PyTorch翻译官网教程-NLP FROM SCRATCH: CLASSIFYING NAMES WITH A CHARACTER-LEVEL RNN

官网链接

NLP From Scratch: Classifying Names with a Character-Level RNN — PyTorch Tutorials 2.0.1+cu117 documentation

使用CHARACTER-LEVEL RNN 对名字分类

我们将建立和训练一个基本的字符级递归神经网络(RNN)来分类单词。本教程以及另外两个“from scratch”的自然语言处理(NLP)教程 NLP From Scratch: Generating Names with a Character-Level RNN 和 NLP From Scratch: Translation with a Sequence to Sequence Network and Attention,演示如何预处理数据以建立NLP模型。特别是，这些教程没有使用torchtext的许多便利功能，因此您可以看到如何简单使用预处理模型NLP。

字符级RNN将单词作为一系列字符来读取 ，每一步输出一个预测和“隐藏状态”，将之前的隐藏状态输入到下一步。我们把最后的预测作为输出，即这个词属于哪个类。

具体来说，我们将训练来自18种语言的几千个姓氏，并根据拼写来预测一个名字来自哪种语言:

$ python predict.py Hinton
(-0.47) Scottish
(-1.52) English
(-3.57) Irish

$ python predict.py Schmidhuber
(-0.19) German
(-2.48) Czech
(-2.68) Dutch

建议准备

在开始本教程之前，建议您安装PyTorch，并对Python编程语言和张量有基本的了解:

• PyTorch 有关安装说明
• Deep Learning with PyTorch: A 60 Minute Blitz 开始使用PyTorch并学习张量的基础知识
• Learning PyTorch with Examples 使用概述
• PyTorch for Former Torch Users 如果您是前Lua Torch用户

了解rnn及其工作原理也很有用:

• The Unreasonable Effectiveness of Recurrent Neural Networks 展示了一些现实生活中的例子
• Understanding LSTM Networks 是专门关于LSTMs的，但也有关于RNNs的信息

准备数据

从这里下载数据并将其解压缩到当前目录。here

“data/names”目录下包含18个文本文件，文件名为“[Language].txt”。每个文件包含一堆名称，每行一个名称，大多数是罗马化的(但我们仍然需要从Unicode转换为ASCII)。

我们最终会得到一个包含每种语言名称列表的字典，{language: [names ...]}。通用变量“category”和“line”(在本例中表示语言和名称)用于以后的可扩展性。

from io import open
import glob
import os

def findFiles(path): return glob.glob(path)

print(findFiles('data/names/*.txt'))

import unicodedata
import string

all_letters = string.ascii_letters + " .,;'"
n_letters = len(all_letters)

# Turn a Unicode string to plain ASCII, thanks to https://stackoverflow.com/a/518232/2809427
def unicodeToAscii(s):
    return ''.join(
        c for c in unicodedata.normalize('NFD', s)
        if unicodedata.category(c) != 'Mn'
        and c in all_letters
    )

print(unicodeToAscii('Ślusàrski'))

# Build the category_lines dictionary, a list of names per language
category_lines = {}
all_categories = []

# Read a file and split into lines
def readLines(filename):
    lines = open(filename, encoding='utf-8').read().strip().split('\n')
    return [unicodeToAscii(line) for line in lines]

for filename in findFiles('data/names/*.txt'):
    category = os.path.splitext(os.path.basename(filename))[0]
    all_categories.append(category)
    lines = readLines(filename)
    category_lines[category] = lines

n_categories = len(all_categories)

输出

['data/names/Arabic.txt', 'data/names/Chinese.txt', 'data/names/Czech.txt', 'data/names/Dutch.txt', 'data/names/English.txt', 'data/names/French.txt', 'data/names/German.txt', 'data/names/Greek.txt', 'data/names/Irish.txt', 'data/names/Italian.txt', 'data/names/Japanese.txt', 'data/names/Korean.txt', 'data/names/Polish.txt', 'data/names/Portuguese.txt', 'data/names/Russian.txt', 'data/names/Scottish.txt', 'data/names/Spanish.txt', 'data/names/Vietnamese.txt']
Slusarski

现在我们有了category_lines，这是一个将每个类别(语言)映射到行(名称)列表的字典。我们还记录了all_categories(只是一个语言列表)和n_categories，以供以后参考。

print(category_lines['Italian'][:5])

输出

['Abandonato', 'Abatangelo', 'Abatantuono', 'Abate', 'Abategiovanni']

把名字变成张量

现在我们已经组织好了所有的名字，我们需要把它们变成张量来使用它们。

为了表示单个字母，我们使用大小为<1 x n_letters> 的 “one-hot vector”。一个独热向量被0填充，除了当前字母所以处是1。例如："b" = <0 1 0 0 0 ...>.

为了组成一个单词，我们将一堆这样的单词连接到一个二维矩阵中<line_length x 1 x n_letters>.

额外的1维度是因为PyTorch假设所有的东西都是分批的——我们在这里只是使用1的批大小。

import torch

# Find letter index from all_letters, e.g. "a" = 0
def letterToIndex(letter):
    return all_letters.find(letter)

# Just for demonstration, turn a letter into a <1 x n_letters> Tensor
def letterToTensor(letter):
    tensor = torch.zeros(1, n_letters)
    tensor[0][letterToIndex(letter)] = 1
    return tensor

# Turn a line into a <line_length x 1 x n_letters>,
# or an array of one-hot letter vectors
def lineToTensor(line):
    tensor = torch.zeros(len(line), 1, n_letters)
    for li, letter in enumerate(line):
        tensor[li][0][letterToIndex(letter)] = 1
    return tensor

print(letterToTensor('J'))

print(lineToTensor('Jones').size())

输出

tensor([[0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
         0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1.,
         0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
         0., 0., 0.]])
torch.Size([5, 1, 57])

创建网络

在autograd之前，在Torch中创建循环神经网络涉及到在几个时间步上克隆一层的参数。图层包含隐藏状态和梯度，现在完全由图形本身处理。这意味着你可以以一种非常“纯粹”的方式实现RNN，作为常规的前馈层。

这个RNN模块(主要是从the PyTorch for Torch users tutorial复制的)只有2个线性层，在输入和隐藏状态上操作，在输出之后有一个LogSoftmax层。

import torch.nn as nn

class RNN(nn.Module):
    def __init__(self, input_size, hidden_size, output_size):
        super(RNN, self).__init__()

        self.hidden_size = hidden_size

        self.i2h = nn.Linear(input_size + hidden_size, hidden_size)
        self.h2o = nn.Linear(hidden_size, output_size)
        self.softmax = nn.LogSoftmax(dim=1)

    def forward(self, input, hidden):
        combined = torch.cat((input, hidden), 1)
        hidden = self.i2h(combined)
        output = self.h2o(hidden)
        output = self.softmax(output)
        return output, hidden

    def initHidden(self):
        return torch.zeros(1, self.hidden_size)

n_hidden = 128
rnn = RNN(n_letters, n_hidden, n_categories)

为了运行这个网络的一个步骤，我们需要传递一个输入(在我们的例子中，是当前字母的张量)和一个先前的隐藏状态(我们一开始将其初始化为零)。我们将返回输出(每种语言的概率)和下一个隐藏状态(我们将其保留到下一步)。

input = letterToTensor('A')
hidden = torch.zeros(1, n_hidden)

output, next_hidden = rnn(input, hidden)

为了提高效率，我们不想为每一步都创建一个新的张量，所以我们将使用lineToTensor而不是letterToTensor并使用切片。这可以通过预计算张量批次来进一步优化。

input = lineToTensor('Albert')
hidden = torch.zeros(1, n_hidden)

output, next_hidden = rnn(input[0], hidden)
print(output)

输出

tensor([[-2.9083, -2.9270, -2.9167, -2.9590, -2.9108, -2.8332, -2.8906, -2.8325,
         -2.8521, -2.9279, -2.8452, -2.8754, -2.8565, -2.9733, -2.9201, -2.8233,
         -2.9298, -2.8624]], grad_fn=<LogSoftmaxBackward0>)

正如您所看到的，输出是一个<1 x n_categories> 张量，其中每个项目是该类别的可能性(越高越有可能)。

训练

训练准备

在开始训练之前，我们应该编写一些辅助函数。首先是解释网络的输出，我们知道这是每个类别的可能性。我们可以用Tensor.topk得到最大值的索引:

def categoryFromOutput(output):
    top_n, top_i = output.topk(1)
    category_i = top_i[0].item()
    return all_categories[category_i], category_i

print(categoryFromOutput(output))

输出

('Scottish', 15)

我们还需要一种快速获取训练示例(名称及其语言)的方法:

import random

def randomChoice(l):
    return l[random.randint(0, len(l) - 1)]

def randomTrainingExample():
    category = randomChoice(all_categories)
    line = randomChoice(category_lines[category])
    category_tensor = torch.tensor([all_categories.index(category)], dtype=torch.long)
    line_tensor = lineToTensor(line)
    return category, line, category_tensor, line_tensor

for i in range(10):
    category, line, category_tensor, line_tensor = randomTrainingExample()
    print('category =', category, '/ line =', line)

输出

category = Chinese / line = Hou
category = Scottish / line = Mckay
category = Arabic / line = Cham
category = Russian / line = V'Yurkov
category = Irish / line = O'Keeffe
category = French / line = Belrose
category = Spanish / line = Silva
category = Japanese / line = Fuchida
category = Greek / line = Tsahalis
category = Korean / line = Chang

训练网络

现在训练这个网络所需要做的就是给它看一堆例子，让它猜测，然后告诉它是否错了。

对于损失函数nn.NLLLoss是合适的，因为RNN的最后一层是nn.LogSoftmax.。

criterion = nn.NLLLoss()

每个训练循环将:

• 创建输入张量和目标张量
• 创建一个零初始隐藏状态
• 读取每个字母

• • 为下一个字母保存隐藏状态

• 将最终输出与目标进行比较
• 反向传播
• 返回输出和损失

learning_rate = 0.005 # If you set this too high, it might explode. If too low, it might not learn

def train(category_tensor, line_tensor):
    hidden = rnn.initHidden()

    rnn.zero_grad()

    for i in range(line_tensor.size()[0]):
        output, hidden = rnn(line_tensor[i], hidden)

    loss = criterion(output, category_tensor)
    loss.backward()

    # Add parameters' gradients to their values, multiplied by learning rate
    for p in rnn.parameters():
        p.data.add_(p.grad.data, alpha=-learning_rate)

    return output, loss.item()

现在我们只需要用一堆例子来运行它。由于train函数返回输出和损失，我们可以打印它的猜测并跟踪损失以便绘制。由于有1000个示例，我们只打印每个print_every示例，并取损失的平均值。

import time
import math

n_iters = 100000
print_every = 5000
plot_every = 1000



# Keep track of losses for plotting
current_loss = 0
all_losses = []

def timeSince(since):
    now = time.time()
    s = now - since
    m = math.floor(s / 60)
    s -= m * 60
    return '%dm %ds' % (m, s)

start = time.time()

for iter in range(1, n_iters + 1):
    category, line, category_tensor, line_tensor = randomTrainingExample()
    output, loss = train(category_tensor, line_tensor)
    current_loss += loss

    # Print ``iter`` number, loss, name and guess
    if iter % print_every == 0:
        guess, guess_i = categoryFromOutput(output)
        correct = '✓' if guess == category else '✗ (%s)' % category
        print('%d %d%% (%s) %.4f %s / %s %s' % (iter, iter / n_iters * 100, timeSince(start), loss, line, guess, correct))

    # Add current loss avg to list of losses
    if iter % plot_every == 0:
        all_losses.append(current_loss / plot_every)
        current_loss = 0

输出

5000 5% (0m 33s) 2.6379 Horigome / Japanese ✓
10000 10% (1m 5s) 2.0172 Miazga / Japanese ✗ (Polish)
15000 15% (1m 39s) 0.2680 Yukhvidov / Russian ✓
20000 20% (2m 12s) 1.8239 Mclaughlin / Irish ✗ (Scottish)
25000 25% (2m 45s) 0.6978 Banh / Vietnamese ✓
30000 30% (3m 18s) 1.7433 Machado / Japanese ✗ (Portuguese)
35000 35% (3m 51s) 0.0340 Fotopoulos / Greek ✓
40000 40% (4m 23s) 1.4637 Quirke / Irish ✓
45000 45% (4m 57s) 1.9018 Reier / French ✗ (German)
50000 50% (5m 30s) 0.9174 Hou / Chinese ✓
55000 55% (6m 2s) 1.0506 Duan / Vietnamese ✗ (Chinese)
60000 60% (6m 35s) 0.9617 Giang / Vietnamese ✓
65000 65% (7m 9s) 2.4557 Cober / German ✗ (Czech)
70000 70% (7m 42s) 0.8502 Mateus / Portuguese ✓
75000 75% (8m 14s) 0.2750 Hamilton / Scottish ✓
80000 80% (8m 47s) 0.7515 Maessen / Dutch ✓
85000 85% (9m 20s) 0.0912 Gan / Chinese ✓
90000 90% (9m 53s) 0.1190 Bellomi / Italian ✓
95000 95% (10m 26s) 0.0137 Vozgov / Russian ✓
100000 100% (10m 59s) 0.7808 Tong / Vietnamese ✓

绘制结果

绘制all_losses的历史损失图显示了网络的学习情况:

import matplotlib.pyplot as plt
import matplotlib.ticker as ticker

plt.figure()
plt.plot(all_losses)

输出

[<matplotlib.lines.Line2D object at 0x7f16606095a0>]

评估结果

为了了解网络在不同类别上的表现如何，我们将创建一个混淆矩阵，表示网络猜测(列)的每种语言(行)。为了计算混淆矩阵，使用evaluate(),在网络中运行一堆样本，这与 train() 去掉反向传播相同。

# Keep track of correct guesses in a confusion matrix
confusion = torch.zeros(n_categories, n_categories)
n_confusion = 10000

# Just return an output given a line
def evaluate(line_tensor):
    hidden = rnn.initHidden()

    for i in range(line_tensor.size()[0]):
        output, hidden = rnn(line_tensor[i], hidden)

    return output

# Go through a bunch of examples and record which are correctly guessed
for i in range(n_confusion):
    category, line, category_tensor, line_tensor = randomTrainingExample()
    output = evaluate(line_tensor)
    guess, guess_i = categoryFromOutput(output)
    category_i = all_categories.index(category)
    confusion[category_i][guess_i] += 1

# Normalize by dividing every row by its sum
for i in range(n_categories):
    confusion[i] = confusion[i] / confusion[i].sum()

# Set up plot
fig = plt.figure()
ax = fig.add_subplot(111)
cax = ax.matshow(confusion.numpy())
fig.colorbar(cax)

# Set up axes
ax.set_xticklabels([''] + all_categories, rotation=90)
ax.set_yticklabels([''] + all_categories)

# Force label at every tick
ax.xaxis.set_major_locator(ticker.MultipleLocator(1))
ax.yaxis.set_major_locator(ticker.MultipleLocator(1))

# sphinx_gallery_thumbnail_number = 2
plt.show()

输出

/var/lib/jenkins/workspace/intermediate_source/char_rnn_classification_tutorial.py:445: UserWarning:

FixedFormatter should only be used together with FixedLocator

/var/lib/jenkins/workspace/intermediate_source/char_rnn_classification_tutorial.py:446: UserWarning:

FixedFormatter should only be used together with FixedLocator

你可以从主轴上挑出亮点，显示它猜错了哪些语言，例如中文猜错了韩语，西班牙语猜错了意大利语。它似乎在希腊语上表现得很好，而在英语上表现得很差(可能是因为与其他语言重叠)。

运行用户输入

def predict(input_line, n_predictions=3):
    print('\n> %s' % input_line)
    with torch.no_grad():
        output = evaluate(lineToTensor(input_line))

        # Get top N categories
        topv, topi = output.topk(n_predictions, 1, True)
        predictions = []

        for i in range(n_predictions):
            value = topv[0][i].item()
            category_index = topi[0][i].item()
            print('(%.2f) %s' % (value, all_categories[category_index]))
            predictions.append([value, all_categories[category_index]])

predict('Dovesky')
predict('Jackson')
predict('Satoshi')

输出

> Dovesky
(-0.57) Czech
(-0.97) Russian
(-3.43) English

> Jackson
(-1.02) Scottish
(-1.49) Russian
(-1.96) English

> Satoshi
(-0.42) Japanese
(-1.70) Polish
(-2.74) Italian

in the Practical PyTorch repo中脚本的最终版本将上述代码拆分为几个文件:

• data.py (加载文件)
• model.py (定义 RNN)
• train.py (执行训练)
• predict.py (运行带有命令行参数的predict() )
• server.py (使用bottle.py作为JSON API提供预测)

运行train.py来训练和保存网络。

运行predict.py并输入一个名称来查看预测:

$ python predict.py Hazaki
(-0.42) Japanese
(-1.39) Polish
(-3.51) Czech

运行server.py 并访问http://localhost:5533/Yourname以获得预测的JSON输出。

练习

尝试使用不同的数据集 -> 类别，例如:

• 任何单词->语言
• 名字->性别
• 角色名称->作家
• 页面标题 -> 博客或社交新闻网站子版块

使用一个更大的和/或更好的形状网络，可以获得更好的结果

• 添加更多线性图层
• 试试 nn.LSTM 和 nn.GRU 网络层
• 将这些RNNs组合成一个更高级的网络