注意力机制和Seq2seq模型

Attention Mechanism

注意力机制借鉴了人类的注意力思维方式,以获得需要重点关注的目标区域

    在 编码器—解码器(seq2seq) 中,解码器在各个时间步依赖相同的背景变量(context vector)来获取输⼊序列信息。解码器输入的语境向量(context vector)不同,每个位置都会计算各自的 attention 输出。 当编码器为循环神经⽹络时,背景变量来⾃它最终时间步的隐藏状态。将源序列输入信息以循环单位状态编码,然后将其传递给解码器以生成目标序列。

    然而这种结构存在着问题,尤其是RNN机制实际中存在长程梯度消失的问题,对于较长的句子,我们很难寄希望于将输入的序列转化为定长的向量而保存所有的有效信息,所以随着所需翻译句子的长度的增加,这种结构的效果会显著下降。与此同时,解码的目标词语可能只与原输入的部分词语有关,而并不是与所有的输入有关。例如,当把 “Hello world” 翻译成 “Bonjour le monde” 时,“Hello” 映射成 “Bonjour”,“world” 映射成 “monde”。

    在 seq2seq 模型中,解码器只能隐式地从编码器的最终状态中选择相应的信息。然而,注意力机制可以将这种选择过程显式地建模。

Image Name

注意力机制框架

Attention 是一种通用的带权池化方法,输入由两部分构成:询问(query)和键值对(key-value pairs)。 kiRdk,viRdvk_i∈R^{d_k}, v_i∈R^{d_v}. Query qRdqq∈R^{d_q} , attention layer 得到输出与value的维度一致 oRdvo∈R^{d_v}. 对于一个query来说,attention layer 会与每一个 key 计算注意力分数并进行权重的归一化,输出的向量 oo 则是 value 的加权求和,而每个 key 计算的权重与 value 一一对应。

为了计算输出,我们首先假设有一个函数α\alpha 用于计算query和key的相似性,然后可以计算所有的 attention scores a1,,ana_1, \ldots, a_n by

ai=α(q,ki). a_i = \alpha(\mathbf q, \mathbf k_i).

我们使用 softmax 函数 获得注意力权重:

b1,,bn=softmax(a1,,an). b_1, \ldots, b_n = \textrm{softmax}(a_1, \ldots, a_n).

最终的输出就是 value 的加权求和:

o=i=1nbivi. \mathbf o = \sum_{i=1}^n b_i \mathbf v_i.

不同的 attetion layer 的区别在于 score 函数的选择

接下来将利用[机器翻译及其相关技术介绍]一文中的(https://blog.csdn.net/RokoBasilisk/article/details/104367653)

介绍两个常用的注意层

  • Dot-product Attention
  • Multilayer Perceptron Attention
import math
import torch 
import torch.nn as nn
# import dataset
import os
os.listdir('path to storaged file of dataset')

工具1: Masked Softmax

# 排除padding位置的影响
def SequenceMask(X, X_len,value=-1e6):
    maxlen = X.size(1)
    #print(X.size(),torch.arange((maxlen),dtype=torch.float)[None, :],'\n',X_len[:, None] )
    # shape as same as X
    mask = torch.arange((maxlen),dtype=torch.float)[None, :] >= X_len[:, None]   
    #print(mask)
    X[mask]=value
    return X
def masked_softmax(X, valid_length):
    # X: 3-D tensor, valid_length: 1-D or 2-D tensor
    softmax = nn.Softmax(dim=-1)
    if valid_length is None:
        return softmax(X)
    else:
        shape = X.shape
        if valid_length.dim() == 1:
            try:
                valid_length = torch.FloatTensor(valid_length.numpy().repeat(shape[1], axis=0))#[2,2,3,3]
            except:
                valid_length = torch.FloatTensor(valid_length.cpu().numpy().repeat(shape[1], axis=0))#[2,2,3,3]
        else:
            valid_length = valid_length.reshape((-1,))
        # fill masked elements with a large negative, whose exp is 0
        X = SequenceMask(X.reshape((-1, shape[-1])), valid_length)
 
        return softmax(X).reshape(shape)

工具2: 超出2维矩阵的乘法

XXYY 是维度分别为(b,n,m)(b,n,m)(b,m,k)(b, m, k)的张量,进行 bb 次二维矩阵乘法后得到 ZZ, 维度为 (b,n,k)(b, n, k)

Z[i,:,:]=dot(X[i,:,:],Y[i,:,:])for i=1,,n . Z[i,:,:] = dot(X[i,:,:], Y[i,:,:])\qquad for\ i= 1,…,n\ .


Dot Product Attention

The dot product 假设query和keys有相同的维度, 即 i,q,kiRd\forall i, q,k_i ∈ R_d. 通过计算 query 和 key 转置的乘积来计算 attention score ,通常还会除去 d\sqrt{d} 减少计算出来的 score 对维度 𝑑 的依赖性,如下

α(q,k)=q,k/d α (q,k)=⟨q,k⟩/ \sqrt{d}

假设 QRm×dQ∈R^{m×d}mm 个query,KRn×dK∈R^{n×d}nn 个 keys. 我们可以通过矩阵运算的方式计算所有 mnmn 个 score:

α(Q,K)=QKT/d α (Q,K)=QK^T/\sqrt{d}

它支持一批查询和键值对。此外,它支持作为正则化随机删除一些注意力权重.

# Save to the d2l package.
class DotProductAttention(nn.Module): 
    def __init__(self, dropout, **kwargs):
        super(DotProductAttention, self).__init__(**kwargs)
        self.dropout = nn.Dropout(dropout)

    # query: (batch_size, #queries, d)
    # key: (batch_size, #kv_pairs, d)
    # value: (batch_size, #kv_pairs, dim_v)
    # valid_length: either (batch_size, ) or (batch_size, xx)
    def forward(self, query, key, value, valid_length=None):
        d = query.shape[-1]
        # set transpose_b=True to swap the last two dimensions of key
        
        scores = torch.bmm(query, key.transpose(1,2)) / math.sqrt(d)
        attention_weights = self.dropout(masked_softmax(scores, valid_length))
        print("attention_weight\n",attention_weights)
        return torch.bmm(attention_weights, value)

测试:

创建了两个批,每个批有一个query和10个key-values对。

通过valid_length指定,对于第一批,只关注前2个键-值对,而对于第二批,检查前6个键-值对

因此,尽管这两个批处理具有相同的查询和键值对,但我们获得的输出是不同的。

atten = DotProductAttention(dropout=0)

keys = torch.ones((2,10,2),dtype=torch.float)
values = torch.arange((40), dtype=torch.float).view(1,10,4).repeat(2,1,1)
atten(torch.ones((2,1,2),dtype=torch.float), keys, values, torch.FloatTensor([2, 6]))

# Result
attention_weight
 tensor([[[0.5000, 0.5000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000,
          0.0000, 0.0000]],

        [[0.1667, 0.1667, 0.1667, 0.1667, 0.1667, 0.1667, 0.0000, 0.0000,
          0.0000, 0.0000]]])
tensor([[[ 2.0000,  3.0000,  4.0000,  5.0000]],

        [[10.0000, 11.0000, 12.0000, 13.0000]]])

Multilayer Porceptron Attentiion

在多层感知器中,我们首先将 query and keys 投影到 RhR^ℎ .为了更具体,我们将可以学习的参数做如下映射
WkRh×dkW_k∈R^{h×d_k} , WqRh×dqW_q∈R^{h×d_q} , and vRhv∈R^h . 将 score 函数定义

α(k,q)=vTtanh(Wkk+Wqq) α(k,q)=v^Ttanh(W_kk+W_qq)
.
然后将key 和 value 在特征的维度上合并(concatenate),然后送至 a single hidden layer perceptron 这层中 hidden layer 为 ℎ and 输出的size为 1 .隐层激活函数为tanh,无偏置.

# Save to the d2l package.
class MLPAttention(nn.Module):  
    def __init__(self, units,ipt_dim,dropout, **kwargs):
        super(MLPAttention, self).__init__(**kwargs)
        # Use flatten=True to keep query's and key's 3-D shapes.
        self.W_k = nn.Linear(ipt_dim, units, bias=False)
        self.W_q = nn.Linear(ipt_dim, units, bias=False)
        self.v = nn.Linear(units, 1, bias=False)
        self.dropout = nn.Dropout(dropout)

    def forward(self, query, key, value, valid_length):
        query, key = self.W_k(query), self.W_q(key)
        #print("size",query.size(),key.size())
        # expand query to (batch_size, #querys, 1, units), and key to
        # (batch_size, 1, #kv_pairs, units). Then plus them with broadcast.
        features = query.unsqueeze(2) + key.unsqueeze(1)
        #print("features:",features.size())  #--------------开启
        scores = self.v(features).squeeze(-1) 
        attention_weights = self.dropout(masked_softmax(scores, valid_length))
        return torch.bmm(attention_weights, value)

测试:

    尽管 MLPAttention 包含一个额外的 MLP 模型,但如果给定相同的输入和相同的键,我们将获得与DotProductAttention相同的输出

atten = MLPAttention(ipt_dim=2,units = 8, dropout=0)
atten(torch.ones((2,1,2), dtype = torch.float), keys, values, torch.FloatTensor([2, 6]))      
#Result
tensor([[[ 2.0000,  3.0000,  4.0000,  5.0000]],

        [[10.0000, 11.0000, 12.0000, 13.0000]]], grad_fn=<BmmBackward>)

比较

在Dot-product Attention中,key与query维度需要一致,在MLP Attention中则不需要。

Seq2seq模型

机器翻译及其相关技术介绍

seq2seq 模型的预测需人为设定终止条件,设定最长序列长度或者输出 [EOS] 结束符号,若不加以限制则可能生成无穷长度序列。

引出:

引入注意力机制的Seq2seq模型

注意力机制本身有高效的并行性,但引入注意力并不能改变seq2seq内部RNN的迭代机制,因此无法加速。

将注意机制添加到 sequence to sequence 模型中,以显式地使用权重聚合 states。

下图展示 encoding 和 decoding 的模型结构,在时间步为 tt 的时候。此刻 attention layer 保存着 encodering 看到的所有信息——即 encoding 的每一步输出。在 decoding 阶段,解码器的 tt 时刻的隐藏状态被当作 query,encoder 的每个时间步的 hidden states 作为 key 和 value 进行 attention 聚合.

Attetion model 的输出当作成上下文信息 context vector,并与解码器输入 DtD_t 拼接起来一起送到解码器:

Image Name

Fig1seqtoseq Fig1具有注意机制的seq-to-seq模型解码的第二步

下图展示了seq2seq机制的所以层的关系,下面展示了encoder和decoder的layer结构

Image Name

Fig2seqtoseq Fig2具有注意机制的seq-to-seq模型中层结构

解码器

由于带有注意机制的 seq2seq 的编码器与之前章节中的 Seq2SeqEncoder 相同,所以在此处我们只关注解码器。

我们添加了一个MLP注意层(MLPAttention),它的隐藏大小与解码器中的LSTM层相同。然后我们通过从编码器传递三个参数来初始化解码器的状态:

  • the encoder outputs of all timesteps:encoder 输出的各个状态,被用于attetion layer 的 memory 部分,有相同的 key 和 values ;

  • the hidden state of the encoder’s final timestep:编码器最后一个时间步的隐藏状态,被用于初始化decoder 的hidden state ;

  • the encoder valid length: 编码器的有效长度,借此,注意层不会考虑编码器输出中的填充标记(Paddings);

在解码的每个时间步,我们使用解码器的最后一个RNN层的输出作为注意层的 query。

然后,将注意力模型的输出与输入嵌入向量连接起来,输入到 RNN 层。虽然 RNN 层隐藏状态也包含来自解码器的历史信息,但是 attention model 的输出显式地选择了 enc_valid_len 以内的编码器输出,这样 attention机制就会尽可能排除其他不相关的信息。

class Seq2SeqAttentionDecoder(d2l.Decoder):
    def __init__(self, vocab_size, embed_size, num_hiddens, num_layers,
                 dropout=0, **kwargs):
        super(Seq2SeqAttentionDecoder, self).__init__(**kwargs)
        self.attention_cell = MLPAttention(num_hiddens,num_hiddens, dropout)
        self.embedding = nn.Embedding(vocab_size, embed_size)
        self.rnn = nn.LSTM(embed_size+ num_hiddens,num_hiddens, num_layers, dropout=dropout)
        self.dense = nn.Linear(num_hiddens,vocab_size)

    def init_state(self, enc_outputs, enc_valid_len, *args):
        outputs, hidden_state = enc_outputs
#         print("first:",outputs.size(),hidden_state[0].size(),hidden_state[1].size())
        # Transpose outputs to (batch_size, seq_len, hidden_size)
        return (outputs.permute(1,0,-1), hidden_state, enc_valid_len)
        #outputs.swapaxes(0, 1)
        
    def forward(self, X, state):
        enc_outputs, hidden_state, enc_valid_len = state
        #("X.size",X.size())
        X = self.embedding(X).transpose(0,1)
#         print("Xembeding.size2",X.size())
        outputs = []
        for l, x in enumerate(X):
#             print(f"\n{l}-th token")
#             print("x.first.size()",x.size())
            # query shape: (batch_size, 1, hidden_size)
            # select hidden state of the last rnn layer as query
            query = hidden_state[0][-1].unsqueeze(1) # np.expand_dims(hidden_state[0][-1], axis=1)
            # context has same shape as query
#             print("query enc_outputs, enc_outputs:\n",query.size(), enc_outputs.size(), enc_outputs.size())
            context = self.attention_cell(query, enc_outputs, enc_outputs, enc_valid_len)
            # Concatenate on the feature dimension
#             print("context.size:",context.size())
            x = torch.cat((context, x.unsqueeze(1)), dim=-1)
            # Reshape x to (1, batch_size, embed_size+hidden_size)
#             print("rnn",x.size(), len(hidden_state))
            out, hidden_state = self.rnn(x.transpose(0,1), hidden_state)
            outputs.append(out)
        outputs = self.dense(torch.cat(outputs, dim=0))
        return outputs.transpose(0, 1), [enc_outputs, hidden_state,
                                        enc_valid_len]
encoder = d2l.Seq2SeqEncoder(vocab_size=10, embed_size=8,
                            num_hiddens=16, num_layers=2)
# encoder.initialize()
decoder = Seq2SeqAttentionDecoder(vocab_size=10, embed_size=8,
                                  num_hiddens=16, num_layers=2)
X = torch.zeros((4, 7),dtype=torch.long)
print("batch size=4\nseq_length=7\nhidden dim=16\nnum_layers=2\n")
print('encoder output size:', encoder(X)[0].size())
print('encoder hidden size:', encoder(X)[1][0].size())
print('encoder memory size:', encoder(X)[1][1].size())
state = decoder.init_state(encoder(X), None)
out, state = decoder(X, state)
out.shape, len(state), state[0].shape, len(state[1]), state[1][0].shape

训练

import zipfile
import torch
import requests
from io import BytesIO
from torch.utils import data
import sys
import collections

class Vocab(object): # This class is saved in d2l.
  def __init__(self, tokens, min_freq=0, use_special_tokens=False):
    # sort by frequency and token
    counter = collections.Counter(tokens)
    token_freqs = sorted(counter.items(), key=lambda x: x[0])
    token_freqs.sort(key=lambda x: x[1], reverse=True)
    if use_special_tokens:
      # padding, begin of sentence, end of sentence, unknown
      self.pad, self.bos, self.eos, self.unk = (0, 1, 2, 3)
      tokens = ['', '', '', '']
    else:
      self.unk = 0
      tokens = ['']
    tokens += [token for token, freq in token_freqs if freq >= min_freq]
    self.idx_to_token = []
    self.token_to_idx = dict()
    for token in tokens:
      self.idx_to_token.append(token)
      self.token_to_idx[token] = len(self.idx_to_token) - 1
      
  def __len__(self):
    return len(self.idx_to_token)
  
  def __getitem__(self, tokens):
    if not isinstance(tokens, (list, tuple)):
      return self.token_to_idx.get(tokens, self.unk)
    else:
      return [self.__getitem__(token) for token in tokens]
    
  def to_tokens(self, indices):
    if not isinstance(indices, (list, tuple)):
      return self.idx_to_token[indices]
    else:
      return [self.idx_to_token[index] for index in indices]

def load_data_nmt(batch_size, max_len, num_examples=1000):
    """Download an NMT dataset, return its vocabulary and data iterator."""
    # Download and preprocess
    def preprocess_raw(text):
        text = text.replace('\u202f', ' ').replace('\xa0', ' ')
        out = ''
        for i, char in enumerate(text.lower()):
            if char in (',', '!', '.') and text[i-1] != ' ':
                out += ' '
            out += char
        return out 


    with open('/home/kesci/input/fraeng6506/fra.txt', 'r') as f:
      raw_text = f.read()


    text = preprocess_raw(raw_text)

    # Tokenize
    source, target = [], []
    for i, line in enumerate(text.split('\n')):
        if i >= num_examples:
            break
        parts = line.split('\t')
        if len(parts) >= 2:
            source.append(parts[0].split(' '))
            target.append(parts[1].split(' '))

    # Build vocab
    def build_vocab(tokens):
        tokens = [token for line in tokens for token in line]
        return Vocab(tokens, min_freq=3, use_special_tokens=True)
    src_vocab, tgt_vocab = build_vocab(source), build_vocab(target)

    # Convert to index arrays
    def pad(line, max_len, padding_token):
        if len(line) > max_len:
            return line[:max_len]
        return line + [padding_token] * (max_len - len(line))

    def build_array(lines, vocab, max_len, is_source):
        lines = [vocab[line] for line in lines]
        if not is_source:
            lines = [[vocab.bos] + line + [vocab.eos] for line in lines]
        array = torch.tensor([pad(line, max_len, vocab.pad) for line in lines])
        valid_len = (array != vocab.pad).sum(1)
        return array, valid_len

    src_vocab, tgt_vocab = build_vocab(source), build_vocab(target)
    src_array, src_valid_len = build_array(source, src_vocab, max_len, True)
    tgt_array, tgt_valid_len = build_array(target, tgt_vocab, max_len, False)
    train_data = data.TensorDataset(src_array, src_valid_len, tgt_array, tgt_valid_len)
    train_iter = data.DataLoader(train_data, batch_size, shuffle=True)
    return src_vocab, tgt_vocab, train_iter
embed_size, num_hiddens, num_layers, dropout = 32, 32, 2, 0.0
batch_size, num_steps = 64, 10
lr, num_epochs, ctx = 0.005, 500, d2l.try_gpu()

src_vocab, tgt_vocab, train_iter = load_data_nmt(batch_size, num_steps)
encoder = d2l.Seq2SeqEncoder(
    len(src_vocab), embed_size, num_hiddens, num_layers, dropout)
decoder = Seq2SeqAttentionDecoder(
    len(tgt_vocab), embed_size, num_hiddens, num_layers, dropout)
model = d2l.EncoderDecoder(encoder, decoder)

预测

d2l.train_s2s_ch9(model, train_iter, lr, num_epochs, ctx)
for sentence in ['Go .', 'Good Night !', "I'm OK .", 'I won !']:
    print(sentence + ' => ' + d2l.predict_s2s_ch9(
        model, sentence, src_vocab, tgt_vocab, num_steps, ctx))
posted @ 2020-02-18 13:16  Roko&Basilisk  阅读(581)  评论(0编辑  收藏  举报