Seq2Seq(Attention)的PyTorch实现

B站视频讲解php

文本主要介绍一下如何使用PyTorch复现Seq2Seq(with Attention),实现简单的机器翻译任务,请先阅读论文Neural Machine Translation by Jointly Learning to Align and Translate,以后花上15分钟阅读个人这两篇文章Seq2Seq 与注意力机制图解Attention,最后再来看文本,方能达到醍醐灌顶,事半功倍的效果html

数据预处理

数据预处理的代码其实就是调用各类API,我不但愿读者被这些不过重要的部分分散了注意力,所以这里我不贴代码,仅口述一下带过便可python

以下图所示,本文使用的是德语→英语数据集,输入是德语,而且输入的每一个句子开头和结尾都带有特殊的标识符。输出是英语,而且输出的每一个句子开头和结尾也都带有特殊标识符git

不论是英语仍是德语,每句话长度都是不固定的,因此我对于每一个batch内的句子,将它们的长度经过加<PAD>变得同样,也就说,一个batch内的句子,长度都是相同的,不一样batch内的句子长度不必定相同。下图维度表示分别是[seq_len, batch_size]github

随便打印一条数据,看一下数据封装的形式markdown

在数据预处理的时候,须要将源句子和目标句子分开构建字典,也就是单独对德语构建一个词库,对英语构建一个词库网络

Encoder

Encoder我是用的单层双向GRUapp

双向GRU的隐藏状态输出由两个向量拼接而成,例如 h 1 = [ h 1 ; h T ] h_1=[\overrightarrow{h_1};\overleftarrow{h_T}] , h 2 = [ h 2 ; h T 1 ] h_2=[\overrightarrow{h_2};\overleftarrow{h}_{T-1}] ......全部时刻的最后一层隐藏状态就构成了GRU的outputdom

o u t p u t = { h 1 , h 2 , . . . h T } output=\{h_1,h_2,...h_T\}

假设这是个m层GRU,那么最后一个时刻全部层中的隐藏状态就构成了GRU的final hidden stateside

h i d d e n = { h T 1 , h T 2 , . . . , h T m } hidden=\{h^1_T,h^2_T,...,h^m_T\}

其中

h T i = [ h T i ; h 1 i ] h^i_T=[\overrightarrow{h^i_T};\overleftarrow{h^i_1}]

因此

h i d d e n = { [ h T 1 ; h 1 1 ] , [ h T 2 ; h 1 2 ] , . . . , [ h T m ; h 1 m ] } hidden=\{[\overrightarrow{h^1_T};\overleftarrow{h^1_1}],[\overrightarrow{h^2_T};\overleftarrow{h^2_1}],...,[\overrightarrow{h^m_T};\overleftarrow{h^m_1}]\}

根据论文,或者你看了个人图解Attention这篇文章就会知道,咱们须要的是hidden的最后一层输出(包括正向和反向),所以咱们能够经过hidden[-2,:,:]hidden[-1,:,:]取出最后一层的hidden states,将它们拼接起来记做 s 0 s_0

最后一个细节之处在于, s 0 s_0 的维度是[batch_size, en_hid_dim*2],即使是没有Attention机制,将 s 0 s_0 做为Decoder的初始隐藏状态也不对,由于维度不匹配,Decoder的初始隐藏状态是三维的,而如今咱们的 s 0 s_0 是二维的,所以须要将 s 0 s_0 的维度转为三维,而且还要调整各个维度上的值。首先我经过一个全链接神经网络,将 s 0 s_0 的维度变为[batch_size, dec_hid_dim]

Encoder的细节就这么多,下面直接上代码,个人代码风格是,注释在上,代码在下

class Encoder(nn.Module):
    def __init__(self, input_dim, emb_dim, enc_hid_dim, dec_hid_dim, dropout):
        super().__init__()
        self.embedding = nn.Embedding(input_dim, emb_dim)
        self.rnn = nn.GRU(emb_dim, enc_hid_dim, bidirectional = True)
        self.fc = nn.Linear(enc_hid_dim * 2, dec_hid_dim)
        self.dropout = nn.Dropout(dropout)
        
    def forward(self, src): 
        ''' src = [src_len, batch_size] '''
        src = src.transpose(0, 1) # src = [batch_size, src_len]
        embedded = self.dropout(self.embedding(src)).transpose(0, 1) # embedded = [src_len, batch_size, emb_dim]
        
        # enc_output = [src_len, batch_size, hid_dim * num_directions]
        # enc_hidden = [n_layers * num_directions, batch_size, hid_dim]
        enc_output, enc_hidden = self.rnn(embedded) # if h_0 is not give, it will be set 0 acquiescently

        # enc_hidden is stacked [forward_1, backward_1, forward_2, backward_2, ...]
        # enc_output are always from the last layer
        
        # enc_hidden [-2, :, : ] is the last of the forwards RNN 
        # enc_hidden [-1, :, : ] is the last of the backwards RNN
        
        # initial decoder hidden is final hidden state of the forwards and backwards 
        # encoder RNNs fed through a linear layer
        # s = [batch_size, dec_hid_dim]
        s = torch.tanh(self.fc(torch.cat((enc_hidden[-2,:,:], enc_hidden[-1,:,:]), dim = 1)))
        
        return enc_output, s
复制代码

Attention

attention无非就是三个公式

E t = t a n h ( a t t n ( s t 1 , H ) ) a t ~ = v E t a t = s o f t m a x ( a t ~ ) E_t=tanh(attn(s_{t-1},H))\\ \tilde{a_t}=vE_t\\ {a_t}=softmax(\tilde{a_t})

其中 s t 1 s_{t-1} 指的就是Encoder中的变量s H H 指的就是Encoder中的变量enc_output a t t n ( ) attn() 其实就是一个简单的全链接神经网络

咱们能够从最后一个公式反推各个变量的维度是什么,或者维度有什么要求

首先 a t a_t 的维度应该是[batch_size, src_len],这是毋庸置疑的,那么 a t ~ \tilde{a_t} 的维度也应该是[batch_size, src_len],或者 a t ~ \tilde{a_t} 是个三维的,可是某个维度值为1,能够经过squeeze()变成两维的。这里咱们先假设 a t ~ \tilde{a_t} 的维度是[batch_size, src_len, 1],等会儿我再解释为何要这样假设

继续往上推,变量 v v 的维度就应该是[?, 1]?表示我暂时不知道它的值应该是多少。 E t E_t 的维度应该是[batch_size, src_len, ?]

如今已知 H H 的维度是[batch_size, src_len, enc_hid_dim*2] s t 1 s_{t-1} 目前的维度是[batch_size, dec_hid_dim],这两个变量须要作拼接,送入全链接神经网络,所以咱们首先须要将 s t 1 s_{t-1} 的维度变成[batch_size, src_len, dec_hid_dim],拼接以后的维度就变成[batch_size, src_len, enc_hid_dim*2+dec_hid_dim],因而 a t t n ( ) attn() 这个函数的输入输出值也就有了

attn = nn.Linear(enc_hid_dim*2+dec_hid_dim, ?)
复制代码

到此为止,除了?部分的值不清楚,其它全部维度都推导出来了。如今咱们回过头思考一下?设置成多少,好像其实并无任何限制,因此咱们能够设置?为任何值(在代码中我设置?dec_hid_dim

Attention细节就这么多,下面给出代码

class Attention(nn.Module):
    def __init__(self, enc_hid_dim, dec_hid_dim):
        super().__init__()
        self.attn = nn.Linear((enc_hid_dim * 2) + dec_hid_dim, dec_hid_dim, bias=False)
        self.v = nn.Linear(dec_hid_dim, 1, bias = False)
        
    def forward(self, s, enc_output):
        
        # s = [batch_size, dec_hid_dim]
        # enc_output = [src_len, batch_size, enc_hid_dim * 2]
        
        batch_size = enc_output.shape[1]
        src_len = enc_output.shape[0]
        
        # repeat decoder hidden state src_len times
        # s = [batch_size, src_len, dec_hid_dim]
        # enc_output = [batch_size, src_len, enc_hid_dim * 2]
        s = s.unsqueeze(1).repeat(1, src_len, 1)
        enc_output = enc_output.transpose(0, 1)
        
        # energy = [batch_size, src_len, dec_hid_dim]
        energy = torch.tanh(self.attn(torch.cat((s, enc_output), dim = 2)))
        
        # attention = [batch_size, src_len]
        attention = self.v(energy).squeeze(2)
        
        return F.softmax(attention, dim=1)
复制代码

Seq2Seq(with Attention)

我调换一下顺序,先讲Seq2Seq,再讲Decoder的部分

传统Seq2Seq是直接将句子中每一个词接二连三输入Decoder进行训练,而引入Attention机制以后,我须要可以人为控制一个词一个词进行输入(由于输入每一个词到Decoder,须要再作一些运算),因此在代码中会看到我使用了for循环,循环trg_len-1次(开头的<SOS>我手动输入,因此循环少一次)

而且训练过程当中我使用了一种叫作Teacher Forcing的机制,保证训练速度的同时增长鲁棒性,若是不了解Teacher Forcing能够看个人这篇文章

思考一下for循环中应该要作哪些事?首先要将变量传入Decoder,因为Attention的计算是在Decoder的内部进行的,因此我须要将dec_inputsenc_output这三个变量传入Decoder,Decoder会返回dec_output以及新的s。以后根据几率对dec_output作Teacher Forcing便可

Seq2Seq细节就这么多,下面给出代码

class Seq2Seq(nn.Module):
    def __init__(self, encoder, decoder, device):
        super().__init__()
        self.encoder = encoder
        self.decoder = decoder
        self.device = device
        
    def forward(self, src, trg, teacher_forcing_ratio = 0.5):
        
        # src = [src_len, batch_size]
        # trg = [trg_len, batch_size]
        # teacher_forcing_ratio is probability to use teacher forcing
        
        batch_size = src.shape[1]
        trg_len = trg.shape[0]
        trg_vocab_size = self.decoder.output_dim
        
        # tensor to store decoder outputs
        outputs = torch.zeros(trg_len, batch_size, trg_vocab_size).to(self.device)
        
        # enc_output is all hidden states of the input sequence, back and forwards
        # s is the final forward and backward hidden states, passed through a linear layer
        enc_output, s = self.encoder(src)
                
        # first input to the decoder is the <sos> tokens
        dec_input = trg[0,:]
        
        for t in range(1, trg_len):
            
            # insert dec_input token embedding, previous hidden state and all encoder hidden states
            # receive output tensor (predictions) and new hidden state
            dec_output, s = self.decoder(dec_input, s, enc_output)
            
            # place predictions in a tensor holding predictions for each token
            outputs[t] = dec_output
            
            # decide if we are going to use teacher forcing or not
            teacher_force = random.random() < teacher_forcing_ratio
            
            # get the highest predicted token from our predictions
            top1 = dec_output.argmax(1) 
            
            # if teacher forcing, use actual next token as next input
            # if not, use predicted token
            dec_input = trg[t] if teacher_force else top1

        return outputs
复制代码

Decoder

Decoder我用的是单向单层GRU

Decoder部分实际上也就是三个公式

c = a t H s t = G R U ( e m b ( y t ) , c , s t 1 ) y t ^ = f ( e m b ( y t ) , c , s t ) c=a_tH\\ s_t=GRU(emb(y_t), c, s_{t-1})\\ \hat{y_t}=f(emb(y_t), c, s_t)

H H 指的是Encoder中的变量enc_output e m b ( y t ) emb(y_t) 指的是将dec_input通过WordEmbedding后获得的结果, f ( ) f() 函数实际上就是为了转换维度,由于须要的输出是TRG_VOCAB_SIZE大小。其中有个细节,GRU的参数只有两个,一个输入,一个隐藏层输入,可是上面的公式有三个变量,因此咱们应该选一个做为隐藏层输入,另外两个"整合"一下,做为输入

咱们从第一个公式正推各个变量的维度是什么

首先在Encoder中最开始先调用一次Attention,获得权重 a t a_t ,它的维度是[batch_size, src_len],而 H H 的维度是[src_len, batch_size, enc_hid_dim*2],它俩要相乘,同时应该保留batch_size这个维度,因此应该先将 a t a_t 扩展一维,而后调换一下 H H 维度的顺序,以后再按照batch相乘(即同一个batch内的矩阵相乘)

a = a.unsqueeze(1) # [batch_size, 1, src_len]
H = H.transpose(0, 1) # [batch_size, src_len, enc_hid_dim*2]
c = torch.bmm(a, h) # [batch_size, 1, enc_hid_dim*2]
复制代码

前面也说了,因为GRU不须要三个变量,因此须要将 e m b ( y t ) emb(y_t) c c 整合一下, y t y_t 实际上就是Seq2Seq类中的dec_input变量,它的维度是[batch_size],所以先将 y t y_t 扩展一个维度,再经过WordEmbedding,这样他就变成[batch_size, 1, emb_dim]。最后对 c c e m b ( y t ) emb(y_t) 进行concat

y = y.unsqueeze(1) # [batch_size, 1]
emb_y = self.emb(y) # [batch_size, 1, emb_dim]
rnn_input = torch.cat((emb_y, c), dim=2) # [batch_size, 1, emb_dim+enc_hid_dim*2]
复制代码

s t 1 s_{t-1} 的维度是[batch_size, dec_hid_dim],因此应该先将其拓展一个维度

rnn_input = rnn_input.transpose(0, 1) # [1, batch_size, emb_dim+enc_hid_dim*2]
s = s.unsqueeze(1) # [batch_size, 1, dec_hid_dim]

# dec_output = [1, batch_size, dec_hid_dim]
# dec_hidden = [1, batch_size, dec_hid_dim] = s (new, is not s previously)
dec_output, dec_hidden = self.rnn(rnn_input, s)
复制代码

最后一个公式,须要将三个变量所有拼接在一块儿,而后经过一个全链接神经网络,获得最终的预测。咱们先分析下这个三个变量的维度, e m b ( y t ) emb(y_t) 的维度是[batch_size, 1, emb_dim] c c 的维度是[batch_size, 1, enc_hid_dim] s t s_t 的维度是[1, batch_size, dec_hid_dim],所以咱们能够像下面这样把他们所有拼接起来

emd_y = emb_y.squeeze(1) # [batch_size, emb_dim]
c = w.squeeze(1) # [batch_size, enc_hid_dim*2]
s = s.squeeze(0) # [batch_size, dec_hid_dim]

fc_input = torch.cat((emb_y, c, s), dim=1) # [batch_size, enc_hid_dim*2+dec_hid_dim+emb_hid] 
复制代码

以上就是Decoder部分的细节,下面给出代码(上面的那些只是示例代码,和下面代码变量名可能不同)

class Decoder(nn.Module):
    def __init__(self, output_dim, emb_dim, enc_hid_dim, dec_hid_dim, dropout, attention):
        super().__init__()
        self.output_dim = output_dim
        self.attention = attention
        self.embedding = nn.Embedding(output_dim, emb_dim)
        self.rnn = nn.GRU((enc_hid_dim * 2) + emb_dim, dec_hid_dim)
        self.fc_out = nn.Linear((enc_hid_dim * 2) + dec_hid_dim + emb_dim, output_dim)
        self.dropout = nn.Dropout(dropout)
        
    def forward(self, dec_input, s, enc_output):
             
        # dec_input = [batch_size]
        # s = [batch_size, dec_hid_dim]
        # enc_output = [src_len, batch_size, enc_hid_dim * 2]
        
        dec_input = dec_input.unsqueeze(1) # dec_input = [batch_size, 1]
        
        embedded = self.dropout(self.embedding(dec_input)).transpose(0, 1) # embedded = [1, batch_size, emb_dim]
        
        # a = [batch_size, 1, src_len] 
        a = self.attention(s, enc_output).unsqueeze(1)
        
        # enc_output = [batch_size, src_len, enc_hid_dim * 2]
        enc_output = enc_output.transpose(0, 1)

        # c = [1, batch_size, enc_hid_dim * 2]
        c = torch.bmm(a, enc_output).transpose(0, 1)

        # rnn_input = [1, batch_size, (enc_hid_dim * 2) + emb_dim]
        rnn_input = torch.cat((embedded, c), dim = 2)
            
        # dec_output = [src_len(=1), batch_size, dec_hid_dim]
        # dec_hidden = [n_layers * num_directions, batch_size, dec_hid_dim]
        dec_output, dec_hidden = self.rnn(rnn_input, s.unsqueeze(0))
        
        # embedded = [batch_size, emb_dim]
        # dec_output = [batch_size, dec_hid_dim]
        # c = [batch_size, enc_hid_dim * 2]
        embedded = embedded.squeeze(0)
        dec_output = dec_output.squeeze(0)
        c = c.squeeze(0)
        
        # pred = [batch_size, output_dim]
        pred = self.fc_out(torch.cat((dec_output, c, embedded), dim = 1))
        
        return pred, dec_hidden.squeeze(0)
复制代码

定义模型

INPUT_DIM = len(SRC.vocab)
OUTPUT_DIM = len(TRG.vocab)
ENC_EMB_DIM = 256
DEC_EMB_DIM = 256
ENC_HID_DIM = 512
DEC_HID_DIM = 512
ENC_DROPOUT = 0.5
DEC_DROPOUT = 0.5

attn = Attention(ENC_HID_DIM, DEC_HID_DIM)
enc = Encoder(INPUT_DIM, ENC_EMB_DIM, ENC_HID_DIM, DEC_HID_DIM, ENC_DROPOUT)
dec = Decoder(OUTPUT_DIM, DEC_EMB_DIM, ENC_HID_DIM, DEC_HID_DIM, DEC_DROPOUT, attn)

model = Seq2Seq(enc, dec, device).to(device)
TRG_PAD_IDX = TRG.vocab.stoi[TRG.pad_token]
criterion = nn.CrossEntropyLoss(ignore_index = TRG_PAD_IDX).to(device)
optimizer = optim.Adam(model.parameters(), lr=1e-3)
复制代码

倒数第二行CrossEntropyLoss()中的参数不多见,ignore_index=TRG_PAD_IDX,这个参数的做用是忽略某一类别,不计算其loss,可是要注意,忽略的是真实值中的类别,例以下面的代码,真实值的类别都是1,而预测值所有预测认为是2(下标从0开始),同时loss function设置忽略第一类的loss,此时会打印出0

label = torch.tensor([1, 1, 1])
pred = torch.tensor([[0.1, 0.2, 0.6], [0.2, 0.1, 0.8], [0.1, 0.1, 0.9]])
loss_fn = nn.CrossEntropyLoss(ignore_index=1)
print(loss_fn(pred, label).item()) # 0
复制代码

若是设置loss function忽略第二类,此时loss并不会为0

label = torch.tensor([1, 1, 1])
pred = torch.tensor([[0.1, 0.2, 0.6], [0.2, 0.1, 0.8], [0.1, 0.1, 0.9]])
loss_fn = nn.CrossEntropyLoss(ignore_index=2)
print(loss_fn(pred, label).item()) # 1.359844
复制代码

最后给出完整代码连接(须要科学的力量) Github项目地址:nlp-tutorial

相关文章
相关标签/搜索