个人笔记

import math
import torch
from torch import nn
from d2l import torch as d2l

#@save
def read_ptb():
    """将PTB数据集加载到文本行的列表中"""
    data_dir = d2l.download_extract('ptb')
    # Readthetrainingset.
    with open(os.path.join(data_dir, 'ptb.train.txt')) as f:
        raw_text = f.read()
    return [line.split() for line in raw_text.split('\n')]

#@save
def subsample(sentences, vocab):
    """下采样高频词"""
    # 排除未知词元'<unk>'
    sentences = [[token for token in line if vocab[token] != vocab.unk]
                 for line in sentences]
    counter = d2l.count_corpus(sentences)
    num_tokens = sum(counter.values())

    # 如果在下采样期间保留词元，则返回True
    def keep(token):
        return(random.uniform(0, 1) <
               math.sqrt(1e-4 / counter[token] * num_tokens))

    return ([[token for token in line if keep(token)] for line in sentences], counter)

#@save
def get_centers_and_contexts(corpus, max_window_size):
    """返回skip-gram模型中的中心词和上下文词"""
    centers, contexts = [], []
    for line in corpus:
        # 要形成“中心词-上下文词”对，每个句子至少需要有2个词
        if len(line) < 2:
            continue
        centers += line
        for i in range(len(line)):  # 上下文窗口中间i
            window_size = random.randint(1, max_window_size)
            indices = list(range(max(0, i - window_size),
                                 min(len(line), i + 1 + window_size)))
            # 从上下文词中排除中心词
            indices.remove(i)
            contexts.append([line[idx] for idx in indices])
    return centers, contexts

#@save
def get_negatives(all_contexts, vocab, counter, K):
    """返回负采样中的噪声词"""
    # 索引为1、2、...（索引0是词表中排除的未知标记）
    sampling_weights = [counter[vocab.to_tokens(i)]**0.75
                        for i in range(1, len(vocab))]
    all_negatives, generator = [], RandomGenerator(sampling_weights)
    for contexts in all_contexts:
        negatives = []
        while len(negatives) < len(contexts) * K:
            neg = generator.draw()
            # 噪声词不能是上下文词
            if neg not in contexts:
                negatives.append(neg)
        all_negatives.append(negatives)
    return all_negatives

#@save
def load_data_ptb(batch_size, max_window_size, num_noise_words):
    """下载PTB数据集，然后将其加载到内存中"""
    num_workers = d2l.get_dataloader_workers() # 设定读取数据的进程数
        sentences = read_ptb()
    vocab = d2l.Vocab(sentences, min_freq=10) # 为语料库构建一个词表
    subsampled, counter = subsample(sentences, vocab) # 下采样高频词
    corpus = [vocab[line] for line in subsampled]
    all_centers, all_contexts = get_centers_and_contexts(
        corpus, max_window_size) # 返回中心词和上下文词
    all_negatives = get_negatives( # 负采样
        all_contexts, vocab, counter, num_noise_words)

    class PTBDataset(torch.utils.data.Dataset):
        def __init__(self, centers, contexts, negatives):
            assert len(centers) == len(contexts) == len(negatives)
            self.centers = centers
            self.contexts = contexts
            self.negatives = negatives

        def __getitem__(self, index):
            return (self.centers[index], self.contexts[index],
                    self.negatives[index])

        def __len__(self):
            return len(self.centers)

    dataset = PTBDataset(all_centers, all_contexts, all_negatives)

    data_iter = torch.utils.data.DataLoader(
        dataset, batch_size, shuffle=True,
        collate_fn=batchify, num_workers=num_workers)
    return data_iter, vocab

batch_size, max_window_size, num_noise_words = 512, 5, 5
data_iter, vocab = d2l.load_data_ptb(batch_size, max_window_size,
                                     num_noise_words)

def skip_gram(center, contexts_and_negatives, embed_v, embed_u):
    """定义前向传播"""
    v = embed_v(center)
    u = embed_u(contexts_and_negatives)
    pred = torch.bmm(v, u.permute(0, 2, 1))
    return pred

class SigmoidBCELoss(nn.Module):
    # 带掩码的二元交叉熵损失
    def __init__(self):
        super().__init__()

    def forward(self, inputs, target, mask=None):
        out = nn.functional.binary_cross_entropy_with_logits(
            inputs, target, weight=mask, reduction="none")
        return out.mean(dim=1)

loss = SigmoidBCELoss()

def sigmd(x): # sigmoid激活函数
    return -math.log(1 / (1 + math.exp(-x)))

# 初始化模型参数
embed_size = 100
net = nn.Sequential(nn.Embedding(num_embeddings=len(vocab),
                                 embedding_dim=embed_size),
                    nn.Embedding(num_embeddings=len(vocab),
                                 embedding_dim=embed_size))

def train(net, data_iter, lr, num_epochs, device=d2l.try_gpu()):
    def init_weights(m):
        if type(m) == nn.Embedding:
            nn.init.xavier_uniform_(m.weight)
    net.apply(init_weights)
    net = net.to(device)
    optimizer = torch.optim.Adam(net.parameters(), lr=lr)
    # 辅助绘图类-Animator定义在第三章
    animator = d2l.Animator(xlabel='epoch', ylabel='loss',
                            xlim=[1, num_epochs])
    # 规范化的损失之和，规范化的损失数
    metric = d2l.Accumulator(2) # 辅助计算类-Accumulator定义在第三章
    for epoch in range(num_epochs):
        timer, num_batches = d2l.Timer(), len(data_iter)
        for i, batch in enumerate(data_iter):
            optimizer.zero_grad()
            center, context_negative, mask, label = [
                data.to(device) for data in batch]

            pred = skip_gram(center, context_negative, net[0], net[1])
            l = (loss(pred.reshape(label.shape).float(), label.float(), mask)
                     / mask.sum(axis=1) * mask.shape[1])
            l.sum().backward()
            optimizer.step()
            metric.add(l.sum(), l.numel())
            if (i + 1) % (num_batches // 5) == 0 or i == num_batches - 1:
                animator.add(epoch + (i + 1) / num_batches,
                             (metric[0] / metric[1],))
    print(f'loss {metric[0] / metric[1]:.3f}, '
          f'{metric[1] / timer.stop():.1f} tokens/sec on {str(device)}')

lr, num_epochs = 0.002, 5
train(net, data_iter, lr, num_epochs) # 训练
# loss 0.410, 346244.5 tokens/sec on cuda:0

def get_similar_tokens(query_token, k, embed):
    W = embed.weight.data
    x = W[vocab[query_token]]
    # 计算余弦相似性。增加1e-9以获得数值稳定性
    cos = torch.mv(W, x) / torch.sqrt(torch.sum(W * W, dim=1) *
                                      torch.sum(x * x) + 1e-9)
    topk = torch.topk(cos, k=k+1)[1].cpu().numpy().astype('int32')
    for i in topk[1:]:  # 删除输入词
        print(f'cosine sim={float(cos[i]):.3f}: {vocab.to_tokens(i)}')
# 应用词向量-寻找最相似的单词Top3
get_similar_tokens('chip', 3, net[0])
# cosine sim=0.695: intel
#  cosine sim=0.627: desktop
# cosine sim=0.627: computer

import os
import random
import torch
from torch import nn
from d2l import torch as d2l

#@save
d2l.DATA_HUB['wikitext-2'] = (
    'https://s3.amazonaws.com/research.metamind.io/wikitext/'
    'wikitext-2-v1.zip', '3c914d17d80b1459be871a5039ac23e752a53cbe')

#@save
def _read_wiki(data_dir):
    """读取数据 wikitext-2并进行简单清洗分割"""
    file_name = os.path.join(data_dir, 'wiki.train.tokens')
    with open(file_name, 'r') 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) # 使用句号作为分隔符来拆分句子
    return paragraphs

#@save
def _get_next_sentence(sentence, next_sentence, paragraphs):
    """50%的概率随机替换到正确的下一句-形成正负例"""
    if random.random() < 0.5:
        is_next = True # 
    else:
        # paragraphs是三重列表的嵌套
        next_sentence = random.choice(random.choice(paragraphs))
        is_next = False
    return sentence, next_sentence, is_next

#@save
def get_tokens_and_segments(tokens_a, tokens_b=None):
    """获取输入序列的词元及其片段索引"""
    tokens = ['<cls>'] + tokens_a + ['<sep>']
    # 0和1分别标记片段A和B
    segments = [0] * (len(tokens_a) + 2)
    if tokens_b is not None:
        tokens += tokens_b + ['<sep>']
        segments += [1] * (len(tokens_b) + 1)
    return tokens, segments

#@save
def _get_nsp_data_from_paragraph(paragraph, paragraphs, vocab, max_len):
    """生成用于下一句预测任务的训练样本"""
    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:
            continue
        tokens, segments = d2l.get_tokens_and_segments(tokens_a, tokens_b)
        nsp_data_from_paragraph.append((tokens, segments, is_next))
    return nsp_data_from_paragraph

#@save
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:
            break
        masked_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_token
        pred_positions_and_labels.append(
            (mlm_pred_position, tokens[mlm_pred_position]))
    return mlm_input_tokens, pred_positions_and_labels

#@save
def _get_mlm_data_from_tokens(tokens, vocab):
    """生成用于遮蔽语言模型任务的训练样本"""
    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>']:
            continue
        candidate_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]

#@save
def _pad_bert_inputs(examples, max_len, vocab):
    """用于填充输入的辅助函数,examples来自两个预训练任务的训练样本生成方法"""
    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)

#@save
class _WikiTextDataset(torch.utils.data.Dataset):
    """定制一个用于训练的`Dataset`类"""
    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 = [d2l.tokenize(
            paragraph, token='word') for paragraph in paragraphs]
        sentences = [sentence for paragraph in paragraphs
                     for sentence in paragraph]
        self.vocab = d2l.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)

#@save
def load_data_wiki(batch_size, max_len):
    """封装以上所有方法，构成加载WikiText-2数据集的函数"""
    num_workers = d2l.get_dataloader_workers()
    data_dir = d2l.download_extract('wikitext-2', '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.vocab

batch_size, max_len = 512, 64
train_iter, vocab = d2l.load_data_wiki(batch_size, max_len)

#@save
class BERTEncoder(nn.Module):
    """BERT编码器"""
    def __init__(self, vocab_size, num_hiddens, norm_shape, ffn_num_input,
                 ffn_num_hiddens, num_heads, num_layers, dropout,
                 max_len=1000, key_size=768, query_size=768, value_size=768,
                 **kwargs):
        super(BERTEncoder, self).__init__(**kwargs)
        self.token_embedding = nn.Embedding(vocab_size, num_hiddens)
        self.segment_embedding = nn.Embedding(2, num_hiddens)
        self.blks = nn.Sequential()
        for i in range(num_layers):
            self.blks.add_module(f"{i}", d2l.EncoderBlock(
                key_size, query_size, value_size, num_hiddens, norm_shape,
                ffn_num_input, ffn_num_hiddens, num_heads, dropout, True))
        # 在BERT中，位置嵌入是可学习的，因此我们创建一个足够长的位置嵌入参数
        self.pos_embedding = nn.Parameter(torch.randn(1, max_len,
                                                      num_hiddens))

    def forward(self, tokens, segments, valid_lens):
        # 在以下代码段中，X的形状保持不变：（批量大小，最大序列长度，num_hiddens）
        X = self.token_embedding(tokens) + self.segment_embedding(segments)
        X = X + self.pos_embedding.data[:, :X.shape[1], :]
        for blk in self.blks:
            X = blk(X, valid_lens)
        return X
#@save
class MaskLM(nn.Module):
    """BERT的掩蔽语言模型任务"""
    def __init__(self, vocab_size, num_hiddens, num_inputs=768, **kwargs):
        super(MaskLM, self).__init__(**kwargs)
        self.mlp = nn.Sequential(nn.Linear(num_inputs, num_hiddens),
                                 nn.ReLU(),
                                 nn.LayerNorm(num_hiddens),
                                 nn.Linear(num_hiddens, vocab_size))

    def forward(self, X, pred_positions):
        num_pred_positions = pred_positions.shape[1]
        pred_positions = pred_positions.reshape(-1)
        batch_size = X.shape[0]
        batch_idx = torch.arange(0, batch_size)
        # 假设batch_size=2，num_pred_positions=3
        # 那么batch_idx是np.array（[0,0,0,1,1,1]）
        batch_idx = torch.repeat_interleave(batch_idx, num_pred_positions)
        masked_X = X[batch_idx, pred_positions]
        masked_X = masked_X.reshape((batch_size, num_pred_positions, -1))
        mlm_Y_hat = self.mlp(masked_X)
        return mlm_Y_hat

#@save
class NextSentencePred(nn.Module):
    """BERT的下一句预测任务"""
    def __init__(self, num_inputs, **kwargs):
        super(NextSentencePred, self).__init__(**kwargs)
        self.output = nn.Linear(num_inputs, 2)

    def forward(self, X):
        # X的形状：(batchsize,num_hiddens)
        return self.output(X)

#@save
class BERTModel(nn.Module):
    """BERT模型：整合以上的三个类"""
    def __init__(self, vocab_size, num_hiddens, norm_shape, ffn_num_input,
                 ffn_num_hiddens, num_heads, num_layers, dropout,
                 max_len=1000, key_size=768, query_size=768, value_size=768,
                 hid_in_features=768, mlm_in_features=768,
                 nsp_in_features=768):
        super(BERTModel, self).__init__()
        self.encoder = BERTEncoder(vocab_size, num_hiddens, norm_shape,
                    ffn_num_input, ffn_num_hiddens, num_heads, num_layers,
                    dropout, max_len=max_len, key_size=key_size,
                    query_size=query_size, value_size=value_size)
        self.hidden = nn.Sequential(nn.Linear(hid_in_features, num_hiddens),
                                    nn.Tanh())
        self.mlm = MaskLM(vocab_size, num_hiddens, mlm_in_features)
        self.nsp = NextSentencePred(nsp_in_features)

    def forward(self, tokens, segments, valid_lens=None,
                pred_positions=None):
        encoded_X = self.encoder(tokens, segments, valid_lens)
        if pred_positions is not None:
            mlm_Y_hat = self.mlm(encoded_X, pred_positions)
        else:
            mlm_Y_hat = None
        # 用于下一句预测的多层感知机分类器的隐藏层，0是“<cls>”标记的索引
        nsp_Y_hat = self.nsp(self.hidden(encoded_X[:, 0, :]))
        return encoded_X, mlm_Y_hat, nsp_Y_hat


net = d2l.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 = d2l.try_all_gpus()
loss = nn.CrossEntropyLoss()

#@save
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):
    """辅助函数，给定训练样本，计算两个任务的对应损失与最终损失"""
    _, mlm_Y_hat, nsp_Y_hat = net(tokens_X, segments_X,
                                  valid_lens_x.reshape(-1),
                                  pred_positions_X)# 前向传播
    # 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_l
    return mlm_l, nsp_l, l

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, d2l.Timer()
    animator = d2l.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 = d2l.Accumulator(4)
    num_steps_reached = False
    while 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 += 1
            if step == num_steps:
                num_steps_reached = True
                break

    print(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, 50)
# MLM loss 5.680, NSP loss 0.770
# 4531.9 sentence pairs/sec on [device(type='cuda', index=0), device(type='cuda', index=1)]

def get_bert_encoding(net, tokens_a, tokens_b=None):
    """辅助函数：获取BERT的输入表示"""
    tokens, segments = d2l.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_X

tokens_a = ['a', 'crane', 'is', 'flying']
encoded_text = get_bert_encoding(net, tokens_a)
encoded_text_cls = encoded_text[:, 0, :] # 输入语句的BERT表示
encoded_text_crane = encoded_text[: 2, :] # 第二个单词的BERT表示