个人主页：道友老李
欢迎加入社区：道友老李的学习社区

介绍

**自然语言处理（Natural Language Processing，NLP）**是计算机科学领域与人工智能领域中的一个重要方向。它研究的是人类（自然）语言与计算机之间的交互。NLP的目标是让计算机能够理解、解析、生成人类语言，并且能够以有意义的方式回应和操作这些信息。

NLP的任务可以分为多个层次，包括但不限于：

1. 词法分析：将文本分解成单词或标记（token），并识别它们的词性（如名词、动词等）。
2. 句法分析：分析句子结构，理解句子中词语的关系，比如主语、谓语、宾语等。
3. 语义分析：试图理解句子的实际含义，超越字面意义，捕捉隐含的信息。
4. 语用分析：考虑上下文和对话背景，理解话语在特定情境下的使用目的。
5. 情感分析：检测文本中表达的情感倾向，例如正面、负面或中立。
6. 机器翻译：将一种自然语言转换为另一种自然语言。
7. 问答系统：构建可以回答用户问题的系统。
8. 文本摘要：从大量文本中提取关键信息，生成简短的摘要。
9. 命名实体识别（NER）：识别文本中提到的特定实体，如人名、地名、组织名等。
10. 语音识别：将人类的语音转换为计算机可读的文字格式。

NLP技术的发展依赖于算法的进步、计算能力的提升以及大规模标注数据集的可用性。近年来，深度学习方法，特别是基于神经网络的语言模型，如BERT、GPT系列等，在许多NLP任务上取得了显著的成功。随着技术的进步，NLP正在被应用到越来越多的领域，包括客户服务、智能搜索、内容推荐、医疗健康等。

BERT 训练之数据集处理

BERT 原理及模型代码实现

【自然语言处理（NLP）】基于Transformer架构的预训练语言模型：BERT 原理及代码实现

数据集处理

导包

import os
import random
import torch
import dltools

加载数据

def _read_wiki(data_dir):file_name = os.path.join(data_dir, 'wiki.train.tokens')with open(file_name, 'r',encoding="utf-8") as f:lines = f.readlines()# 大写字母转换为小写字母paragraphs = [line.strip().lower().split(' . ') for line in lines if len(line.split(' . ')) >= 2]random.shuffle(paragraphs)return paragraphs_read_wiki('./wikitext-2')

生成下一句预测任务的数据

def _get_next_sentence(sentence, next_sentence, paragraphs):if random.random() < 0.5:is_next = Trueelse:# paragraphs是三重列表的嵌套next_sentence = random.choice(random.choice(paragraphs))is_next = Falsereturn sentence, next_sentence, is_next

从段落中获取nsp数据

def _get_nsp_data_from_paragraph(paragraph, paragraphs, vocab, max_len):nsp_data_from_paragraph = []for i in range(len(paragraph) - 1):tokens_a, tokens_b, is_next = _get_next_sentence(paragraph[i], paragraph[i + 1], paragraphs)# 考虑1个'<cls>'词元和2个'<sep>'词元if len(tokens_a) + len(tokens_b) + 3 > max_len:continuetokens, segments = dltools.get_tokens_and_segments(tokens_a, tokens_b)nsp_data_from_paragraph.append((tokens, segments, is_next))return nsp_data_from_paragraph

生成遮蔽语言模型任务的数据

1. 为遮蔽语言模型的输入创建新的词元副本，其中输入可能包含替换的mask或随机词元
2. 打乱后用于在遮蔽语言模型任务中获取15%的随机词元进行预测
3. 80%的时间：将词替换为mask词元
4. 10%的时间：保持词不变
5. 10%的时间：用随机词替换该词

def _replace_mlm_tokens(tokens, candidate_pred_positions, num_mlm_preds,vocab):# 为遮蔽语言模型的输入创建新的词元副本，其中输入可能包含替换的“<mask>”或随机词元mlm_input_tokens = [token for token in tokens]pred_positions_and_labels = []# 打乱后用于在遮蔽语言模型任务中获取15%的随机词元进行预测random.shuffle(candidate_pred_positions)for mlm_pred_position in candidate_pred_positions:if len(pred_positions_and_labels) >= num_mlm_preds:breakmasked_token = None# 80%的时间：将词替换为“<mask>”词元if random.random() < 0.8:masked_token = '<mask>'else:# 10%的时间：保持词不变if random.random() < 0.5:masked_token = tokens[mlm_pred_position]# 10%的时间：用随机词替换该词else:masked_token = random.choice(vocab.idx_to_token)mlm_input_tokens[mlm_pred_position] = masked_tokenpred_positions_and_labels.append((mlm_pred_position, tokens[mlm_pred_position]))return mlm_input_tokens, pred_positions_and_labels

从token中获取mlm数据

在遮蔽语言模型任务中不会预测特殊词元

def _get_mlm_data_from_tokens(tokens, vocab):candidate_pred_positions = []# tokens是一个字符串列表for i, token in enumerate(tokens):# 在遮蔽语言模型任务中不会预测特殊词元if token in ['<cls>', '<sep>']:continuecandidate_pred_positions.append(i)# 遮蔽语言模型任务中预测15%的随机词元num_mlm_preds = max(1, round(len(tokens) * 0.15))mlm_input_tokens, pred_positions_and_labels = _replace_mlm_tokens(tokens, candidate_pred_positions, num_mlm_preds, vocab)pred_positions_and_labels = sorted(pred_positions_and_labels,key=lambda x: x[0])pred_positions = [v[0] for v in pred_positions_and_labels]mlm_pred_labels = [v[1] for v in pred_positions_and_labels]return vocab[mlm_input_tokens], pred_positions, vocab[mlm_pred_labels]

将文本转换为预训练数据集

1. valid_lens不包括’'的计数
2. 填充词元的预测将通过乘以0权重在损失中过滤掉

def _pad_bert_inputs(examples, max_len, vocab):max_num_mlm_preds = round(max_len * 0.15)all_token_ids, all_segments, valid_lens,  = [], [], []all_pred_positions, all_mlm_weights, all_mlm_labels = [], [], []nsp_labels = []for (token_ids, pred_positions, mlm_pred_label_ids, segments,is_next) in examples:all_token_ids.append(torch.tensor(token_ids + [vocab['<pad>']] * (max_len - len(token_ids)), dtype=torch.long))all_segments.append(torch.tensor(segments + [0] * (max_len - len(segments)), dtype=torch.long))# valid_lens不包括'<pad>'的计数valid_lens.append(torch.tensor(len(token_ids), dtype=torch.float32))all_pred_positions.append(torch.tensor(pred_positions + [0] * (max_num_mlm_preds - len(pred_positions)), dtype=torch.long))# 填充词元的预测将通过乘以0权重在损失中过滤掉all_mlm_weights.append(torch.tensor([1.0] * len(mlm_pred_label_ids) + [0.0] * (max_num_mlm_preds - len(pred_positions)),dtype=torch.float32))all_mlm_labels.append(torch.tensor(mlm_pred_label_ids + [0] * (max_num_mlm_preds - len(mlm_pred_label_ids)), dtype=torch.long))nsp_labels.append(torch.tensor(is_next, dtype=torch.long))return (all_token_ids, all_segments, valid_lens, all_pred_positions,all_mlm_weights, all_mlm_labels, nsp_labels)

创建Dataset

1. 输入paragraphs[i]是代表段落的句子字符串列表
2. 而输出paragraphs[i]是代表段落的句子列表，其中每个句子都是词元列表
3. 获取下一句子预测任务的数据
4. 获取遮蔽语言模型任务的数据
5. 填充输入

class _WikiTextDataset(torch.utils.data.Dataset):def __init__(self, paragraphs, max_len):# 输入paragraphs[i]是代表段落的句子字符串列表；# 而输出paragraphs[i]是代表段落的句子列表，其中每个句子都是词元列表paragraphs = [dltools.tokenize(paragraph, token='word') for paragraph in paragraphs]sentences = [sentence for paragraph in paragraphsfor sentence in paragraph]self.vocab = dltools.Vocab(sentences, min_freq=5, reserved_tokens=['<pad>', '<mask>', '<cls>', '<sep>'])# 获取下一句子预测任务的数据examples = []for paragraph in paragraphs:examples.extend(_get_nsp_data_from_paragraph(paragraph, paragraphs, self.vocab, max_len))# 获取遮蔽语言模型任务的数据examples = [(_get_mlm_data_from_tokens(tokens, self.vocab)+ (segments, is_next))for tokens, segments, is_next in examples]# 填充输入(self.all_token_ids, self.all_segments, self.valid_lens,self.all_pred_positions, self.all_mlm_weights,self.all_mlm_labels, self.nsp_labels) = _pad_bert_inputs(examples, max_len, self.vocab)def __getitem__(self, idx):return (self.all_token_ids[idx], self.all_segments[idx],self.valid_lens[idx], self.all_pred_positions[idx],self.all_mlm_weights[idx], self.all_mlm_labels[idx],self.nsp_labels[idx])def __len__(self):return len(self.all_token_ids)

加载WikiText-2数据集

def load_data_wiki(batch_size, max_len):"""加载WikiText-2数据集"""num_workers = dltools.get_dataloader_workers()data_dir = "./wikitext-2/"paragraphs = _read_wiki(data_dir)train_set = _WikiTextDataset(paragraphs, max_len)train_iter = torch.utils.data.DataLoader(train_set, batch_size,shuffle=True, num_workers=num_workers)return train_iter, train_set.vocabbatch_size, max_len = 512, 64
train_iter, vocab = load_data_wiki(batch_size, max_len)for (tokens_X, segments_X, valid_lens_x, pred_positions_X, mlm_weights_X,mlm_Y, nsp_y) in train_iter:print(tokens_X.shape, segments_X.shape, valid_lens_x.shape,pred_positions_X.shape, mlm_weights_X.shape, mlm_Y.shape,nsp_y.shape)break

torch.Size([512, 64]) torch.Size([512, 64]) torch.Size([512]) torch.Size([512, 10]) torch.Size([512, 10]) torch.Size([512, 10]) torch.Size([512])

len(vocab)

20256

BERT 训练代码实现

导包

import torch
from torch import nn
import dltools

加载数据

dltools中加载本地wiki文件，请自行修改路径 ./data/wikitext-2

batch_size, max_len = 1, 64
# dltools中加载本地wiki文件，请自行修改路径 ./data/wikitext-2
train_iter, vocab = dltools.load_data_wiki(batch_size, max_len)# tokens, segments, valid_lens, pred_positions, mlm_weights,mlm, nsp
for i in train_iter:break
i

构建BERT模型

net = dltools.BERTModel(len(vocab), num_hiddens=128, norm_shape=[128],ffn_num_input=128, ffn_num_hiddens=256, num_heads=2,num_layers=2, dropout=0.2, key_size=128, query_size=128,value_size=128, hid_in_features=128, mlm_in_features=128,nsp_in_features=128)
devices = dltools.try_all_gpus()

模型损失

1. 前向传播
2. 计算遮蔽语言模型损失
3. 计算下一句子预测任务的损失

loss = nn.CrossEntropyLoss()def _get_batch_loss_bert(net, loss, vocab_size, tokens_X,segments_X, valid_lens_x,pred_positions_X, mlm_weights_X,mlm_Y, nsp_y):# 前向传播_, mlm_Y_hat, nsp_Y_hat = net(tokens_X, segments_X,valid_lens_x.reshape(-1),pred_positions_X)# 计算遮蔽语言模型损失mlm_l = loss(mlm_Y_hat.reshape(-1, vocab_size), mlm_Y.reshape(-1)) * mlm_weights_X.reshape(-1, 1)mlm_l = mlm_l.sum() / (mlm_weights_X.sum() + 1e-8)# 计算下一句子预测任务的损失nsp_l = loss(nsp_Y_hat, nsp_y)l = mlm_l + nsp_lreturn mlm_l, nsp_l, l

训练

遮蔽语言模型损失的和，下一句预测任务损失的和，句子对的数量，计数

def train_bert(train_iter, net, loss, vocab_size, devices, num_steps):net = nn.DataParallel(net, device_ids=devices).to(devices[0])trainer = torch.optim.Adam(net.parameters(), lr=0.01)step, timer = 0, dltools.Timer()animator = dltools.Animator(xlabel='step', ylabel='loss',xlim=[1, num_steps], legend=['mlm', 'nsp'])# 遮蔽语言模型损失的和，下一句预测任务损失的和，句子对的数量，计数metric = dltools.Accumulator(4)num_steps_reached = Falsewhile step < num_steps and not num_steps_reached:for tokens_X, segments_X, valid_lens_x, pred_positions_X,mlm_weights_X, mlm_Y, nsp_y in train_iter:tokens_X = tokens_X.to(devices[0])segments_X = segments_X.to(devices[0])valid_lens_x = valid_lens_x.to(devices[0])pred_positions_X = pred_positions_X.to(devices[0])mlm_weights_X = mlm_weights_X.to(devices[0])mlm_Y, nsp_y = mlm_Y.to(devices[0]), nsp_y.to(devices[0])trainer.zero_grad()timer.start()mlm_l, nsp_l, l = _get_batch_loss_bert(net, loss, vocab_size, tokens_X, segments_X, valid_lens_x,pred_positions_X, mlm_weights_X, mlm_Y, nsp_y)l.backward()trainer.step()metric.add(mlm_l, nsp_l, tokens_X.shape[0], 1)timer.stop()animator.add(step + 1,(metric[0] / metric[3], metric[1] / metric[3]))step += 1if step == num_steps:num_steps_reached = Truebreakprint(f'MLM loss {metric[0] / metric[3]:.3f}, 'f'NSP loss {metric[1] / metric[3]:.3f}')print(f'{metric[2] / timer.sum():.1f} sentence pairs/sec on 'f'{str(devices)}')train_bert(train_iter, net, loss, len(vocab), devices, 500)

获取BERT编码器

def get_bert_encoding(net, tokens_a, tokens_b=None):tokens, segments = dltools.get_tokens_and_segments(tokens_a, tokens_b)token_ids = torch.tensor(vocab[tokens], device=devices[0]).unsqueeze(0)segments = torch.tensor(segments, device=devices[0]).unsqueeze(0)valid_len = torch.tensor(len(tokens), device=devices[0]).unsqueeze(0)encoded_X, _, _ = net(token_ids, segments, valid_len)return encoded_Xtokens_a = ['a', 'crane', 'is', 'flying']
encoded_text = get_bert_encoding(net, tokens_a)
# 词元：'<cls>','a','crane','is','flying','<sep>'
encoded_text_cls = encoded_text[:, 0, :]
encoded_text_crane = encoded_text[:, 2, :]
encoded_text.shape, encoded_text_cls.shape, encoded_text_crane[0][:3]

(torch.Size([1, 6, 128]),torch.Size([1, 128]),tensor([-1.0005,  0.8355,  0.2930], grad_fn=<SliceBackward0>))

tokens_a, tokens_b = ['a', 'crane', 'driver', 'came'], ['he', 'just', 'left']
encoded_pair = get_bert_encoding(net, tokens_a, tokens_b)
# 词元：'<cls>','a','crane','driver','came','<sep>','he','just',
# 'left','<sep>'
encoded_pair_cls = encoded_pair[:, 0, :]
encoded_pair_crane = encoded_pair[:, 2, :]
encoded_pair.shape, encoded_pair_cls.shape, encoded_pair_crane[0][:3]

(torch.Size([1, 10, 128]),torch.Size([1, 128]),tensor([-1.0168,  0.8235,  0.2141], grad_fn=<SliceBackward0>))