第N5周：Pytorch文本分类入门

🕗 发布于 2024-11-04 14:52 pytorch 分类 人工智能

🍨 本文为🔗365天深度学习训练营中的学习记录博客
🍖 原作者：K同学啊

本周任务：

了解文本分类的基本流程
学习常用数据清洗方法
学习如何使用jieba实现英文分词
学习如何构建文本向量

前期准备

加载数据

import torch
import torch.nn as nn
import torchvision
from torchvision import transforms, datasets
import os,PIL,pathlib,warnings

warnings.filterwarnings('ignore')

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print(device)

加载AG News数据集

from torchtext.datasets import AG_NEWS

train_iter = AG_NEWS(split='train')

构建词典

from torchtext.data.utils import get_tokenizer
from torchtext.vocab import build_vocab_from_iterator

tokenizer = get_tokenizer('basic_english')

def yield_tokens(data_iter):
    for _, text in data_iter:
        yield tokenizer(text)

vocab = build_vocab_from_iterator(yield_tokens(train_iter), specials=['<unk'])
vocab.set_default_index(vocab['<unk'])

vocab(['here','is','an','example'])

text_pipeline = lambda x: vocab(tokenizer(x))
label_pipeline = lambda x: int(x) - 1

text_pipeline('here is the an example')

label_pipeline('10')

生成数据批次和迭代器

from torch.utils.data import DataLoader

def collate_batch(batch):
    label_list, text_list, offsets = [],[],[0]

    for (_label, _text) in batch:
        label_list.append(label_pipeline(_label))

        processed_text = torch.tensor(text_pipeline(_text), dtype=torch.int64)
        text_list.append(processed_text)

        offsets.append(processed_text.size(0))

    label_list = torch.tensor(label_list, dtype=torch.int64)
    text_list = torch.cat(text_list)
    offsets = torch.tensor(offsets[:-1]).cumsum(dim=0)

    return label_list.to(device), text_list.to(device), offsets.to(device)

datalodaer = DataLoader(train_iter, batch_size=8, shuffle=False, collate_fn=collate_batch)

准备模型

定义模型

from torch import nn

class TextClassificationModel(nn.Module):

    def __init__(self, vocab_size, embed_dim, num_class):
        super(TextClassificationModel, self).__init__()

        self.embedding = nn.EmbeddingBag(vocab_size, embed_dim, sparse=False)

        self.fc = nn.Linear(embed_dim, num_class)
        self.init_weights()

    def init_weights(self):
        initrange = 0.5
        self.embedding.weight.data.uniform_(-initrange,initrange)
        self.fc.weight.data.uniform_(-initrange,initrange)
        self.fc.bias.data.zero_()

    def forward(self,text,offsets):
        embedded = self.embedding(text,offsets)
        return self.fc(embedded)

定义实例

num_class = len(set([label for (label,text) in train_iter]))
vocab_size = len(vocab)
em_size = 64
model = TextClassificationModel(vocab_size,em_size,num_class).to(device)

定义训练函数与评估函数

import time

def train(dataloader):
    model.train()
    total_acc, train_loss, total_count = 0,0,0
    log_interval = 500
    start_time = time.time()

    for idx, (label,text,offsets) in enumerate(dataloader):

        predicted_label = model(text, offsets)

        optimizer.zero_grad()
        loss = criterion(predicted_label, label)
        loss.backward()
        optimizer.step()

        total_acc += (predicted_label.argmax(1) == label).sum().item()
        train_loss += loss.item()
        total_count += label.size(0)

        if idx % log_interval == 0 and idx > 0:
            elapsed = time.time() - start_time
            print('| epoch {:1d} | {:4d}/{:4d} batches'
                  '| train_acc {:4.3f} train_loss {:4.5f}'.format(epoch, idx, len(dataloader),
                  total_acc/total_count, train_loss/total_count))
            
            total_acc, train_loss, total_count = 0,0,0
            start_time = time.time()
    
def evaluate(dataloader):
    model.eval()
    total_acc,train_loss, total_count = 0,0,0

    with torch.no_grad():
        for idx, (label,text,offsets) in enumerate(dataloader):
            predicted_label = model(text,offsets)

            loss = criterion(predicted_label, label)

            total_acc += (predicted_label.argmax(1) == label).sum().item()
            train_loss += loss.item()
            total_count += label.size(0)

    return total_acc/total_count, train_loss/total_count

训练模型

拆分数据集并运行模型

from torch.utils.data.dataset import random_split
from torchtext.data.functional import to_map_style_dataset

EPOCHS = 10
LR = 5
BATCH_SIZE = 64

criterion = torch.nn.CrossEntropyLoss()
optimizer = torch.optim.SGD(model.parameters(), lr=LR)
scheduler = torch.optim.lr_scheduler.StepLR(optimizer, 1.0, gamma=0.1)
total_accu = None

train_iter, test_iter = AG_NEWS()
train_dataset = to_map_style_dataset(train_iter)
test_dataset = to_map_style_dataset(test_iter)
num_train = int(len(train_dataset) * 0.95)

split_train_, split_valid_ = random_split(train_dataset, [num_train, len(train_dataset) - num_train])

train_dataloader = DataLoader(split_train_, batch_size=BATCH_SIZE, shuffle=True, collate_fn=collate_batch)
valid_dataloader = DataLoader(split_valid_, batch_size=BATCH_SIZE, shuffle=True, collate_fn=collate_batch)
test_dataloader = DataLoader(test_dataset, batch_size=BATCH_SIZE, shuffle=True, collate_fn=collate_batch)

for epoch in range(1,EPOCHS + 1):
    epoch_start_time = time.time()
    train(train_dataloader)
    val_acc, val_loss = evaluate(valid_dataloader)

    if total_accu is not None and total_accu > val_acc:
        scheduler.step()
    else:
        total_accu = val_acc
    print('-' * 69)
    print('|epoch {:1d} | time: {:4.2f}s |'
          'valid_acc {:4.3f} valid_loss {:4.3f}'.format(epoch, time.time() - epoch_start_time, val_acc, val_loss))

    print('-' * 69)

| epoch 1 |  500/1782 batches| train_acc 0.914 train_loss 0.00397
| epoch 1 | 1000/1782 batches| train_acc 0.917 train_loss 0.00385
| epoch 1 | 1500/1782 batches| train_acc 0.913 train_loss 0.00402
---------------------------------------------------------------------
|epoch 1 | time: 9.01s |valid_acc 0.920 valid_loss 0.004
---------------------------------------------------------------------
| epoch 2 |  500/1782 batches| train_acc 0.924 train_loss 0.00356
| epoch 2 | 1000/1782 batches| train_acc 0.925 train_loss 0.00346
| epoch 2 | 1500/1782 batches| train_acc 0.923 train_loss 0.00349
---------------------------------------------------------------------
|epoch 2 | time: 10.16s |valid_acc 0.913 valid_loss 0.004
---------------------------------------------------------------------
| epoch 3 |  500/1782 batches| train_acc 0.941 train_loss 0.00284
| epoch 3 | 1000/1782 batches| train_acc 0.945 train_loss 0.00271
| epoch 3 | 1500/1782 batches| train_acc 0.943 train_loss 0.00273
---------------------------------------------------------------------
|epoch 3 | time: 8.85s |valid_acc 0.924 valid_loss 0.004
---------------------------------------------------------------------
| epoch 4 |  500/1782 batches| train_acc 0.945 train_loss 0.00268
| epoch 4 | 1000/1782 batches| train_acc 0.945 train_loss 0.00267
| epoch 4 | 1500/1782 batches| train_acc 0.946 train_loss 0.00265
---------------------------------------------------------------------
|epoch 4 | time: 8.88s |valid_acc 0.925 valid_loss 0.004
---------------------------------------------------------------------
| epoch 5 |  500/1782 batches| train_acc 0.945 train_loss 0.00269
| epoch 5 | 1000/1782 batches| train_acc 0.948 train_loss 0.00257
| epoch 5 | 1500/1782 batches| train_acc 0.945 train_loss 0.00265
---------------------------------------------------------------------
|epoch 5 | time: 9.23s |valid_acc 0.922 valid_loss 0.004
---------------------------------------------------------------------
| epoch 6 |  500/1782 batches| train_acc 0.948 train_loss 0.00257
| epoch 6 | 1000/1782 batches| train_acc 0.950 train_loss 0.00249
| epoch 6 | 1500/1782 batches| train_acc 0.947 train_loss 0.00259
---------------------------------------------------------------------
|epoch 6 | time: 9.30s |valid_acc 0.925 valid_loss 0.004
---------------------------------------------------------------------
| epoch 7 |  500/1782 batches| train_acc 0.949 train_loss 0.00251
| epoch 7 | 1000/1782 batches| train_acc 0.946 train_loss 0.00264
| epoch 7 | 1500/1782 batches| train_acc 0.950 train_loss 0.00245
---------------------------------------------------------------------
|epoch 7 | time: 8.93s |valid_acc 0.925 valid_loss 0.004
---------------------------------------------------------------------
| epoch 8 |  500/1782 batches| train_acc 0.949 train_loss 0.00251
| epoch 8 | 1000/1782 batches| train_acc 0.946 train_loss 0.00260
| epoch 8 | 1500/1782 batches| train_acc 0.950 train_loss 0.00249
---------------------------------------------------------------------
|epoch 8 | time: 8.79s |valid_acc 0.925 valid_loss 0.004
---------------------------------------------------------------------
| epoch 9 |  500/1782 batches| train_acc 0.947 train_loss 0.00254
| epoch 9 | 1000/1782 batches| train_acc 0.950 train_loss 0.00250
| epoch 9 | 1500/1782 batches| train_acc 0.948 train_loss 0.00258
---------------------------------------------------------------------
|epoch 9 | time: 8.84s |valid_acc 0.925 valid_loss 0.004
---------------------------------------------------------------------
| epoch 10 |  500/1782 batches| train_acc 0.949 train_loss 0.00248
| epoch 10 | 1000/1782 batches| train_acc 0.947 train_loss 0.00256
| epoch 10 | 1500/1782 batches| train_acc 0.951 train_loss 0.00249
---------------------------------------------------------------------
|epoch 10 | time: 9.80s |valid_acc 0.925 valid_loss 0.004
---------------------------------------------------------------------

使用测试数据评估模型

print('Checking the results of test dataset.')
test_acc, test_loss = evaluate(test_dataloader)
print('test accuracy {:8.3f}'.format(test_acc))

Checking the results of test dataset.
test accuracy    0.909

总结

文本分类常见流程为
- 准备好原始文本：AG News是广泛用来进行文本分类的数据集
- 文本清洗：AG News属于已经清洗好的数据集
- 分词：torchtext库的get_tokenizer()是用于将文本数据分词的函数，它返回一个分词器函数，可以将一个字符串转换为一个单词的列表
- 文本向量化：其实就是上周的词嵌入过程，这里使用EmbeddingBag方式进行嵌入，将离散的单词映射为固定大小的连续向量。这些向量能较好地捕捉单词间的语义关系。
- 建模：我们定义的是TextClassificationMode模型，它首先对文本进行嵌入，然后对嵌入结果进行均值聚合

原文地址：https://blog.csdn.net/a536723241/article/details/143393096

免责声明：本站文章内容转载自网络资源，如本站内容侵犯了原著者的合法权益，可联系本站删除。更多内容请关注自学内容网（zxcms.com）！

上一篇：配置BGP与IGP交互和路由自动聚合示例
下一篇：VuePress文档初始化请求过多问题探讨

npm error code ETIMEDOUT 简单排查
今天突然没到一个仓库的ius问题。改完之后想发布npm包出现下面的场景npm login遇到错误时，通常是网络连接问题，可能与代理设置、网络环境或 NPM 配置有关。
阅读更多2024-11-30
Qt入门4——常用控件2
一篇文章教你使用Qt中常用的显示类控件和输入类控件！
阅读更多2024-11-30
网络爬虫的原理
爬虫的原理
阅读更多2024-11-30
@antv/x6 再vue中，自定义图形，画流程图、数据建模、er图等图形
X6 是基于 HTML 和 SVG 的图编辑引擎，提供低成本的定制能力和开箱即用的内置扩展，方便我们快速搭建 DAG 图、ER 图、流程图、血缘图等应用。
阅读更多2024-11-30
高效数据集成：旺店通与金蝶云自动化处理调拨出库单
在某些特殊场景下，标准转换逻辑可能无法满足业务需求。此时，可以利用自定义脚本功能，实现更复杂的数据处理逻辑。"label":"库存状态",通过以上步骤，我们可以
阅读更多2024-11-30
SprinBoot整合KafKa的使用(详解)
前言1.高吞吐量（High Throughput）Kafka 设计的一个核心特性是高吞吐量。它能够每秒处理百万级别的消息，适合需要高频次、低延迟消息传递的场景。即使在大规模分布式环境下，它也能保持很高
阅读更多2024-11-30
【一文读懂】大语言模型
介绍了目前最火的大语言模型，一文带你理解大语言模型相关基础知识。
阅读更多2024-11-30
Permute for Mac 媒体文件格式转换软件安装教程【音视频图像文件转换，简单操作，轻松转换，提高效率】
Permute for Mac 媒体文件格式转换软件安装教程【音视频图像文件转换，简单操作，轻松转换，提高效率】
阅读更多2024-11-30
Linux Docker 部署 Jenkins 详解教程
**Jenkins** 是当前最流行的持续集成和持续交付（CI/CD）工具之一，它能够帮助开发团队实现自动化构建、测试和部署。在现代 DevOps 流程中，Jenkins 是不可或缺的一部分。结合 D
阅读更多2024-11-30
上海迪士尼奇幻冬日巡游：IP营销如何出圈？
从领航花车上的达菲和朋友们，到《冰雪奇缘》主题花车上的艾莎、安娜和雪宝，再到《疯狂动物城》冰川镇上的尼克和朱迪，每一个细节都充满了迪士尼的经典IP元素，让游客仿佛置身于一个梦幻的童话世界。而其推出的“
阅读更多2024-11-30

第N5周：Pytorch文本分类入门

前期准备

加载数据

加载AG News数据集

构建词典

生成数据批次和迭代器

准备模型

定义模型

定义实例

定义训练函数与评估函数

训练模型

拆分数据集并运行模型

使用测试数据评估模型

总结

相关文章