>- **🍨 本文为[🔗365天深度学习训练营](https://mp.weixin.qq.com/s/0dvHCaOoFnW8SCp3JpzKxg) 中的学习记录博客**
>- **🍖 原作者：[K同学啊](https://mtyjkh.blog.csdn.net/)**
任务：加载txt文件，并用Embeddingbag&Embedding完成词嵌入
一、导入必要的包和数据
import pandas as pd
import torch
from torch import nn
import jieba as jb
data_dir = "/kaggle/input/text-data1/.txt"
data = pd.read_csv(data_dir, header=None)
data结果：
二、使用jieba分词：
tokenized_texts = [list(jb.cut(text)) for text in data[0]]
print(tokenized_texts)输出分词结果：
三、制作词汇表：
word_index = {}
index_word = {}
for i, word in enumerate(set([word for text in tokenized_texts for word in text])):
word_index[word] = i
index_word[i] = word
vocab_size = len(word_index)
print(f"{vocab_size} / {word_index}")词汇表大小：87
词汇表结果：
四、将文字转换为整数序列：
sequences1 = torch.tensor([word_index[word] for word in tokenized_texts[0]], dtype=torch.long)
sequences2 = torch.tensor([word_index[word] for word in tokenized_texts[1]], dtype=torch.long)五、用embadding将句子转换为词嵌入向量：
embedding = nn.Embedding(vocab_size, 4)
embedding_sequences1 = embedding(sequences1)
embedding_sequences2 = embedding(sequences2)
print(embedding_sequences1)
print(embedding_sequences2)embadding结果：
tensor([[-2.2421, 0.4768, 1.5981, -0.1372],
[ 0.2662, 0.6506, -0.3813, 0.3465],
[ 0.2144, 0.6271, 1.4468, 0.2856],
[ 0.1479, 0.4894, -0.3237, -2.0757],
[-0.5640, -1.1852, 0.3003, 1.1728],
[ 0.4459, -0.7934, -0.1811, 1.3251],
[-1.4543, -1.3130, 1.0980, 0.8700],
[ 0.2644, -0.8209, 2.3713, -1.3098],
[ 0.2144, 0.6271, 1.4468, 0.2856],
[ 1.9804, -0.0224, -1.0361, 0.1449],
[-0.8875, 1.1185, 0.5886, -1.3458],
[-0.0645, 1.1644, -0.7693, -0.4122],
[ 0.7435, -0.4162, 1.2579, -1.7981],
[-0.2672, 0.4423, -1.7335, -2.1201],
[ 1.4095, -0.1287, -0.2783, 0.1564],
[ 1.2758, 0.7741, -1.6845, 0.0374],
[ 0.7291, 2.4709, -1.0184, 0.9997],
[ 0.3255, 0.1719, -1.2049, -0.7580],
[-1.1748, 1.2276, -0.2742, 1.5316],
[ 0.2144, 0.6271, 1.4468, 0.2856],
[ 0.4989, 0.2639, -1.3275, 1.4525],
[-0.2672, 0.4423, -1.7335, -2.1201],
[-1.6770, 0.5937, -0.4984, -0.0159],
[-0.1202, 0.6479, 0.8591, -1.0248],
[ 0.1757, -1.5443, 0.4225, -0.1653],
[-0.4791, -0.4607, -1.1054, -1.2024],
[-0.0297, 0.5840, 0.5421, 0.1201],
[ 0.6148, -1.5387, 1.0232, 0.2355],
[-1.2716, -0.9989, 1.1541, -1.3458],
[ 0.7014, 0.5741, -0.2984, 0.7311],
[-0.1746, -1.4164, 0.0412, 0.3107],
[ 0.6628, -0.1708, 0.2175, -1.2665],
[ 0.3255, 0.1719, -1.2049, -0.7580],
[-1.1748, 1.2276, -0.2742, 1.5316],
[-0.0645, 1.1644, -0.7693, -0.4122],
[ 1.2490, -0.8628, 0.3805, -0.6328],
[-0.2672, 0.4423, -1.7335, -2.1201],
[-0.3621, -0.1040, 0.9854, 0.6555],
[ 0.1479, 0.4894, -0.3237, -2.0757],
[-0.9348, -0.9895, 1.6729, -0.0273],
[ 1.2758, 0.7741, -1.6845, 0.0374],
[ 0.4989, 0.2639, -1.3275, 1.4525],
[-0.2672, 0.4423, -1.7335, -2.1201],
[ 0.5002, 0.2031, 0.5255, 0.2732],
[-0.9373, -0.8043, -0.7819, 0.6784],
[ 0.1671, -0.9603, -0.9574, -1.0333],
[-0.4921, 1.0120, -1.5466, -0.4126],
[-0.0189, -0.6197, 0.2134, 0.6700],
[ 0.3798, 1.1016, 0.9914, -1.7548],
[ 1.6996, 1.1439, 0.3383, -1.5814],
[ 0.6642, -0.8293, -0.0363, -0.6464],
[-1.1748, 1.2276, -0.2742, 1.5316],
[-0.2030, 0.4043, -0.4269, 2.0898],
[-0.9348, -0.9895, 1.6729, -0.0273],
[ 0.6830, 1.4578, -1.4581, -1.2760],
[ 0.0985, 2.2091, 0.4579, 0.0645],
[-0.8245, 0.7995, 1.2056, -0.8458],
[ 0.2116, 0.3476, 0.3539, -0.6649],
[-1.1390, -1.7131, 0.4298, -0.5274],
[-0.2672, 0.4423, -1.7335, -2.1201],
[ 1.8577, 0.3560, 0.3905, 0.1583],
[-0.3970, 0.4146, 0.0542, 0.7691],
[-0.8477, -0.1442, -1.3092, 0.3221],
[-1.1390, -1.7131, 0.4298, -0.5274],
[ 0.1671, -0.9603, -0.9574, -1.0333],
[-0.5640, -1.1852, 0.3003, 1.1728],
[-0.4246, 0.8655, -0.5637, 0.8559],
[-0.9348, -0.9895, 1.6729, -0.0273],
[ 0.2144, 0.6271, 1.4468, 0.2856],
[-0.3495, -2.5871, -0.7908, -1.3156],
[-0.4246, 0.8655, -0.5637, 0.8559],
[ 1.7937, 1.3631, -0.0138, -0.6104],
[-2.0114, 1.8538, 0.2406, -0.0205],
[ 0.2144, 0.6271, 1.4468, 0.2856],
[-0.0645, 1.1644, -0.7693, -0.4122]], grad_fn=<EmbeddingBackward0>)
tensor([[-0.3867, -0.0058, -0.7481, 0.4434],
[-0.4892, 1.1163, 1.5090, 0.9548],
[-0.3621, -0.1040, 0.9854, 0.6555],
[ 0.4989, 0.2639, -1.3275, 1.4525],
[-0.2672, 0.4423, -1.7335, -2.1201],
[ 1.9659, 0.6739, -0.7447, 2.0967],
[-0.8151, 0.5109, -0.5044, 0.2174],
[ 1.7830, -0.6902, 0.3569, 0.8761],
[-1.3112, 0.5679, 0.2629, 0.6237],
[ 1.1008, 0.3024, 2.0047, 1.1738],
[ 0.9640, -0.2232, 1.7577, 0.1672],
[ 0.6152, 0.7374, 0.0465, -0.4756],
[-0.1858, 0.1699, -0.6322, -0.4958],
[ 0.1092, -1.2058, -1.2382, 1.7523],
[ 1.5649, -1.1317, 0.6957, 0.4157],
[ 0.9640, -0.2232, 1.7577, 0.1672],
[ 1.7874, 1.4637, -1.2369, 0.4733],
[-0.0645, 1.1644, -0.7693, -0.4122],
[ 1.7830, -0.6902, 0.3569, 0.8761],
[-1.3112, 0.5679, 0.2629, 0.6237],
[ 1.1008, 0.3024, 2.0047, 1.1738],
[ 0.9640, -0.2232, 1.7577, 0.1672],
[ 0.2144, 0.6271, 1.4468, 0.2856],
[-0.4225, 0.2887, 1.0175, 0.1776],
[-0.3821, 0.1820, 0.1650, -0.6111],
[-0.5640, -1.1852, 0.3003, 1.1728],
[ 0.4481, 0.4344, -0.2955, -0.9457],
[ 1.5302, 1.3884, 0.0516, -0.4676],
[-1.1748, 1.2276, -0.2742, 1.5316],
[ 0.8415, -0.1924, -0.7211, -0.2267],
[-2.1970, 0.7031, -0.1178, -0.4499],
[ 1.2758, 0.7741, -1.6845, 0.0374],
[ 1.5492, -1.2600, -0.5016, 0.2411],
[-0.2672, 0.4423, -1.7335, -2.1201],
[ 0.2468, 0.4953, 0.2391, -0.4277],
[-1.8871, 0.4976, 0.3504, 1.2424],
[ 1.2758, 0.7741, -1.6845, 0.0374],
[-0.6616, -1.0267, -0.0144, 1.3585],
[ 0.2144, 0.6271, 1.4468, 0.2856],
[-0.7786, 0.8371, 0.0190, 1.0264],
[-0.6832, -0.7319, -0.8668, 0.6007],
[ 0.1757, -1.5443, 0.4225, -0.1653],
[-0.2672, 0.4423, -1.7335, -2.1201],
[ 0.3108, 0.0142, 0.3639, -0.9848],
[-0.6616, -1.0267, -0.0144, 1.3585],
[ 0.2144, 0.6271, 1.4468, 0.2856],
[-0.7786, 0.8371, 0.0190, 1.0264],
[-0.6832, -0.7319, -0.8668, 0.6007],
[ 0.6202, -0.6972, 0.9012, 0.0262],
[-0.0645, 1.1644, -0.7693, -0.4122],
[ 0.6907, 0.2946, -0.8080, -1.1677],
[-0.2672, 0.4423, -1.7335, -2.1201],
[ 1.5302, 1.3884, 0.0516, -0.4676],
[-1.1748, 1.2276, -0.2742, 1.5316],
[-0.2030, 0.4043, -0.4269, 2.0898],
[ 0.5002, 0.2031, 0.5255, 0.2732],
[ 1.5477, -2.2033, -0.3226, 0.1326],
[-0.2892, -2.0367, 0.9881, -0.8522],
[ 0.2144, 0.6271, 1.4468, 0.2856],
[-0.2672, 0.4423, -1.7335, -2.1201],
[-0.4246, 0.8655, -0.5637, 0.8559],
[-0.9348, -0.9895, 1.6729, -0.0273],
[ 0.0985, 2.2091, 0.4579, 0.0645],
[-0.8245, 0.7995, 1.2056, -0.8458],
[ 0.2116, 0.3476, 0.3539, -0.6649],
[-1.1390, -1.7131, 0.4298, -0.5274],
[-0.0645, 1.1644, -0.7693, -0.4122],
[ 0.7435, -0.4162, 1.2579, -1.7981],
[-0.2672, 0.4423, -1.7335, -2.1201],
[ 1.4095, -0.1287, -0.2783, 0.1564],
[ 0.3255, 0.1719, -1.2049, -0.7580],
[-1.1748, 1.2276, -0.2742, 1.5316],
[ 0.2144, 0.6271, 1.4468, 0.2856],
[ 0.0812, 1.0305, 1.2780, 0.4558],
[-0.2672, 0.4423, -1.7335, -2.1201],
[-1.6770, 0.5937, -0.4984, -0.0159],
[-0.0591, -1.0665, -0.5901, 0.2102],
[ 1.1762, 0.1607, 1.5622, 0.0729],
[ 0.2144, 0.6271, 1.4468, 0.2856],
[ 1.7830, -0.6902, 0.3569, 0.8761],
[-1.3112, 0.5679, 0.2629, 0.6237],
[ 1.1008, 0.3024, 2.0047, 1.1738],
[ 0.9640, -0.2232, 1.7577, 0.1672],
[ 0.9006, 0.0673, 0.6713, -1.2377]], grad_fn=<EmbeddingBackward0>)
六、embedding_bag的均值方法将句子转换成词嵌入向量：
embedding_bag = nn.EmbeddingBag(vocab_size, 30, mode = "mean")
input_sequences = torch.cat([sequences1,
sequences2])
offsets = torch.tensor([0, len(sequences1)], dtype=torch.long)
eb_seq = embedding_bag(input_sequences, offsets)
print(eb_seq)embadding bag的结果：
七、总结：
通过这次练习加上上一次的实践，除整数序列以外又学习了词嵌入操作，理解了文字转换数值的一些方法，希望能在接下来的任务中能够熟练运用这些方法并加深理解吧。
扫一扫
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
