Transformer核心源码阅读笔记

一.参考链接

• Transformer理论解读：https://blog.csdn.net/ACM_hades/article/details/88898313
• 代码参考连接：https://github.com/graykode/nlp-tutorial

二.代码

import numpy as np
import torch
import torch.nn as nn
import torch.optim as optim
from torch.autograd import Variable
import matplotlib.pyplot as plt

dtype = torch.FloatTensor
# S: Symbol that shows starting of decoding input (decoder 输入开始符)
# E: Symbol that shows starting of decoding output (decoder 输出结束符)
# P: Symbol that will fill in blank sequence if current batch data size is short than time steps(填充符)
sentences = ['ich mochte ein bier P', 'S i want a beer', 'i want a beer E']

# Transformer Parameters
# Padding Should be Zero
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}
number_dict = {i: w for i, w in enumerate(tgt_vocab)} #字典生成式
tgt_vocab_size = len(tgt_vocab)

src_len = 5
tgt_len = 5

d_model = 512  # Embedding Size
d_ff = 2048 # FeedForward dimension
d_k = d_v = 64  # dimension of K(=Q), V（k，q，v向量维度一致）
n_layers = 6  # number of Encoder of Decoder Layer
n_heads = 8  # number of heads in Multi-Head Attention

#将句子list转变为Variable()
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 Variable(torch.LongTensor(input_batch)), Variable(torch.LongTensor(output_batch)), Variable(torch.LongTensor(target_batch))

#生成不可训练位置embeding矩阵。
def get_sinusoid_encoding_table(n_position, d_model):
    """
    :param n_position: 句子最大长度
    :param d_model: 模型维度
    :return:位置embeding矩阵
    """
    def cal_angle(position, hid_idx):
        return position / np.power(10000, 2 * (hid_idx // 2) / d_model)
    def get_posi_angle_vec(position):#生成当前为position的位置向量，维度为d_model
        return [cal_angle(position, hid_j) for hid_j in range(d_model)]

    sinusoid_table = np.array([get_posi_angle_vec(pos_i) for pos_i in range(n_position)])
    sinusoid_table[:, 0::2] = np.sin(sinusoid_table[:, 0::2])  # dim 2i
    sinusoid_table[:, 1::2] = np.cos(sinusoid_table[:, 1::2])  # dim 2i+1
    return torch.FloatTensor(sinusoid_table)

#将被询问的词(seq_k)中的填充词mask掉
#使得seq_q中的词不在询问seq_k中填充词
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
    #seq_k=[batch_size,len_k],判断与0是否相等，再在维度1上增加一个维度
    pad_attn_mask = seq_k.data.eq(0).unsqueeze(1)  # batch_size x 1 x len_k(=len_q), one is masking
    #在维度1上扩展len_q份
    return pad_attn_mask.expand(batch_size, len_q, len_k)  # batch_size x len_q x len_k

#生成以后上三角的mask矩阵
def get_attn_subsequent_mask(seq):
    attn_shape = [seq.size(0), seq.size(1), seq.size(1)]
    subsequent_mask = np.triu(np.ones(attn_shape), k=1)#上三角矩阵
    subsequent_mask = torch.from_numpy(subsequent_mask).byte()
    return subsequent_mask


#根据Q, K, V, attn_mask矩阵生成上下文矩阵context
class ScaledDotProductAttention(nn.Module):
    def __init__(self):
        super(ScaledDotProductAttention, self).__init__()

    def forward(self, Q, K, V, attn_mask):
        scores = torch.matmul(Q, K.transpose(-1, -2)) / np.sqrt(d_k) # scores : [batch_size x n_heads x len_q(=len_k) x len_k(=len_q)]
        #将需要mask掉的位置的分数令为负无穷，这样经过softmax后概率变为0
        scores.masked_fill_(attn_mask, -1e9) # Fills elements of self tensor with value where mask is one.
        attn = nn.Softmax(dim=-1)(scores)
        context = torch.matmul(attn, V)
        return context, attn

#输入：Q=K=V=[batch_size,len,d_model], attn_mask 矩阵输出 [batch_size,len,d_model]
class MultiHeadAttention(nn.Module):
    def __init__(self):
        super(MultiHeadAttention, self).__init__()
        self.W_Q = nn.Linear(d_model, d_k * n_heads)#定义Q，K，V向量的线性映射
        self.W_K = nn.Linear(d_model, d_k * n_heads)
        self.W_V = nn.Linear(d_model, d_v * n_heads)

    def forward(self, Q, K, V, attn_mask):
        # 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)
        # (B, S, D) -proj-> (B, S, D) -split-> (B, S, H, W) -trans-> (B, H, S, W)
        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]
        #在head所在维度扩展mask矩阵
        attn_mask = attn_mask.unsqueeze(1).repeat(1, n_heads, 1, 1) # attn_mask : [batch_size x n_heads x len_q x len_k]

        # context: [batch_size x n_heads x len_q x d_v], attn: [batch_size x n_heads x len_q(=len_k) x len_k(=len_q)]
        context, attn = ScaledDotProductAttention()(q_s, k_s, v_s, attn_mask)
        #将多个head上是上下文向量进行拼接
        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]
        #再经过一个线性映射将上下文向量维度恢复到d_model
        output = nn.Linear(n_heads * d_v, d_model)(context)
        #残差连接+正则化
        return nn.LayerNorm(d_model)(output + residual), attn # output: [batch_size x len_q x d_model]
#实现：FFN+Add+normalization
class PoswiseFeedForwardNet(nn.Module):
    def __init__(self):
        super(PoswiseFeedForwardNet, self).__init__()
        #[bact_size,d_model,len_q]——>[bact_size,d_ff,len_q]——>[bact_size,d_model,len_q]
        #kernel_size=1相当与中间维度进行了一个线性映射
        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)

    def forward(self, inputs):
        residual = inputs # inputs : [batch_size, len_q, d_model]
        output = nn.ReLU()(self.conv1(inputs.transpose(1, 2)))#先转置
        output = self.conv2(output).transpose(1, 2)#再转回来
        return nn.LayerNorm(d_model)(output + residual)#残差连接与正则化

class EncoderLayer(nn.Module):
    def __init__(self):
        super(EncoderLayer, self).__init__()
        self.enc_self_attn = MultiHeadAttention()#定义多头self-attention子层
        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_inputs to same Q,K,V
        enc_outputs = self.pos_ffn(enc_outputs) # enc_outputs: [batch_size x len_q x d_model]
        return enc_outputs, attn

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 Encoder(nn.Module):
    def __init__(self):
        super(Encoder, self).__init__()#调用父类的构造函数。
        self.src_emb = nn.Embedding(src_vocab_size, d_model) #实现word embedding
        #创建位置embeding，并设置其为不可训练的。
        self.pos_emb = nn.Embedding.from_pretrained(get_sinusoid_encoding_table(src_len+1, d_model),freeze=True)
        self.layers = nn.ModuleList([EncoderLayer() for _ in range(n_layers)])

    def forward(self, enc_inputs): # enc_inputs : [batch_size , source_len]
        enc_outputs = self.src_emb(enc_inputs) + self.pos_emb(torch.LongTensor([[1,2,3,4,0]]))#[batch_size,q_len,d_model]
        enc_self_attn_mask = get_attn_pad_mask(enc_inputs, enc_inputs)#生成mask矩阵 batch_size x len x len
        enc_self_attns = []
        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 Decoder(nn.Module):
    def __init__(self):
        super(Decoder, self).__init__()
        self.tgt_emb = nn.Embedding(tgt_vocab_size, d_model)
        self.pos_emb = nn.Embedding.from_pretrained(get_sinusoid_encoding_table(tgt_len+1, d_model),freeze=True)
        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) + self.pos_emb(torch.LongTensor([[5,1,2,3,4]]))
        dec_self_attn_pad_mask = get_attn_pad_mask(dec_inputs, dec_inputs)#pad的mask
        dec_self_attn_subsequent_mask = get_attn_subsequent_mask(dec_inputs)#翻译时后面词的mask
        #先两个mask矩阵相加再逐元素与0比较大于0为：1；反之为：0
        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)#在encoder-decoder attention子层中使用的mask矩阵。

        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()
        self.decoder = Decoder()
        self.projection = nn.Linear(d_model, tgt_vocab_size, bias=False)#将decoder的输出映射得到logits score
    def forward(self, enc_inputs, dec_inputs):
        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_logits = self.projection(dec_outputs) # dec_logits : [batch_size x src_vocab_size x tgt_vocab_size]
        return dec_logits.view(-1, dec_logits.size(-1)), enc_self_attns, dec_self_attns, dec_enc_attns

model = Transformer()

criterion = nn.CrossEntropyLoss()#交叉熵损失函数
optimizer = optim.Adam(model.parameters(), lr=0.001)


for epoch in range(20):
    optimizer.zero_grad()
    enc_inputs, dec_inputs, target_batch = make_batch(sentences)#句子转化为数字batch
    outputs, enc_self_attns, dec_self_attns, dec_enc_attns = model(enc_inputs, dec_inputs)
    loss = criterion(outputs, target_batch.contiguous().view(-1))#交叉熵损失
    print('Epoch:', '%04d' % (epoch + 1), 'cost =', '{:.6f}'.format(loss))
    loss.backward()
    optimizer.step()

# Test
predict, _, _, _ = model(enc_inputs, dec_inputs)
predict = predict.data.max(1, keepdim=True)[1]
print(sentences[0], '->', [number_dict[n.item()] for n in predict.squeeze()])