赞
踩
深度学习论文: Attention is All You Need及其PyTorch实现
Attention is All You Need
PDF:https://arxiv.org/abs/1706.03762.pdf
PyTorch: https://github.com/shanglianlm0525/PyTorch-Networks
大多数先进的神经序列转换模型采用编码器-解码器结构,其中编码器将输入符号序列转换为连续表示,解码器则基于这些表示逐个生成输出符号序列。在每个步骤中,模型采用自回归方式,将先前生成的符号作为额外输入来生成下一个符号。
编码器采用N=6个结构相同的层堆叠而成。每一层包含两个子层,第一个是多头自注意力机制,第二个则是简单的位置全连接前馈网络。为提升性能,在每个子层之间引入了残差连接,并实施了层归一化。具体来说,每个子层的输出经过LayerNorm(x + Sublayer(x))计算得出,其中Sublayer(x)代表子层自身的功能实现。为了确保残差连接的顺畅进行,编码器中的所有子层以及嵌入层均生成维度为dmodel=512的输出。
def clones(module, N): "Produce N identical layers." return nn.ModuleList([copy.deepcopy(module) for _ in range(N)]) class LayerNorm(nn.Module): "Construct a layernorm module (See citation for details)." def __init__(self, features, eps=1e-6): super(LayerNorm, self).__init__() self.a_2 = nn.Parameter(torch.ones(features)) self.b_2 = nn.Parameter(torch.zeros(features)) self.eps = eps def forward(self, x): mean = x.mean(-1, keepdim=True) std = x.std(-1, keepdim=True) return self.a_2 * (x - mean) / (std + self.eps) + self.b_2 class Encoder(nn.Module): "Core encoder is a stack of N layers" def __init__(self, layer, N): super(Encoder, self).__init__() self.layers = clones(layer, N) self.norm = LayerNorm(layer.size) def forward(self, x, mask): "Pass the input (and mask) through each layer in turn." for layer in self.layers: x = layer(x, mask) return self.norm(x) class SublayerConnection(nn.Module): """ A residual connection followed by a layer norm. Note for code simplicity the norm is first as opposed to last. """ def __init__(self, size, dropout): super(SublayerConnection, self).__init__() self.norm = LayerNorm(size) self.dropout = nn.Dropout(dropout) def forward(self, x, sublayer): "Apply residual connection to any sublayer with the same size." return x + self.dropout(sublayer(self.norm(x))) class EncoderLayer(nn.Module): "Encoder is made up of self-attn and feed forward (defined below)" def __init__(self, size, self_attn, feed_forward, dropout): super(EncoderLayer, self).__init__() self.self_attn = self_attn self.feed_forward = feed_forward self.sublayer = clones(SublayerConnection(size, dropout), 2) self.size = size def forward(self, x, mask): "Follow Figure 1 (left) for connections." x = self.sublayer[0](x, lambda x: self.self_attn(x, x, x, mask)) return self.sublayer[1](x, self.feed_forward)
解码器同样由N=6个结构相同的层堆叠而成。除了编码器层中的两个子层,解码器还额外引入了一个第三子层,专门用于对编码器堆叠的输出执行多头注意力机制。与编码器设计相似,解码器中的每个子层也采用残差连接与层归一化。此外,还对解码器堆叠中的自注意力子层进行了优化,确保在生成输出时,位置i的预测仅依赖于位置小于i的已知输出,这通过掩码和输出嵌入偏移一个位置的方式实现。
class Decoder(nn.Module): "Generic N layer decoder with masking." def __init__(self, layer, N): super(Decoder, self).__init__() self.layers = clones(layer, N) self.norm = LayerNorm(layer.size) def forward(self, x, memory, src_mask, tgt_mask): for layer in self.layers: x = layer(x, memory, src_mask, tgt_mask) return self.norm(x) class DecoderLayer(nn.Module): "Decoder is made of self-attn, src-attn, and feed forward (defined below)" def __init__(self, size, self_attn, src_attn, feed_forward, dropout): super(DecoderLayer, self).__init__() self.size = size self.self_attn = self_attn self.src_attn = src_attn self.feed_forward = feed_forward self.sublayer = clones(SublayerConnection(size, dropout), 3) def forward(self, x, memory, src_mask, tgt_mask): "Follow Figure 1 (right) for connections." m = memory x = self.sublayer[0](x, lambda x: self.self_attn(x, x, x, tgt_mask)) x = self.sublayer[1](x, lambda x: self.src_attn(x, m, m, src_mask)) return self.sublayer[2](x, self.feed_forward)
Transformer巧妙地运用了三种不同的多头注意力机制:
在“编码器-解码器注意力”层中,查询源于前一解码器层,而键和值则汲取自编码器的输出。这种设计使得解码器中的任意位置都能关注输入序列的每一个细节,完美模拟了序列到序列模型中典型的编码器-解码器注意力机制。
编码器内部则嵌入了自注意力层。在这个自注意力层中,键、值和查询均来自编码器前一层的输出,确保编码器中的每个位置都能全面捕捉到前一层中的信息。
解码器同样采用了自注意力层,允许解码器中的每个位置关注到包括该位置在内的解码器内部所有位置的信息。为了确保自回归属性的保持,我们精心设计了一个机制:在缩放点积注意力内部,将对应非法连接的softmax输入值设为−∞,从而有效屏蔽了向左的信息流动。
Scaled Dot-Product Attention 缩放点积注意力,输入包括维度为
d
k
d_{k}
dk的查询和键,以及维度为
d
v
d_{v}
dv的值。计算查询与所有键的点积,每个都除以
d
k
\sqrt{d_{k} }
dk
,然后应用softmax函数以获得值的权重。
def attention(query, key, value, mask=None, dropout=None):
"Compute 'Scaled Dot Product Attention'"
d_k = query.size(-1)
scores = torch.matmul(query, key.transpose(-2, -1)) \
/ math.sqrt(d_k)
if mask is not None:
scores = scores.masked_fill(mask == 0, -1e9)
p_attn = F.softmax(scores, dim = -1)
if dropout is not None:
p_attn = dropout(p_attn)
return torch.matmul(p_attn, value), p_attn
与其仅使用具有dmodel维度的键、值和查询来执行单一的注意力函数,将查询、键和值分别通过不同的、学习得到的线性投影,各自线性投影
h
h
h次,分别映射到
d
k
d_{k}
dk、
d
k
d_{k}
dk和
d
v
d_{v}
dv维度后效果更好。随后在这些投影后的查询、键和值上并行地执行注意力函数,从而得到
d
v
d_{v}
dv维度的输出值。最后,我们将这些输出值进行拼接,并再次通过投影得到最终的值。这种方法能够更充分地利用输入信息,提升模型的性能。
在这项研究中,我们采用了h=8个并行的注意力层,也称之为注意力头。对于每一个注意力头,我们都设定其维度为
d
k
d_{k}
dk =
d
v
d_{v}
dv =
d
m
o
d
e
l
/
h
d_{model} / h
dmodel/h=64。由于每个注意力头的维度有所降低,因此其整体计算成本与使用完整维度的单头注意力相近。
class MultiHeadedAttention(nn.Module): def __init__(self, h, d_model, dropout=0.1): "Take in model size and number of heads." super(MultiHeadedAttention, self).__init__() assert d_model % h == 0 # We assume d_v always equals d_k self.d_k = d_model // h self.h = h self.linears = clones(nn.Linear(d_model, d_model), 4) self.attn = None self.dropout = nn.Dropout(p=dropout) def forward(self, query, key, value, mask=None): "Implements Figure 2" if mask is not None: # Same mask applied to all h heads. mask = mask.unsqueeze(1) nbatches = query.size(0) # 1) Do all the linear projections in batch from d_model => h x d_k query, key, value = \ [l(x).view(nbatches, -1, self.h, self.d_k).transpose(1, 2) for l, x in zip(self.linears, (query, key, value))] # 2) Apply attention on all the projected vectors in batch. x, self.attn = attention(query, key, value, mask=mask, dropout=self.dropout) # 3) "Concat" using a view and apply a final linear. x = x.transpose(1, 2).contiguous() \ .view(nbatches, -1, self.h * self.d_k) return self.linears[-1](x)
除了注意力子层,编码器和解码器的每一层还包括一个全连接前馈网络,它独立且相同地应用于每个位置,包含两个线性变换和一个ReLU激活函数。
这里的输入和输出的维度是
d
m
o
d
e
l
=
512
d_{model}=512
dmodel=512,而内层的维度是
d
f
f
=
2048
d_{ff}=2048
dff=2048。
class PositionwiseFeedForward(nn.Module):
"Implements FFN equation."
def __init__(self, d_model, d_ff, dropout=0.1):
super(PositionwiseFeedForward, self).__init__()
self.w_1 = nn.Linear(d_model, d_ff)
self.w_2 = nn.Linear(d_ff, d_model)
self.dropout = nn.Dropout(dropout)
def forward(self, x):
return self.w_2(self.dropout(F.relu(self.w_1(x))))
与其他序列转换模型相似,这里使用学习嵌入将输入标记和输出标记转换为 d m o d e l d_{model} dmodel维向量。解码器输出通过线性变换和softmax函数转换为预测概率。在我们的模型中,嵌入层和softmax前的线性变换共享权重矩阵,并在嵌入层中将权重乘以 d m o d e l \sqrt{d_{model}} dmodel 。
class Embeddings(nn.Module):
def __init__(self, d_model, vocab):
super(Embeddings, self).__init__()
self.lut = nn.Embedding(vocab, d_model)
self.d_model = d_model
def forward(self, x):
return self.lut(x) * math.sqrt(self.d_model)
在编码器和解码器堆栈底部的输入嵌入中添加“位置编码”。位置编码的维度 d m o d e l d_{model} dmodel与嵌入的维度相同,以便可以将两者相加。位置编码有很多选择,包括学习和固定的。
在这里,选用不同频率的正弦和余弦函数:
其中pos是位置,i是维度。也就是说,位置编码的每个维度都对应于一个正弦波。波长从2π到10000·2π形成几何级数。选择这个函数是因为它可以使模型容易学习通过相对位置进行关注,因为对于任何固定的偏移量k,
P
E
p
o
s
+
k
PE_{pos+k}
PEpos+k都可以表示为
P
E
p
o
s
PE_{pos}
PEpos的线性函数。
此外,在编码器和解码器堆栈中都对嵌入和位置编码的和使用了Dropout。对于基础模型,使用的Dropout ratio 为 P d r o p P_{drop} Pdrop=0.1。
class PositionalEncoding(nn.Module): "Implement the PE function." def __init__(self, d_model, dropout, max_len=5000): super(PositionalEncoding, self).__init__() self.dropout = nn.Dropout(p=dropout) # Compute the positional encodings once in log space. pe = torch.zeros(max_len, d_model) position = torch.arange(0, max_len).unsqueeze(1) div_term = torch.exp(torch.arange(0, d_model, 2) * -(math.log(10000.0) / d_model)) pe[:, 0::2] = torch.sin(position * div_term) pe[:, 1::2] = torch.cos(position * div_term) pe = pe.unsqueeze(0) self.register_buffer('pe', pe) def forward(self, x): x = x + Variable(self.pe[:, :x.size(1)], requires_grad=False) return self.dropout(x)
参考资料:
1 Attention is All You Need
2 The Illustrated Transformer
3 The Annotated Transformer
Copyright © 2003-2013 www.wpsshop.cn 版权所有,并保留所有权利。