Provided proper attribution is provided, Google hereby grants permission to reproduce the tables and figures in this paper solely for use in journalistic or scholarly works. 只要注明出处，Google 特此允许复制本文中的表格和数字，仅供新闻或学术著作使用。
The dominant sequence transduction models are based on complex recurrent or convolutional neural networks that include an encoder and a decoder. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 Englishto-German translation task, improving over the existing best results, including ensembles, by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data. 主流的序列转换模型基于复杂的递归或卷积神经网络，其中包括一个编码器和一个解码器。性能最好的模型还通过注意力机制连接编码器和解码器。我们提出了一种新的简单网络架构--"转换器"（Transformer），它完全基于注意力机制，无需递归和卷积。在两项机器翻译任务上的实验表明，这些模型的质量更优，同时可并行化程度更高，所需的训练时间也大大减少。在 WMT 2014 英德翻译任务中，我们的模型达到了 28.4 BLEU，比包括集合在内的现有最佳结果提高了 2 BLEU 以上。在 WMT 2014 英法翻译任务中，我们的模型在 8 个 GPU 上进行了 3.5 天的训练后，取得了 41.8 分的单模型最新 BLEU 分数，这只是文献中最佳模型训练成本的一小部分。我们将 Transformer 成功地应用于英语选区解析，并同时使用大量和有限的训练数据，从而证明 Transformer 可以很好地推广到其他任务中。
1 Introduction 1 引言
Recurrent neural networks, long short-term memory [13] and gated recurrent [7] neural networks in particular, have been firmly established as state of the art approaches in sequence modeling and transduction problems such as language modeling and machine translation [35, 2, 5]. Numerous efforts have since continued to push the boundaries of recurrent language models and encoder-decoder architectures [38, 24, 15]. 递归神经网络，尤其是长短期记忆[13]和门控递归[7]神经网络，已被牢固确立为语言建模和机器翻译等序列建模和转译问题的最先进方法[35, 2, 5]。自此以后，许多人继续努力推动递归语言模型和编码器-解码器架构的发展[38, 24, 15]。
Recurrent models typically factor computation along the symbol positions of the input and output sequences. Aligning the positions to steps in computation time, they generate a sequence of hidden states , as a function of the previous hidden state and the input for position . This inherently sequential nature precludes parallelization within training examples, which becomes critical at longer sequence lengths, as memory constraints limit batching across examples. Recent work has achieved significant improvements in computational efficiency through factorization tricks [21] and conditional computation [32], while also improving model performance in case of the latter. The fundamental constraint of sequential computation, however, remains. 递归模型通常沿着输入和输出序列的符号位置进行计算。它们将位置与计算时间的步长对齐，生成隐藏状态序列 ，作为前一个隐藏状态 和位置 的输入的函数。这种固有的顺序性质排除了训练实例内的并行化，而在序列长度较长的情况下，这一点变得至关重要，因为内存约束限制了跨实例的批处理。最近的工作通过因式分解技巧[21]和条件计算[32]显著提高了计算效率，同时也改善了后者的模型性能。然而，顺序计算的基本限制仍然存在。
Attention mechanisms have become an integral part of compelling sequence modeling and transduction models in various tasks, allowing modeling of dependencies without regard to their distance in the input or output sequences [2, 19]. In all but a few cases [27], however, such attention mechanisms are used in conjunction with a recurrent network. 在各种任务中，注意机制已成为引人注目的序列建模和转导模型的一个组成部分，它可以对依赖关系进行建模，而不必考虑它们在输入或输出序列中的距离[2, 19]。然而，除了少数情况[27]，这种注意机制都是与递归网络结合使用的。
In this work we propose the Transformer, a model architecture eschewing recurrence and instead relying entirely on an attention mechanism to draw global dependencies between input and output. The Transformer allows for significantly more parallelization and can reach a new state of the art in translation quality after being trained for as little as twelve hours on eight P100 GPUs. 在这项工作中，我们提出了 Transformer 模型架构，它摒弃了递归，而是完全依赖注意力机制来绘制输入和输出之间的全局依赖关系。Transformer 可以大大提高并行化程度，在 8 个 P100 GPU 上只需训练 12 个小时，翻译质量就能达到新的水平。
2 Background 2 背景
The goal of reducing sequential computation also forms the foundation of the Extended Neural GPU [16], ByteNet [18] and ConvS2S [9], all of which use convolutional neural networks as basic building block, computing hidden representations in parallel for all input and output positions. In these models, the number of operations required to relate signals from two arbitrary input or output positions grows in the distance between positions, linearly for ConvS2S and logarithmically for ByteNet. This makes it more difficult to learn dependencies between distant positions [12]. In the Transformer this is reduced to a constant number of operations, albeit at the cost of reduced effective resolution due to averaging attention-weighted positions, an effect we counteract with Multi-Head Attention as described in section 3.2 减少顺序计算的目标也是扩展神经 GPU[16]、ByteNet[18]和 ConvS2S[9]的基础，它们都使用卷积神经网络作为基本构建模块，并行计算所有输入和输出位置的隐藏表示。在这些模型中，将来自两个任意输入或输出位置的信号联系起来所需的运算次数随位置间距离的增加而增加，ConvS2S 是线性增加，ByteNet 是对数增加。这就增加了学习远距离位置之间依赖关系的难度[12]。在 Transformer 中，这将被减少到一个恒定的操作数，尽管代价是由于平均注意力加权位置而降低了有效分辨率。
Self-attention, sometimes called intra-attention is an attention mechanism relating different positions of a single sequence in order to compute a representation of the sequence. Self-attention has been used successfully in a variety of tasks including reading comprehension, abstractive summarization, textual entailment and learning task-independent sentence representations [4, 27, 28, 22]. 自我注意（有时也称为内部注意）是一种注意机制，它将单个序列的不同位置联系起来，以计算序列的表征。自我注意已成功应用于多种任务中，包括阅读理解、抽象概括、文本引申和学习与任务无关的句子表征 [4,27,28,22]。
End-to-end memory networks are based on a recurrent attention mechanism instead of sequencealigned recurrence and have been shown to perform well on simple-language question answering and language modeling tasks [34]. 端到端记忆网络基于递归注意机制，而不是序列对齐递归，在简单语言问题解答和语言建模任务中表现出色 [34]。
To the best of our knowledge, however, the Transformer is the first transduction model relying entirely on self-attention to compute representations of its input and output without using sequencealigned RNNs or convolution. In the following sections, we will describe the Transformer, motivate self-attention and discuss its advantages over models such as [17, 18] and [9]. 然而，据我们所知，Transformer 是第一个完全依靠自我注意来计算输入和输出表示而不使用序列对齐 RNN 或卷积的转导模型。在下面的章节中，我们将描述 Transformer，激发自我注意，并讨论它与 [17, 18] 和 [9] 等模型相比的优势。
3 Model Architecture 3 模型结构
Most competitive neural sequence transduction models have an encoder-decoder structure [5, 2, 35]. Here, the encoder maps an input sequence of symbol representations to a sequence of continuous representations . Given , the decoder then generates an output sequence of symbols one element at a time. At each step the model is auto-regressive [10], consuming the previously generated symbols as additional input when generating the next. 大多数竞争性神经序列转换模型都具有编码器-解码器结构 [5, 2, 35]。在这里，编码器将输入的符号表示序列 映射到连续表示序列 。给定 后，解码器每次生成一个元素符号的输出序列 。在每一步中，该模型都是自动回归的 [10]，在生成下一步时，会消耗之前生成的符号作为额外输入。
Figure 1: The Transformer - model architecture. 图 1：变压器--模型结构。
The Transformer follows this overall architecture using stacked self-attention and point-wise, fully connected layers for both the encoder and decoder, shown in the left and right halves of Figure 1 , respectively. 图 1 的左半部分和右半部分分别显示了编码器和解码器采用堆叠式自关注和点式全连接层的整体架构。
3.1 Encoder and Decoder Stacks 3.1 编码器和解码器堆栈
Encoder: The encoder is composed of a stack of identical layers. Each layer has two sub-layers. The first is a multi-head self-attention mechanism, and the second is a simple, positionwise fully connected feed-forward network. We employ a residual connection [11] around each of the two sub-layers, followed by layer normalization [1]. That is, the output of each sub-layer is LayerNorm , where Sublayer is the function implemented by the sub-layer itself. To facilitate these residual connections, all sub-layers in the model, as well as the embedding layers, produce outputs of dimension . 编码器：编码器由 相同的层堆叠组成。每一层都有两个子层。第一个是多头自注意机制，第二个是简单的位置全连接前馈网络。我们在两个子层的每个周围都采用了残差连接[11]，然后进行层归一化[1]。也就是说，每个子层的输出都是 LayerNorm ，其中子层 是子层本身实现的函数。为了方便这些残差连接，模型中的所有子层以及嵌入层都会产生维数为 的输出。
Decoder: The decoder is also composed of a stack of identical layers. In addition to the two sub-layers in each encoder layer, the decoder inserts a third sub-layer, which performs multi-head attention over the output of the encoder stack. Similar to the encoder, we employ residual connections around each of the sub-layers, followed by layer normalization. We also modify the self-attention sub-layer in the decoder stack to prevent positions from attending to subsequent positions. This masking, combined with fact that the output embeddings are offset by one position, ensures that the predictions for position can depend only on the known outputs at positions less than . 解码器：解码器也由 相同的层堆叠组成。除了每个编码器层中的两个子层外，解码器还插入了第三个子层，对编码器堆栈的输出执行多头关注。与编码器类似，我们在每个子层周围采用残差连接，然后进行层归一化。我们还修改了解码器堆栈中的自我关注子层，以防止位置关注到后续位置。这种屏蔽，再加上输出嵌入偏移一个位置的事实，确保了对位置 的预测只能依赖于小于 位置的已知输出。
3.2 Attention 3.2 注意事项
An attention function can be described as mapping a query and a set of key-value pairs to an output, where the query, keys, values, and output are all vectors. The output is computed as a weighted sum 注意力函数可以描述为将一个查询和一组键值对映射到一个输出，其中查询、键、值和输出都是向量。输出计算为加权和
Figure 2: (left) Scaled Dot-Product Attention. (right) Multi-Head Attention consists of several attention layers running in parallel. 图 2：（左）缩放点积注意。(右图）多头注意由多个并行运行的注意层组成。
of the values, where the weight assigned to each value is computed by a compatibility function of the query with the corresponding key. 的权重，其中分配给每个值的权重是通过查询与相应密钥的兼容性函数计算得出的。
3.2.1 Scaled Dot-Product Attention 3.2.1 按比例点积注意
We call our particular attention "Scaled Dot-Product Attention" (Figure 2). The input consists of queries and keys of dimension , and values of dimension . We compute the dot products of the query with all keys, divide each by , and apply a softmax function to obtain the weights on the values. 我们将这种特殊的注意力称为 "缩放点积注意力"（图 2）。输入包括查询和维度为 的键，以及维度为 的值。我们计算查询与所有密钥的点积，将每个点积除以 ，然后应用软最大函数获得值的权重。
In practice, we compute the attention function on a set of queries simultaneously, packed together into a matrix . The keys and values are also packed together into matrices and . We compute the matrix of outputs as: 在实践中，我们会同时计算一组查询的注意力函数，并将其打包到一个矩阵 中。键和值也被打包到矩阵 和 中。我们计算的输出矩阵为
The two most commonly used attention functions are additive attention [2], and dot-product (multiplicative) attention. Dot-product attention is identical to our algorithm, except for the scaling factor of . Additive attention computes the compatibility function using a feed-forward network with a single hidden layer. While the two are similar in theoretical complexity, dot-product attention is much faster and more space-efficient in practice, since it can be implemented using highly optimized matrix multiplication code. 最常用的两种注意力函数是加法注意力[2]和点积（乘法）注意力。点积注意力与我们的算法相同，只是缩放因子为 。加法注意使用单隐层前馈网络计算相容函数。虽然两者在理论复杂度上相似，但点积注意力在实践中速度更快，空间效率更高，因为它可以使用高度优化的矩阵乘法代码来实现。
While for small values of the two mechanisms perform similarly, additive attention outperforms dot product attention without scaling for larger values of [3]. We suspect that for large values of , the dot products grow large in magnitude, pushing the softmax function into regions where it has extremely small gradients To counteract this effect, we scale the dot products by . 对于 的较小值，这两种机制的表现类似，但对于 [3]，加法注意在不缩放的情况下优于点积注意。我们怀疑，对于 的较大值，点积的幅度会变大，从而将软最大值函数推向梯度极小的区域 为了抵消这种影响，我们通过 对点积进行缩放。
3.2.2 Multi-Head Attention 3.2.2 多头关注
Instead of performing a single attention function with dimensional keys, values and queries, we found it beneficial to linearly project the queries, keys and values times with different, learned linear projections to and dimensions, respectively. On each of these projected versions of queries, keys and values we then perform the attention function in parallel, yielding -dimensional output values. These are concatenated and once again projected, resulting in the final values, as depicted in Figure 2 我们发现，与其使用 维度的键、值和查询来执行单个注意力函数，不如将查询、键和值 ，分别线性投影到 和 维度的不同学习线性投影上。然后，我们在每个查询、键和值的投影版本上并行执行注意力函数，得到 维的输出值。如图 2 所示，这些值被串联起来并再次投影，从而得到最终值
Multi-head attention allows the model to jointly attend to information from different representation subspaces at different positions. With a single attention head, averaging inhibits this. 多头注意力允许模型在不同位置共同关注来自不同表征子空间的信息。而在单注意头的情况下，平均化会抑制这一点。
Where the projections are parameter matrices and . 其中，投影是参数矩阵 和 。
In this work we employ parallel attention layers, or heads. For each of these we use . Due to the reduced dimension of each head, the total computational cost is similar to that of single-head attention with full dimensionality. 在这项工作中，我们采用了 并行注意力层或头。我们使用 。由于减少了每个头的维度，总计算成本与全维度的单头注意力相似。
3.2.3 Applications of Attention in our Model 3.2.3 注意力在模型中的应用
The Transformer uses multi-head attention in three different ways: 变形金刚通过三种不同方式实现多头关注：
In "encoder-decoder attention" layers, the queries come from the previous decoder layer, and the memory keys and values come from the output of the encoder. This allows every position in the decoder to attend over all positions in the input sequence. This mimics the typical encoder-decoder attention mechanisms in sequence-to-sequence models such as [38, 2, 9]. 在 "编码器-解码器注意 "层中，查询来自前一个解码器层，而记忆键和记忆值则来自编码器的输出。这使得解码器中的每个位置都能关注输入序列中的所有位置。这模仿了序列到序列模型（如 [38, 2, 9]）中典型的编码器-解码器注意机制。
The encoder contains self-attention layers. In a self-attention layer all of the keys, values and queries come from the same place, in this case, the output of the previous layer in the encoder. Each position in the encoder can attend to all positions in the previous layer of the encoder. 编码器包含自注意层。在自注意层中，所有的键、值和查询都来自同一个地方，在这种情况下，就是编码器中上一层的输出。编码器中的每个位置都可以关注编码器上一层的所有位置。
Similarly, self-attention layers in the decoder allow each position in the decoder to attend to all positions in the decoder up to and including that position. We need to prevent leftward information flow in the decoder to preserve the auto-regressive property. We implement this inside of scaled dot-product attention by masking out (setting to ) all values in the input of the softmax which correspond to illegal connections. See Figure 2 同样，解码器中的自关注层允许解码器中的每个位置关注解码器中包括该位置在内的所有位置。我们需要防止解码器中的信息向左流动，以保持自动回归特性。我们通过屏蔽（设置为 ）软极大值输入中所有与非法连接相对应的值，在缩放点积关注中实现这一点。参见图 2
In addition to attention sub-layers, each of the layers in our encoder and decoder contains a fully connected feed-forward network, which is applied to each position separately and identically. This consists of two linear transformations with a ReLU activation in between. 除了注意力子层外，我们的编码器和解码器中的每一层都包含一个全连接的前馈网络，该网络分别对每个位置进行相同的处理。这包括两个线性变换，中间有一个 ReLU 激活。
While the linear transformations are the same across different positions, they use different parameters from layer to layer. Another way of describing this is as two convolutions with kernel size 1 . The dimensionality of input and output is , and the inner-layer has dimensionality . 虽然不同位置的线性变换相同，但各层使用的参数不同。另一种描述方法是两个核大小为 1 的卷积。输入和输出的维度为 ，内层的维度为 。
3.4 Embeddings and Softmax 3.4 嵌入与 Softmax
Similarly to other sequence transduction models, we use learned embeddings to convert the input tokens and output tokens to vectors of dimension . We also use the usual learned linear transformation and softmax function to convert the decoder output to predicted next-token probabilities. In our model, we share the same weight matrix between the two embedding layers and the pre-softmax linear transformation, similar to [30]. In the embedding layers, we multiply those weights by . 与其他序列转换模型类似，我们使用学习到的嵌入将输入标记和输出标记转换成维数为 的向量。我们还使用通常的学习线性变换和 softmax 函数将解码器输出转换为预测的下一个标记概率。在我们的模型中，我们在两个嵌入层和预软最大线性变换之间共享相同的权重矩阵，这与 [30] 相似。在嵌入层中，我们将这些权重乘以 。
Table 1: Maximum path lengths, per-layer complexity and minimum number of sequential operations for different layer types. is the sequence length, is the representation dimension, is the kernel size of convolutions and the size of the neighborhood in restricted self-attention. 是序列长度， 是表示维度， 是卷积的内核大小， 是受限自注意的邻域大小。
Layer Type 层类型
Complexity per Layer 每层的复杂性
顺序操作
Sequential
Operations
Maximum Path Length 最大路径长度
Self-Attention 自我关注
Recurrent
Convolutional 卷积
Self-Attention (restricted) 自我关注（受限）
3.5 Positional Encoding 3.5 位置编码
Since our model contains no recurrence and no convolution, in order for the model to make use of the order of the sequence, we must inject some information about the relative or absolute position of the tokens in the sequence. To this end, we add "positional encodings" to the input embeddings at the bottoms of the encoder and decoder stacks. The positional encodings have the same dimension as the embeddings, so that the two can be summed. There are many choices of positional encodings, learned and fixed [9]. 由于我们的模型不包含递归和卷积，为了让模型能够利用序列的顺序，我们必须注入一些关于序列中标记的相对或绝对位置的信息。为此，我们在编码器和解码器堆栈底部的输入嵌入中添加了 "位置编码"。位置编码的维度 与嵌入式编码相同，因此两者可以相加。位置编码有多种选择，包括学习编码和固定编码 [9]。
In this work, we use sine and cosine functions of different frequencies: 在这项工作中，我们使用了不同频率的正弦和余弦函数：
where pos is the position and is the dimension. That is, each dimension of the positional encoding corresponds to a sinusoid. The wavelengths form a geometric progression from to . We chose this function because we hypothesized it would allow the model to easily learn to attend by relative positions, since for any fixed offset can be represented as a linear function of . 其中，pos 是位置， 是维度。也就是说，位置编码的每个维度都对应一个正弦波。波长形成一个从 到 的几何级数。我们之所以选择这个函数，是因为我们假设它可以让模型轻松学会按相对位置进行关注，因为对于任何固定偏移量， 都可以表示为 的线性函数。
We also experimented with using learned positional embeddings [9] instead, and found that the two versions produced nearly identical results (see Table 3 row (E)). We chose the sinusoidal version because it may allow the model to extrapolate to sequence lengths longer than the ones encountered during training. 我们还尝试使用学习到的位置嵌入[9]来代替，结果发现两个版本产生的结果几乎相同（见表 3 第（E）行）。我们之所以选择正弦波版本，是因为它可以让模型推断出比训练时遇到的序列长度更长的序列。
4 Why Self-Attention 4 为什么要自我关注
In this section we compare various aspects of self-attention layers to the recurrent and convolutional layers commonly used for mapping one variable-length sequence of symbol representations to another sequence of equal length , with , such as a hidden layer in a typical sequence transduction encoder or decoder. Motivating our use of self-attention we consider three desiderata. 在本节中，我们将比较自注意层与递归层和卷积层的各个方面，前者通常用于将一个可变长度的符号表示序列 映射到另一个等长序列 ，后者则是 ，如典型序列转换编码器或解码器中的隐藏层。我们使用自注意的动机有三个。
One is the total computational complexity per layer. Another is the amount of computation that can be parallelized, as measured by the minimum number of sequential operations required. 一个是每层的总计算复杂度。另一个是可并行化的计算量，以所需的最小顺序运算次数来衡量。
The third is the path length between long-range dependencies in the network. Learning long-range dependencies is a key challenge in many sequence transduction tasks. One key factor affecting the ability to learn such dependencies is the length of the paths forward and backward signals have to traverse in the network. The shorter these paths between any combination of positions in the input and output sequences, the easier it is to learn long-range dependencies [12]. Hence we also compare the maximum path length between any two input and output positions in networks composed of the different layer types. 第三是网络中长距离依赖关系之间的路径长度。学习长程依赖关系是许多序列转导任务的关键挑战。影响学习此类依赖关系能力的一个关键因素是前向和后向信号在网络中必须穿越的路径长度。输入和输出序列中任意位置组合之间的路径越短，学习远距离依赖关系就越容易[12]。因此，我们还比较了由不同层类型组成的网络中任意两个输入和输出位置之间的最大路径长度。
As noted in Table 1. a self-attention layer connects all positions with a constant number of sequentially executed operations, whereas a recurrent layer requires sequential operations. In terms of computational complexity, self-attention layers are faster than recurrent layers when the sequence 如表 1 所示，自注意层连接所有位置的连续操作数不变，而递归层需要 连续操作。就计算复杂度而言，当序列为"...... "时，自注意层比递归层更快。
length is smaller than the representation dimensionality , which is most often the case with sentence representations used by state-of-the-art models in machine translations, such as word-piece [38] and byte-pair [31] representations. To improve computational performance for tasks involving very long sequences, self-attention could be restricted to considering only a neighborhood of size in the input sequence centered around the respective output position. This would increase the maximum path length to . We plan to investigate this approach further in future work. 长度 小于表示维度 ，这通常是机器翻译中最先进的模型所使用的句子表示，如单词片[38]和字节对[31]表示。为了提高涉及超长序列任务的计算性能，可以限制自我关注只考虑输入序列中以各自输出位置为中心的大小为 的邻域。这将把最大路径长度增加到 。我们计划在今后的工作中进一步研究这种方法。
A single convolutional layer with kernel width does not connect all pairs of input and output positions. Doing so requires a stack of convolutional layers in the case of contiguous kernels, or in the case of dilated convolutions [18], increasing the length of the longest paths between any two positions in the network. Convolutional layers are generally more expensive than recurrent layers, by a factor of . Separable convolutions [6], however, decrease the complexity considerably, to . Even with , however, the complexity of a separable convolution is equal to the combination of a self-attention layer and a point-wise feed-forward layer, the approach we take in our model. 内核宽度为 的单一卷积层无法连接所有输入和输出位置对。如果采用连续内核，则需要堆叠 卷积层；如果采用扩张卷积，则需要堆叠 [18]，从而增加了网络中任意两个位置之间最长路径的长度。卷积层通常比递归层昂贵，系数为 。不过，可分离卷积[6]可将复杂度大大降低到 。不过，即使有 ，可分离卷积的复杂度也相当于自注意层和点式前馈层的组合，而我们在模型中采用的正是这种方法。
As side benefit, self-attention could yield more interpretable models. We inspect attention distributions from our models and present and discuss examples in the appendix. Not only do individual attention heads clearly learn to perform different tasks, many appear to exhibit behavior related to the syntactic and semantic structure of the sentences. 作为附带的好处，自我关注可以产生更多可解释的模型。我们从我们的模型中检查了注意力分布，并在附录中介绍和讨论了一些例子。各个注意力头不仅明显学会执行不同的任务，而且许多注意力头似乎还表现出与句子的句法和语义结构有关的行为。
5 Training 5 培训
This section describes the training regime for our models. 本节将介绍我们模型的训练机制。
5.1 Training Data and Batching 5.1 训练数据和批处理
We trained on the standard WMT 2014 English-German dataset consisting of about 4.5 million sentence pairs. Sentences were encoded using byte-pair encoding [3], which has a shared sourcetarget vocabulary of about 37000 tokens. For English-French, we used the significantly larger WMT 2014 English-French dataset consisting of 36M sentences and split tokens into a 32000 word-piece vocabulary [38]. Sentence pairs were batched together by approximate sequence length. Each training batch contained a set of sentence pairs containing approximately 25000 source tokens and 25000 target tokens. 我们在 WMT 2014 英德标准数据集上进行了训练，该数据集包含约 450 万个句子对。句子使用字节对编码[3]进行编码，其中有大约 37000 个共享的源目标词汇。对于英语-法语，我们使用了规模更大的 WMT 2014 英语-法语数据集，该数据集包含 3600 万个句子，并将标记拆分为 32000 个词块词汇[38]。句子对按近似序列长度分组。每个训练批包含一组句对，其中包含约 25000 个源词块和 25000 个目标词块。
5.2 Hardware and Schedule 5.2 硬件和时间表
We trained our models on one machine with 8 NVIDIA P100 GPUs. For our base models using the hyperparameters described throughout the paper, each training step took about 0.4 seconds. We trained the base models for a total of 100,000 steps or 12 hours. For our big models,(described on the bottom line of table 3), step time was 1.0 seconds. The big models were trained for 300,000 steps (3.5 days). 我们在一台配备 8 个英伟达 P100 GPU 的机器上训练模型。对于我们的基础模型，使用本文所述的超参数，每个训练步骤耗时约 0.4 秒。我们总共训练了 100,000 步或 12 个小时的基本模型。对于我们的大型模型（如表 3 底行所述），每步训练时间为 1.0 秒。大型模型的训练时间为 300,000 步（3.5 天）。
5.3 Optimizer 5.3 优化器
We used the Adam optimizer [20] with and . We varied the learning rate over the course of training, according to the formula: 我们使用了亚当优化器[20] 和 。在训练过程中，我们根据公式改变了学习率：
This corresponds to increasing the learning rate linearly for the first warmup_steps training steps, and decreasing it thereafter proportionally to the inverse square root of the step number. We used warmup_steps . 这相当于在第一个 warmup_steps 训练步数中线性增加学习率，之后则按步数的平方反比例降低学习率。我们使用 warmup_steps 。
5.4 Regularization 5.4 正则化
We employ three types of regularization during training: 在训练过程中，我们采用了三种正则化方法：
Table 2: The Transformer achieves better BLEU scores than previous state-of-the-art models on the English-to-German and English-to-French newstest2014 tests at a fraction of the training cost. 表 2：在英语到德语和英语到法语的 newstest2014 测试中，Transformer 的 BLEU 分数优于之前的先进模型，而训练成本仅为之前的一小部分。
Residual Dropout We apply dropout [33] to the output of each sub-layer, before it is added to the sub-layer input and normalized. In addition, we apply dropout to the sums of the embeddings and the positional encodings in both the encoder and decoder stacks. For the base model, we use a rate of . 残差滤波 我们对每个子层的输出进行滤波 [33]，然后将其添加到子层输入中并进行归一化处理。此外，我们还对编码器和解码器堆栈中的嵌入和位置编码之和进行了滤除。在基础模型中，我们使用的速率为 。
Label Smoothing During training, we employed label smoothing of value [36]. This hurts perplexity, as the model learns to be more unsure, but improves accuracy and BLEU score. 标签平滑 在训练过程中，我们使用了值为 [36] 的标签平滑。这样做会增加模型的不确定性，从而降低复杂度，但却能提高准确度和 BLEU 得分。
6 Results 6 项成果
6.1 Machine Translation 6.1 机器翻译
On the WMT 2014 English-to-German translation task, the big transformer model (Transformer (big) in Table 2) outperforms the best previously reported models (including ensembles) by more than 2.0 BLEU, establishing a new state-of-the-art BLEU score of 28.4. The configuration of this model is listed in the bottom line of Table 3 . Training took 3.5 days on 8 P100 GPUs. Even our base model surpasses all previously published models and ensembles, at a fraction of the training cost of any of the competitive models. 在 WMT 2014 英译德翻译任务中，大转换器模型（表 2 中的转换器（big））的 BLEU 值比之前报道的最佳模型（包括集合）高出 2.0 以上，达到了 28.4 的新的最先进 BLEU 值。该模型的配置见表 3 底行。在 8 个 P100 GPU 上的训练耗时 3.5 天。即使是我们的基本模型，也超越了之前发布的所有模型和集合，而训练成本只是任何竞争模型的一小部分。
On the WMT 2014 English-to-French translation task, our big model achieves a BLEU score of 41.0, outperforming all of the previously published single models, at less than the training cost of the previous state-of-the-art model. The Transformer (big) model trained for English-to-French used dropout rate , instead of 0.3 . 在 WMT 2014 英语到法语的翻译任务中，我们的大模型获得了 41.0 的 BLEU 分数，超过了之前发布的所有单一模型，而训练成本还不到之前最先进模型的 。为英译法训练的 Transformer（大）模型使用了辍学率 ，而不是 0.3。
For the base models, we used a single model obtained by averaging the last 5 checkpoints, which were written at 10 -minute intervals. For the big models, we averaged the last 20 checkpoints. We used beam search with a beam size of 4 and length penalty [38]. These hyperparameters were chosen after experimentation on the development set. We set the maximum output length during inference to input length +50 , but terminate early when possible [38]. 对于基本模型，我们使用的是通过平均最近 5 个检查点得到的单一模型，这些检查点以 10 分钟的间隔写入。对于大型模型，我们取最后 20 个检查点的平均值。我们使用波束搜索，波束大小为 4，长度惩罚为 [38]。这些超参数是在开发集上实验后选择的。我们将推理过程中的最大输出长度设定为输入长度 +50 ，但尽可能提前终止 [38]。
Table 2 summarizes our results and compares our translation quality and training costs to other model architectures from the literature. We estimate the number of floating point operations used to train a model by multiplying the training time, the number of GPUs used, and an estimate of the sustained single-precision floating-point capacity of each GPU 表 2 总结了我们的结果，并将我们的翻译质量和训练成本与文献中的其他模型架构进行了比较。我们通过将训练时间、使用的 GPU 数量和每个 GPU 的持续单精度浮点运算能力的估计值相乘，估算出训练一个模型所使用的浮点运算次数。
6.2 Model Variations 6.2 模型变化
To evaluate the importance of different components of the Transformer, we varied our base model in different ways, measuring the change in performance on English-to-German translation on the 为了评估 Transformer 不同组成部分的重要性，我们以不同的方式改变了基础模型，并测量了英译德的翻译性能变化。
Table 3: Variations on the Transformer architecture. Unlisted values are identical to those of the base model. All metrics are on the English-to-German translation development set, newstest2013. Listed perplexities are per-wordpiece, according to our byte-pair encoding, and should not be compared to per-word perplexities. 表 3：变压器架构的各种变化。未列出的值与基础模型的值相同。所有指标均基于英德翻译开发集 newstest2013。根据我们的字节对编码，列出的困惑度为每字片段的困惑度，不应与每字困惑度进行比较。
训练步骤
train
steps
PPL
(dev)
BLEU
params
base
6
512
2048
8
64
64
0.1
0.1
4.92
25.8
65
(A)
1
512
512
5.29
24.9
4
128
128
5.00
25.5
16
32
32
4.91
25.8
32
16
16
5.01
25.4
(B)
16
5.16
25.1
58
32
5.01
25.4
60
(C)
2
6.11
23.7
36
4
5.19
25.3
50
8
4.88
25.5
80
256
32
32
5.75
24.5
28
1024
128
128
4.66
26.0
168
1024
5.12
25.4
53
4096
4.75
26.2
90
(D)
0.0
5.77
24.6
0.2
4.95
25.5
0.0
4.67
25.3
0.2
5.47
25.7
(E)
positional embedding instead of sinusoids 位置嵌入代替正弦波
4.92
25.7
big
6
1024
4096
16
0.3
4.33
26.4
213
development set, newstest2013. We used beam search as described in the previous section, but no checkpoint averaging. We present these results in Table 3 开发集 newstest2013。我们使用了上一节所述的波束搜索，但没有使用检查点平均法。表 3 列出了这些结果
In Table 3 rows (A), we vary the number of attention heads and the attention key and value dimensions, keeping the amount of computation constant, as described in Section 3.2.2 While single-head attention is worse than the best setting, quality also drops off with too many heads. 如第 3.2 节所述，在表 3 行（A）中，我们改变了注意头的数量以及注意键和值的维度，计算量保持不变。
In Table 3 rows (B), we observe that reducing the attention key size hurts model quality. This suggests that determining compatibility is not easy and that a more sophisticated compatibility function than dot product may be beneficial. We further observe in rows (C) and (D) that, as expected, bigger models are better, and dropout is very helpful in avoiding over-fitting. In row (E) we replace our sinusoidal positional encoding with learned positional embeddings [9], and observe nearly identical results to the base model. 在表 3 行（B）中，我们发现减少关注键大小 会影响模型质量。这表明，确定兼容性并不容易，比点积更复杂的兼容性函数可能更有益处。在第（C）行和第（D）行中，我们进一步观察到，正如我们所预期的那样，模型越大越好，而 dropout 对避免过度拟合很有帮助。在第（E）行中，我们用学习到的位置嵌入[9]替换了正弦位置编码，观察到的结果与基础模型几乎完全相同。
6.3 English Constituency Parsing 6.3 英语选区解析
To evaluate if the Transformer can generalize to other tasks we performed experiments on English constituency parsing. This task presents specific challenges: the output is subject to strong structural constraints and is significantly longer than the input. Furthermore, RNN sequence-to-sequence models have not been able to attain state-of-the-art results in small-data regimes [37]. 为了评估转换器是否能推广到其他任务，我们进行了英语选区解析实验。这项任务具有特殊的挑战性：输出结果受到强大的结构约束，而且比输入结果要长得多。此外，RNN 序列到序列模型在小数据环境下也无法达到最先进的结果 [37]。
We trained a 4-layer transformer with on the Wall Street Journal (WSJ) portion of the Penn Treebank [25], about 40K training sentences. We also trained it in a semi-supervised setting, using the larger high-confidence and BerkleyParser corpora from with approximately sentences [37]. We used a vocabulary of tokens for the WSJ only setting and a vocabulary of tokens for the semi-supervised setting. 我们使用 对宾夕法尼亚州树库（Penn Treebank）[25] 中的《华尔街日报》（WSJ）部分（约 4 万个训练句子）进行了 4 层转换器训练。我们还在半监督环境下使用更大的高置信度语料库和 BerkleyParser 语料库（约有 句子）[37] 对其进行了训练。在仅使用 WSJ 的情况下，我们使用的词汇量为 ，而在半监督情况下，我们使用的词汇量为 。
We performed only a small number of experiments to select the dropout, both attention and residual (section 5.4), learning rates and beam size on the Section 22 development set, all other parameters remained unchanged from the English-to-German base translation model. During inference, we 我们仅在第 22 节开发集上进行了少量实验，以选择注意力和残差（第 5.4 节）、学习率和波束大小，所有其他参数均与英译德基础翻译模型保持一致。在推理过程中，我们
Table 4: The Transformer generalizes well to English constituency parsing (Results are on Section 23 of WSJ) 表 4：转换器在英语选区解析中的通用性很好（结果见《WSJ》第 23 节）
Parser
Training
WSJ 23 F1
Vinyals & Kaiser el al. (2014) [37] Vinyals & Kaiser el al. (2014) [37]
Vinyals & Kaiser el al. (2014) [37] Vinyals & Kaiser el al. (2014) [37]
semi-supervised 半监督
92.1
Transformer (4 layers) 变压器（4 层）
semi-supervised 半监督
92.7
Luong et al. (2015) [23] Luong 等人（2015）[23］
multi-task
93.0
Dyer et al. (2016) [8]
generative
93.3
increased the maximum output length to input length +300 . We used a beam size of 21 and for both WSJ only and the semi-supervised setting. 最大输出长度增加到输入长度 +300 。我们在仅 WSJ 和半监督设置中分别使用了 21 和 的波束大小。
Our results in Table 4 show that despite the lack of task-specific tuning our model performs surprisingly well, yielding better results than all previously reported models with the exception of the Recurrent Neural Network Grammar [8]. 表 4 中的结果表明，尽管缺乏针对特定任务的调整，但我们的模型表现出人意料的好，其结果优于除递归神经网络语法 [8] 以外的所有以前报道过的模型。
In contrast to RNN sequence-to-sequence models [37], the Transformer outperforms the BerkeleyParser [29] even when training only on the WSJ training set of 40K sentences. 与 RNN 序列到序列模型[37]相比，即使只在由 40K 个句子组成的 WSJ 训练集上进行训练，Transformer 的表现也优于 BerkeleyParser [29]。
7 Conclusion 7 结论
In this work, we presented the Transformer, the first sequence transduction model based entirely on attention, replacing the recurrent layers most commonly used in encoder-decoder architectures with multi-headed self-attention. 在这项工作中，我们提出了 "变形器"，这是第一个完全基于注意力的序列转换模型，用多头自我注意力取代了编码器-解码器架构中最常用的递归层。
For translation tasks, the Transformer can be trained significantly faster than architectures based on recurrent or convolutional layers. On both WMT 2014 English-to-German and WMT 2014 English-to-French translation tasks, we achieve a new state of the art. In the former task our best model outperforms even all previously reported ensembles. 对于翻译任务，Transformer 的训练速度明显快于基于递归层或卷积层的架构。在 WMT 2014 英译德和 WMT 2014 英译法翻译任务中，我们都达到了新的技术水平。在前一项任务中，我们的最佳模型甚至优于之前报告的所有集合。
We are excited about the future of attention-based models and plan to apply them to other tasks. We plan to extend the Transformer to problems involving input and output modalities other than text and to investigate local, restricted attention mechanisms to efficiently handle large inputs and outputs such as images, audio and video. Making generation less sequential is another research goals of ours. 我们对基于注意力的模型的未来充满期待，并计划将其应用于其他任务。我们计划将 Transformer 扩展到涉及文本以外的输入和输出模式的问题上，并研究局部的、受限的注意力机制，以有效处理大型输入和输出，如图像、音频和视频。我们的另一个研究目标是减少生成的顺序。
The code we used to train and evaluate our models is available at https://github.com// tensorflow/tensor2tensor. 我们用于训练和评估模型的代码可在 https://github.com// tensorflow/tensor2tensor 上获取。
Acknowledgements We are grateful to Nal Kalchbrenner and Stephan Gouws for their fruitful comments, corrections and inspiration. 感谢 Nal Kalchbrenner 和 Stephan Gouws 富有成效的评论、纠正和启发。
References 参考资料
[1] Jimmy Lei Ba, Jamie Ryan Kiros, and Geoffrey E Hinton. Layer normalization. arXiv preprint arXiv:1607.06450, 2016. [1] Jimmy Lei Ba, Jamie Ryan Kiros, and Geoffrey E Hinton.Layer normalization. ArXiv preprint arXiv:1607.06450, 2016.
[2] Dzmitry Bahdanau, Kyunghyun Cho, and Yoshua Bengio. Neural machine translation by jointly learning to align and translate. CoRR, abs/1409.0473, 2014. [2] Dzmitry Bahdanau、Kyunghyun Cho 和 Yoshua Bengio.通过联合学习对齐和翻译的神经机器翻译。CoRR，abs/1409.0473，2014.
[3] Denny Britz, Anna Goldie, Minh-Thang Luong, and Quoc V. Le. Massive exploration of neural machine translation architectures. CoRR, abs/1703.03906, 2017. [3] Denny Britz、Anna Goldie、Minh-Thang Luong 和 Quoc V. Le。神经机器翻译架构的大规模探索。CoRR，abs/1703.03906，2017.
[4] Jianpeng Cheng, Li Dong, and Mirella Lapata. Long short-term memory-networks for machine reading. arXiv preprint arXiv:1601.06733, 2016. [4] Jianpeng Cheng, Li Dong, and Mirella Lapata.机器阅读的长短期记忆网络》，arXiv preprint arXiv:1601.06733, 2016.
[5] Kyunghyun Cho, Bart van Merrienboer, Caglar Gulcehre, Fethi Bougares, Holger Schwenk, and Yoshua Bengio. Learning phrase representations using rnn encoder-decoder for statistical machine translation. CoRR, abs/1406.1078, 2014. [5] Kyunghyun Cho、Bart van Merrienboer、Caglar Gulcehre、Fethi Bougares、Holger Schwenk 和 Yoshua Bengio.使用rnn编码器-解码器学习短语表征，用于统计机器翻译。CoRR，abs/1406.1078，2014。
[6] Francois Chollet. Xception: Deep learning with depthwise separable convolutions. arXiv preprint arXiv:1610.02357, 2016. [6] Francois Chollet.Xception：ArXiv preprint arXiv:1610.02357, 2016.
[7] Junyoung Chung, Çaglar Gülçehre, Kyunghyun Cho, and Yoshua Bengio. Empirical evaluation of gated recurrent neural networks on sequence modeling. CoRR, abs/1412.3555, 2014. [7] Junyoung Chung, Çaglar Gülçehre, Kyunghyun Cho, and Yoshua Bengio.门控递归神经网络在序列建模上的经验评估。CoRR，abs/1412.3555，2014.
[8] Chris Dyer, Adhiguna Kuncoro, Miguel Ballesteros, and Noah A. Smith. Recurrent neural network grammars. In Proc. of NAACL, 2016. [8] Chris Dyer, Adhiguna Kuncoro, Miguel Ballesteros, and Noah A. Smith.递归神经网络语法。In Proc. of NAACL, 2016.
[9] Jonas Gehring, Michael Auli, David Grangier, Denis Yarats, and Yann N. Dauphin. Convolutional sequence to sequence learning. arXiv preprint arXiv:1705.03122v2, 2017. [9] Jonas Gehring, Michael Auli, David Grangier, Denis Yarats, and Yann N. Dauphin.卷积序列到序列学习。arXiv preprint arXiv:1705.03122v2, 2017.
[10] Alex Graves. Generating sequences with recurrent neural networks. arXiv preprint arXiv:1308.0850, 2013. [10] Alex Graves.用递归神经网络生成序列。arXiv preprint arXiv:1308.0850, 2013.
[11] Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 770-778, 2016. [11] 何开明、张翔宇、任绍清、孙健。图像识别的深度残差学习。In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 770-778, 2016.
[12] Sepp Hochreiter, Yoshua Bengio, Paolo Frasconi, and Jürgen Schmidhuber. Gradient flow in recurrent nets: the difficulty of learning long-term dependencies, 2001. [12] Sepp Hochreiter, Yoshua Bengio, Paolo Frasconi, and Jürgen Schmidhuber.递归网中的梯度流：学习长期依赖关系的困难》，2001 年。
[13] Sepp Hochreiter and Jürgen Schmidhuber. Long short-term memory. Neural computation, 9(8):1735-1780, 1997. [13] Sepp Hochreiter 和 Jürgen Schmidhuber.Long short-term memory.神经计算，9（8）：1735-1780，1997.
[14] Zhongqiang Huang and Mary Harper. Self-training PCFG grammars with latent annotations across languages. In Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing, pages 832-841. ACL, August 2009. [14] 黄中强和玛丽-哈珀。使用跨语言潜在注释自我训练 PCFG 语法。2009 年自然语言处理经验方法大会论文集》，第 832-841 页。ACL，2009 年 8 月。
[15] Rafal Jozefowicz, Oriol Vinyals, Mike Schuster, Noam Shazeer, and Yonghui Wu. Exploring the limits of language modeling. arXiv preprint arXiv:1602.02410, 2016. [15] Rafal Jozefowicz、Oriol Vinyals、Mike Schuster、Noam Shazeer 和 Yonghui Wu.探索语言建模的极限。arXiv preprint arXiv:1602.02410, 2016.
[16] Łukasz Kaiser and Samy Bengio. Can active memory replace attention? In Advances in Neural Information Processing Systems, (NIPS), 2016. [16] Łukasz Kaiser 和 Samy Bengio.主动记忆能否取代注意力？神经信息处理系统进展》（NIPS），2016 年。
[17] Łukasz Kaiser and Ilya Sutskever. Neural GPUs learn algorithms. In International Conference on Learning Representations (ICLR), 2016. [17] Łukasz Kaiser 和 Ilya Sutskever.神经 GPU 学习算法。国际学习表征会议（ICLR），2016 年。
[18] Nal Kalchbrenner, Lasse Espeholt, Karen Simonyan, Aaron van den Oord, Alex Graves, and Koray Kavukcuoglu. Neural machine translation in linear time. arXiv preprint v2, 2017. [18] Nal Kalchbrenner, Lasse Espeholt, Karen Simonyan, Aaron van den Oord, Alex Graves, and Koray Kavukcuoglu.线性时间内的神经机器翻译。ArXiv preprint v2, 2017.
[19] Yoon Kim, Carl Denton, Luong Hoang, and Alexander M. Rush. Structured attention networks. In International Conference on Learning Representations, 2017. [19] Yoon Kim, Carl Denton, Luong Hoang, and Alexander M. Rush.结构化注意力网络。国际学习表征会议，2017 年。
[20] Diederik Kingma and Jimmy Ba. Adam: A method for stochastic optimization. In ICLR, 2015. [20] Diederik Kingma 和 Jimmy Ba.亚当：一种随机优化方法。In ICLR, 2015.
[21] Oleksii Kuchaiev and Boris Ginsburg. Factorization tricks for LSTM networks. arXiv preprint arXiv:1703.10722, 2017. [21] Oleksii Kuchaiev 和 Boris Ginsburg.LSTM 网络的因式分解技巧》，arXiv preprint arXiv:1703.10722, 2017.
[22] Zhouhan Lin, Minwei Feng, Cicero Nogueira dos Santos, Mo Yu, Bing Xiang, Bowen Zhou, and Yoshua Bengio. A structured self-attentive sentence embedding. arXiv preprint arXiv:1703.03130, 2017. [22] Zhouhan Lin, Minwei Feng, Cicero Nogueira dos Santos, Mo Yu, Bing Xiang, Bowen Zhou, and Yoshua Bengio.ArXiv preprint arXiv:1703.03130, 2017.
[23] Minh-Thang Luong, Quoc V. Le, Ilya Sutskever, Oriol Vinyals, and Lukasz Kaiser. Multi-task sequence to sequence learning. arXiv preprint arXiv:1511.06114, 2015. [23] Minh-Thang Luong, Quoc V. Le, Ilya Sutskever, Oriol Vinyals, and Lukasz Kaiser.ArXiv preprint arXiv:1511.06114, 2015.
[24] Minh-Thang Luong, Hieu Pham, and Christopher D Manning. Effective approaches to attentionbased neural machine translation. arXiv preprint arXiv:1508.04025, 2015. [24] Minh-Thang Luong, Hieu Pham, and Christopher D Manning.基于注意力的神经机器翻译的有效方法。arXiv preprint arXiv:1508.04025, 2015.
[25] Mitchell P Marcus, Mary Ann Marcinkiewicz, and Beatrice Santorini. Building a large annotated corpus of english: The penn treebank. Computational linguistics, 19(2):313-330, 1993. [25] Mitchell P Marcus, Mary Ann Marcinkiewicz, and Beatrice Santorini.建立大型英语注释语料库：The penn treebank.Computational linguistics, 19(2):313-330, 1993.
[26] David McClosky, Eugene Charniak, and Mark Johnson. Effective self-training for parsing. In Proceedings of the Human Language Technology Conference of the NAACL, Main Conference, pages 152-159. ACL, June 2006. [26] David McClosky、Eugene Charniak 和 Mark Johnson。解析的有效自我训练。In Proceedings of the Human Language Technology Conference of the NAACL, Main Conference, pages 152-159.ACL，2006 年 6 月。
[27] Ankur Parikh, Oscar Täckström, Dipanjan Das, and Jakob Uszkoreit. A decomposable attention model. In Empirical Methods in Natural Language Processing, 2016. [27] Ankur Parikh, Oscar Täckström, Dipanjan Das, and Jakob Uszkoreit.可分解注意力模型。自然语言处理实证方法》，2016 年。
[28] Romain Paulus, Caiming Xiong, and Richard Socher. A deep reinforced model for abstractive summarization. arXiv preprint arXiv:1705.04304, 2017. [28] Romain Paulus, Caiming Xiong, and Richard Socher.用于抽象概括的深度强化模型》，arXiv preprint arXiv:1705.04304, 2017.
[29] Slav Petrov, Leon Barrett, Romain Thibaux, and Dan Klein. Learning accurate, compact, and interpretable tree annotation. In Proceedings of the 21 st International Conference on Computational Linguistics and 44th Annual Meeting of the ACL, pages 433-440. ACL, July 2006. [29] Slav Petrov、Leon Barrett、Romain Thibaux 和 Dan Klein.学习准确、紧凑和可解释的树注释。第 21 届计算语言学国际会议暨第 44 届 ACL 年会论文集》，第 433-440 页。ACL，2006 年 7 月。
[30] Ofir Press and Lior Wolf. Using the output embedding to improve language models. arXiv preprint arXiv:1608.05859, 2016. [30] Ofir Press 和 Lior Wolf.使用输出嵌入改进语言模型》，arXiv preprint arXiv:1608.05859, 2016.
[31] Rico Sennrich, Barry Haddow, and Alexandra Birch. Neural machine translation of rare words with subword units. arXiv preprint arXiv:1508.07909, 2015. [31] Rico Sennrich、Barry Haddow 和 Alexandra Birch.用子词单元进行罕见词的神经机器翻译。arXiv preprint arXiv:1508.07909, 2015.
[32] Noam Shazeer, Azalia Mirhoseini, Krzysztof Maziarz, Andy Davis, Quoc Le, Geoffrey Hinton, and Jeff Dean. Outrageously large neural networks: The sparsely-gated mixture-of-experts layer. arXiv preprint arXiv:1701.06538, 2017. [32] Noam Shazeer、Azalia Mirhoseini、Krzysztof Maziarz、Andy Davis、Quoc Le、Geoffrey Hinton 和 Jeff Dean。超大型神经网络：ArXiv preprint arXiv:1701.06538, 2017.
[33] Nitish Srivastava, Geoffrey E Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov. Dropout: a simple way to prevent neural networks from overfitting. Journal of Machine Learning Research, 15(1):1929-1958, 2014. [33] Nitish Srivastava, Geoffrey E Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov.Dropout：防止神经网络过度拟合的简单方法。机器学习研究期刊》，15（1）：1929-1958，2014 年。
[34] Sainbayar Sukhbaatar, Arthur Szlam, Jason Weston, and Rob Fergus. End-to-end memory networks. In C. Cortes, N. D. Lawrence, D. D. Lee, M. Sugiyama, and R. Garnett, editors, Advances in Neural Information Processing Systems 28, pages 2440-2448. Curran Associates, Inc., 2015. [34] Sainbayar Sukhbaatar、Arthur Szlam、Jason Weston 和 Rob Fergus。端到端内存网络。在 C. Cortes、N. D. Lawrence、D. D. Lee、M. Sugiyama 和 R. Garnett 编辑的《神经信息处理系统进展 28》中，第 2440-2448 页。Curran Associates, Inc., 2015.
[35] Ilya Sutskever, Oriol Vinyals, and Quoc VV Le. Sequence to sequence learning with neural networks. In Advances in Neural Information Processing Systems, pages 3104-3112, 2014. [35] Ilya Sutskever、Oriol Vinyals 和 Quoc VV Le。神经网络的序列到序列学习。神经信息处理系统进展》，第 3104-3112 页，2014 年。
[36] Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jonathon Shlens, and Zbigniew Wojna. Rethinking the inception architecture for computer vision. CoRR, abs/1512.00567, 2015. [36] Christian Szegedy、Vincent Vanhoucke、Sergey Ioffe、Jonathon Shlens 和 Zbigniew Wojna.重新思考计算机视觉的初始架构。CoRR，abs/1512.00567，2015.
[37] Vinyals & Kaiser, Koo, Petrov, Sutskever, and Hinton. Grammar as a foreign language. In Advances in Neural Information Processing Systems, 2015. [37] Vinyals & Kaiser，Koo，Petrov，Sutskever，and Hinton.语法是一门外语。In Advances in Neural Information Processing Systems, 2015.
[38] Yonghui Wu, Mike Schuster, Zhifeng Chen, Quoc V Le, Mohammad Norouzi, Wolfgang Macherey, Maxim Krikun, Yuan Cao, Qin Gao, Klaus Macherey, et al. Google's neural machine translation system: Bridging the gap between human and machine translation. arXiv preprint arXiv:1609.08144, 2016. [38] Yonghui Wu, Mike Schuster, Zhifeng Chen, Quoc V Le, Mohammad Norouzi, Wolfgang Macherey, Maxim Krikun, Yuan Cao, Qin Gao, Klaus Macherey, et al. 谷歌的神经机器翻译系统：ArXiv preprint arXiv:1609.08144, 2016.
[39] Jie Zhou, Ying Cao, Xuguang Wang, Peng Li, and Wei Xu. Deep recurrent models with fast-forward connections for neural machine translation. CoRR, abs/1606.04199, 2016. [39] Jie Zhou, Ying Cao, Xuguang Wang, Peng Li, and Wei Xu.用于神经机器翻译的具有快进连接的深度递归模型。CoRR，abs/1606.04199，2016.
[40] Muhua Zhu, Yue Zhang, Wenliang Chen, Min Zhang, and Jingbo Zhu. Fast and accurate shift-reduce constituent parsing. In Proceedings of the 51st Annual Meeting of the ACL (Volume 1: Long Papers), pages 434-443. ACL, August 2013. [40] Muhua Zhu，Yue Zhang，Wenliang Chen，Min Zhang，and Jingbo Zhu.快速准确的移位还原成分解析。第 51 届 ACL 年会论文集（第 1 卷：长篇论文），第 434-443 页。ACL，2013 年 8 月。
Attention Visualizations 注意力可视化
Figure 3: An example of the attention mechanism following long-distance dependencies in the encoder self-attention in layer 5 of 6 . Many of the attention heads attend to a distant dependency of the verb 'making', completing the phrase 'making...more difficult'. Attentions here shown only for the word 'making'. Different colors represent different heads. Best viewed in color. 图 3：第 6 层中第 5 层编码器自我注意中长距离依赖关系的注意机制示例。许多注意头都会注意动词 "making "的远距离依赖关系，从而完成短语 "making...more difficult"。这里只显示了 "制造 "一词的注意力。不同的颜色代表不同的注意头。最好用彩色观看。
Figure 4: Two attention heads, also in layer 5 of 6 , apparently involved in anaphora resolution. Top: Full attentions for head 5. Bottom: Isolated attentions from just the word 'its' for attention heads 5 and 6. Note that the attentions are very sharp for this word. 图 4：同样位于第 5 层 6 的两个注意头显然参与了拟声词的解析。上图：注意头 5 的全部注意力。下图第 5 和第 6 注意头对 "its "一词的单独注意。请注意，该词的注意力非常敏锐。
Figure 5: Many of the attention heads exhibit behaviour that seems related to the structure of the sentence. We give two such examples above, from two different heads from the encoder self-attention at layer 5 of 6 . The heads clearly learned to perform different tasks. 图 5：许多注意头的行为似乎与句子结构有关。我们在上面给出了两个这样的例子，它们分别来自第 6 层第 5 个编码器自我注意中的两个不同的注意头。这些注意头显然学会了执行不同的任务。
Equal contribution. Listing order is random. Jakob proposed replacing RNNs with self-attention and started the effort to evaluate this idea. Ashish, with Illia, designed and implemented the first Transformer models and has been crucially involved in every aspect of this work. Noam proposed scaled dot-product attention, multi-head attention and the parameter-free position representation and became the other person involved in nearly every detail. Niki designed, implemented, tuned and evaluated countless model variants in our original codebase and tensor2tensor. Llion also experimented with novel model variants, was responsible for our initial codebase, and efficient inference and visualizations. Lukasz and Aidan spent countless long days designing various parts of and implementing tensor2tensor, replacing our earlier codebase, greatly improving results and massively accelerating our research. 同等贡献。列表顺序随机。Jakob 提议用自我关注取代 RNN，并开始了对这一想法的评估工作。Ashish 和 Illia 一起设计并实现了第一个 Transformer 模型，并在这项工作的各个方面发挥了关键作用。诺姆（Noam）提出了缩放点积注意力、多头注意力和无参数位置表示法，并成为参与几乎所有细节的另一个人。尼基在我们的原始代码库和 tensor2tensor 中设计、实施、调整和评估了无数模型变体。Llion 也尝试了新的模型变体，负责我们的初始代码库以及高效推理和可视化。Lukasz 和 Aidan 花了无数个漫长的日子，设计并实现了 tensor2tensor 的各个部分，取代了我们早期的代码库，极大地改进了结果，并大大加快了我们的研究速度。
Work performed while at Google Brain. 在 Google Brain 工作期间完成的工作。
Work performed while at Google Research. 在 Google Research 工作期间完成的工作。
To illustrate why the dot products get large, assume that the components of and are independent random variables with mean 0 and variance 1 . Then their dot product, , has mean 0 and variance . 为了说明点积为什么会变大，假设 和 的分量是均值为 0、方差为 1 的独立随机变量。那么它们的点积 的均值为 0，方差为 。
We used values of 2.8, 3.7, 6.0 and 9.5 TFLOPS for K80, K40, M40 and P100, respectively. 我们为 K80、K40、M40 和 P100 分别使用了 2.8、3.7、6.0 和 9.5 TFLOPS 的数值。
