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 BLEUon the WMT 2014 English- to-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.

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 ht, as a function of the previous hidden state ht?1 and the input for position t. 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. ? ?

Encoder:?The encoder is composed of a stack of?N?= 6 identical layers. Each layer has two sub-layers. The first is a multi-head self-attention mechanism, and the second is a simple, position- wise 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(x?+ Sublayer(x)),

where Sublayer(x) 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?dmodel?= 512.

編碼器：編碼器由N=6個相同層組成。每層有兩個子層。

第一層是多頭multi-head?self-attention機制，

第二層是一個簡單的、位置全連接的前饋網絡。我們在兩子層的每一個子層周圍使用一個殘差連接[11]，然后是層規范化[1]。也就是說，每個子層的輸出是

LayerNorm(x?+ Sublayer(x))，

其中Sublayer(x) 是子層本身實現的功能。為了方便這些殘差連接，模型中的所有子層以及嵌入層都會生成維d model=512的輸出。

We call our particular attention "Scaled Dot-Product Attention" (Figure 2).?The input consists of queries and keys of dimension?dk?, and values of dimension?dv?. 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.

In practice, we compute the attention function on a set of queries simultaneously, packed together into a matrix?Q. The keys and values are also packed together into matrices?K?and?V?. We compute the matrix of outputs as:

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. Additive attention computes the compatibility function using a feed-forward network witha single hidden layer. While the two are similar in theoretical complexity, dot-product attention ismuch faster and more space-efficient in practice, since it can be implemented using highly optimized matrix multiplication code.

我們發現，不同于使用dmodel維度的鍵、值和查詢來執行單一的注意功能，將查詢、鍵和值分別以不同的線性投影h次線性投影到dk、dk和dv維度是有益的。

然后，在這些查詢、鍵和值的投影版本中，我們并行地執行注意功能，生成dv維輸出值。如圖2所示，這些被連接起來并再次投影，從而產生最終值。Multi-head attention允許模型共同關注來自不同位置的不同表示子空間的信息。只有一個注意力集中的頭腦，平均化可以抑制這種情況。

The Transformer uses multi-head attention in three different ways:

• 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].
• 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.
• 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.

• 在“編碼器-解碼器-注意”層中，查詢來自上一個解碼器層，而內存鍵和值來自編碼器的輸出。這使得解碼器中的每個位置都可以參與輸入序列中的所有位置。這模仿了典型的編碼器-解碼器-注意機制的順序對序列模型，如[38，2，9]。
• 編碼器包含self-attention層。在自關注層中，所有鍵、值和查詢都來自同一個位置，在本例中，是編碼器中前一層的輸出。編碼器中的每個位置都可以處理編碼器前一層中的所有位置。
• 類似地，解碼器中的 self-attention層允許解碼器中的每個位置關注解碼器中直到并包括該位置的所有位置。為了保持解碼器的自回歸特性，需要防止解碼器中的信息向左流動。我們通過屏蔽softmax輸入中所有與非法連接相對應的值（設置為–∞）來實現這個內標度點積注意。見圖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.
FFN(x) = max(0, xW1 + b1)W2 + b2 (2)

除了注意力子層之外，我們的編碼器和解碼器中的每一層都包含一個完全連接的前饋網絡，該網絡分別且相同地應用于每個位置。。這包括兩個線性變換，中間有一個ReLU激活。

FFN（x）=max（0，xW1+b1）W2+b2

與其他序列轉換模型類似，我們使用學習的嵌入來將輸入標記和輸出標記轉換為維度dmodel的向量。我們還使用通常的學習線性變換和softmax函數將解碼器輸出轉換為預測的下一個令牌概率。在我們的模型中，我們在兩個嵌入層和pre-softmax線性變換之間共享相同的權重矩陣，類似于[30]。在嵌入層中，我們將這些權重乘以√dmodel。

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 dmodel as the embeddings, so that the two can be summed. There are many choices of positional encodings, learned and fixed [9].

Table 1: Maximum path lengths, per-layer complexity and minimum number of sequential operations for different layer types. n is the sequence length, d is the representation dimension, k is the kernel size of convolutions and r the size of the neighborhood in restricted self-attention.

表1:不同層類型的最大路徑長度、每層復雜度和最小順序操作數。

n為序列長度，

d為表示維數，

k為卷積的核大小，

r為受限自我注意的鄰域大小。

在這項工作中，我們使用不同頻率的正弦和余弦函數：
其中pos是位置，i是維度。

也就是說，位置編碼的每個維度對應于一個正弦。波長形成了從2π到10000·2π的幾何級數。我們選擇這個函數是因為我們假設它可以讓模型很容易地學會通過相對位置來參與，因為對于任何固定偏移量k，P-Epos+k可以表示為P-Epos的線性函數。

In this section we compare various aspects of self-attention layers to the recurrent and convolu- tional layers commonly used for mapping one variable-length sequence of symbol representations (x1, ..., xn) to another sequence of equal length (z1, ..., zn), with xi, zi ∈ Rd, such as a hidden layer in a typical sequence transduction encoder or decoder. Motivating our use of self-attention we consider three desiderata.

• 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.

• 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.

• 一個是每層的總計算復雜度。
• 另一個是可以并行化的計算量，用所需的最少順序操作數來衡量。
• 第三個是網絡中長距離依賴關系之間的路徑長度。在許多序列轉導任務中，學習長程依賴性是一個關鍵的挑戰。影響學習這種依賴關系的能力的一個關鍵因素是網絡中向前和向后信號必須經過的路徑的長度。輸入和輸出序列中任何位置組合之間的這些路徑越短，就越容易學習長期依賴性[12]。因此，我們還比較了在由不同層類型組成的網絡中的任意兩個輸入和輸出位置之間的最大路徑長度。

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 O(n) sequential operations. In terms of computational complexity, self-attention layers are faster than recurrent layers when the sequence length n is smaller than the representation dimensionality d, 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 r in the input sequence centered around the respective output position. This would increase the maximum path length to O(n/r). We plan to investigate this approach further in future work.

A single convolutional layer with kernel width k < n does not connect all pairs of input and output positions. Doing so requires a stack of O(n/k) convolutional layers in the case of contiguous kernels, or O(logk (n)) 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 k. Separable convolutions [6], however, decrease the complexity considerably, to O(k · n · d + n · d2). Even with k = n, 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.

我們使用標準的WMT 2014英德數據集進行訓練，該數據集由大約450萬個句子對組成。句子使用字節對編碼[3]進行編碼，它有一個大約37000個標記的共享源-目標詞匯表。

對于英語-法語，我們使用了更大的WMT 2014英語-法語數據集，該數據集包含3600萬個句子，并將標記拆分為32000個詞條詞匯[38]。句子對由近似的序列長度組合在一起。每個訓練批次包含一組句子對，其中包含大約25000個源令牌和25000個目標令牌。

We used the Adam optimizer [20] with β1 = 0.9, β2 = 0.98 and E = 10?9. We varied the learning rate over the course of training, according to the formula:

• 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 Pdrop = 0.1.

• 在將Dropout[33]添加到子層輸入并規范化之前，我們對每個子層的輸出應用Dropout[33]。
• 此外，我們還將dropout應用于編碼器和解碼器堆棧中的嵌入和位置編碼的總和。
• 對于基本模型，我們使用Pdrop=0.1的速率。

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.

Label Smoothing During training, we employed label smoothing of value Els = 0.1 [36]. This hurts perplexity, as the model learns to be more unsure, but improves accuracy and BLEU score.

表2:Transformer在英語到德語和英語到法語的newstest2014測試中取得了比以前最先進的模式更好的BLEU分數，只需花費很少的訓練成本。

在訓練中，我們使用Els = 0.1[36]的值進行標簽平滑。這傷害了perplexity，因為模型學會了更不確定，但提高了準確性和BLEU分數。

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.

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 1/4 the training cost of the previous state-of-the-art model. The Transformer (big) model trained for English-to-French used dropout rate Pdrop = 0.1, instead of 0.3.

在WMT 2014英德翻譯任務中，big transformer模型（表2中的transformer（big））比之前報告的最佳模型（包括集成）的表現超過2.0 BLEU，新的最新BLEU分數為28.4。此型號的配置列于表3的最后一行。訓練時間為3.5天，平均成績為8 P100。即使是我們的基礎模型也超過了所有之前發布的模型和集合，只是任何競爭模型的訓練成本的一小部分。

在WMT 2014英法翻譯任務中，我們的大模型達到了41.0的BLEU分數，超過了所有之前發布的單一模型，在不到1/4的訓練成本的前一個最先進的模式。為英語到法語培訓的Transformer（big）模型使用的是輟學率Pdrop=0.1，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 α = 0.6 [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].

We trained a 4-layer transformer with dmodel = 1024 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 17M sentences [37]. We used a vocabulary of 16K tokens for the WSJ only setting and a vocabulary of 32K tokens for the semi-supervised setting.