Transformer (Code)

Based on Yu-Hsiang Huang’s implement, graykode’s implement and DASOU’s Video.

import numpy as np
import torch
import torch.nn as nn
import torch.optim as optim
import matplotlib.pyplot as plt
import math


def make_batch(sentences):
    input_batch = [[src_vocab[n] for n in sentences[0].split()]]
    output_batch = [[tgt_vocab[n] for n in sentences[1].split()]]
    target_batch = [[tgt_vocab[n] for n in sentences[2].split()]]
    return torch.LongTensor(input_batch), torch.LongTensor(output_batch), torch.LongTensor(target_batch)


def get_attn_subsequent_mask(seq):
    attn_shape = [seq.size(0), seq.size(1), seq.size(1)]
    # attn_shape: [batch_size, tgt_len, tgt_len]
    subsequence_mask = np.triu(np.ones(attn_shape), k=1)
    subsequence_mask = torch.from_numpy(subsequence_mask).byte()
    return subsequence_mask  # [batch_size, tgt_len, tgt_len]


def get_attn_pad_mask(seq_q, seq_k):
    batch_size, len_q = seq_q.size()
    batch_size, len_k = seq_k.size()
    # eq(zero) is PAD token
    pad_attn_mask = seq_k.data.eq(0).unsqueeze(1)
    # batch_size x 1 x len_k, one is masking
    return pad_attn_mask.expand(batch_size, len_q, len_k)
    # batch_size x len_q x len_k


class ScaledDotProductAttention(nn.Module):
    def __init__(self):
        super(ScaledDotProductAttention, self).__init__()

    def forward(self, Q, K, V, attn_mask):
        # Q: [batch_size x n_heads x len_q x d_k]
        # K: [batch_size x n_heads x len_k x d_k]
        # V: [batch_size x n_heads x len_k x d_v]
       
        scores = torch.matmul(Q, K.transpose(-1, -2)) / np.sqrt(d_k)

        # 把被mask的地方置为无限小-1e9
        scores.masked_fill_(attn_mask, -1e9)
        attn = nn.Softmax(dim=-1)(scores)
        context = torch.matmul(attn, V)
        return context, attn


class MultiHeadAttention(nn.Module):
    def __init__(self):
        super(MultiHeadAttention, self).__init__()
        # 使用映射linear做一个映射得到参数矩阵Wq, Wk,Wv
        self.W_Q = nn.Linear(d_model, d_k * n_heads)
        self.W_K = nn.Linear(d_model, d_k * n_heads)
        self.W_V = nn.Linear(d_model, d_v * n_heads)
        self.linear = nn.Linear(n_heads * d_v, d_model)
        self.layer_norm = nn.LayerNorm(d_model)

    def forward(self, Q, K, V, attn_mask):
        # Q, K, V分别输入而不是只输入一个，因decoder中有qkv不同的情况出现

        # Q: [batch_size x len_q x d_model]
        # K: [batch_size x len_k x d_model]
        # V: [batch_size x len_k x d_model]
        residual, batch_size = Q, Q.size(0)
        
        # 下面这个就是先映射，后分头；q和k分头之后维度一致
        q_s = self.W_Q(Q).view(batch_size, -1, n_heads, d_k).transpose(1,2)
        # q_s: [batch_size x n_heads x len_q x d_k]
        k_s = self.W_K(K).view(batch_size, -1, n_heads, d_k).transpose(1,2)
        # k_s: [batch_size x n_heads x len_k x d_k]
        v_s = self.W_V(V).view(batch_size, -1, n_heads, d_v).transpose(1,2)
        # v_s: [batch_size x n_heads x len_k x d_v]

        # 把pad信息重复了n个头上
        attn_mask = attn_mask.unsqueeze(1).repeat(1, n_heads, 1, 1)

        # context: [batch_size x n_heads x len_q x d_v]
        # attn: [batch_size x n_heads x len_q x len_k]
        context, attn = ScaledDotProductAttention()(q_s, k_s, v_s, attn_mask)
        context = context.transpose(1, 2).contiguous().view(batch_size, -1, n_heads * d_v)
        # context: [batch_size x len_q x n_heads * d_v]
        output = self.linear(context)
        return self.layer_norm(output + residual), attn
        # output: [batch_size x len_q x d_model]


class PoswiseFeedForwardNet(nn.Module):
    def __init__(self):
        super(PoswiseFeedForwardNet, self).__init__()
        self.conv1 = nn.Conv1d(in_channels=d_model, out_channels=d_ff, kernel_size=1)
        self.conv2 = nn.Conv1d(in_channels=d_ff, out_channels=d_model, kernel_size=1)
        self.layer_norm = nn.LayerNorm(d_model)

    def forward(self, inputs):
        residual = inputs 
        # inputs : [batch_size, len_q, d_model]
        output = nn.ReLU()(self.conv1(inputs.transpose(1, 2)))
        output = self.conv2(output).transpose(1, 2)
        return self.layer_norm(output + residual)


class PositionalEncoding(nn.Module):
    def __init__(self, d_model, dropout=0.1, max_len=5000):
        super(PositionalEncoding, self).__init__()
        self.dropout = nn.Dropout(p=dropout)

        pe = torch.zeros(max_len, d_model)
        position = torch.arange(0, max_len, dtype=torch.float).unsqueeze(1)
        div_term = torch.exp(torch.arange(0, d_model, 2).float() * (-math.log(10000.0) / d_model))
        # div_term = e^(-2i/d_model * log(10000))
        pe[:, 0::2] = torch.sin(position * div_term)
        pe[:, 1::2] = torch.cos(position * div_term)
        # 用列表索引切片步长为2取出奇数位和偶数位
        # pe: [max_len*d_model]

        pe = pe.unsqueeze(0).transpose(0, 1)
        # pe: [max_len*1*d_model]

        self.register_buffer('pe', pe)
        # 缓冲区，参数不更新就可以

    def forward(self, x):
        # x: [seq_len, batch_size, d_model]
        # x：word embedding
        x += self.pe[:x.size(0), :]
        return self.dropout(x)


class EncoderLayer(nn.Module):
    def __init__(self):
        super(EncoderLayer, self).__init__()
        self.enc_self_attn = MultiHeadAttention()
        self.pos_ffn = PoswiseFeedForwardNet()

    def forward(self, enc_inputs, enc_self_attn_mask):
        enc_outputs, attn = self.enc_self_attn(enc_inputs, enc_inputs, enc_inputs, enc_self_attn_mask)
        enc_outputs = self.pos_ffn(enc_outputs)
        # enc_outputs: [batch_size x len_q x d_model]
        return enc_outputs, attn


class Encoder(nn.Module):
    def __init__(self):
        super(Encoder, self).__init__()
        self.src_emb = nn.Embedding(src_vocab_size, d_model)
        # 生成一个src_vocab_size * d_model大小的矩阵 
        self.pos_emb = PositionalEncoding(d_model)
        # 位置编码
        self.layers = nn.ModuleList([EncoderLayer() for _ in range(n_layers)]) 
        # 使用ModuleList对多个encoder进行堆叠，因为后续的encoder并没有使用词向量和位置编码，所以抽离出来

    def forward(self, enc_inputs):
        # enc_inputs：[batch_size x source_len]
        # 下面这个代码通过src_emb，进行索引定位，enc_outputs输出形状是[batch_size, src_len, d_model]
        enc_outputs = self.src_emb(enc_inputs)
        enc_outputs = self.pos_emb(enc_outputs.transpose(0, 1)).transpose(0, 1)

        # get_attn_pad_mask是为了得到句子中pad的位置信息，给到模型后面，在计算自注意力和交互注意力的时候去掉pad符号的影响
        enc_self_attn_mask = get_attn_pad_mask(enc_inputs, enc_inputs)
        enc_self_attns = []
        
        # 实现六个block堆叠
        for layer in self.layers:
            enc_outputs, enc_self_attn = layer(enc_outputs, enc_self_attn_mask)
            enc_self_attns.append(enc_self_attn)
        return enc_outputs, enc_self_attns


class DecoderLayer(nn.Module):
    def __init__(self):
        super(DecoderLayer, self).__init__()
        self.dec_self_attn = MultiHeadAttention()
        self.dec_enc_attn = MultiHeadAttention()
        self.pos_ffn = PoswiseFeedForwardNet()

    def forward(self, dec_inputs, enc_outputs, dec_self_attn_mask, dec_enc_attn_mask):
        dec_outputs, dec_self_attn = self.dec_self_attn(dec_inputs, dec_inputs, dec_inputs, dec_self_attn_mask)
        dec_outputs, dec_enc_attn = self.dec_enc_attn(dec_outputs, enc_outputs, enc_outputs, dec_enc_attn_mask)
        dec_outputs = self.pos_ffn(dec_outputs)
        return dec_outputs, dec_self_attn, dec_enc_attn


class Decoder(nn.Module):
    def __init__(self):
        super(Decoder, self).__init__()
        self.tgt_emb = nn.Embedding(tgt_vocab_size, d_model)
        self.pos_emb = PositionalEncoding(d_model)
        self.layers = nn.ModuleList([DecoderLayer() for _ in range(n_layers)])

    def forward(self, dec_inputs, enc_inputs, enc_outputs): 
        # dec_inputs: [batch_size x target_len]
        
        dec_outputs = self.tgt_emb(dec_inputs)
        # [batch_size, tgt_len, d_model]
        dec_outputs = self.pos_emb(dec_outputs.transpose(0, 1)).transpose(0, 1)
        # [batch_size, tgt_len, d_model]

        # mask住pad部分
        dec_self_attn_pad_mask = get_attn_pad_mask(dec_inputs, dec_inputs)

        # mask住当前单词之后看不到的部分，实际为上三角为1的矩阵
        dec_self_attn_subsequent_mask = get_attn_subsequent_mask(dec_inputs)

        # 叠加两个mask的符号矩阵，分别为mask当前单词向后的部分和pad部分
        dec_self_attn_mask = torch.gt((dec_self_attn_pad_mask + dec_self_attn_subsequent_mask), 0)
        
        dec_enc_attn_mask = get_attn_pad_mask(dec_inputs, enc_inputs)

        dec_self_attns, dec_enc_attns = [], []
        for layer in self.layers:
            dec_outputs, dec_self_attn, dec_enc_attn = layer(dec_outputs, enc_outputs, dec_self_attn_mask, dec_enc_attn_mask)
            dec_self_attns.append(dec_self_attn)
            dec_enc_attns.append(dec_enc_attn)
        return dec_outputs, dec_self_attns, dec_enc_attns


class Transformer(nn.Module):
    def __init__(self):
        super(Transformer, self).__init__()
        self.encoder = Encoder()  # 编码层
        self.decoder = Decoder()  # 解码层
        self.projection = nn.Linear(d_model, tgt_vocab_size, bias=False) # 输出层
        # tgt_vocab_size：词表大小
        # 将输出结果放缩到tgt_vocab_size大小
    
    def forward(self, enc_inputs, dec_inputs):
        # enc_inputs: [batch_size, src_len]
        enc_outputs, enc_self_attns = self.encoder(enc_inputs)
        
        dec_outputs, dec_self_attns, dec_enc_attns = self.decoder(dec_inputs, enc_inputs, enc_outputs)

        # dec_outputs做映射到词表大小
        # dec_logits : [batch_size x src_vocab_size x tgt_vocab_size]
        dec_logits = self.projection(dec_outputs)
        return dec_logits.view(-1, dec_logits.size(-1)), enc_self_attns, dec_self_attns, dec_enc_attns



if __name__ == '__main__':

    ## 句子的输入部分，
    sentences = ['ich mochte ein bier P', 'S i want a beer', 'i want a beer E']
    #			encode input			decoder input		decoder output
    # P: pad	S: start	E: end

    
    # 构建词表
    src_vocab = {'P': 0, 'ich': 1, 'mochte': 2, 'ein': 3, 'bier': 4}
    src_vocab_size = len(src_vocab)

    tgt_vocab = {'P': 0, 'i': 1, 'want': 2, 'a': 3, 'beer': 4, 'S': 5, 'E': 6}
    tgt_vocab_size = len(tgt_vocab)

    src_len = 5 # length of source
    tgt_len = 5 # length of target

    # 模型参数
    d_model = 512  # Embedding Size
    d_ff = 2048  # FeedForward dimension
    d_k = d_v = 64  # dimension of K(=Q), V
    n_layers = 6  # number of Encoder of Decoder Layer
    n_heads = 8  # number of heads in Multi-Head Attention

    model = Transformer()

    criterion = nn.CrossEntropyLoss()
    optimizer = optim.Adam(model.parameters(), lr=0.001)

    enc_inputs, dec_inputs, target_batch = make_batch(sentences)

    for epoch in range(20):
        optimizer.zero_grad()
        outputs, enc_self_attns, dec_self_attns, dec_enc_attns = model(enc_inputs, dec_inputs)
        loss = criterion(outputs, target_batch.contiguous().view(-1))
        print('Epoch:', '%04d' % (epoch + 1), 'cost =', '{:.6f}'.format(loss))
        loss.backward()
        optimizer.step()