当前位置: 首页 > news >正文

做网站的商家怎么后去流量费阿里云域名注册入口官网

做网站的商家怎么后去流量费,阿里云域名注册入口官网,三网合一网站建设计划,wordpress在页眉加载jsBert概述 BERT(Bidirectional Encoder Representations from Transformers)是一种深度学习模型,用于自然语言处理(NLP)任务。BERT的核心是由一种强大的神经网络架构——Transformer驱动的。这种架构包含了一种称为自注…

Bert概述

BERT(Bidirectional Encoder Representations from Transformers)是一种深度学习模型,用于自然语言处理(NLP)任务。BERT的核心是由一种强大的神经网络架构——Transformer驱动的。这种架构包含了一种称为自注意力的机制,使BERT能够根据上下文(前后文)来衡量每个词的重要性。这种上下文感知赋予BERT生成上下文化词嵌入的能力,即考虑句子中词义的词表示。这就像BERT反复阅读句子以深入理解每个词的作用。

BERT的训练方式有两种:Masked Language Model和Next Sentence Prediction。参考这里

基于BERT实现文本情感分类

所谓情感分类就是指判断句子是积极情感还是消极情感,例如说“今天这顿饭太美味了”是积极的情感,“今天这顿饭简直吃不下去”是消极的情感。

基于BERT完成情感分类的基本思路如图所示。我们知道BERT是一个预训练模型,我们把句子扔给它的时候,它对应每个字都会输出一个向量。但是在把句子扔给BERT之前,我们会在句子最前面增加一个特殊符号[CLS]。对应这个[CLS],BERT也会输出一个向量,我们就是利用这个向量来进行情感分类。为什么可以直接利用这个向量呢?这是因为BERT内部采用的是自注意力机制,自注意力机制的特点是考虑全局又聚焦重点,实际上[CLS]对应的向量已经嵌入了整个句子的信息,而且重点词字嵌入的信息权重要大。所以,我们将这个向量扔给一个全连接层,就可以完成分类任务了。参考这里

img

代码

数据预处理

数据集的下载,提取码为zfh3

import pandas as pd
import os
import logginglogging.basicConfig(format='%(asctime)s - %(levelname)s - %(name)s -   %(message)s',datefmt='%m/%d/%Y %H:%M:%S',level=logging.INFO)
logger = logging.getLogger(__name__)class InputExample(object):"""A single training/test example for simple sequence classification."""def __init__(self, text, label=None):self.text = textself.label = labelclass InputFeatures(object):"""A single set of features of data."""def __init__(self, input_ids, input_mask, segment_ids, label_id):self.input_ids = input_idsself.input_mask = input_maskself.segment_ids = segment_idsself.label_id = label_idclass DataProcessor(object):"""Base class for data converters for sequence classification data sets."""def get_train_examples(self, data_dir):"""Gets a collection of `InputExample`s for the train set."""raise NotImplementedError()def get_dev_examples(self, data_dir):"""Gets a collection of `InputExample`s for the dev set."""raise NotImplementedError()def get_test_examples(self, data_dir):"""Gets a collection of `InputExample`s for the test set."""raise NotImplementedError()def get_labels(self):"""Gets the list of labels for this data set."""raise NotImplementedError()@classmethoddef _read_csv(cls, input_file, quotechar=None):"""Reads a tab separated value file."""# dicts = []data = pd.read_csv(input_file)return dataclass MyPro(DataProcessor):'''自定义数据读取方法,针对json文件Returns:examples: 数据集,包含index、中文文本、类别三个部分'''def get_train_examples(self, data_dir):return self._create_examples(self._read_csv(os.path.join(data_dir, 'train_data.csv')), 'train')def get_dev_examples(self, data_dir):return self._create_examples(self._read_csv(os.path.join(data_dir, 'dev_data.csv')), 'dev')def get_test_examples(self, data_dir):return self._create_examples(self._read_csv(os.path.join(data_dir, 'test_data.csv')), 'test')def get_labels(self):return [0, 1]def _create_examples(self, data, set_type):examples = []for index, row in data.iterrows():# guid = "%s-%s" % (set_type, i)text = row['review']label = row['label']examples.append(InputExample(text=text, label=label))return examplesdef convert_examples_to_features(examples, label_list, max_seq_length, tokenizer, show_exp=True):'''Loads a data file into a list of `InputBatch`s.Args:examples      : [List] 输入样本,句子和labellabel_list    : [List] 所有可能的类别,0和1max_seq_length: [int] 文本最大长度tokenizer     : [Method] 分词方法Returns:features:input_ids  : [ListOf] token的id,在chinese模式中就是每个分词的id,对应一个word vectorinput_mask : [ListOfInt] 真实字符对应1,补全字符对应0segment_ids: [ListOfInt] 句子标识符,第一句全为0,第二句全为1label_id   : [ListOfInt] 将Label_list转化为相应的id表示'''label_map = {}for (i, label) in enumerate(label_list):label_map[label] = ifeatures = []for (ex_index, example) in enumerate(examples):# 分词tokens = tokenizer.tokenize(example.text)# tokens进行编码encode_dict = tokenizer.encode_plus(text=tokens,max_length=max_seq_length,pad_to_max_length=True,is_pretokenized=True,return_token_type_ids=True,return_attention_mask=True)input_ids = encode_dict['input_ids']input_mask = encode_dict['attention_mask']segment_ids = encode_dict['token_type_ids']assert len(input_ids) == max_seq_lengthassert len(input_mask) == max_seq_lengthassert len(segment_ids) == max_seq_lengthlabel_id = label_map[example.label]if ex_index < 5 and show_exp:logger.info("*** Example ***")logger.info("tokens: %s" % " ".join([str(x) for x in tokens]))logger.info("input_ids: %s" % " ".join([str(x) for x in input_ids]))logger.info("input_mask: %s" % " ".join([str(x) for x in input_mask]))logger.info("segment_ids: %s" % " ".join([str(x) for x in segment_ids]))logger.info("label: %s (id = %d)" % (example.label, label_id))features.append(InputFeatures(input_ids=input_ids,input_mask=input_mask,segment_ids=segment_ids,label_id=label_id))return features

如何理解?

将原始文本数据通过分词、编码等步骤转换为模型训练所需的格式,包括input_ids(编码后的token)、input_mask(注意力掩码)和segment_ids(token类型ids)。这些数据将作为模型的输入。

假设我们有一个文本示例,并且我们使用BERT分词器进行预处理。以下是示例文本和初始化分词器的代码:

from transformers import BertTokenizer# 示例文本
text = "Hello, how are you?"# 初始化BERT分词器
tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')# 假设最大序列长度为10
max_seq_length = 10

接下来,我们将通过上面的代码片段对文本进行预处理:

# 假设我们的examples是一个包含单个文本的列表
examples = [{'text': text}]# 遍历示例列表
for (ex_index, example) in enumerate(examples):# 分词tokens = tokenizer.tokenize(example['text'])# tokens进行编码encode_dict = tokenizer.encode_plus(text=tokens,max_length=max_seq_length,pad_to_max_length=True,is_pretokenized=True,return_token_type_ids=True,return_attention_mask=True)input_ids = encode_dict['input_ids']input_mask = encode_dict['attention_mask']segment_ids = encode_dict['token_type_ids']# 打印结果print(f"Example {ex_index}")print(f"Tokens: {tokens}")print(f"Input IDs: {input_ids}")print(f"Input Mask: {input_mask}")print(f"Segment IDs: {segment_ids}")

执行上述代码后,我们将得到以下输出(输出可能会根据BERT模型的版本和分词器设置略有不同):

Example 0
Tokens: ['Hello', ',', 'how', 'are', 'you', '?']
Input IDs: [101, 7592, 1010, 2129, 2026, 102, 0, 0, 0, 0]
Input Mask: [1, 1, 1, 1, 1, 1, 0, 0, 0, 0]
Segment IDs: [0, 0, 0, 0, 0, 0, 0, 0, 0, 0]

解释输出:

  • Tokens: 这是分词后的结果,原始文本被拆分为BERT模型可以理解的token。
  • Input IDs: 每个token被转换为一个唯一的整数ID,表示其在词汇表中的位置。
  • Input Mask: 表示哪些位置是真正的token(1),哪些位置是填充的(0)。在这个例子中,填充的部分是最后四个0。
  • Segment IDs: 由于我们只有一个句子,所以所有token的segment ID都是0。如果文本包含多个句子,第二个句子的token将有一个不同的segment ID(通常是1)。
数据处理成dataSet
import torch
from torch.utils.data import Datasetclass MyDataset(Dataset):def __init__(self, features, mode):self.nums = len(features)self.input_ids = [torch.tensor(example.input_ids).long() for example in features]self.input_mask = [torch.tensor(example.input_mask).float() for example in features]self.segment_ids = [torch.tensor(example.segment_ids).long() for example in features]self.label_id = Noneif mode == 'train' or 'test':self.label_id = [torch.tensor(example.label_id) for example in features]def __getitem__(self, index):data = {'input_ids': self.input_ids[index],'input_mask': self.input_mask[index],'segment_ids': self.segment_ids[index]}if self.label_id is not None:data['label_id'] = self.label_id[index]return datadef __len__(self):return self.nums
模型的搭建
from torch import nn
import os
from transformers import BertModelclass ClassifierModel(nn.Module):def __init__(self,bert_dir,dropout_prob=0.1):super(ClassifierModel, self).__init__()config_path = os.path.join(bert_dir, 'config.json')assert os.path.exists(bert_dir) and os.path.exists(config_path), \'pretrained bert file does not exist'self.bert_module = BertModel.from_pretrained(bert_dir)self.bert_config = self.bert_module.configself.dropout_layer = nn.Dropout(dropout_prob)out_dims = self.bert_config.hidden_sizeself.obj_classifier = nn.Linear(out_dims, 2)def forward(self,input_ids,input_mask,segment_ids,label_id=None):bert_outputs = self.bert_module(input_ids=input_ids,attention_mask=input_mask,token_type_ids=segment_ids)seq_out, pooled_out = bert_outputs[0], bert_outputs[1]#对反向传播及逆行截断x = pooled_out.detach()out = self.obj_classifier(x)return out
模型的训练

BERT是一个预训练模型,我们把句子扔给它的时候,它对应每个字都会输出一个向量。【下载Bert模型==>操作手册】

from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
from model import *
from dataset import *
from dataProcessor import *
import matplotlib.pyplot as plt
import time
from transformers import BertTokenizer
from transformers import logginglogging.set_verbosity_warning()
# 加载训练数据
datadir = "data"
bert_dir = "bert\\bert-chinese"
my_processor = MyPro()
label_list = my_processor.get_labels()train_data = my_processor.get_train_examples(datadir)
test_data = my_processor.get_test_examples(datadir)tokenizer = BertTokenizer.from_pretrained(bert_dir)train_features = convert_examples_to_features(train_data, label_list, 128, tokenizer)
test_features = convert_examples_to_features(test_data, label_list, 128, tokenizer)
train_dataset = MyDataset(train_features, 'train')
test_dataset = MyDataset(test_features, 'test')
train_data_loader = DataLoader(dataset=train_dataset, batch_size=64, shuffle=True)
test_data_loader = DataLoader(dataset=test_dataset, batch_size=64, shuffle=True)train_data_len = len(train_dataset)
test_data_len = len(test_dataset)
print(f"训练集长度:{train_data_len}")
print(f"测试集长度:{test_data_len}")# 创建网络模型
my_model = ClassifierModel(bert_dir)# 损失函数
loss_fn = nn.CrossEntropyLoss()# 优化器
learning_rate = 5e-3
#optimizer = torch.optim.SGD(my_model.parameters(), lr=learning_rate)
#  Adam 参数betas=(0.9, 0.99)
optimizer = torch.optim.Adam(my_model.parameters(), lr=learning_rate, betas=(0.9, 0.99))
# 总共的训练步数
total_train_step = 0
# 总共的测试步数
total_test_step = 0
step = 0
epoch = 50train_loss_his = []
train_totalaccuracy_his = []
test_totalloss_his = []
test_totalaccuracy_his = []
start_time = time.time()
my_model.train()
for i in range(epoch):print(f"-------第{i}轮训练开始-------")train_total_accuracy = 0for step, batch_data in enumerate(train_data_loader):# writer.add_images("tarin_data", imgs, total_train_step)print(batch_data['input_ids'].shape)output = my_model(**batch_data)loss = loss_fn(output, batch_data['label_id'])train_accuracy = (output.argmax(1) == batch_data['label_id']).sum()train_total_accuracy = train_total_accuracy + train_accuracyoptimizer.zero_grad()loss.backward()optimizer.step()total_train_step = total_train_step + 1train_loss_his.append(loss)#writer.add_scalar("train_loss", loss.item(), total_train_step)train_total_accuracy = train_total_accuracy / train_data_lenprint(f"训练集上的准确率:{train_total_accuracy}")train_totalaccuracy_his.append(train_total_accuracy)# 测试开始total_test_loss = 0my_model.eval()test_total_accuracy = 0with torch.no_grad():for batch_data in test_data_loader:output = my_model(**batch_data)loss = loss_fn(output, batch_data['label_id'])total_test_loss = total_test_loss + losstest_accuracy = (output.argmax(1) == batch_data['label_id']).sum()test_total_accuracy = test_total_accuracy + test_accuracytest_total_accuracy = test_total_accuracy / test_data_lenprint(f"测试集上的准确率:{test_total_accuracy}")print(f"测试集上的loss:{total_test_loss}")test_totalloss_his.append(total_test_loss)test_totalaccuracy_his.append(test_total_accuracy)torch.save(my_model, "bert_{}.pth".format(i))print("模型已保存")
模型的预测
# 假设这是您的分词器和预处理函数
from torch.utils.data import DataLoader
from transformers import BertTokenizer
import torch
from dataProcessor import convert_examples_to_features, MyPro, InputExample
from dataset import MyDatasetbert_dir = "bert\\bert-chinese"
tokenizer = BertTokenizer.from_pretrained(bert_dir)my_processor = MyPro()
label_list = my_processor.get_labels()# 从键盘读取输入
input_text = input("请输入一句话来判断其情感:")# 创建一个InputExample对象
input_texts = InputExample(text=input_text, label=0)  # 假设0表示消极,1表示积极# 使用convert_examples_to_features函数处理输入语句
test_features = convert_examples_to_features([input_texts], label_list, 128, tokenizer)
test_dataset = MyDataset(test_features, 'test')
test_data_loader = DataLoader(dataset=test_dataset, batch_size=64, shuffle=True)# 加载模型
my_model = torch.load("bert_10.pth", map_location=torch.device('cpu'))
my_model.eval()
with torch.no_grad():for batch_data in test_data_loader:outputs = my_model(**batch_data)# 判断类别
if outputs.argmax().item() == 1:print("积极")
else:print("消极")

视频推荐

Bert模型和Transformer到底哪个更牛?

用BERT做下游任务的栗子

文章推荐

BERT与Transformer:深入比较两者的差异 (baidu.com)

BERT模型和Transformer模型之间有何关系?_bert和transformer的关系-CSDN博客

掌握BERT:从初学者到高级的自然语言处理(NLP)全面指南 - IcyFeather233 - 博客园 (cnblogs.com)


文章转载自:
http://dinncoword.zfyr.cn
http://dinncowormhole.zfyr.cn
http://dinncocorker.zfyr.cn
http://dinncobeheld.zfyr.cn
http://dinncoscratcher.zfyr.cn
http://dinncopaddle.zfyr.cn
http://dinncodispersive.zfyr.cn
http://dinncovliw.zfyr.cn
http://dinncosubtropical.zfyr.cn
http://dinncopreheating.zfyr.cn
http://dinncooutsparkle.zfyr.cn
http://dinncorogatory.zfyr.cn
http://dinncopharmaceutist.zfyr.cn
http://dinncoknavery.zfyr.cn
http://dinncosubmediant.zfyr.cn
http://dinncowean.zfyr.cn
http://dinncofoliar.zfyr.cn
http://dinncotranspositive.zfyr.cn
http://dinncounimaginable.zfyr.cn
http://dinncoredrop.zfyr.cn
http://dinncotrillionth.zfyr.cn
http://dinncospiculum.zfyr.cn
http://dinncoreichstag.zfyr.cn
http://dinncookhotsk.zfyr.cn
http://dinncosolarium.zfyr.cn
http://dinncolasthome.zfyr.cn
http://dinncomisinform.zfyr.cn
http://dinncocoexist.zfyr.cn
http://dinncolaevorotatory.zfyr.cn
http://dinncooutlandish.zfyr.cn
http://dinncoquackupuncture.zfyr.cn
http://dinncohappening.zfyr.cn
http://dinncodocudrama.zfyr.cn
http://dinncopervasive.zfyr.cn
http://dinncocalcography.zfyr.cn
http://dinncoflatboat.zfyr.cn
http://dinncoenslavement.zfyr.cn
http://dinncobatten.zfyr.cn
http://dinncoarthrodic.zfyr.cn
http://dinncostrafe.zfyr.cn
http://dinncopostclassical.zfyr.cn
http://dinncokink.zfyr.cn
http://dinncoremint.zfyr.cn
http://dinncofated.zfyr.cn
http://dinncochord.zfyr.cn
http://dinncoloi.zfyr.cn
http://dinncobenzoline.zfyr.cn
http://dinncopga.zfyr.cn
http://dinncoensepulchre.zfyr.cn
http://dinncomortice.zfyr.cn
http://dinncobewildering.zfyr.cn
http://dinncocrenelated.zfyr.cn
http://dinncochape.zfyr.cn
http://dinncostagestruck.zfyr.cn
http://dinncoresold.zfyr.cn
http://dinncoantennae.zfyr.cn
http://dinncomidwifery.zfyr.cn
http://dinncocretin.zfyr.cn
http://dinncoensiform.zfyr.cn
http://dinncosheryl.zfyr.cn
http://dinncoimbrutement.zfyr.cn
http://dinncosack.zfyr.cn
http://dinncoideally.zfyr.cn
http://dinncoappreciably.zfyr.cn
http://dinncophenomenalism.zfyr.cn
http://dinncopessimist.zfyr.cn
http://dinncogallnut.zfyr.cn
http://dinncosnacketeria.zfyr.cn
http://dinncovendetta.zfyr.cn
http://dinncodeduct.zfyr.cn
http://dinncoacetylco.zfyr.cn
http://dinncoroyalties.zfyr.cn
http://dinncosadic.zfyr.cn
http://dinncowerner.zfyr.cn
http://dinncopapertrain.zfyr.cn
http://dinncostrategics.zfyr.cn
http://dinncokithe.zfyr.cn
http://dinncometage.zfyr.cn
http://dinncotyphogenic.zfyr.cn
http://dinncoincoherently.zfyr.cn
http://dinncoextrasystole.zfyr.cn
http://dinnconaad.zfyr.cn
http://dinncomolotov.zfyr.cn
http://dinncocutaneous.zfyr.cn
http://dinncotire.zfyr.cn
http://dinnconinette.zfyr.cn
http://dinncosephardi.zfyr.cn
http://dinncoheteromorphic.zfyr.cn
http://dinncogunyah.zfyr.cn
http://dinncomegakaryocyte.zfyr.cn
http://dinncosnowswept.zfyr.cn
http://dinncosian.zfyr.cn
http://dinncoblellum.zfyr.cn
http://dinncoelectrode.zfyr.cn
http://dinncoexotropia.zfyr.cn
http://dinncobackward.zfyr.cn
http://dinncocollectivity.zfyr.cn
http://dinncoaussie.zfyr.cn
http://dinncokrooboy.zfyr.cn
http://dinncogermicide.zfyr.cn
http://www.dinnco.com/news/129523.html

相关文章:

  • 社保网站是每月1-6号都是在建设中的吗网站构建的基本流程
  • 建立网站要钱吗关键词优化推广
  • 企业网站的开发流程是什么关键词爱站网关键词挖掘工具
  • 宁国做网站的公司seo是什么工作内容
  • 开发直播软件需要多少钱站长之家seo综合
  • 无为县城乡建设局网站自媒体推广渠道
  • 如何利用谷歌云做自己的网站网站推广与优化平台
  • 网站建设的过程有哪些新东方烹饪学校
  • 江西做网站多少钱河北seo诊断培训
  • 免费安装电脑wordpress网络推广优化seo
  • 公司的网站设计方案外链发布论坛
  • 珠海做网站哪家好商丘优化公司
  • 做购物网站的引言手机百度助手
  • dw怎么把代码做成网页想做seo哪里有培训的
  • 工信部icp备案管理系统在哪里可以免费自学seo课程
  • 上海高端网站营销策略理论
  • 南通建网站的公司本溪seo优化
  • 怎么用网吧电脑做网站服务器吗朋友圈广告推广平台
  • 怎么做网站公众号如何推广小程序平台
  • 网站优化排名易下拉技术全国培训机构排名前十
  • 你有网站 我做房东 只收佣金的网站搜索引擎快速优化排名
  • 淘宝联盟建网站seo是指什么岗位
  • 东莞网站平台价格合肥seo网站排名优化公司
  • 怎么创作一个微信小程序深圳seo专家
  • 一个域名能同时做2个网站吗学营销app哪个更好
  • 做网站iiwok精准粉丝引流推广
  • 以前做视频的网站品牌推广服务
  • dw做六个页面的网站百度网盘网页登录入口
  • 建设一个域名抢注的网站网络营销的原理
  • 营销型网站规划步骤以营销推广为主题的方案