【nlp入门实战】中文文本分类-Pytorch实现

Banana0840

已于 2024-08-16 10:07:48 修改

阅读量180

点赞数 9

文章标签：自然语言处理分类 pytorch

于 2024-08-15 19:27:35 首次发布

本文链接：https://blog.csdn.net/Banana0840/article/details/141230011

版权

🍨 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

在这里插入图片描述

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

warnings.filterwarnings('ignore')

device = torch.device('cpu')

import pandas as pd

train_data = pd.read_csv('./data/n5_train.csv', sep='\t', header=None)
train_data.head()


     0	1
0	还有双鸭山到淮阴的汽车票吗13号的	Travel-Query
1	从这里怎么回家	Travel-Query
2	随便播放一首专辑阁楼里的佛里的歌	Music-Play
3	给看一下墓王之王嘛	FilmTele-Play
4	我想看挑战两把s686打突变团竞的游戏视频	Video-Play

def coustom_data_iter(texts, labels):
    for x, y in zip(texts, labels):
        yield x, y


train_iter = coustom_data_iter(train_data[0].values[:], train_data[1].values[:])

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

# 中文分词
tokenizer = jieba.lcut


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>']) # 设置默认索引，如果找不到单词，则选择默认索引

Building prefix dict from the default dictionary …
Loading model from cache C:\Users\dlt_t\AppData\Local\Temp\jieba.cache
Loading model cost 1.055 seconds.
Prefix dict has been built successfully.

label_name = list(set(train_data[1].values[:]))
label_name

[‘Calendar-Query’,
‘Music-Play’,
‘FilmTele-Play’,
‘Weather-Query’,
‘Travel-Query’,
‘Video-Play’,
‘TVProgram-Play’,
‘Audio-Play’,
‘Other’,
‘Alarm-Update’,
‘HomeAppliance-Control’,
‘Radio-Listen’]

text_pipeline = lambda x: vocab(tokenizer(x))
label_pipeline = lambda x: label_name.index(x)
print(text_pipeline('随便播放一首歌'))
print(label_pipeline('Music-Play'))

[173, 4, 1499]
1

from torch.utils.data import DataLoader


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

    for (_text, _label) in batch:
        # 标签列表
        label_list.append(label_pipeline(_label))

        # 文本列表
        processed_text = torch.tensor(text_pipeline(_text))
        text_list.append(processed_text)

        # 偏移量，语句总词汇量
        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 text_list.to(device), label_list.to(device), offsets.to(device)


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

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,  # 词典大小
                                         embed_dim,  # 嵌入维度
                                         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)  # 初始化权重
        self.fc.weight.data.uniform_(-initrange, initrange)
        self.fc.bias.data.zero_()  # 偏置值归零

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

self.embedding.weight.data.uniform_(-initrange, initrange)这段代码是在 PyTorch 框架下用于初始化神经网络的词嵌入层（embedding layer）权重的一种方法。这里使用了均匀分布的随机值来初始化权重，具体来说，其作用如下：

self.embedding: 这是神经网络中的词嵌入层（embedding layer）。词嵌入层的作用是将离散的单词表示（通常为整数索引）映射为固定大小的连续向量。这些向量捕捉了单词之间的语义关系，并作为网络的输入。

self.embedding.weight: 这是词嵌入层的权重矩阵，它的形状为 (vocab_size, embedding_dim)，其中 vocab_size 是词汇表的大小，embedding_dim 是嵌入向量的维度。

self.embedding.weight.data: 这是权重矩阵的数据部分，我们可以在这里直接操作其底层的张量。

.uniform_(-initrange, initrange): 这是一个原地操作（in-place operation），用于将权重矩阵的值用一个均匀分布进行初始化。均匀分布的范围为 [-initrange, initrange]，其中 initrange 是一个正数。

通过这种方式初始化词嵌入层的权重，可以使得模型在训练开始时具有一定的随机性，有助于避免梯度消失或梯度爆炸等问题。在训练过程中，这些权重将通过优化算法不断更新，以捕捉到更好的单词表示。

num_class = len(label_name)
vocab_size = len(vocab)
em_size = 64
model = TextClassificationModel(vocab_size, em_size, num_class).to(device)

import time

def train(dataloader):
    model.train()  # 切换为训练模式
    total_acc, train_loss, total_count = 0, 0, 0
    log_interval = 50
    start_time   = time.time()

    for idx, (text,label,offsets) in enumerate(dataloader):
        
        predicted_label = model(text, offsets)
        
        optimizer.zero_grad()                    # grad属性归零
        loss = criterion(predicted_label, label) # 计算网络输出和真实值之间的差距，label为真实值
        loss.backward()                          # 反向传播
        torch.nn.utils.clip_grad_norm_(model.parameters(), 0.1) # 梯度裁剪
        optimizer.step()  # 每一步自动更新
        
        # 记录acc与loss
        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()  # 切换为测试模式
    total_acc, train_loss, total_count = 0, 0, 0

    with torch.no_grad():
        for idx, (text,label,offsets) in enumerate(dataloader):
            predicted_label = model(text, offsets)
            
            loss = criterion(predicted_label, label)  # 计算loss值
            # 记录测试数据
            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

from torch.utils.data.dataset import random_split
from torchtext.data.functional import to_map_style_dataset
# 超参数
EPOCHS     = 10 # epoch
LR         = 5  # 学习率
BATCH_SIZE = 64 # batch size for training

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

# 构建数据集
train_iter = coustom_data_iter(train_data[0].values[:], train_data[1].values[:])
train_dataset = to_map_style_dataset(train_iter)

split_train_, split_valid_ = random_split(train_dataset,
                                          [int(len(train_dataset)*0.8),int(len(train_dataset)*0.2)])

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)

for epoch in range(1, EPOCHS + 1):
    epoch_start_time = time.time()
    train(train_dataloader)
    val_acc, val_loss = evaluate(valid_dataloader)
    
    # 获取当前的学习率
    lr = optimizer.state_dict()['param_groups'][0]['lr']
    
    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} | lr {:4.6f}'.format(epoch,
                                           time.time() - epoch_start_time,
                                           val_acc,val_loss,lr))

    print('-' * 69)

| epoch 1 |   50/ 152 batches | train_acc 0.456 train_loss 0.03034
| epoch 1 |  100/ 152 batches | train_acc 0.722 train_loss 0.01876
| epoch 1 |  150/ 152 batches | train_acc 0.761 train_loss 0.01404
---------------------------------------------------------------------
| epoch 1 | time: 2.81s | valid_acc 0.803 valid_loss 0.012 | lr 5.000000
---------------------------------------------------------------------
| epoch 2 |   50/ 152 batches | train_acc 0.813 train_loss 0.01089
| epoch 2 |  100/ 152 batches | train_acc 0.838 train_loss 0.00895
| epoch 2 |  150/ 152 batches | train_acc 0.851 train_loss 0.00815
---------------------------------------------------------------------
| epoch 2 | time: 2.16s | valid_acc 0.848 valid_loss 0.008 | lr 5.000000
---------------------------------------------------------------------
| epoch 3 |   50/ 152 batches | train_acc 0.880 train_loss 0.00675
| epoch 3 |  100/ 152 batches | train_acc 0.877 train_loss 0.00674
| epoch 3 |  150/ 152 batches | train_acc 0.895 train_loss 0.00585
---------------------------------------------------------------------
| epoch 3 | time: 2.75s | valid_acc 0.875 valid_loss 0.007 | lr 5.000000
---------------------------------------------------------------------
| epoch 4 |   50/ 152 batches | train_acc 0.904 train_loss 0.00549
| epoch 4 |  100/ 152 batches | train_acc 0.917 train_loss 0.00480
| epoch 4 |  150/ 152 batches | train_acc 0.920 train_loss 0.00438
---------------------------------------------------------------------
| epoch 4 | time: 2.67s | valid_acc 0.890 valid_loss 0.006 | lr 5.000000
---------------------------------------------------------------------
| epoch 5 |   50/ 152 batches | train_acc 0.931 train_loss 0.00386
| epoch 5 |  100/ 152 batches | train_acc 0.932 train_loss 0.00389
| epoch 5 |  150/ 152 batches | train_acc 0.935 train_loss 0.00373
---------------------------------------------------------------------
| epoch 5 | time: 2.93s | valid_acc 0.896 valid_loss 0.006 | lr 5.000000
---------------------------------------------------------------------
| epoch 6 |   50/ 152 batches | train_acc 0.946 train_loss 0.00319
| epoch 6 |  100/ 152 batches | train_acc 0.953 train_loss 0.00290
| epoch 6 |  150/ 152 batches | train_acc 0.950 train_loss 0.00294
---------------------------------------------------------------------
| epoch 6 | time: 2.74s | valid_acc 0.902 valid_loss 0.005 | lr 5.000000
---------------------------------------------------------------------
| epoch 7 |   50/ 152 batches | train_acc 0.967 train_loss 0.00237
| epoch 7 |  100/ 152 batches | train_acc 0.962 train_loss 0.00240
| epoch 7 |  150/ 152 batches | train_acc 0.956 train_loss 0.00243
---------------------------------------------------------------------
| epoch 7 | time: 2.86s | valid_acc 0.905 valid_loss 0.005 | lr 5.000000
---------------------------------------------------------------------
| epoch 8 |   50/ 152 batches | train_acc 0.971 train_loss 0.00190
| epoch 8 |  100/ 152 batches | train_acc 0.972 train_loss 0.00190
| epoch 8 |  150/ 152 batches | train_acc 0.971 train_loss 0.00189
---------------------------------------------------------------------
| epoch 8 | time: 2.60s | valid_acc 0.902 valid_loss 0.005 | lr 5.000000
---------------------------------------------------------------------
| epoch 9 |   50/ 152 batches | train_acc 0.983 train_loss 0.00146
| epoch 9 |  100/ 152 batches | train_acc 0.983 train_loss 0.00137
| epoch 9 |  150/ 152 batches | train_acc 0.983 train_loss 0.00142
---------------------------------------------------------------------
| epoch 9 | time: 2.51s | valid_acc 0.907 valid_loss 0.005 | lr 0.500000
---------------------------------------------------------------------
| epoch 10 |   50/ 152 batches | train_acc 0.984 train_loss 0.00143
| epoch 10 |  100/ 152 batches | train_acc 0.987 train_loss 0.00130
| epoch 10 |  150/ 152 batches | train_acc 0.983 train_loss 0.00136
---------------------------------------------------------------------
| epoch 10 | time: 2.52s | valid_acc 0.909 valid_loss 0.005 | lr 0.500000
---------------------------------------------------------------------

test_acc, test_loss = evaluate(valid_dataloader)
print('模型准确率为：{:5.4f}'.format(test_acc))

模型准确率为：0.9087

def predict(text, text_pipeline):
    with torch.no_grad():
        text = torch.tensor(text_pipeline(text))
        output = model(text, torch.tensor([0]))
        return output.argmax(1).item()

# ex_text_str = "随便播放一首专辑阁楼里的佛里的歌"
ex_text_str = "还有双鸭山到淮阴的汽车票吗13号的"

model = model.to("cpu")

print("该文本的类别是：%s" %label_name[predict(ex_text_str, text_pipeline)])