\n\ntokenizer<\/code><\/pre>\nBertTokenizerFast(name_or_path='..\/models\/rbt3', vocab_size=21128, model_max_length=1000000000000000019884624838656, is_fast=True, padding_side='right', truncation_side='right', special_tokens={'unk_token': '[UNK]', 'sep_token': '[SEP]', 'pad_token': '[PAD]', 'cls_token': '[CLS]', 'mask_token': '[MASK]'}, clean_up_tokenization_spaces=True), added_tokens_decoder={\n 0: AddedToken(\"[PAD]\", rstrip=False, lstrip=False, single_word=False, normalized=False, special=True),\n 100: AddedToken(\"[UNK]\", rstrip=False, lstrip=False, single_word=False, normalized=False, special=True),\n 101: AddedToken(\"[CLS]\", rstrip=False, lstrip=False, single_word=False, normalized=False, special=True),\n 102: AddedToken(\"[SEP]\", rstrip=False, lstrip=False, single_word=False, normalized=False, special=True),\n 103: AddedToken(\"[MASK]\", rstrip=False, lstrip=False, single_word=False, normalized=False, special=True),\n}<\/code><\/pre>\n#\u7b2c6\u7ae0\/\u8bd5\u7f16\u7801\u53e5\u5b50\ntokenizer.batch_encode_plus(\n ['\u660e\u6708\u88c5\u9970\u4e86\u4f60\u7684\u7a97\u5b50', '\u4f60\u88c5\u9970\u4e86\u522b\u4eba\u7684\u68a6'],\n truncation=True,\n)<\/code><\/pre>\nAsking to truncate to max_length but no maximum length is provided and the model has no predefined maximum length. Default to no truncation.\n\n{'input_ids': [[101, 3209, 3299, 6163, 7652, 749, 872, 4638, 4970, 2094, 102], [101, 872, 6163, 7652, 749, 1166, 782, 4638, 3457, 102]], 'token_type_ids': [[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0]], 'attention_mask': [[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1], [1, 1, 1, 1, 1, 1, 1, 1, 1, 1]]}<\/code><\/pre>\n#\u7b2c6\u7ae0\/\u4ece\u78c1\u76d8\u52a0\u8f7d\u6570\u636e\u96c6\nfrom datasets import load_from_disk\n\ndataset = load_from_disk('.\/data\/ChnSentiCorp')\n\n#\u7f29\u5c0f\u6570\u636e\u89c4\u6a21\uff0c\u4fbf\u4e8e\u6d4b\u8bd5\ndataset['train'] = dataset['train'].shuffle().select(range(2000))\ndataset['test'] = dataset['test'].shuffle().select(range(100))\n\ndataset<\/code><\/pre>\nDatasetDict({\n train: Dataset({\n features: ['text', 'label'],\n num_rows: 2000\n })\n validation: Dataset({\n features: ['text', 'label'],\n num_rows: 0\n })\n test: Dataset({\n features: ['text', 'label'],\n num_rows: 100\n })\n})<\/code><\/pre>\n\n- \u5728\u8fd9\u6bb5\u4ee3\u7801\u4e2d\uff0c\u5bf9\u6570\u636e\u96c6\u8fdb\u884c\u4e86\u91c7\u6837\uff0c\u76ee\u7684\u6709\u4ee5\u4e0b\u4e24\u65b9\u9762\uff1a<\/li>\n<\/ul>\n
\n- \u4fbf\u4e8e\u6d4b\u8bd5\uff1b<\/li>\n
- \u6a21\u62df\u5728\u8bad\u7ec3\u96c6\u7684\u4f53\u91cf\u8f83\u5c0f\u7684\u60c5\u51b5\uff0c\u4ee5\u9a8c\u8bc1\u5373\u4f7f\u662f\u5c0f\u7684\u6570\u636e\u96c6\uff0c\u4e5f\u80fd\u901a\u8fc7\u8fc1\u79fb\u5b66\u4e60\u5f97\u5230\u4e00\u4e2a\u8f83\u597d\u7684\u8bad\u7ec3\u7ed3\u679c<\/li>\n<\/ol>\n
\n- \u53ef\u89c1\u8bad\u7ec3\u96c6\u7684\u6570\u91cf\u4ec5\u67092000\u6761\uff0c\u6d4b\u8bd5\u96c6\u7684\u6570\u91cf\u67091000\u6761\u3002<\/li>\n
- \u73b0\u5728\u7684\u6570\u636e\u96c6\u8fd8\u662f\u6587\u672c\u6570\u636e\uff0c\u4f7f\u7528\u7f16\u7801\u5de5\u5177\u628a\u8fd9\u4e9b\u62bd\u8c61\u7684\u6587\u5b57\u7f16\u7801\u6210\u8ba1\u7b97\u673a\u5584\u4e8e\u5904\u7406\u7684\u6570\u5b57\uff0c\u4ee3\u7801\u5982\u4e0b\uff1a<\/li>\n<\/ul>\n
#\u7b2c6\u7ae0\/\u7f16\u7801\ndef f(data):\n return tokenizer.batch_encode_plus(data['text'], truncation=True)\n\ndataset = dataset.map(f,\n batched=True,\n batch_size=1000,\n num_proc=4,\n remove_columns=['text'])\n\ndataset<\/code><\/pre>\nMap (num_proc=4): 0%| | 0\/2000 [00:00<\/pre>\n\u5728\u8fd9\u6bb5\u4ee3\u7801\u4e2d\uff0c\u4f7f\u7528\u4e86\u6279\u91cf\u5904\u7406\u7684\u6280\u5de7\uff0c\u80fd\u591f\u52a0\u5feb\u8ba1\u7b97\u7684\u901f\u5ea6\u3002<\/p>\n
\n- \u53c2\u6570batched=True: \u8868\u660e\u4f7f\u7528\u6279\u91cf\u6765\u5904\u7406\u6570\u636e\uff0c\u800c\u4e0d\u662f\u4e00\u6761\u4e00\u6761\u7684\u5904\u7406\uff1b<\/li>\n
- \u53c2\u6570batch_size=1000: \u8868\u660e\u6bcf\u4e2a\u6279\u6b21\u67091000\u6761\u6570\u636e\uff1b<\/li>\n
- \u53c2\u6570num_proc=4: \u8868\u660e\u4f7f\u75284\u4e2a\u7ebf\u7a0b\u8fdb\u884c\u64cd\u4f5c\uff1b<\/li>\n
- \u53c2\u6570remove_columns=['text']\uff1a\u8868\u660e\u6620\u5c04\u7ed3\u675f\u540e\u5220\u9664\u6570\u636e\u53ca\u4e2d\u7684text\u5b57\u6bb5<\/li>\n<\/ol>\n
#\u7b2c6\u7ae0\/\u79fb\u9664\u592a\u957f\u7684\u53e5\u5b50\ndef f(data):\n return [len(i) <= 512 for i in data['input_ids']]\n\ndataset = dataset.filter(f, batched=True, batch_size=1000, num_proc=4)\n\ndataset<\/code><\/pre>\nFilter (num_proc=4): 0%| | 0\/2000 [00:00<\/pre>\n\u5b9a\u4e49\u6a21\u578b\u548c\u8bad\u7ec3\u5de5\u5177<\/h1>\n#\u7b2c6\u7ae0\/\u52a0\u8f7d\u9884\u8bad\u7ec3\u6a21\u578b\nfrom transformers import AutoModelForSequenceClassification\n\nmodel = AutoModelForSequenceClassification.from_pretrained('..\/models\/rbt3',\n num_labels=2)\n\n#\u7edf\u8ba1\u6a21\u578b\u53c2\u6570\u91cf\nsum([i.nelement() for i in model.parameters()]) \/ 10000<\/code><\/pre>\nSome weights of BertForSequenceClassification were not initialized from the model checkpoint at ..\/models\/rbt3 and are newly initialized: ['classifier.bias', 'classifier.weight']\nYou should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n\n3847.8338<\/code><\/pre>\n\u5982\u524d\u6240\u8ff0\uff0c\u6b64\u5904\u52a0\u8f7d\u7684\u6a21\u578b\u5e94\u8be5\u548c\u7f16\u7801\u5de5\u5177\u914d\u5bf9\u4f7f\u7528\uff0c\u6240\u4ee5\u6b64\u5904\u52a0\u8f7d\u7684\u6a21\u578b\u4e3ahfl\/rb3\u6a21\u578b\uff0c\u8be5\u6a21\u578b\u7531\u54c8\u5c14\u6ee8\u5de5\u4e1a\u5927\u5b66\u8baf\u98de\u8054\u5408\u5b9e\u9a8c\u5ba4(HFL)\u5206\u4eab\u5230HuggingFace\u6a21\u578b\u5e93\uff0c\u8fd9\u662f\u4e00\u4e2a\u57fa\u4e8e\u4e2d\u6587\u6587\u672c\u6570\u636e\u8bad\u7ec3\u7684BERT\u6a21\u578b\u3002\u540e\u7eed\u5c06\u4f7f\u7528\u51c6\u5907\u597d\u7684\u6570\u636e\u96c6\u5bf9\u8be5\u6a21\u578b\u8fdb\u884c\u518d\u8bad\u7ec3\uff0c\u518d\u4ee3\u7801\u7684\u6700\u540e\u4e00\u884c\u7edf\u8ba1\u4e86\u8be5\u6a21\u578b\u7684\u53c2\u6570\u91cf\uff0c\u4ee5\u5927\u81f4\u8861\u91cf\u4e00\u4e2a\u6a21\u578b\u7684\u4f53\u91cf\u5927\u5c0f\u3002\u8be5\u6a21\u578b\u7684\u53c2\u6570\u91cf\u7ea6\u4e3a3800\u4e07\u4e2a\uff0c\u8fd9\u662f\u4e00\u4e2a\u8f83\u5c0f\u7684\u6a21\u578b\u3002<\/p>\n
#\u7b2c6\u7ae0\/\u6a21\u578b\u8bd5\u7b97\nimport torch\n\n#\u6a21\u62df\u4e00\u6279\u6570\u636e\ndata = {\n 'input_ids': torch.ones(4, 10, dtype=torch.long),\n 'token_type_ids': torch.ones(4, 10, dtype=torch.long),\n 'attention_mask': torch.ones(4, 10, dtype=torch.long),\n 'labels': torch.ones(4, dtype=torch.long)\n}\n\n#\u6a21\u578b\u8bd5\u7b97\nout = model(**data)\n\nout['loss'], out['logits'].shape<\/code><\/pre>\n(tensor(0.8869, grad_fn=), torch.Size([4, 2]))<\/code><\/pre>\n\u6a21\u578b\u7684\u8f93\u51fa\u4e3b\u8981\u5305\u62ec\u4e24\u4e2a\u90e8\u5206\uff0c\u4e00\u90e8\u5206\u662floss\uff0c\u53e6\u4e00\u90e8\u5206\u662flogits\u3002\u5bf9\u4e8e\u4e0d\u540c\u7684\u6a21\u578b\uff0c\u8f93\u51fa\u7684\u5185\u5bb9\u4e5f\u4f1a\u4e0d\u4e00\u6837\uff0c\u4f46\u4e00\u822c\u4f1a\u5305\u62eclos\uff0c\u6240\u4ee5\u5728\u4f7f\u7528HuggingFace\u6a21\u578b\u65f6\uff0c\u4e0d\u9700\u8981\u81ea\u884c\u8ba1\u7b97loss\uff0c\u800c\u65f6\u6a21\u578b\u81ea\u884c\u5c01\u88c5\uff0c\u8fd9\u65b9\u4fbf\u4e86\u6a21\u578b\u7684\u518d\u8bad\u7ec3\u3002<\/p>\n
# #\u7b2c6\u7ae0\/\u52a0\u8f7d\u8bc4\u4ef7\u6307\u6807\n# import os\n# os.environ["HF_ENDPOINT"] = "https:\/\/hf-mirror.com"\n# from datasets import load_metric\n\n# metric = load_metric('accuracy')<\/code><\/pre>\n\u7531\u4e8e\u6a21\u578b\u8ba1\u7b97\u7684\u8f93\u51fa\u548c\u8bc4\u4ef7\u6307\u6807\u8981\u6c42\u7684\u8f93\u5165\u8fd8\u6709\u5dee\u522b\uff0c\u6240\u4ee5\u9700\u8981\u5b9a\u4e49\u4e00\u4e2a\u8f6c\u6362\u51fd\u6570\uff0c\u628a\u6a21\u578b\u8ba1\u7b97\u7684\u8f93\u51fa\u8f6c\u6362\u6210\u8bc4\u4ef7\u6307\u6807\u53ef\u4ee5\u8ba1\u7b97\u7684\u6570\u636e\u7c7b\u578b\uff0c\u8fd9\u4e2a\u51fd\u6570\u5c31\u662f\u5728\u8bad\u7ec3\u8fc7\u7a0b\u771f\u6b63\u5f97\u5230\u7684\u8bc4\u4ef7\u51fd\u6570\uff0c\u4ee3\u7801\u5982\u4e0b\uff1a<\/p>\n
#\u7b2c6\u7ae0\/\u5b9a\u4e49\u8bc4\u4ef7\u51fd\u6570\nimport numpy as np\nfrom transformers.trainer_utils import EvalPrediction\n\ndef compute_metrics(eval_pred):\n logits, labels = eval_pred\n logits = logits.argmax(axis=1)\n return {'accuracy': (logits == labels).sum() \/ len(labels)}\n #return metric.compute(predictions=logits, references=labels)\n\n#\u6a21\u62df\u8f93\u51fa\neval_pred = EvalPrediction(\n predictions=np.array([[0, 1], [2, 3], [4, 5], [6, 7]]),\n label_ids=np.array([1, 1, 0, 1]),\n)\n\ncompute_metrics(eval_pred)<\/code><\/pre>\n{'accuracy': 0.75}<\/code><\/pre>\n\u5b9a\u4e49\u8bad\u7ec3\u8d85\u53c2\u6570<\/h2>\n
\u5728\u5f00\u59cb\u8bad\u7ec3\u4e4b\u524d\uff0c\u9700\u8981\u5b9a\u4e49\u597d\u8d85\u53c2\u6570\uff0cHuggingFace\u4f7f\u7528TrainingArgument\u5bf9\u8c61\u6765\u5c01\u88c5\u53c2\u6570\uff0c\u4ee3\u7801\u5982\u4e0b\uff1a<\/p>\n
#\u7b2c6\u7ae0\/\u5b9a\u4e49\u8bad\u7ec3\u53c2\u6570\nfrom transformers import TrainingArguments\nimport tensorflow as tf\n#\u5b9a\u4e49\u8bad\u7ec3\u53c2\u6570\nargs = TrainingArguments(\n #\u5b9a\u4e49\u4e34\u65f6\u6570\u636e\u4fdd\u5b58\u8def\u5f84\n output_dir='.\/output_dir',\n\n #\u5b9a\u4e49\u6d4b\u8bd5\u6267\u884c\u7684\u7b56\u7565\uff0c\u53ef\u53d6\u503cno\u3001epoch\u3001steps\n evaluation_strategy='steps',\n\n #\u5b9a\u4e49\u6bcf\u9694\u591a\u5c11\u4e2astep\u6267\u884c\u4e00\u6b21\u6d4b\u8bd5\n eval_steps=30,\n\n #\u5b9a\u4e49\u6a21\u578b\u4fdd\u5b58\u7b56\u7565\uff0c\u53ef\u53d6\u503cno\u3001epoch\u3001steps\n save_strategy='steps',\n\n #\u5b9a\u4e49\u6bcf\u9694\u591a\u5c11\u4e2astep\u4fdd\u5b58\u4e00\u6b21\n save_steps=30,\n\n #\u5b9a\u4e49\u5171\u8bad\u7ec3\u51e0\u4e2a\u8f6e\u6b21\n num_train_epochs=1,\n\n #\u5b9a\u4e49\u5b66\u4e60\u7387\n learning_rate=1e-4,\n\n #\u52a0\u5165\u53c2\u6570\u6743\u91cd\u8870\u51cf\uff0c\u9632\u6b62\u8fc7\u62df\u5408\n weight_decay=1e-2,\n\n #\u5b9a\u4e49\u6d4b\u8bd5\u548c\u8bad\u7ec3\u65f6\u7684\u6279\u6b21\u5927\u5c0f\n per_device_eval_batch_size=16,\n per_device_train_batch_size=16,\n\n #\u5b9a\u4e49\u662f\u5426\u8981\u4f7f\u7528gpu\u8bad\u7ec3\n no_cuda=False,\n)<\/code><\/pre>\n2024-07-13 00:25:45.296917: I tensorflow\/core\/util\/port.cc:113] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable TF_ENABLE_ONEDNN_OPTS=0<\/code>.\n2024-07-13 00:25:45.548390: E external\/local_xla\/xla\/stream_executor\/cuda\/cuda_dnn.cc:9261] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered\n2024-07-13 00:25:45.548432: E external\/local_xla\/xla\/stream_executor\/cuda\/cuda_fft.cc:607] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered\n2024-07-13 00:25:45.590491: E external\/local_xla\/xla\/stream_executor\/cuda\/cuda_blas.cc:1515] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered\n2024-07-13 00:25:45.678285: I tensorflow\/core\/platform\/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.\nTo enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.\n2024-07-13 00:25:46.644895: W tensorflow\/compiler\/tf2tensorrt\/utils\/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT\n\/root\/anaconda3\/lib\/python3.11\/site-packages\/transformers\/training_args.py:1494: FutureWarning: evaluation_strategy<\/code> is deprecated and will be removed in version 4.46 of \ud83e\udd17 Transformers. Use eval_strategy<\/code> instead\n warnings.warn(<\/code><\/pre>\nTrainingArguments\u5bf9\u8c61\u4e2d\u53ef\u4ee5\u5c01\u88c5\u7684\u8d85\u53c2\u6570\u5f88\u591a\uff0c\u4f46\u9664\u4e86output_dir\u4e4b\u5916\u7684\u5176\u4ed6\u8d85\u53c2\u6570\u5747\u6709\u9ed8\u8ba4\u503c\uff0c\u5728\u4e0a\u9762\u7684\u793a\u4f8b\u4ee3\u7801\u4e2d\u53ea\u7ed9\u51fa\u4e86\u5e38\u7528\u7684\u53c2\u6570\uff0c\u5bf9\u4e8e\u521d\u5b66\u8005\u5efa\u8bae\u4ece\u8fd9\u4e9b\u7b80\u5355\u7684\u53c2\u6570\u5f00\u59cb\u8c03\u8bd5\uff0c\u5b8c\u6574\u7684\u53c2\u6570\u5217\u8868\u53ef\u4ee5\u53c2\u7167HuggingFace\u5b98\u65b9\u6587\u6863\u3002<\/p>\n
\u5b9a\u4e49\u8bad\u7ec3\u5668<\/h2>\n
\u5b8c\u6210\u4e86\u4e0a\u9762\u7684\u51c6\u5907\u5de5\u4f5c\uff0c\u73b0\u5728\u53ef\u4ee5\u5b9a\u4e49\u8bad\u7ec3\u5668\uff0c\u4ee3\u7801\u5982\u4e0b\uff1a<\/p>\n
#\u7b2c6\u7ae0\/\u5b9a\u4e49\u8bad\u7ec3\u5668\nfrom transformers import Trainer\nfrom transformers.data.data_collator import DataCollatorWithPadding\n\n#\u5b9a\u4e49\u8bad\u7ec3\u5668\ntrainer = Trainer(\n model=model,\n args=args,\n train_dataset=dataset['train'],\n eval_dataset=dataset['test'],\n compute_metrics=compute_metrics,\n data_collator=DataCollatorWithPadding(tokenizer),\n)<\/code><\/pre>\n\u5b9a\u4e49\u8bad\u7ec3\u5668\u9700\u8981\u4f20\u9012\u7684\u6a21\u578b\uff0c\u8d85\u53c2\u6570\u5bf9\u8c61\uff0c\u8bad\u7ec3\u548c\u9a8c\u8bc1\u6570\u636e\u96c6\uff0c\u8bc4\u4ef7\u51fd\u6570\uff0c\u4ee5\u53ca\u6570\u636e\u6574\u7406\u51fd\u6570\u3002<\/p>\n
\u6574\u7406\u51fd\u6570\u4ecb\u7ecd<\/h2>\n
\u6570\u636e\u6574\u7406\u51fd\u6570\u4f7f\u7528\u4e86HuggingFace\u63d0\u4f9b\u7684DataCollatorWithPadding\u5bf9\u8c61\uff0c\u5b83\u80fd\u628a\u4e00\u4e2a\u6279\u6b21\u4e2d\u957f\u77ed\u4e0d\u4e00\u7684\u53e5\u5b50\uff0c\u8865\u5145\u6210\u7edf\u4e00\u7684\u957f\u5ea6\uff0c\u957f\u5ea6\u53d6\u51b3\u4e8e\u8fd9\u4e2a\u6279\u6b21\u4e2d\u6700\u957f\u7684\u53e5\u5b50\u6709\u591a\u957f\uff0c\u6240\u6709\u6570\u636e\u7684\u957f\u5ea6\u4e00\u81f4\u540e\u5373\u53ef\u8f6c\u6362\u6210\u77e9\u9635\uff0c\u6a21\u578b\u671f\u5f85\u7684\u6570\u636e\u7c7b\u578b\u4e5f\u662f\u77e9\u9635\uff0c\u6240\u4ee5\u7ecf\u8fc7\u6570\u636e\u6574\u7406\u51fd\u6570\u7684\u5904\u7406\u4e4b\u540e\uff0c\u6570\u636e\u5373\u88ab\u6574\u7406\u6210\u6a21\u578b\u53ef\u4ee5\u76f4\u63a5\u8ba1\u7b97\u7684\u77e9\u9635\u683c\u5f0f\u3002\u53ef\u4ee5\u901a\u8fc7\u4e0b\u9762\u7684\u4f8b\u5b50\u9a8c\u8bc1\uff0c\u4ee3\u7801\u5982\u4e0b\uff1a<\/p>\n
#\u7b2c6\u7ae0\/\u6d4b\u8bd5\u6570\u636e\u6574\u7406\u51fd\u6570\ndata_collator = DataCollatorWithPadding(tokenizer)\n\n#\u83b7\u53d6\u4e00\u6279\u6570\u636e\ndata = dataset['train'][:5]\n\n#\u8f93\u51fa\u8fd9\u4e9b\u53e5\u5b50\u7684\u957f\u5ea6\nfor i in data['input_ids']:\n print(len(i))\n\n#\u8c03\u7528\u6570\u636e\u6574\u7406\u51fd\u6570\ndata = data_collator(data)\n\n#\u67e5\u770b\u6574\u7406\u540e\u7684\u6570\u636e\nfor k, v in data.items():\n print(k, v.shape)<\/code><\/pre>\n82\n186\n326\n47\n71\ninput_ids torch.Size([5, 326])\ntoken_type_ids torch.Size([5, 326])\nattention_mask torch.Size([5, 326])\nlabels torch.Size([5])<\/code><\/pre>\n\u5728\u8fd9\u6bb5\u4ee3\u7801\u4e2d\uff0c\u9996\u5148\u521d\u59cb\u5316\u4e86\u4e00\u4e2aDataCollatorWithPadding\u5bf9\u8c61\u4f5c\u4e3a\u6570\u636e\u6574\u7406\u51fd\u6570\uff0c\u7136\u540e\u4ece\u8bad\u7ec3\u96c6\u4e2d\u83b7\u53d6\u4e865\u6761\u6570\u636e\u4f5c\u4e3a\u4e00\u6279\u6570\u636e\uff0c\u4ece\u8f93\u51fa\u53ef\u4ee5\u770b\u51fa\u8fd9\u4e9b\u53e5\u5b50\u6709\u957f\u6709\u77ed\uff0c\u4e4b\u540e\u4f7f\u7528\u6570\u636e\u6574\u7406\u51fd\u6570\u5904\u7406\u8fd9\u6279\u6570\u636e\uff0c\u5f97\u5230\u7684\u7ed3\u679c\u518d\u8f93\u51fa\u5f62\u72b6\uff0c\u53ef\u4ee5\u770b\u5230\u8fd9\u4e9b\u6570\u636e\u5df2\u7ecf\u88ab\u6574\u7406\u6210\u540c\u610f\u7684\u957f\u5ea6\uff0c\u957f\u5ea6\u53d6\u51b3\u4e8e\u8fd9\u6279\u53e5\u5b50\u4e2d\u6700\u957f\u7684\u53e5\u5b50\uff0c\u5e76\u4e14\u88ab\u8f6c\u6362\u4e3a\u77e9\u9635\u5f62\u5f0f\u3002<\/p>\n
tokenizer.decode(data['input_ids'][0])<\/code><\/pre>\n'[CLS] \u6b64 \u6b3e \u673a \u5b50 \u4e0d \u4e3a windows xp \u8bbe \u8ba1 \u3002 \u88c5 \u673a ( xp ) \u6bd4 \u8f83 \u9ebb \u70e6 \uff0c \u6298 \u817e \u4e86 \u4e00 \u5929 \u603b \u7b97 \u641e \u5b9a \u3002 \u4e0d \u8fc7 \u8fd8 \u662f \u7559 \u6709 \u4e0d \u8db3 \uff0c \u5c31 \u662f \u58f0 \u97f3 \u6709 \u65f6 \u6709 [UNK] \u6251 \u6251 [UNK] \u58f0 \u54cd \u3002 \u4e3b \u8981 \u8fd8 \u662f \u4e0d \u517c \u5bb9 \u7684 \u539f \u56e0 \uff0c \u73b0 \u5728 \u5c31 \u8fd9 \u4e48 \u7528 \u7740 \u5427 \uff0c \u4ee5 \u540e \u518d \u5347 \u7ea7 \u3002 [SEP] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD]'<\/code><\/pre>\n\u53ef\u4ee5\u770b\u5230\uff0c\u6570\u636e\u6574\u7406\u51fd\u6570\u662f\u901a\u8fc7\u5bf9\u53e5\u5b50\u7684\u5c3e\u90e8\u8865\u5145PAD\u6765\u5bf9\u53e5\u5b50\u8865\u957f\u7684\u3002<\/p>\n
\u8bad\u7ec3\u548c\u6d4b\u8bd5<\/h1>\n1.\u8bad\u7ec3\u6a21\u578b<\/h2>\n
\u5728\u5f00\u59cb\u8bad\u7ec3\u4e4b\u524d\uff0c\u4e0d\u59a8\u76f4\u63a5\u5bf9\u6a21\u578b\u8fdb\u884c\u4e00\u6b21\u6d4b\u8bd5\uff0c\u5148\u5b9a\u4e0b\u8bad\u7ec3\u524d\u7684\u57fa\u51c6\uff0c\u5728\u8bad\u7ec3\u7ed3\u675f\u540e\u518d\u5bf9\u6bd4\u8fd9\u91cc\u5f97\u5230\u7684\u57fa\u51c6\uff0c\u4ee5\u9a8c\u8bc1\u8bad\u7ec3\u7684\u6709\u6548\u6027\uff0c\u4ee3\u7801\u5982\u4e0b\uff1a<\/p>\n
#\u7b2c6\u7ae0\/\u8bc4\u4ef7\u6a21\u578b\ntrainer.evaluate()<\/code><\/pre>\n![\"image-20240713005301033\"](\"https:\/\/lemon-guess.oss-cn-hangzhou.aliyuncs.com\/img\/image-20240713005301033.png\")
<\/p>\n
{'eval_loss': 0.702571451663971,\n 'eval_accuracy': 0.5360824742268041,\n 'eval_runtime': 0.6406,\n 'eval_samples_per_second': 151.43,\n 'eval_steps_per_second': 10.928}<\/code><\/pre>\n\u53ef\u89c1\u6a21\u578b\u5728\u8bad\u7ec3\u4e4b\u524d\uff0c\u670941%\u7684\u51c6\u786e\u7387\u3002\u7531\u4e8e\u4f7f\u7528\u7684\u4e8c\u5206\u7c7b\u8bad\u7ec3\u96c6\uff0c\u6240\u4ee541%\u7684\u6b63\u786e\u7387\u8fd1\u4e4e\u4e8e\u778e\u731c\u3002\u8fd9\u7b26\u5408\u9884\u671f\uff0c\u56e0\u4e3a\u6a21\u578b\u8fd8\u6ca1\u6709\u8bad\u7ec3\uff0c\u63a5\u4e0b\u6765\u5bf9\u6a21\u578b\u8fdb\u884c\u8bad\u7ec3\uff0c\u671f\u5f85\u5b83\u80fd\u8d85\u8fc7\u6b64\u5904\u5f97\u5230\u7684\u6210\u7ee9\u3002
\n\u5bf9\u6a21\u578b\u8fdb\u884c\u8bad\u7ec3\uff0c\u4ee3\u7801\u5982\u4e0b\uff1a<\/p>\n
#\u7b2c6\u7ae0\/\u8bad\u7ec3\ntrainer.train()<\/code><\/pre>\n![\"image-20240713004803422\"](\"https:\/\/lemon-guess.oss-cn-hangzhou.aliyuncs.com\/img\/image-20240713004803422.png\")
<\/p>\n
TrainOutput(global_step=124, training_loss=0.4607375975578062, metrics={'train_runtime': 18.6753, 'train_samples_per_second': 105.648, 'train_steps_per_second': 6.64, 'total_flos': 69630349836324.0, 'train_loss': 0.4607375975578062, 'epoch': 1.0})<\/code><\/pre>\n#\u7b2c6\u7ae0\/\u4ece\u67d0\u4e2a\u5b58\u6863\u7ee7\u7eed\u8bad\u7ec3\ntrainer.train(resume_from_checkpoint='.\/output_dir\/checkpoint-90')<\/code><\/pre>\n![\"image-20240713004829520\"](\"https:\/\/lemon-guess.oss-cn-hangzhou.aliyuncs.com\/img\/image-20240713004829520.png\")
<\/p>\n
TrainOutput(global_step=124, training_loss=0.09972453886462797, metrics={'train_runtime': 5.0666, 'train_samples_per_second': 389.415, 'train_steps_per_second': 24.474, 'total_flos': 69630349836324.0, 'train_loss': 0.09972453886462797, 'epoch': 1.0})<\/code><\/pre>\n#\u7b2c6\u7ae0\/\u8bc4\u4ef7\u6a21\u578b\ntrainer.evaluate()<\/code><\/pre>\n![\"image-20240713004855829\"](\"https:\/\/lemon-guess.oss-cn-hangzhou.aliyuncs.com\/img\/image-20240713004855829.png\")
<\/p>\n
{'eval_loss': 0.4022093713283539,\n 'eval_accuracy': 0.8247422680412371,\n 'eval_runtime': 0.2265,\n 'eval_samples_per_second': 428.175,\n 'eval_steps_per_second': 30.899,\n 'epoch': 1.0}<\/code><\/pre>\n#\u7b2c6\u7ae0\/\u624b\u52a8\u4fdd\u5b58\u6a21\u578b\u53c2\u6570\ntrainer.save_model(output_dir='.\/output_dir\/save_model')<\/code><\/pre>\n#\u7b2c6\u7ae0\/\u624b\u52a8\u52a0\u8f7d\u6a21\u578b\u53c2\u6570\nimport torch\nfrom safetensors.torch import load_file\n# model.load_state_dict(torch.load('.\/output_dir\/save_model\/model.safetensors'))\n# \u52a0\u8f7dsafetensors\u6587\u4ef6\nstate_dict = load_file('.\/output_dir\/save_model\/model.safetensors')\n\n# \u521b\u5efa\u6a21\u578b\u5b9e\u4f8b\uff08\u8fd9\u91cc\u5047\u8bbe\u4f60\u5df2\u7ecf\u5b9a\u4e49\u4e86\u6a21\u578b\u7c7b\uff09\n#\u3000model = YourModelClass()\n\n# \u5c06\u72b6\u6001\u5b57\u5178\u52a0\u8f7d\u5230\u6a21\u578b\u4e2d\nmodel.load_state_dict(state_dict)\n\n# \u786e\u4fdd\u6a21\u578b\u5728\u6b63\u786e\u7684\u8bbe\u5907\u4e0a\nmodel = model.to('cuda') if torch.cuda.is_available() else model<\/code><\/pre>\n2.\u6d4b\u8bd5\u6a21\u578b<\/h2>\n
\u5728\u4e0b\u9762\u7684\u4ee3\u7801\u4e2d\uff0c\u9996\u5148\u628a\u6a21\u578b\u5207\u6362\u5230\u8fd0\u884c\u6a21\u578b\uff0c\u7136\u540e\u4ece\u6d4b\u8bd5\u96c6\u4e2d\u83b7\u53d6\u4e00\u4e2a\u6279\u6b21\u7684\u6570\u636e\u7528\u4e8e\u9884\u6d4b\uff0c\u4e4b\u540e\u628a\u8fd9\u6279\u6570\u636e\u8f93\u5165\u6a21\u578b\u8fdb\u884c\u8ba1\u7b97\uff0c\u5f97\u51fa\u7684\u7ed3\u679c\u5373\u4e3a\u6a21\u578b\u9884\u6d4b\u7684\u7ed3\u679c\uff0c\u6700\u540e\u8f93\u51fa\u524d\u56db\u53e5\u7684\u7ed3\u679c\uff0c\u5e76\u4e0e\u771f\u5b9e\u7684label\u8fdb\u884c\u6bd4\u8f83\u3002<\/p>\n
#\u7b2c6\u7ae0\/\u6d4b\u8bd5\nmodel.eval()\n\nfor i, data in enumerate(trainer.get_eval_dataloader()):\n break\n\nfor k, v in data.items():\n data[k] = v.to('cuda')\n\nout = model(**data)\nout = out['logits'].argmax(dim=1)\n\nfor i in range(16):\n print(tokenizer.decode(data['input_ids'][i], skip_special_tokens=True))\n print('label=', data['labels'][i].item())\n print('predict=', out[i].item())<\/code><\/pre>\n\u4f4f \u7684 \u662f \u6c5f \u666f \u623f \uff0c \u623f \u95f4 \u8d85 \u5927 \uff0c \u73af \u5883 \u4e00 \u6d41 \uff0c \u4ef7 \u94b1 \u4e0d \u7b97 \u9ad8 \uff08 258 \u5143 \u5305 2 \u65e9 \uff09 \uff0c \u7f8e \u4e2d \u4e0d \u8db3 \u662f \u670d \u52a1 \u6709 \u70b9 \u8ddf \u4e0d \u4e0a \uff0c \u603b \u4f53 \u4e0d \u9519 \u3002\nlabel= 1\npredict= 1\n\u62a2 \u8d2d \u5230 \u7684 \u672c \u672c \uff0c \u6027 \u4ef7 \u6bd4 \u8d85 \u9ad8 \uff0c \u56e0 \u4e3a \u81ea \u5df1 \u5df2 \u6709 \u672c \u672c \uff0c \u4e70 \u4e2a \u4eba \u60c5 \uff0c \u8ba9 \u7ed9 \u4e86 \u540c \u4e8b \u3002\nlabel= 1\npredict= 1\n\u96c6 \u6210 \u663e \u5361, \u73a9 \u4f7f \u547d \u53ec \u5524 5 \u7b49 \u8981 \u6c42 \u5f88 \u9ad8 \u7684 \u6e38 \u620f \u6bd4 \u8f83 \u5361, \u6ca1 \u6709 \u9ad8 \u6e05 \u8f93 \u51fa, \u7528 \u6599 \u6ca1 \u6709 \u4ee5 \u524d \u8001 \u6b3e \u7ed3 \u5b9e \u4e86\nlabel= 0\npredict= 0\n\u6cb3 \u5357 \u4eba \u5c31 \u662f \u6cb3 \u5357 \u4eba \uff01 \u4e0d \u88ab \u61f5 \u4e00 \u6b21 \uff0c \u90a3 \u7b97 \u767d \u6765 \u4e00 \u56de \uff1b \u5728 \u5168 \u56fd \u4e43 \u81f3 \u5168 \u4e16 \u754c \u90a3 \u79cd \u53eb \u505a \u7ecf \u6d4e \u623f \u7684 \u623f \u95f4 \uff0c \u5728 \u6cb3 \u5357 \u6d4e \u6e90 \u96c5 \u58eb \u8fbe \u53eb \u505a \u8d35 \u5bbe \u697c \uff0c \u771f \u591f \u9017 \u7684 \uff01 \uff01 \uff01\nlabel= 0\npredict= 0\n\u623f \u95f4 \u88c5 \u4fee \u5dee \u8bbe \u65bd \u4e5f \u4e0d \u597d, \u65e9 \u9910 \u5f88 \u4e00 \u822c. \u4f4d \u7f6e \u504f \u50fb.\nlabel= 0\npredict= 0\n\u4ee5 \u524d \u5728 \u300a \u5973 \u4eba \u6211 \u6700 \u5927 \u300b \u8282 \u76ee \u4e2d \u6700 \u7231 \u770b kevin \u8001 \u5e08 \u53c2 \u4e88 \u7684 \u8282 \u76ee \uff0c \u6bcf \u6b21 \u770b \u4ed6 \u5bf9 \u5f69 \u5986 \u7684 \u8bb2 \u89e3 \uff0c \u81ea \u5df1 \u603b \u6709 \u6536 \u83b7 \u3002 \u800c \u4e14 \u542c \u4ed6 \u5bf9 \u5f69 \u5986 \u7684 \u4ecb \u7ecd \uff0c \u603b \u662f \u6d45 \u663e \u6613 \u61c2 \uff0c \u800c \u4ed6 \u7ed9 \u4eba \u7684 \u611f \u89c9 \u662f \u90a3 \u4e48 \u70ed \u5fc3 \u5730 \u8981 \u628a \u81ea \u5df1 \u6240 \u77e5 \u9053 \u7684 \u62a4 \u80a4 \u548c \u5f69 \u5986 \u7684 \u77e5 \u8bc6 \u6beb \u4e0d \u541d \u556c \u5730 \u6559 \u7ed9 \u5927 \u5bb6 \u3002 \u6628 \u5929 \u4e00 \u62ff \u5230 \u4e66 \u6211 \u5c31 \u8feb \u4e0d \u53ca \u5f85 \u5730 \u5f00 \u59cb \u4e86 \u65b0 \u4e00 \u8f6e \u7684 \u5b66 \u4e60 \uff0c \u5475 \u5475 \u53c8 \u6536 \u83b7 \u4e86 \u4e00 \u4e9b \u62a4 \u80a4 \u7684 \u5c0f \u77e5 \u8bc6 \u3002 \u800c \u4e14 \u6211 \u975e \u5e38 \u559c \u6b22 \u6574 \u672c \u4e66 \u7684 \u7f16 \u6392 \uff0c \u5f53 \u7136 \u8981 \u662f kevin \u8001 \u5e08 \u518d \u51fa \u4e00 \u4e2a dvd \u7279 \u8f91 \u8ddf \u4e66 \u4e00 \u8d77 \u5356 \u5c31 \u66f4 \u597d \u4e86 \u3002\nlabel= 1\npredict= 1\n\u6211 \u7684 \u8ba2 \u5355 \u662f 7 \u6708 17 \u65e5 \u5c31 \u53d1 \u51fa \u6765 \u4e86 \uff0c \u4f46 \u662f \u73b0 \u5728 \u662f 8 \u6708 2 \u65e5 \uff0c \u6211 \u8fd8 \u6ca1 \u6709 \u6536 \u5230 \u8d27 \u3002 \u600e \u4e48 \u53d1 \u8bc4 \u8bba \u5462 \uff1f \u8ba2 \u5355 \u4e0a \u6709 \u7535 \u8bdd \u5440 \uff0c \u5373 \u4f7f \u627e \u4e0d \u5230 \u5730 \u5740 \uff0c \u6253 \u4e2a \u7535 \u8bdd \u95ee \u4e00 \u4e0b \u5440 \uff0c \u5230 \u5e95 \u662f \u51fa \u5728 \u4ec0 \u4e48 \u5730 \u65b9 \uff0c \u5e0c \u671b \u8d35 \u7f51 \u597d \u597d \u67e5 \u4e00 \u4e0b \uff0c \u907f \u514d \u4ee5 \u540e \u518d \u51fa \u73b0 \u8fd9 \u6837 \u7684 \u95ee \u9898 \u3002\nlabel= 0\npredict= 0\n\u6211 \u4e8e 6 \u6708 1 \u65e5 \u518d \u6b21 \u5165 \u4f4f, \u4f4f \u7684 \u662f 1312 \u623f, \u9996 \u5148 \u4ef7 \u683c \u7531 238 \u5143 \u6da8 \u5230 278 \u5143, \u662f \u6574 \u4e2a \u4e4c \u9c81 \u6728 \u9f50 \u9152 \u5e97 \u65fa \u5b63 \u90fd \u6da8 \u4e86, \u636e \u8bf4 \u6bcf \u5e74 7 \u6708 \u548c 8 \u6708 \u4e4c \u9c81 \u6728 \u9f50 \u9152 \u5e97 \u90fd \u8fd8 \u8981 \u6da8, \u8fd9 \u4e5f \u53ef \u4ee5 \u7406 \u89e3, \u6bd5 \u7adf \u5c31 \u90a3 \u4e48 \u51e0 \u4e2a \u6708 \u53ef \u4ee5 \u8d5a \u94b1, \u51ac \u5929 \u90fd \u662f \u96f6 \u4e0b 20 \u51e0 \u5ea6, \u4eba \u5f88 \u5c11. \u4f46 \u662f \u6211 \u53d1 \u73b0 \u623f \u95f4 \u7684 \u8ff7 \u4f60 \u5427 \u64a4 \u4e86, \u53ea \u6709 \u7559 \u4e86 \u4e24 \u74f6 \u77ff \u6cc9 \u6c34. \u536b\nlabel= 0\npredict= 0\n\u65e0 \u9884 \u88c5 \u7cfb \u7edf \uff0c \u4e14 \u4e3b \u677f bios \u6ca1 \u6709 \u6253 \u5165 slic \u8bc1 \u4e66 \uff0c \u56e0 \u6b64 \u4e0d \u80fd \u6fc0 \u6d3b oem \u7248 vista \uff0c \u9020 \u6210 vista \u65e0 \u6cd5 \u4f7f \u7528 \u3002 \u6362 \u4e86 \u51e0 \u4e2a \u7248 \u672c vista \u548c \u6fc0 \u6d3b \u5de5 \u5177 \u90fd \u6ca1 \u6709 \u62ff \u4e0b vista \uff0c \u6211 \u4f1a \u518d \u8bd5 \uff0c \u4e5f \u8bf7 \u6210 \u529f \u5b89 \u88c5 vista \u7684 \u670b \u53cb \u6307 \u6559\nlabel= 0\npredict= 0\n\u4e3a \u4e86 \u767d \u5f66 \u7684 \u6b7b \uff0c \u4f24 \u5fc3 \u4e86 \u597d \u4e45 \u3002 \u5f88 \u4e45 \u6ca1 \u6709 \u4e3a \u4e86 \u4e00 \u672c \u4e66 \u3002 \u4e3a \u4e86 \u4e00 \u672c \u4e66 \u91cc \u90a3 \u4e9b \u5982 \u6b64 \u8d34 \u8fd1 \u751f \u6d3b \u7684 \u6587 \u5b57 \u800c \u611f \u52a8 \u4f24 \u5fc3 \u4e86 \u3002 \u559c \u6b22 \u5b81 \u9ed8 \u7684 \u6f47 \u6d12 \uff0c \u521a \u5f00 \u59cb \u7684 \u65f6 \u5019 \u8fd8 \u5728 \u4e3a \u4ed6 \u4eec \u4e24 \u4e2a \u5728 \u9762 \u5bf9 \u7231 \u60c5 \u65f6 \u7684 \u6e05 \u9192 \u800c \u4e0d \u89e3 \u3002 \u6700 \u540e \u76f4 \u5230 \u767d \u5f66 \u610f \u5916 \u6b7b \u540e \u51fa \u73b0 \u7684 \u5730 \u56fe \uff0c \u5b81 \u9ed8 \u7684 \u53cd \u5e94 \u3002 \u624d \u77e5 \u9053 \u539f \u6765 \u4ed6 \u4eec \u90fd \u7231 \u5f7c \u6b64 \u7231 \u7684 \u90a3 \u4e48 \u6df1 \u3002 \u3002 \u3002 \u3002 \u60f3 \u8d77 \u4e86 \u5b81 \u9ed8 \u90a3 \u53e5 \u8bdd \uff0c \u5979 \u8ddf \u6881 \u8473 \u8473 \u8bf4 \u5979 \u5b81 \u613f \u8f93 \u7ed9 \u5979 \uff0c \u8f93 \u7ed9 \u4efb \u4f55 \u4eba \u4e5f \u4e0d \u60f3 \u8f93 \u7ed9 \u751f \u79bb \u6b7b \u522b \u3002 \u3002 \u3002 \u3002\nlabel= 1\npredict= 1\n\uff19\uff18 \u5e74 \u8fd9 \u5bb6 \u9152 \u5e97 \u5f00 \u4e1a \u65f6 \u4f4f \u8fc7 \uff0c \u611f \u89c9 \u5f88 \u597d \uff0c \u4eca \u5e74 \u53bb \u9752 \u5c9b \u8fd8 \u662f \u9009 \u62e9 \u4e86 \u4f4f \u5728 \u8fd9 \u91cc \uff0e \uff11\uff10 \u5e74 \u8fc7 \u53bb \u4e86 \uff0c \u867d \u7136 \u786c \u4ef6 \u7565 \u663e \u9648 \u65e7 \u4f46 \u670d \u52a1 \u6c34 \u5e73 \u4f9d \u65e7 \u8ba9 \u6211 \u611f \u89c9 \u5f88 \u6ee1 \u610f \uff0e\nlabel= 1\npredict= 1\n\u8fd8 \u7a0d \u5fae \u91cd \u4e86 \u70b9 \uff0c \u53ef \u80fd \u662f \u786c \u76d8 \u5927 \u7684 \u539f \u6545 \uff0c \u8fd8 \u8981 \u518d \u8f7b \u534a \u65a4 \u5c31 \u597d \u4e86 \u3002 \u5176 \u4ed6 \u8981 \u8fdb \u4e00 \u6b65 \u9a8c \u8bc1 \u3002 \u8d34 \u7684 \u51e0 \u79cd \u819c \u6c14 \u6ce1 \u8f83 \u591a \uff0c \u7528 \u4e0d \u4e86 \u591a \u4e45 \u5c31 \u8981 \u66f4 \u6362 \u4e86 \uff0c \u5c4f \u5e55 \u819c \u7a0d \u597d \u70b9 \uff0c \u4f46 \u6bd4 \u6ca1 \u6709 \u8981 \u5f3a \u591a \u4e86 \u3002 \u5efa \u8bae \u914d \u8d60 \u51e0 \u5f20 \u819c \u8ba9 \u7528 \u7528 \u6237 \u81ea \u5df1 \u8d34 \u3002\nlabel= 0\npredict= 0\n\u770b \u5230 \u8fd9 \u672c \u4e66 \u7ed9 \u6211 \u7684 \u7b2c \u4e00 \u611f \u89c9 \u662f \uff1a \u76f8 \u89c1 \u6068 \u665a \uff01 \u4ee5 \u524d \u603b \u4ee5 \u4e3a \u81ea \u5df1 \u4ec0 \u4e48 \u90fd \u770b \u5f97 \u5f00 \u3001 \u4ec0 \u4e48 \u90fd \u5f88 \u61c2 \u3001 \u4ec0 \u4e48 \u90fd \u4e0d \u5728 \u4e4e \u3002 \u53ef \u5728 \u770b \u8fd9 \u672c \u4e66 \u7684 \u8fc7 \u7a0b \u4e2d \uff0c \u6211 \u611f \u5230 \u81ea \u5df1 \u771f \u5f97 \u662f \u4ec0 \u4e48 \u90fd \u4e0d \u61c2 \uff0c \u7279 \u522b \u662f \u5728 \u505a \u4eba \u65b9 \u9762 \u5f88 \u6b20 \u7f3a \u3002 \u94b1 \u6559 \u6388 \u4ee5 \u524d \u5728 \u767e \u5bb6 \u8bb2 \u575b \u8bb2 \u89e3 \u7384 \u85cf \u65f6 \uff0c \u6211 \u5c31 \u611f \u5230 \u4ed6 \u5f88 \u4eb2 \u5207 \uff0c \u7279 \u522b \u662f \u4e00 \u4e9b \u9053 \u7406 \u8bb2 \u5f97 \u5f88 \u901a \u4fd7 \u6613 \u61c2 \u3002 \u8fd9 \u6b21 \u8bb2 \u89e3 \u4e09 \u5b57 \u7ecf \u66f4 \u662f \u4e00 \u4e2a \u96be \u5f97 \u7684 \u5b66 \u4e60 \u673a \u4f1a \uff0c \u867d \u7136 \u8fd9 \u672c \u4e66 \u6211 \u8fd8 \u6ca1 \u6709 \u770b \u5b8c \uff0c \u4f46 \u6211 \u5df2 \u7ecf \u88ab \u6df1 \u6df1 \u5730 \u5438 \u5f15 \u4f4f \u4e86 \uff0c \u5e0c \u671b \u5927 \u5bb6 \u4e5f \u80fd \u4ece \u4e2d \u6709 \u6240 \u611f \u609f \u3002\nlabel= 1\npredict= 1\n\u505a \u5de5 \u624e \u5b9e \uff0c \u6309 \u952e \u8bbe \u8ba1 \u5f88 \u4eba \u6027 \u5316 \uff0c \u6563 \u70ed \u5f88 \u597d \uff0c \u5e26 \u6709 \u6b63 \u7248 \u7cfb \u7edf \u8ba9 \u4eba \u7528 \u7740 \u6709 \u70b9 \u540d \u6b63 \u8a00 \u987a \u7684 \u611f \u89c9 \u3002 \u5b8c \u7f8e \u5c4f \uff0c \u5f88 \u50cf \u955c \u9762 \u5c4f \u3002 \u8fd9 \u6b3e \u662f \u5e26 \u6709 \u9ea6 \u514b \u98ce \u7684 \uff0c \u597d \u50cf \u53c2 \u6570 \u91cc \u6ca1 \u5199 \uff0c \u6709 \u70b9 \u60ca \u559c \u3002 \u89e3 \u538b \u7cfb \u7edf \u5f88 \u5feb \uff0c \u52a0 \u88c5 \u5185 \u5b58 \u4e5f \u65b9 \u4fbf \uff0c \u62c6 \u4e86 \u4e24 \u4e2a \u87ba \u4e1d \u62a0 \u4e0b \u540e \u76d6 \u5c31 \u53ef \u4ee5 \u63d2 \u4e86 \uff0c \u8bf4 \u660e \u4e66 \u91cc \u6709 \u56fe \u89e3 \u3002 \u8d5e \u4e00 \u4e0b \u4eac \u4e1c \u9001 \u5230 \u8d60 \u54c1 \uff1a thinkpad \u539f \u88c5 \u5305 \u5305 \u6211 \u8d85 \u559c \u6b22 \uff0c \u7b2c \u4e00 \u6b21 \u62e5 \u6709 \u8fd9 \u4e48 \u8bbe \u8ba1 \u4eba \u6027 \u5316 \u7684 \u5305 \u5305 \uff1b \u4e09 \u661f \u5185 \u5b58 \u6761 1g \uff0c \u5f88 \u5bb9 \u6613 \u5b89 \u88c5 \uff0c \u5f25 \u8865 \u4e86 \u672c \u672c \u5185 \u5b58 \u5c0f \u7684 \u7f3a \u9677 \uff1b thinkpad \u5c0f \u9ed1 \u9f20 \u6807 \uff0c \u597d \u7528 \u53ef \u7231 \uff1b\nlabel= 1\npredict= 1\n\u8d3e \u5fd7 \u521a \u662f \u6709 \u624d \u7684 \uff0c \u6240 \u4ee5 \u5199 \u4e86 \u6625 \u79cb \uff0c \u4f46 \u4eba \u561b \u5c31 \u5411 \u6625 \u79cb \u90a3 \u6837 \u4e71 \uff0c \u4e0d \u5bf9 \uff0c \u8bf4 \u9519 \u4e86 \uff01 \u5e94 \u8be5 \u8bf4 \uff0c \u4eba \u5fc3 \u561b \uff0c \u5c31 \u50cf \u6625 \u79cb \u90a3 \u6837 \u4e71 \u3002 \u6240 \u4ee5 \u300a \u6625 \u79cb \u300b \u4e00 \u51fa \uff0c \u8d5e \u5f39 \u7686 \u6709 \uff0c \u5176 \u5b9e \u8d3e \u5fd7 \u521a \u8bf4 \u5f97 \u5bf9 \uff0c \u4e0d \u770b \u6625 \u79cb \uff0c \u8fde \u7956 \u5b97 \u59d3 \u4ec0 \u4e48 \u90fd \u4e0d \u77e5 \u9053 \u5462 \uff01 \u867d \u7136 \u4e66 \u4e2d \u53e5 \u5b50 \u7565 \u6709 \u4e0d \u96c5 \uff0c \u5c31 \u662f \u4fd7 \u4e86 \u70b9 \uff0c \u4f46 \u4fd7 \u5bb9 \u6613 \u770b \u660e \u767d \u3002 \u5982 \u679c \u5199 \u4e66 \u7684 \u90fd \u6574 \u70b9 \u6587 \u8a00 \u6587 \u6216 \u8005 \u4e3a \u4e0d \u4fd7 \uff0c \u6545 \u610f \u8f6c \u5f2f \u6ca1 \u89d2 \uff0c \u6625 \u79cb \u4e0d \u7528 \u5199 \u4e86 \uff0c \u4e5f \u4e0d \u7528 \u770b \u4e86 \uff0c \u7956 \u5b97 \u4e5f \u4e0d \u7528 \u7406 \u4e86 \uff01 \u6240 \u4ee5 \u6625 \u79cb \u8981 \u770b \uff0c \u4e0d \u7ba1 \u4ed6 \u5199 \u5f97 \u4fd7 \u4e0d \u4fd7 \u3002\nlabel= 1\npredict= 0\n\u786c \u4ef6 \u5f88 \u597d, \u670d \u52a1 \u8d28 \u91cf \u4e5f \u5f88 \u597d, \u7279 \u522b \u662f \u9152 \u5e97 \u5468 \u56f4 \u73af \u5883 \u975e \u5e38 \u4f18 \u96c5. \u65e9 \u9910 \u4e5f \u4e0d \u9519, \u4e0b \u6b21 \u53bb \u8fd8 \u4f1a \u5165 \u4f4f\nlabel= 1\npredict= 1<\/code><\/pre>\n\u4ece\u6d4b\u8bd5\u7ed3\u679c\u4e0a\u6765\u770b\uff0c\u6709\u4e00\u4e9b\u9519\u8bef\uff0c\u4f46\u5927\u90e8\u5206\u9884\u6d4b\u662f\u6b63\u786e\u7684\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"
\u4f7f\u7528\u8bad\u7ec3\u5de5\u5177 \u5728\u672c\u7ae0\u4e2d\u5c06\u4f7f\u7528\u4e00\u4e2a\u60c5\u611f\u5206\u7c7b\u4efb\u52a1\u7684\u4f8b\u5b50\u6765\u518d\u8bad\u7ec3\u4e00\u4e2a\u6a21\u578b\uff0c\u4ee5\u6b64\u6765\u8bb2\u89e3HuggingFace\u8bad\u7ec3\u5de5\u5177\u7684\u4f7f\u7528\u65b9\u6cd5\u3002 #\u7b2c6\u7ae0\/\u52a0\u8f7d \u2026 …<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ngg_post_thumbnail":0},"categories":[278],"tags":[279],"_links":{"self":[{"href":"\/index.php?rest_route=\/wp\/v2\/posts\/9373"}],"collection":[{"href":"\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=9373"}],"version-history":[{"count":3,"href":"\/index.php?rest_route=\/wp\/v2\/posts\/9373\/revisions"}],"predecessor-version":[{"id":9376,"href":"\/index.php?rest_route=\/wp\/v2\/posts\/9373\/revisions\/9376"}],"wp:attachment":[{"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=9373"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=9373"},{"taxonomy":"post_tag","embeddable":true,"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=9373"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}