3atv精品不卡视频,97人人超碰国产精品最新,中文字幕av一区二区三区人妻少妇,久久久精品波多野结衣,日韩一区二区三区精品

歡迎訪問 生活随笔!

生活随笔

當前位置: 首頁 > 编程资源 > 编程问答 >内容正文

编程问答

Paper:2017年的Google机器翻译团队《Transformer:Attention Is All You Need》翻译并解读

發布時間:2025/3/21 编程问答 37 豆豆
生活随笔 收集整理的這篇文章主要介紹了 Paper:2017年的Google机器翻译团队《Transformer:Attention Is All You Need》翻译并解读 小編覺得挺不錯的,現在分享給大家,幫大家做個參考.

Paper:2017年的Google機器翻譯團隊《Transformer:Attention Is All You Need》翻譯并解讀

目錄

論文評價

1、Motivation:

2、創新點:

Abstract ??

1、Introduction

2、Background

3、Model Architecture

3.1、Encoder and Decoder Stacks

3.2、Attention?

3.2.1、Scaled Dot-Product Attention

3.2.2、Multi-Head Attention? ?

3.2.3、Applications of Attention in our Model

3.3、Position-wise Feed-Forward Networks

3.4、Embeddings and Softmax

3.5、Positional Encoding

4、Why Self-Attention

5、Training

5.1、Training Data and Batching

5.2、Hardware and Schedule

5.3、Optimizer

5.4、Regularization?

6、Results

6.1、Machine Translation

6.2、Model Variations

6.3、English Constituency Parsing

7、Conclusion

References


論文評價

? ? ? ?2017年,Google機器翻譯團隊發表的《Attention is all you need》中大量使用了自注意力(self-attention)機制來學習文本表示。

參考文章:《attention is all you need》解讀

1、Motivation:

  • 靠attention機制,不使用rnn和cnn,并行度高
  • 通過attention,抓長距離依賴關系比rnn強

2、創新點:

  • 通過self-attention,自己和自己做attention,使得每個詞都有全局的語義信息(長依賴
  • 由于 Self-Attention 是每個詞和所有詞都要計算 Attention,所以不管他們中間有多長距離,最大的路徑長度也都只是 1。可以捕獲長距離依賴關系
  • 提出multi-head attention,可以看成attention的ensemble版本,不同head學習不同的子空間語義。

論文地址
鏈接:《Attention Is All You Need》
PDF:《Attention Is All You Need》

Abstract ??

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.

主要的序列轉導模型是基于復雜的遞歸或卷積神經網絡,包括編碼器和解碼器。性能最好的模型還需通過注意機制連接編碼器和解碼器。我們提出了一種新的簡單的網絡結構,Transformer,它完全基于注意機制,完全不需要遞歸和卷積。在兩個機器翻譯任務上的實驗表明,這些模型具有更好的并行性更少的訓練時間。我們的模型在WMT 2014英語到德語翻譯任務中達到28.4的BLEU,改進現有的最佳結果,包括集成,超過2 BLEU。在WMT 2014英法翻譯任務中,我們的模型在8個GPU上訓練3.5天后,建立了一個新的單模型最新的BLEU分數41.8,這是文獻中最好模型的訓練成本的一小部分。我們將該Transformer成功地應用于具有大量和有限訓練數據的英語選區句法分析中,證明了該變換器對其他任務具有良好的推廣效果。

BLEU是一種文本評估算法,即Bilingual Evaluation Understudy,它是用來評估機器翻譯專業人工翻譯之間的對應關系,核心思想就是機器翻譯越接近專業人工翻譯,質量就越好,經過BLEU算法得出的分數可以作為機器翻譯質量的其中一個指標。

1、Introduction

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].
?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. ??Work performed while at Google Brain. ??Work performed while at Google Research.
遞歸神經網絡,特別是LSTM[13]和GRU[7]神經網絡,已經作為序列建模和轉導問題(如語言建模和機器翻譯[35,2,5])的最新方法被牢固地建立起來。此后,無數的努力繼續推進遞歸語言模型和編碼器-解碼器體系結構的界限[38,24,15]。
*同等貢獻。列表順序是隨機的。Jakob建議用self-attention取代RNNs,并開始努力評估這個想法。Ashish和Illia一起設計并實現了第一個Transformer 模型,并在這項工作的各個方面都有重要的參與。Noam提出了標度點積注意、多頭注意和無參數位置表示,成為幾乎涉及每個細節的另一個人。Niki在我們最初的代碼庫和tensor2tensor中設計、實現、調整和評估了無數的模型變體。Llion還試驗了新的模型變體,負責我們的初始代碼庫,以及有效的推理和可視化。Lukasz和Aidan花了無數天的時間設計和實現Tensor2Sensor的各個部分,取代了我們早期的代碼庫,極大地改進了結果,極大地加速了我們的研究。在谷歌大腦工作期間完成的工作。?在谷歌研究所工作期間完成的工作。

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. ? ?

遞歸模型通常沿輸入和輸出序列的符號位置進行因子計算。將位置與計算時間中的步驟對齊,它們生成隱藏狀態的序列ht,作為先前隱藏狀態ht-1和位置t的輸入的函數。這種固有的序列性質排除了訓練示例中的并行化,而在較長的序列長度下,并行化變得至關重要,因為內存限制限制了跨示例的批處理。最近的工作通過因子分解技巧[21]和條件計算[32]在計算效率方面取得了顯著的提高,同時也提高了后者的模型性能。然而,順序計算的基本約束仍然存在。 ??
Attention mechanisms have become an integral part of compelling sequence modeling and transduc- tion 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. ? ?注意機制已經成為各種任務中強制序列建模和轉換模型的一個組成部分,允許不考慮輸入或輸出序列中的距離的依賴關系建模[2,19]。然而,在除少數情況外的所有情況下[27],這種注意機制都與一個遞歸網絡結合使用。 ??
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.在這項工作中,我們提出了Transformer,這是一種避免重復出現的模型架構,而完全依賴于注意機制來繪制輸入和輸出之間的全局依賴關系。該Transformer 允許更顯著的并行化,并可以達到一個新的水平,在翻譯質量后,在8個P100 gpu訓練了12小時。

2、Background

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.減少順序計算的目的也形成了擴展神經GPU〔16〕、ByteNet〔18〕和VusS2S〔9〕的基礎,所有這些都使用卷積神經網絡作為基本構建塊,并行計算所有輸入和輸出位置的隱藏表示。在這些模型中,將來自兩個任意輸入或輸出位置的信號關聯起來所需的操作數在位置之間的距離中增長,convs2為線性增長,ByteNet為對數增長。這使得學習遠距離位置之間的依賴性變得更加困難[12]。在Transformer 中,這被減少到一個恒定的操作數,盡管由于平均注意加權位置而降低了有效分辨率,但是我們用多頭注意抵消了這一影響,如第3.2節所述。
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].Self-attention,有時稱為注意內注意,是一種注意機制,它將單個序列的不同位置聯系起來,以計算序列的表示Self-attention在閱讀理解、抽象概括、語篇蘊涵和學習任務無關的句子表征等任務中得到了成功的運用[4,27,28,22]。
End-to-end memory networks are based on a recurrent attention mechanism instead of sequence- aligned recurrence and have been shown to perform well on simple-language question answering and language modeling tasks [34].端到端的記憶網絡是基于一種循環注意機制而不是順序排列的循環,并且已經被證明在簡單的語言問答和語言建模任務中表現良好[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 sequence- aligned 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,據我們所知,Transformer 是第一個完全依賴于self-attention來計算其輸入和輸出表示的傳導模型,而不使用序列對齊的RNN或卷積。在下面的章節中,我們將描述Transformer ,激發self-attention,并討論其相對于[17,18]和[9]等模型的優勢。

3、Model Architecture

Most competitive neural sequence transduction models have an encoder-decoder structure [5, 2, 35]. Here, the encoder maps an input sequence of symbol representations (x1, ..., xn) to a sequence of continuous representations z = (z1, ..., zn). Given z, the decoder then generates an output sequence (y1, ..., ym) 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.大多數競爭性神經序列轉導模型都具有編碼器-解碼器結構[5,2,35]。這里,編碼器將符號表示的輸入序列(x1,…,xn)映射到連續表示的序列z=(z1,…,zn)。給定z,解碼器然后一次生成一個符號的輸出序列(y1,…,ym)。在每一步模型都是自回歸的[10],在生成下一步時,將先前生成的符號作為附加輸入
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.Transformer 遵循這個總體架構,使用堆疊的自關注層和點方式的完全連接層,分別用于編碼器和解碼器,如圖1的左半部分和右半部分所示。

? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? Figure 1: The Transformer - model architecture.

3.1、Encoder and Decoder Stacks

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。

Decoder:?The decoder is also composed of a stack of?N?= 6 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?i?can depend only on the known outputs at positions less than?i.解碼器:解碼器也由N=6個相同層組成。除了每個編碼器層中的兩個子層之外,解碼器還插入第三個子層,該子層對編碼器堆棧的輸出執行multi-head attention。與編碼器類似,我們在每個子層周圍使用殘差連接,然后進行層規范化。我們還修改解碼器堆棧中的自關注子層,以防止位置關注后續位置。這種掩蔽加上輸出嵌入偏移一個位置的事實,確保位置i的預測只能依賴于小于i的位置處的已知輸出。

3.2、Attention?

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 of the values, where the weight assigned to each value is computed by a compatibility function of the query with the corresponding key.一個注意函數可以描述為將查詢和一組鍵值對映射到輸出,其中query, keys, valuesoutput 都是向量。輸出被計算為值的加權和,其中分配給每個值的權重由查詢的兼容函數和相應的鍵計算。

Figure 2: (left) Scaled Dot-Product Attention. (right) Multi-Head Attention consists of several attention layers running in parallel.
圖2:(左)縮放點積注意力機制? ? ? ? ? (右圖)多頭注意力機制由幾個平行運行的注意力層組成。

3.2.1、Scaled Dot-Product Attention

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.

我們稱特別注意為“縮放點積注意”(圖2)。輸入由維度dk的查詢和鍵以及維度dv的值組成。我們使用所有鍵計算查詢的點積,用√dk除以每個鍵,然后應用softmax函數獲得值的權重
?

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:

在實際應用中,我們同時計算一組查詢的注意函數,將它們組合成一個矩陣Q,并將鍵和值組合成矩陣K和V。我們將輸出矩陣計算為:

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.

最常用的兩個注意函數是additive attention[2]和dot-product?attention(乘法)。dot-product?attention與我們的算法相同,除了比例因子。additive attention使用具有單個隱藏層的前饋網絡計算兼容性函數。雖然這兩種方法在理論復雜度上是相似的,但是由于它可以用高度優化的矩陣乘法碼來實現,所以在實際應用中,dot-product?attention要快得多,空間也更有效
?
While for small values of dk the two mechanisms perform similarly, additive attention outperforms dot product attention without scaling for larger values of dk [3]. We suspect that for large values of dk , the dot products grow large in magnitude, pushing the softmax function into regions where it has extremely small gradients 4. To counteract this effect, we scale the dot products by而對于dk的小值,兩種機制的表現相似,additive attention優于dot product attention,而對于dk的大值則沒有標度[3]。我們懷疑,對于dk的大值,點積在數量級上增長很大,將softmax函數推到梯度非常小的區域4。為了抵消這一影響,我們將 dot products積按

3.2.2、Multi-Head Attention? ?

Instead of performing a single attention function with dmodel-dimensional keys, values and queries, we found it beneficial to linearly project the queries, keys and values h times with different, learned linear projections to dk , dk and dv dimensions, respectively. On each of these projected versions of queries, keys and values we then perform the attention function in parallel, yielding dv -dimensional output values. These are concatenated and once again projected, resulting in the final values, as depicted in Figure 2.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.

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

In this work we employ h = 8 parallel attention layers, or heads. For each of these we use dk = dv = dmodel/h = 64. Due to the reduced dimension of each head, the total computational cost is similar to that of single-head attention with full dimensionality.在這項工作中,我們使用了h=8個平行的注意層,或者說頭部。對于每一個,我們使用dk=dv=dmodel/h=64。由于每個頭部的維數減少,總的計算成本與全維度的單頭部注意的計算成本相似。

3.2.3、Applications of Attention in our Model

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.
Transformer 以三種不同的方式使用multi-head
  • 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。

3.3、Position-wise Feed-Forward Networks

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

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 dmodel = 512, and the inner-layer has dimensionality df f = 2048.雖然不同位置的線性變換是相同的,但它們在不同的層之間使用不同的參數。另一種描述方法是兩個核大小為1的卷積。輸入輸出的維數為dmodel=512,內層的維數為df f=2048。

3.4、Embeddings and Softmax

Similarly to other sequence transduction models, we use learned embeddings to convert the input tokens and output tokens to vectors of dimension dmodel. We also use the usual learned linear transfor- mation 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 √dmodel.

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。

3.5、Positional Encoding

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].

由于我們的模型不包含遞歸和卷積,為了使模型能夠利用序列的順序,我們必須注入一些關于序列中標記的相對或絕對位置的信息。為此,我們將“位置編碼”添加到編碼器和解碼器堆棧底部的輸入嵌入中。位置編碼與嵌入編碼具有相同的維度dmodel,因此可以將它們相加。有許多位置編碼的選擇,學習和固定的[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。

In this work, we use sine and cosine functions of different frequencies:

where pos is the position and i is the dimension. That is, each dimension of the positional encoding corresponds to a sinusoid. The wavelengths form a geometric progression from 2π to 10000 · 2π. 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 k, P Epos+k can be represented as a linear function of P Epos.

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。

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.我們還嘗試使用學習的位置嵌入[9],發現這兩個版本產生了幾乎相同的結果(見表3第(E)行)。我們之所以選擇正弦曲線,是因為它可能允許模型外推到比訓練中遇到的序列長度更長的序列。

4、Why Self-Attention

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.

在本節中,我們將self-attention層的各個方面與通常用于映射一個可變長度的符號表示序列(x1,…,xn)到另一個等長序列(Z1,…,Zn)的遞歸和卷積層進行比較。例如典型的序列轉換編碼器或解碼器中的隱藏層。激發我們使用self-attention,我們考慮三個目的。
?
  • 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.

如表1所示,self-attention層用固定數量的順序執行的操作連接所有位置,而遞歸層需要O(n)順序操作。在計算復雜性方面,當序列長度N小于表示維度D時,自關注層比遞歸層更快,這是最常用的機器翻譯中最先進的模型所使用的句子表示的情況,例如Word(38)和字節對(31)表示。為了提高包含很長序列的任務的計算性能,可以將self-attention限制為僅考慮以各自輸出位置為中心的輸入序列中r大小的鄰域。這將增加到O(n/r)的最大路徑長度。我們計劃在今后的工作中進一步研究這種方法。
?

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.

核寬度k<n的單個卷積層不能連接所有輸入和輸出位置對。這樣做需要一堆O(n/k)卷積層(對于相鄰的核),或者O(logk(n))卷積層(對于擴展卷積)[18],增加網絡中任意兩個位置之間最長路徑的長度。卷積層通常比遞歸層更昂貴,由K的可分離卷積〔6〕,但是,大大降低了復雜度,達到O(k·n·d+n·d2)。然而,即使具有k= n,可分離卷積的復雜性等于自關注層和點前饋層的組合,我們在模型中采用的方法
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.作為附帶好處,self-attention可以產生更多可解釋的模型。我們從我們的模型中檢查注意分布,并在附錄中給出和討論示例。不僅個體的注意頭清楚地學會了執行不同的任務,許多似乎表現出與句子的句法和語義結構有關的行為

5、Training

This section describes the training regime for our models.本節介紹我們的模型的訓練制度。

5.1、Training Data and Batching

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 source- target 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

5.2、Hardware and Schedule—8臺+12小時/3.5天

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).我們用8臺NVIDIA P100 gpu在一臺機器上訓練我們的模型。對于使用本文描述的超參數的基礎模型,每個訓練步驟大約需要0.4秒。我們對基礎模型進行了總共100000步或12小時的訓練。對于我們的大型模型(見表3的底線),步進時間為1.0秒。大模型接受了30萬步(3.5天)的訓練。

5.3、Optimizer—Adam優化器

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:

我們使用Adam優化器[20],beta1=0.9,beta2=0.98,E=10-9。我們在整個訓練過程中根據以下公式調整學習率:
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 = 4000.這相當于線性地增加第一個warmup_steps 的學習率,然后與步驟數的平方根成比例地降低學習率。我們采用了warmup_steps =4000。

5.4、Regularization?—3種正則化

We employ three types of regularization during training:
  • 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.
在訓練期間,我們采用3種正則化
  • 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。

6、Results

6.1、Machine Translation

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].

對于基本模型,我們使用一個通過平均最后5個檢查點(以10分鐘為間隔寫入)獲得的模型。對于大型模型,我們平均使用了最后20個檢查點。我們使用波束搜索,波束大小為4,長度懲罰α=0.6[38]。這些超參數是在開發集上進行實驗后選擇的。在輸入長度為50時,我們設定最大輸出長度,但在可能時終止(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 5.表2總結了我們的結果,并將我們的翻譯質量和訓練成本與文獻中的其他模型架構進行了比較。我們通過乘以訓練時間、使用的GPU數量和每個GPU 5的持續單精度浮點容量,來估計用于訓練模型的浮點操作的數量。


6.2、Model Variations

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 development set, newstest2013. We used beam search as described in the previous section, but no checkpoint averaging. We present these results in Table 3.為了評估Transformer不同部件的重要性,我們以不同的方式改變了我們的基本模型,測量了開發集newstest2013中英德語翻譯的性能變化。我們使用前一節中描述的波束搜索,但沒有檢查點平均。這些結果見表3。

Table 4: The Transformer generalizes well to English constituency parsing (Results are on Section 23 of WSJ)
表4:Transformer很好地概括了英語選民解析(結果見《華爾街日報》第23章)

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 0.9 BLEU worse than the best setting, quality also drops off with too many heads.
在表3行(A)中,我們改變了注意頭的數量、注意鍵和值維度,保持計算量不變,如3.2.2節所述。單頭注意力比最佳設置差0.9個BLEU,但如果頭太多,質量也會下降

6.3、English Constituency Parsing

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].為了評估轉換程序是否可以推廣到其他任務,我們進行了英語選區解析的實驗。這項任務提出了具體的挑戰:產出受到強烈的結構限制,而且明顯長于投入。此外,RNN序列到序列模型無法在小數據區獲得最新的結果[37]。
?

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.

我們在《華爾街日報》(Wall Street Journal)的Penn Treebank[25]部分訓練了一個dmodel=1024的4層transformer ,大約4萬個訓練句子。我們還訓練了它在半監督設置,使用較大的高置信度和BerkleyParser語料庫從大約17M句[37 ]。我們只對WSJ設置使用了16K令牌的詞匯表,對半監督設置使用了32K令牌的詞匯表。
?
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 increased the maximum output length to input length + 300. We used a beam size of 21 and α = 0.3 for both WSJ only and the semi-supervised setting.我們只進行了少量的實驗來選擇22段開發集上的dropout、attention 和residual (第5.4節)、學習率和beam 大小,所有其他參數從英語到德語的基礎翻譯模型保持不變。在推理過程中,我們將最大輸出長度增加到輸入長度+ 300。我們使用21和α=0.3的光束尺寸,僅用于WSJ和半監督設置。
?
Our results in Table 4 show that despite the lack of task-specific tuning our model performs sur- prisingly well, yielding better results than all previously reported models with the exception of the Recurrent Neural Network Grammar [8].
In contrast to RNN sequence-to-sequence models [37], the Transformer outperforms the Berkeley- Parser [29] even when training only on the WSJ training set of 40K sentences.
我們在表4中的結果表明,盡管缺乏特定于任務的調整,我們的模型仍然運行得很好,產生的結果比以前報告的所有模型都要好,除了遞歸神經網絡語法[8]。
與RNN的序列到序列模型[37]相比,Transformer甚至在僅接受華爾街日報的40K句子訓練集訓練時,也比Berkeley- Parser[29]表現更好

7、Conclusion

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.在這項工作中,我們提出了第一個完全基于注意的序列轉換模型Transformer,它用多頭self-attention取代了編碼器-解碼器架構中最常用的遞歸層。
?
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.對于轉換任務,可以比基于遞歸層或卷積層的體系結構更快地訓練Transformer 。在WMT 2014英語-德語和WMT 2014英語-法語翻譯任務中,我們達到了一個新的水平。在前一個任務中,我們的最佳模型甚至優于所有先前報告的集成模型
?
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.我們對基于注意力的模型的未來感到興奮,并計劃將其應用到其他任務中。我們計劃將轉換器擴展到涉及文本以外的輸入和輸出模式的問題,并研究局部的、受限的注意機制,以有效地處理圖像、音頻和視頻等大型輸入和輸出。減少世代的連續性是我們的另一個研究目標。
?
The code we used to train and evaluate our models is available at https://github.com/ tensorflow/tensor2tensor.
Acknowledgements We are grateful to Nal Kalchbrenner and Stephan Gouws for their fruitful comments, corrections and inspiration.
我們用來訓練和評估模型的代碼可以在https://github.com/tensorflow/tensor2tensor上找到。感謝Nal Kalchbrenner和Stephan Gouws的富有成效的評論、更正和啟發。

References

[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.

[3] Denny Britz, Anna Goldie, Minh-Thang Luong, and Quoc V. Le. Massive exploration of neural machine translation architectures.?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.

[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.

[6] Francois Chollet. Xception: Deep learning with depthwise separable convolutions.?arXiv preprint?arXiv:1610.02357, 2016.

[7] Junyoung Chung, ?aglar Gül?ehre, Kyunghyun Cho, and Yoshua Bengio. Empirical evaluation of gated recurrent neural networks on sequence modeling.?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.

[9] Jonas Gehring, Michael Auli, David Grangier, Denis Yarats, and Yann N. Dauphin. Convolu- tional sequence to sequence learning.?arXiv preprint?arXiv:1705.03122v2, 2017.

[10] Alex Graves. Generating sequences with recurrent neural networks.?arXiv preprint?arXiv:1308.0850, 2013.

[11] Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for im- age recognition. 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.

[13] Sepp Hochreiter and Jürgen Schmidhuber. Long short-term memory.?Neural computation, 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.

[15] Rafal Jozefowicz, Oriol Vinyals, Mike Schuster, Noam Shazeer, and Yonghui Wu. Exploring the limits of language modeling.?arXiv preprint?arXiv:1602.02410, 2016.

[16] ?ukasz Kaiser and Samy Bengio. Can active memory replace attention? In?Advances in Neural Information Processing Systems, (NIPS), 2016.

[17] ?ukasz Kaiser and Ilya Sutskever. Neural GPUs learn algorithms. In?International Conference on Learning Representations (ICLR), 2016.

[18] Nal Kalchbrenner, Lasse Espeholt, Karen Simonyan, Aaron van den Oord, Alex Graves, and Ko- ray Kavukcuoglu. Neural machine translation in linear time.?arXiv preprint?arXiv:1610.10099v2, 2017.

[19] Yoon Kim, Carl Denton, Luong Hoang, and Alexander M. Rush. Structured attention networks.
In?International Conference on Learning Representations, 2017.

[20] Diederik Kingma and Jimmy Ba. Adam: A method for stochastic optimization. In?ICLR, 2015. [21] Oleksii Kuchaiev and Boris Ginsburg. Factorization tricks for LSTM networks.?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.

[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.

[24] Minh-Thang Luong, Hieu Pham, and Christopher D Manning. Effective approaches to attention- based neural machine translation.?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.

[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.

[27] Ankur Parikh, Oscar T?ckstr?m, Dipanjan Das, and Jakob Uszkoreit. A decomposable attention model. In?Empirical Methods in Natural Language Processing, 2016.

[28] Romain Paulus, Caiming Xiong, and Richard Socher. A deep reinforced model for abstractive summarization.?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 21st International Conference on Computational Linguistics and 44th Annual Meeting of the ACL, pages 433–440. ACL, July 2006.

[30] Ofir Press and Lior Wolf. Using the output embedding to improve language models.?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.

[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.

[33] Nitish Srivastava, Geoffrey E Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdi- nov. Dropout: a simple way to prevent neural networks from overfitting.?Journal of Machine Learning Research, 15(1):1929–1958, 2014.

[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.

[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.

[36] Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jonathon Shlens, and Zbigniew Wojna.?Rethinking the inception architecture for computer vision.?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.

[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.

[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.

[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.

《新程序員》:云原生和全面數字化實踐50位技術專家共同創作,文字、視頻、音頻交互閱讀

總結

以上是生活随笔為你收集整理的Paper:2017年的Google机器翻译团队《Transformer:Attention Is All You Need》翻译并解读的全部內容,希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網站內容還不錯,歡迎將生活随笔推薦給好友。

国产成人精品一区二区在线小狼 | 性史性农村dvd毛片 | 亚洲 高清 成人 动漫 | 久久综合香蕉国产蜜臀av | 久久久久久久人妻无码中文字幕爆 | 日本精品少妇一区二区三区 | 精品国产麻豆免费人成网站 | 久久99精品国产麻豆蜜芽 | 少妇性l交大片欧洲热妇乱xxx | 天堂а√在线地址中文在线 | 精品亚洲成av人在线观看 | 中文字幕色婷婷在线视频 | 亚洲精品国偷拍自产在线观看蜜桃 | 亚洲精品一区三区三区在线观看 | 国产精品理论片在线观看 | 国产凸凹视频一区二区 | 日日天日日夜日日摸 | 国产热a欧美热a在线视频 | 少妇无码一区二区二三区 | 色综合久久久无码网中文 | 亚洲日韩乱码中文无码蜜桃臀网站 | 久久久久成人片免费观看蜜芽 | 波多野结衣一区二区三区av免费 | 国产99久久精品一区二区 | 欧美xxxx黑人又粗又长 | 扒开双腿疯狂进出爽爽爽视频 | 亚洲色成人中文字幕网站 | 亚洲人成网站色7799 | 亚洲自偷自拍另类第1页 | 国产偷抇久久精品a片69 | 国产超碰人人爽人人做人人添 | 东京热男人av天堂 | 男女超爽视频免费播放 | 久久亚洲国产成人精品性色 | 免费无码午夜福利片69 | 久久亚洲中文字幕精品一区 | 亚洲欧美综合区丁香五月小说 | 国产另类ts人妖一区二区 | a片免费视频在线观看 | 日日麻批免费40分钟无码 | 成 人 网 站国产免费观看 | 色情久久久av熟女人妻网站 | 性啪啪chinese东北女人 | 亚洲乱码中文字幕在线 | 九九在线中文字幕无码 | 搡女人真爽免费视频大全 | 久久精品中文闷骚内射 | 丰腴饱满的极品熟妇 | 国产午夜无码视频在线观看 | 任你躁在线精品免费 | 国产精品久久久午夜夜伦鲁鲁 | 成人无码精品一区二区三区 | 午夜精品一区二区三区在线观看 | 精品国产精品久久一区免费式 | 精品无码av一区二区三区 | 日韩欧美中文字幕在线三区 | 性啪啪chinese东北女人 | 国产激情艳情在线看视频 | 国产免费久久久久久无码 | 精品国产一区二区三区四区在线看 | 无码成人精品区在线观看 | 婷婷综合久久中文字幕蜜桃三电影 | 亚洲另类伦春色综合小说 | 丝袜足控一区二区三区 | 亚洲精品成人福利网站 | 激情爆乳一区二区三区 | 亚洲 高清 成人 动漫 | 亚洲一区av无码专区在线观看 | 好男人www社区 | 国产精品高潮呻吟av久久4虎 | 成人影院yy111111在线观看 | 熟女少妇人妻中文字幕 | 无码国产乱人伦偷精品视频 | 亚洲日韩精品欧美一区二区 | 色婷婷香蕉在线一区二区 | 国产国语老龄妇女a片 | 人妻体内射精一区二区三四 | 亚洲中文字幕乱码av波多ji | 夜夜高潮次次欢爽av女 | 亚洲精品国产第一综合99久久 | 久久97精品久久久久久久不卡 | 中文字幕无线码 | 久久www免费人成人片 | 精品国偷自产在线 | 欧美日韩在线亚洲综合国产人 | 帮老师解开蕾丝奶罩吸乳网站 | 精品国偷自产在线视频 | 日本丰满熟妇videos | 精品 日韩 国产 欧美 视频 | 曰韩无码二三区中文字幕 | 国产人妻精品午夜福利免费 | 亚洲综合久久一区二区 | 极品尤物被啪到呻吟喷水 | 久久精品99久久香蕉国产色戒 | 久青草影院在线观看国产 | √8天堂资源地址中文在线 | 亚洲色大成网站www | 内射巨臀欧美在线视频 | 18无码粉嫩小泬无套在线观看 | 天海翼激烈高潮到腰振不止 | 久久zyz资源站无码中文动漫 | 又大又硬又黄的免费视频 | 草草网站影院白丝内射 | 日本大乳高潮视频在线观看 | 男女下面进入的视频免费午夜 | 色情久久久av熟女人妻网站 | 又湿又紧又大又爽a视频国产 | 国产一区二区不卡老阿姨 | 一本精品99久久精品77 | 色婷婷综合激情综在线播放 | 性生交大片免费看女人按摩摩 | 免费无码一区二区三区蜜桃大 | 成人aaa片一区国产精品 | 久久久精品456亚洲影院 | 无码任你躁久久久久久久 | 小鲜肉自慰网站xnxx | 国产成人无码一二三区视频 | 午夜福利一区二区三区在线观看 | 国产精品无码久久av | 一本久久a久久精品vr综合 | 国产精品视频免费播放 | 亚洲日韩中文字幕在线播放 | 少妇人妻av毛片在线看 | 国产真实伦对白全集 | 青草视频在线播放 | 亚洲欧洲日本无在线码 | 人妻少妇被猛烈进入中文字幕 | 精品久久久久香蕉网 | 亚洲色成人中文字幕网站 | 鲁鲁鲁爽爽爽在线视频观看 | 狠狠亚洲超碰狼人久久 | 久久综合香蕉国产蜜臀av | 中文字幕人成乱码熟女app | 日韩亚洲欧美精品综合 | 亚洲国产精华液网站w | 夜先锋av资源网站 | 蜜臀av无码人妻精品 | 精品 日韩 国产 欧美 视频 | 人妻少妇精品久久 | 免费网站看v片在线18禁无码 | 色五月五月丁香亚洲综合网 | 国产香蕉97碰碰久久人人 | 国产手机在线αⅴ片无码观看 | 久久99精品久久久久久动态图 | 亚洲第一网站男人都懂 | 亚洲成av人综合在线观看 | 激情爆乳一区二区三区 | 国产乱人伦偷精品视频 | 亚洲欧美国产精品久久 | 亚洲小说春色综合另类 | 精品一区二区三区无码免费视频 | 18禁止看的免费污网站 | 亚洲成av人影院在线观看 | aa片在线观看视频在线播放 | 桃花色综合影院 | 亚洲中文字幕在线无码一区二区 | 久久精品人人做人人综合试看 | 亚洲一区二区三区偷拍女厕 | 性生交大片免费看女人按摩摩 | www国产亚洲精品久久网站 | 麻豆精品国产精华精华液好用吗 | 丰满岳乱妇在线观看中字无码 | 又大又硬又爽免费视频 | 福利一区二区三区视频在线观看 | 性欧美疯狂xxxxbbbb | 老头边吃奶边弄进去呻吟 | 给我免费的视频在线观看 | 国产美女极度色诱视频www | 图片区 小说区 区 亚洲五月 | 中文字幕无码视频专区 | 妺妺窝人体色www婷婷 | 黑人巨大精品欧美黑寡妇 | 亚拍精品一区二区三区探花 | 给我免费的视频在线观看 | 色婷婷综合中文久久一本 | 成年美女黄网站色大免费全看 | 久久综合九色综合97网 | 天下第一社区视频www日本 | 成年美女黄网站色大免费视频 | 国产精品久久久久久亚洲毛片 | 日日天干夜夜狠狠爱 | 好爽又高潮了毛片免费下载 | 欧美日韩一区二区三区自拍 | 国产手机在线αⅴ片无码观看 | 伊人久久大香线焦av综合影院 | 天堂а√在线地址中文在线 | 成人aaa片一区国产精品 | 午夜理论片yy44880影院 | 波多野结衣av在线观看 | 色窝窝无码一区二区三区色欲 | 丝袜美腿亚洲一区二区 | 亚洲中文字幕va福利 | 国产精品亚洲综合色区韩国 | 久久99精品久久久久婷婷 | 国产乱码精品一品二品 | 少妇被粗大的猛进出69影院 | 精品国产aⅴ无码一区二区 | 99久久婷婷国产综合精品青草免费 | 亚洲 日韩 欧美 成人 在线观看 | 日本成熟视频免费视频 | 成 人 免费观看网站 | 国产性猛交╳xxx乱大交 国产精品久久久久久无码 欧洲欧美人成视频在线 | 欧美日韩视频无码一区二区三 | 色诱久久久久综合网ywww | 精品一区二区不卡无码av | 大地资源网第二页免费观看 | 无码人妻丰满熟妇区毛片18 | 97精品国产97久久久久久免费 | 人人澡人摸人人添 | 成人精品一区二区三区中文字幕 | 亚洲国产av精品一区二区蜜芽 | 久激情内射婷内射蜜桃人妖 | 亚洲色欲色欲天天天www | 丰满少妇人妻久久久久久 | 狠狠亚洲超碰狼人久久 | 西西人体www44rt大胆高清 | 麻豆成人精品国产免费 | 国产精品久久久久久亚洲影视内衣 | 国产又爽又猛又粗的视频a片 | 老司机亚洲精品影院无码 | 精品欧洲av无码一区二区三区 | 亚洲精品一区二区三区四区五区 | 久久久久成人精品免费播放动漫 | 波多野结衣aⅴ在线 | 日产精品高潮呻吟av久久 | 中国女人内谢69xxxx | 成年女人永久免费看片 | 免费观看又污又黄的网站 | 成人综合网亚洲伊人 | 亚洲国产高清在线观看视频 | 亚洲精品美女久久久久久久 | 性做久久久久久久久 | 国产精品无码一区二区三区不卡 | 纯爱无遮挡h肉动漫在线播放 | 日日夜夜撸啊撸 | 亚洲aⅴ无码成人网站国产app | 丝袜美腿亚洲一区二区 | 久久精品女人天堂av免费观看 | 亚洲综合无码一区二区三区 | 麻豆精品国产精华精华液好用吗 | 国产特级毛片aaaaaa高潮流水 | 男人扒开女人内裤强吻桶进去 | 久久人妻内射无码一区三区 | 无码福利日韩神码福利片 | 亚洲一区二区三区 | 一个人看的视频www在线 | 无码国产色欲xxxxx视频 | 国产精品亚洲一区二区三区喷水 | 精品人妻中文字幕有码在线 | 精品久久久久久亚洲精品 | 亚洲精品综合五月久久小说 | 国产高清av在线播放 | 西西人体www44rt大胆高清 | 国产欧美精品一区二区三区 | 一本久道高清无码视频 | 亚洲va中文字幕无码久久不卡 | 欧美精品国产综合久久 | 少妇激情av一区二区 | 成人精品天堂一区二区三区 | 男女猛烈xx00免费视频试看 | 妺妺窝人体色www婷婷 | 丰满肥臀大屁股熟妇激情视频 | 国产在线精品一区二区高清不卡 | 亚洲精品欧美二区三区中文字幕 | 国产精品美女久久久久av爽李琼 | 黑森林福利视频导航 | 久久精品国产99精品亚洲 | 蜜臀av在线播放 久久综合激激的五月天 | 99久久久无码国产aaa精品 | 精品熟女少妇av免费观看 | 女人被男人躁得好爽免费视频 | 婷婷综合久久中文字幕蜜桃三电影 | 噜噜噜亚洲色成人网站 | 国产成人综合美国十次 | 成熟妇人a片免费看网站 | 大肉大捧一进一出视频出来呀 | 香港三级日本三级妇三级 | 1000部夫妻午夜免费 | www国产亚洲精品久久久日本 | 亚洲日韩av一区二区三区四区 | 又大又黄又粗又爽的免费视频 | 亚洲va中文字幕无码久久不卡 | 免费中文字幕日韩欧美 | 欧美午夜特黄aaaaaa片 | 日韩精品无码免费一区二区三区 | 人人妻人人澡人人爽人人精品 | 波多野结衣乳巨码无在线观看 | 国内精品九九久久久精品 | 骚片av蜜桃精品一区 | 亚洲国产精品毛片av不卡在线 | 亚洲成色www久久网站 | 国产在线精品一区二区高清不卡 | 精品久久久久久亚洲精品 | 亚洲精品综合一区二区三区在线 | 国产精品-区区久久久狼 | 久久亚洲国产成人精品性色 | 丰满人妻被黑人猛烈进入 | 夜夜影院未满十八勿进 | 精品一二三区久久aaa片 | 精品久久久久久人妻无码中文字幕 | 精品国偷自产在线 | 欧美黑人巨大xxxxx | 色综合久久久久综合一本到桃花网 | 亚洲 高清 成人 动漫 | 伊人久久大香线焦av综合影院 | 好屌草这里只有精品 | 人妻人人添人妻人人爱 | 国产xxx69麻豆国语对白 | 色欲综合久久中文字幕网 | 性欧美熟妇videofreesex | 精品乱子伦一区二区三区 | 日本一卡2卡3卡4卡无卡免费网站 国产一区二区三区影院 | 国产手机在线αⅴ片无码观看 | 成人免费视频一区二区 | 欧美日韩视频无码一区二区三 | 人妻无码久久精品人妻 | √天堂资源地址中文在线 | 日韩精品无码免费一区二区三区 | 又黄又爽又色的视频 | 久久久成人毛片无码 | 国产av无码专区亚洲a∨毛片 | 在线观看国产一区二区三区 | 亚洲欧美中文字幕5发布 | 黑人玩弄人妻中文在线 | 一本久久伊人热热精品中文字幕 | 丰满诱人的人妻3 | 精品亚洲韩国一区二区三区 | 动漫av网站免费观看 | 国产精品va在线播放 | 国产超级va在线观看视频 | 国产av无码专区亚洲a∨毛片 | 初尝人妻少妇中文字幕 | 97精品国产97久久久久久免费 | 国产精品久久国产精品99 | 人妻少妇被猛烈进入中文字幕 | 男女作爱免费网站 | 久久久精品欧美一区二区免费 | 亚洲精品www久久久 | 国产av人人夜夜澡人人爽麻豆 | 一本久道久久综合婷婷五月 | 亚洲中文字幕在线观看 | 久久午夜无码鲁丝片秋霞 | 国产成人无码a区在线观看视频app | 在线观看国产一区二区三区 | 久久久久免费精品国产 | 国产成人亚洲综合无码 | 老头边吃奶边弄进去呻吟 | 欧美精品免费观看二区 | 国产后入清纯学生妹 | 国产午夜无码视频在线观看 | 欧美丰满熟妇xxxx | 精品夜夜澡人妻无码av蜜桃 | 久久精品99久久香蕉国产色戒 | 无码av最新清无码专区吞精 | 午夜理论片yy44880影院 | 亚洲精品综合五月久久小说 | 精品一区二区三区无码免费视频 | 久久人妻内射无码一区三区 | 久久精品女人天堂av免费观看 | 人妻熟女一区 | 国产成人人人97超碰超爽8 | 少妇性俱乐部纵欲狂欢电影 | 久久久无码中文字幕久... | 亚洲人成人无码网www国产 | av无码久久久久不卡免费网站 | 久久zyz资源站无码中文动漫 | 欧美 日韩 亚洲 在线 | 午夜精品久久久内射近拍高清 | 性生交片免费无码看人 | 无码人妻av免费一区二区三区 | 2019午夜福利不卡片在线 | 国产超级va在线观看视频 | 日韩av无码一区二区三区不卡 | 精品无码国产一区二区三区av | 国产亚洲精品久久久久久久久动漫 | 亚洲国产精品无码一区二区三区 | 国内丰满熟女出轨videos | 麻豆国产97在线 | 欧洲 | 图片小说视频一区二区 | 国产激情一区二区三区 | 波多野结衣av一区二区全免费观看 | 久久久av男人的天堂 | 日韩无码专区 | 131美女爱做视频 | 国产精品无码mv在线观看 | 国产特级毛片aaaaaaa高清 | 婷婷综合久久中文字幕蜜桃三电影 | 免费人成网站视频在线观看 | 夜夜影院未满十八勿进 | 少女韩国电视剧在线观看完整 | aⅴ亚洲 日韩 色 图网站 播放 | 国产人妻精品一区二区三区 | 国产精品久久福利网站 | 国产精品va在线观看无码 | 欧美乱妇无乱码大黄a片 | 精品无码一区二区三区爱欲 | 99久久99久久免费精品蜜桃 | 天堂在线观看www | 欧洲熟妇精品视频 | 无码帝国www无码专区色综合 | 欧美刺激性大交 | 激情内射日本一区二区三区 | 在线精品国产一区二区三区 | 人妻互换免费中文字幕 | 欧美猛少妇色xxxxx | 2020最新国产自产精品 | 一本一道久久综合久久 | 欧美刺激性大交 | 久久亚洲国产成人精品性色 | 国产精华av午夜在线观看 | 亚洲成色www久久网站 | 国产性生交xxxxx无码 | 亚洲无人区午夜福利码高清完整版 | 欧美 丝袜 自拍 制服 另类 | 娇妻被黑人粗大高潮白浆 | 久久精品人妻少妇一区二区三区 | 欧美怡红院免费全部视频 | 国产无遮挡又黄又爽又色 | 精品日本一区二区三区在线观看 | 亚洲无人区午夜福利码高清完整版 | 麻豆国产丝袜白领秘书在线观看 | 全黄性性激高免费视频 | 国产乱人伦av在线无码 | 亚洲国产精品一区二区美利坚 | 自拍偷自拍亚洲精品被多人伦好爽 | 成人女人看片免费视频放人 | 精品久久久久久人妻无码中文字幕 | 亚洲成a人片在线观看无码 | 在线精品亚洲一区二区 | 免费无码一区二区三区蜜桃大 | 日本精品少妇一区二区三区 | 高清国产亚洲精品自在久久 | 色爱情人网站 | 亚洲高清偷拍一区二区三区 | 欧美freesex黑人又粗又大 | 国产精品视频免费播放 | 中文字幕色婷婷在线视频 | 国产成人无码av片在线观看不卡 | 成人一在线视频日韩国产 | 国产免费观看黄av片 | 欧美日韩综合一区二区三区 | 久久久国产一区二区三区 | 国产内射老熟女aaaa | 一二三四社区在线中文视频 | 精品国偷自产在线 | 亚洲精品国产a久久久久久 | 国产精品-区区久久久狼 | 伊人久久大香线蕉亚洲 | 蜜桃av蜜臀av色欲av麻 999久久久国产精品消防器材 | 亚洲伊人久久精品影院 | 玩弄少妇高潮ⅹxxxyw | 国产精品久久福利网站 | 粉嫩少妇内射浓精videos | 乱码午夜-极国产极内射 | 亚洲色无码一区二区三区 | 亚洲成a人片在线观看日本 | 人妻有码中文字幕在线 | 四虎国产精品免费久久 | 久久久精品国产sm最大网站 | 久久久av男人的天堂 | 日韩亚洲欧美中文高清在线 | 国产精品人人爽人人做我的可爱 | 久久久久久a亚洲欧洲av冫 | 久久久久久久久888 | 欧美一区二区三区视频在线观看 | 久久久www成人免费毛片 | 亚洲午夜无码久久 | 在线观看国产午夜福利片 | 人人妻人人藻人人爽欧美一区 | 国产综合久久久久鬼色 | 国内精品一区二区三区不卡 | 亚洲国产高清在线观看视频 | 久久精品一区二区三区四区 | 好男人社区资源 | 日韩欧美中文字幕在线三区 | 国产av剧情md精品麻豆 | 麻豆国产丝袜白领秘书在线观看 | 精品国产一区二区三区av 性色 | 亚洲熟妇色xxxxx亚洲 | 美女扒开屁股让男人桶 | 欧美日本免费一区二区三区 | 丰满妇女强制高潮18xxxx | 国产绳艺sm调教室论坛 | 97久久精品无码一区二区 | 少妇厨房愉情理9仑片视频 | 午夜精品一区二区三区在线观看 | 领导边摸边吃奶边做爽在线观看 | 久热国产vs视频在线观看 | 精品一区二区不卡无码av | 最新版天堂资源中文官网 | 少妇无码av无码专区在线观看 | 奇米综合四色77777久久 东京无码熟妇人妻av在线网址 | 无码人妻久久一区二区三区不卡 | 国内少妇偷人精品视频免费 | 特大黑人娇小亚洲女 | 少妇厨房愉情理9仑片视频 | 国产成人无码av片在线观看不卡 | 国产精品福利视频导航 | 亚无码乱人伦一区二区 | 国产精品视频免费播放 | 成人无码影片精品久久久 | 人人妻人人澡人人爽欧美一区九九 | 99久久精品午夜一区二区 | 熟妇激情内射com | 亚洲爆乳无码专区 | 无码纯肉视频在线观看 | 国产成人精品三级麻豆 | 97久久精品无码一区二区 | 人妻少妇精品久久 | 麻花豆传媒剧国产免费mv在线 | 欧美丰满熟妇xxxx性ppx人交 | 又色又爽又黄的美女裸体网站 | 日韩成人一区二区三区在线观看 | 亚洲国产欧美日韩精品一区二区三区 | 全黄性性激高免费视频 | 亚洲国精产品一二二线 | 亚洲中文字幕av在天堂 | 久久亚洲精品中文字幕无男同 | 精品国产青草久久久久福利 | 亚洲欧美日韩综合久久久 | 性欧美疯狂xxxxbbbb | 国产亚洲精品精品国产亚洲综合 | 又大又黄又粗又爽的免费视频 | 大胆欧美熟妇xx | 国产亚洲人成在线播放 | 亚洲理论电影在线观看 | 久久精品中文字幕一区 | 亚洲日韩乱码中文无码蜜桃臀网站 | 丰满人妻翻云覆雨呻吟视频 | 老熟女重囗味hdxx69 | 无码午夜成人1000部免费视频 | 丰满人妻一区二区三区免费视频 | 天海翼激烈高潮到腰振不止 | 国内少妇偷人精品视频 | 啦啦啦www在线观看免费视频 | 久久97精品久久久久久久不卡 | 国产艳妇av在线观看果冻传媒 | 成人无码视频在线观看网站 | 久久久无码中文字幕久... | 国产亚洲人成在线播放 | 国产电影无码午夜在线播放 | 国产乱人伦偷精品视频 | 天堂亚洲2017在线观看 | 乌克兰少妇xxxx做受 | 青青青手机频在线观看 | 欧美国产日韩亚洲中文 | 狠狠色丁香久久婷婷综合五月 | 国产在线无码精品电影网 | 狠狠色丁香久久婷婷综合五月 | 国产精品久久久久久无码 | 亚洲乱码日产精品bd | 国产精品二区一区二区aⅴ污介绍 | 2019午夜福利不卡片在线 | 亚洲熟妇色xxxxx欧美老妇y | 亚洲自偷自拍另类第1页 | 亚洲日韩精品欧美一区二区 | 亚洲 a v无 码免 费 成 人 a v | 欧美野外疯狂做受xxxx高潮 | 国产亚洲欧美日韩亚洲中文色 | 男人的天堂av网站 | 婷婷色婷婷开心五月四房播播 | 欧美日韩视频无码一区二区三 | 初尝人妻少妇中文字幕 | 最新国产麻豆aⅴ精品无码 | 日产精品高潮呻吟av久久 | 亚洲精品久久久久久一区二区 | 中文字幕无码日韩专区 | 色一情一乱一伦一区二区三欧美 | 国产黄在线观看免费观看不卡 | 亚洲男女内射在线播放 | 欧美人妻一区二区三区 | 牲欲强的熟妇农村老妇女视频 | 久久精品国产99精品亚洲 | 在线亚洲高清揄拍自拍一品区 | 狠狠躁日日躁夜夜躁2020 | 亚洲区小说区激情区图片区 | 波多野结衣 黑人 | 人妻少妇精品无码专区二区 | 中文无码成人免费视频在线观看 | 国产va免费精品观看 | 久久亚洲精品中文字幕无男同 | 久久精品国产99精品亚洲 | 久久久久久亚洲精品a片成人 | 欧美日本免费一区二区三区 | 色窝窝无码一区二区三区色欲 | 无码国产乱人伦偷精品视频 | 无遮挡国产高潮视频免费观看 | 狠狠cao日日穞夜夜穞av | 好男人社区资源 | 乱中年女人伦av三区 | 欧美激情内射喷水高潮 | 亚洲国产精品久久久久久 | 亚洲熟妇色xxxxx欧美老妇 | 成人av无码一区二区三区 | 免费无码午夜福利片69 | 亚洲精品午夜国产va久久成人 | 亚洲va欧美va天堂v国产综合 | 亚洲欧美国产精品久久 | 乌克兰少妇xxxx做受 | 一二三四在线观看免费视频 | 亚洲大尺度无码无码专区 | 中文字幕无码av波多野吉衣 | 亚洲一区二区三区播放 | 强辱丰满人妻hd中文字幕 | 国产精品鲁鲁鲁 | 日本精品少妇一区二区三区 | 欧美人与动性行为视频 | 少妇性l交大片欧洲热妇乱xxx | 欧美老妇交乱视频在线观看 | 日本爽爽爽爽爽爽在线观看免 | 99精品视频在线观看免费 | 99久久精品国产一区二区蜜芽 | 无码人妻少妇伦在线电影 | 精品国产一区av天美传媒 | 国产人妻人伦精品 | 国产人妻人伦精品1国产丝袜 | 久久久精品成人免费观看 | 国产精品久久久久久亚洲影视内衣 | 夜先锋av资源网站 | 中文字幕无码免费久久9一区9 | 国产香蕉尹人综合在线观看 | 国产97在线 | 亚洲 | 一本精品99久久精品77 | 国产一区二区三区精品视频 | 伊人久久大香线蕉亚洲 | 精品无人区无码乱码毛片国产 | 国产精品免费大片 | 国产麻豆精品精东影业av网站 | 一本久久伊人热热精品中文字幕 | 老太婆性杂交欧美肥老太 | 亚洲自偷自偷在线制服 | 久久久久99精品国产片 | 久久国产劲爆∧v内射 | 最近的中文字幕在线看视频 | 免费男性肉肉影院 | 四虎永久在线精品免费网址 | 亚洲男人av香蕉爽爽爽爽 | 日韩欧美中文字幕在线三区 | 人人妻人人藻人人爽欧美一区 | 国产亚洲视频中文字幕97精品 | 理论片87福利理论电影 | 精品一区二区不卡无码av | 精品久久久中文字幕人妻 | 亚洲欧美精品伊人久久 | 亚洲小说春色综合另类 | 亚洲精品一区二区三区四区五区 | 国产人妻人伦精品1国产丝袜 | 国产精品香蕉在线观看 | 动漫av网站免费观看 | √天堂资源地址中文在线 | 久久精品国产一区二区三区 | 日日天干夜夜狠狠爱 | 国产真人无遮挡作爱免费视频 | aⅴ在线视频男人的天堂 | 亚洲啪av永久无码精品放毛片 | 久久99精品久久久久婷婷 | 亚洲日韩一区二区 | 国产农村妇女高潮大叫 | 少妇无码吹潮 | 欧美老熟妇乱xxxxx | 亚洲成av人在线观看网址 | 一个人看的www免费视频在线观看 | 亚洲无人区一区二区三区 | 在线天堂新版最新版在线8 | 久久午夜夜伦鲁鲁片无码免费 | 成 人 网 站国产免费观看 | 97夜夜澡人人爽人人喊中国片 | 精品久久久久久人妻无码中文字幕 | 国产性猛交╳xxx乱大交 国产精品久久久久久无码 欧洲欧美人成视频在线 | 国产精品亚洲а∨无码播放麻豆 | 天堂无码人妻精品一区二区三区 | 国产精品欧美成人 | 久久久久成人精品免费播放动漫 | 国产真人无遮挡作爱免费视频 | 激情综合激情五月俺也去 | 精品久久久久久人妻无码中文字幕 | 伊人久久大香线蕉午夜 | 午夜精品一区二区三区在线观看 | 全黄性性激高免费视频 | www国产亚洲精品久久久日本 | 免费中文字幕日韩欧美 | 日本一卡2卡3卡四卡精品网站 | 国产日产欧产精品精品app | 亚洲中文字幕在线无码一区二区 | 亚洲一区二区三区国产精华液 | 蜜桃视频插满18在线观看 | 97夜夜澡人人爽人人喊中国片 | 粗大的内捧猛烈进出视频 | 亚洲精品久久久久久一区二区 | 久久亚洲精品成人无码 | 人人澡人人妻人人爽人人蜜桃 | 国产在线精品一区二区高清不卡 | 国产 浪潮av性色四虎 | 扒开双腿疯狂进出爽爽爽视频 | 久久久久人妻一区精品色欧美 | 午夜精品久久久久久久久 | 中文字幕无码热在线视频 | 丰满人妻被黑人猛烈进入 | 在线精品亚洲一区二区 | 丝袜美腿亚洲一区二区 | 四十如虎的丰满熟妇啪啪 | 国产激情艳情在线看视频 | 国产精品免费大片 | 天堂无码人妻精品一区二区三区 | 亚洲 日韩 欧美 成人 在线观看 | 永久免费观看国产裸体美女 | 人人妻人人澡人人爽欧美一区 | 亚洲欧美国产精品久久 | 亚洲va中文字幕无码久久不卡 | 亚洲乱码中文字幕在线 | 国产在线精品一区二区三区直播 | 黑人巨大精品欧美一区二区 | 久久久www成人免费毛片 | 久久国产自偷自偷免费一区调 | 婷婷五月综合激情中文字幕 | 奇米影视7777久久精品 | 最近中文2019字幕第二页 | 狠狠色欧美亚洲狠狠色www | 亚洲 欧美 激情 小说 另类 | 久久国产36精品色熟妇 | 国产一区二区不卡老阿姨 | 丰满少妇女裸体bbw | 国产精品无码一区二区三区不卡 | 国产精品人人妻人人爽 | 99久久精品无码一区二区毛片 | 18禁黄网站男男禁片免费观看 | 亚洲综合无码久久精品综合 | 久久综合网欧美色妞网 | 亚洲日本一区二区三区在线 | 内射白嫩少妇超碰 | 小sao货水好多真紧h无码视频 | 国产精品国产三级国产专播 | 精品国偷自产在线视频 | 日韩亚洲欧美精品综合 | 中文字幕中文有码在线 | 色综合久久88色综合天天 | 青青青手机频在线观看 | 无码任你躁久久久久久久 | 岛国片人妻三上悠亚 | 97久久国产亚洲精品超碰热 | ass日本丰满熟妇pics | 成人精品视频一区二区 | 永久免费精品精品永久-夜色 | 日本精品久久久久中文字幕 | 国产内射爽爽大片视频社区在线 | 男女爱爱好爽视频免费看 | 丁香花在线影院观看在线播放 | 中文字幕久久久久人妻 | 国产热a欧美热a在线视频 | 亚洲熟妇色xxxxx欧美老妇y | 国产精品怡红院永久免费 | 亚洲呦女专区 | а√资源新版在线天堂 | 99久久久无码国产aaa精品 | 国产麻豆精品一区二区三区v视界 | 欧美一区二区三区视频在线观看 | 四虎影视成人永久免费观看视频 | 国产成人一区二区三区别 | 纯爱无遮挡h肉动漫在线播放 | 国产一区二区三区日韩精品 | 亚洲一区av无码专区在线观看 | 18无码粉嫩小泬无套在线观看 | 午夜熟女插插xx免费视频 | 久久综合香蕉国产蜜臀av | 四虎国产精品免费久久 | 久久久无码中文字幕久... | 高潮毛片无遮挡高清免费 | 无码帝国www无码专区色综合 | 亚洲精品一区二区三区婷婷月 | 丝袜人妻一区二区三区 | 久久亚洲精品中文字幕无男同 | 夜精品a片一区二区三区无码白浆 | 国产精品99爱免费视频 | 麻豆国产人妻欲求不满 | 老太婆性杂交欧美肥老太 | 无套内射视频囯产 | av在线亚洲欧洲日产一区二区 | 欧美日韩色另类综合 | 老子影院午夜伦不卡 | 女人被男人爽到呻吟的视频 | 亚洲人成网站色7799 | 乱码av麻豆丝袜熟女系列 | 国产成人一区二区三区在线观看 | 欧美真人作爱免费视频 | 岛国片人妻三上悠亚 | 久久国产精品二国产精品 | 国产va免费精品观看 | 久久国产36精品色熟妇 | 久久亚洲精品成人无码 | 少妇无码一区二区二三区 | 日韩亚洲欧美中文高清在线 | 无码人妻精品一区二区三区不卡 | a片免费视频在线观看 | 国产麻豆精品一区二区三区v视界 | 欧美喷潮久久久xxxxx | av无码久久久久不卡免费网站 | 性色av无码免费一区二区三区 | 日本丰满熟妇videos | 国产艳妇av在线观看果冻传媒 | 午夜肉伦伦影院 | 波多野结衣av在线观看 | 亚洲熟妇色xxxxx欧美老妇y | 麻豆md0077饥渴少妇 | 日本饥渴人妻欲求不满 | 国产精品久久久久久亚洲影视内衣 | 亚洲精品美女久久久久久久 | 亚洲午夜福利在线观看 | 偷窥日本少妇撒尿chinese | 正在播放东北夫妻内射 | 亚洲小说春色综合另类 | 久久国产劲爆∧v内射 | 久久综合香蕉国产蜜臀av | 人妻夜夜爽天天爽三区 | 天天av天天av天天透 | 高清无码午夜福利视频 | а√资源新版在线天堂 | 成人av无码一区二区三区 | 亚洲精品国偷拍自产在线观看蜜桃 | 国内综合精品午夜久久资源 | 野外少妇愉情中文字幕 | 夜夜高潮次次欢爽av女 | 熟妇人妻激情偷爽文 | 一本久道久久综合狠狠爱 | 夜夜躁日日躁狠狠久久av | 国产亚洲欧美日韩亚洲中文色 | 精品久久久无码人妻字幂 | 午夜精品久久久内射近拍高清 | 日本熟妇大屁股人妻 | 一本色道久久综合亚洲精品不卡 | 午夜精品一区二区三区的区别 | 亚欧洲精品在线视频免费观看 | 久久精品中文闷骚内射 | 成人欧美一区二区三区黑人 | 99久久精品日本一区二区免费 | 丁香花在线影院观看在线播放 | 永久免费精品精品永久-夜色 | 国产av一区二区三区最新精品 | 男女猛烈xx00免费视频试看 | 一本久道久久综合婷婷五月 | 亚洲国产精品无码久久久久高潮 | 亚洲日韩av片在线观看 | 一本大道伊人av久久综合 | 国产麻豆精品一区二区三区v视界 | 伦伦影院午夜理论片 | 国产一区二区三区影院 | 激情内射日本一区二区三区 | 国产xxx69麻豆国语对白 | 图片小说视频一区二区 | 久久伊人色av天堂九九小黄鸭 | 99riav国产精品视频 | 久久99久久99精品中文字幕 | 久久99精品久久久久久 | 亚洲天堂2017无码中文 | 亚洲国产日韩a在线播放 | 国产精品美女久久久 | 亚洲男人av天堂午夜在 | 午夜精品久久久久久久久 | 99久久精品午夜一区二区 | 亚洲日韩一区二区三区 | 国内精品一区二区三区不卡 | 牲欲强的熟妇农村老妇女 | 日本精品人妻无码免费大全 | 欧美日韩视频无码一区二区三 | 人妻无码αv中文字幕久久琪琪布 | 国产人妻精品一区二区三区不卡 | 亚洲无人区午夜福利码高清完整版 | 亚洲精品国产品国语在线观看 | 性开放的女人aaa片 | 国产精品久久久午夜夜伦鲁鲁 | 久久精品丝袜高跟鞋 | 性欧美videos高清精品 | 亚洲色欲色欲欲www在线 | 亚洲欧美国产精品专区久久 | 荫蒂被男人添的好舒服爽免费视频 | 久久精品女人天堂av免费观看 | 国产精品无码成人午夜电影 | 人妻无码久久精品人妻 | 精品国产福利一区二区 | 国内精品人妻无码久久久影院蜜桃 | 久久久久亚洲精品中文字幕 | 精品国精品国产自在久国产87 | 夜先锋av资源网站 | 午夜精品一区二区三区在线观看 | 领导边摸边吃奶边做爽在线观看 | 丰满人妻一区二区三区免费视频 | 国产人成高清在线视频99最全资源 | 成年美女黄网站色大免费全看 | 国产国语老龄妇女a片 | 成人欧美一区二区三区黑人免费 | 女人高潮内射99精品 | 国产精品igao视频网 | 国产乱码精品一品二品 | 性生交大片免费看女人按摩摩 | 欧美xxxxx精品 | 国产精品对白交换视频 | 精品无码av一区二区三区 | 九九热爱视频精品 | 99久久人妻精品免费一区 | 国语精品一区二区三区 | 性做久久久久久久免费看 | 久久午夜夜伦鲁鲁片无码免费 | 丰满肥臀大屁股熟妇激情视频 | 国产性猛交╳xxx乱大交 国产精品久久久久久无码 欧洲欧美人成视频在线 | 久久精品女人天堂av免费观看 | 免费乱码人妻系列无码专区 | 国产两女互慰高潮视频在线观看 | 成人精品一区二区三区中文字幕 | 麻豆精品国产精华精华液好用吗 | 精品少妇爆乳无码av无码专区 | 动漫av一区二区在线观看 | 欧美人与牲动交xxxx | 爽爽影院免费观看 | 欧美黑人性暴力猛交喷水 | 99精品国产综合久久久久五月天 | 奇米影视7777久久精品 | 无码人中文字幕 | a国产一区二区免费入口 | 18黄暴禁片在线观看 | 久久99国产综合精品 | 人人澡人人透人人爽 | 无码人妻久久一区二区三区不卡 | 国产一区二区不卡老阿姨 | 久久精品国产99久久6动漫 | 久久国内精品自在自线 | av无码不卡在线观看免费 | 国产成人一区二区三区别 | 国产午夜手机精彩视频 | 亚洲精品久久久久久一区二区 | 国产成人无码专区 | 国产亚洲精品久久久久久 | 久久久久久久女国产乱让韩 | 久久人人爽人人爽人人片av高清 | 熟妇女人妻丰满少妇中文字幕 | 亚洲精品www久久久 | 国产国语老龄妇女a片 | 99久久精品午夜一区二区 | 国产极品视觉盛宴 | 性啪啪chinese东北女人 | 成人动漫在线观看 | 亚洲熟妇色xxxxx欧美老妇 | 中文字幕无码人妻少妇免费 | 国内老熟妇对白xxxxhd | 亚洲精品国偷拍自产在线观看蜜桃 | 在线天堂新版最新版在线8 | аⅴ资源天堂资源库在线 | 亚洲精品中文字幕久久久久 | 少妇人妻av毛片在线看 | 国产成人av免费观看 | 久久久久免费精品国产 | 久精品国产欧美亚洲色aⅴ大片 | 麻豆精品国产精华精华液好用吗 | 激情五月综合色婷婷一区二区 | 中文字幕无码av激情不卡 | 成熟女人特级毛片www免费 | 国产超级va在线观看视频 | 日韩亚洲欧美中文高清在线 | 少妇高潮一区二区三区99 | 强辱丰满人妻hd中文字幕 | a国产一区二区免费入口 | 性做久久久久久久免费看 | 国产又爽又猛又粗的视频a片 | 国产偷抇久久精品a片69 | 中文字幕无码日韩专区 | 影音先锋中文字幕无码 | 伊人久久大香线蕉午夜 | 高潮毛片无遮挡高清免费视频 | 亚洲小说图区综合在线 | 一本色道久久综合狠狠躁 | 国产精品爱久久久久久久 | 亚洲中文字幕乱码av波多ji | 国产一区二区三区精品视频 | 亚洲色大成网站www | 亚洲人亚洲人成电影网站色 | 青草视频在线播放 | 日韩精品无码一本二本三本色 | 国产精品久久久久7777 | 任你躁在线精品免费 | 激情国产av做激情国产爱 | 免费国产黄网站在线观看 | 国产超级va在线观看视频 | 国产97在线 | 亚洲 | 国产亲子乱弄免费视频 | 天堂在线观看www | 日韩亚洲欧美精品综合 | 日本高清一区免费中文视频 | 爆乳一区二区三区无码 | 中文字幕乱妇无码av在线 | 国产精品无码一区二区桃花视频 | 亚洲熟熟妇xxxx | 亚洲欧洲无卡二区视頻 | 久久熟妇人妻午夜寂寞影院 | 在教室伦流澡到高潮hnp视频 | 免费无码av一区二区 | 伊人久久婷婷五月综合97色 | 又粗又大又硬毛片免费看 | 性色欲网站人妻丰满中文久久不卡 | 国产av久久久久精东av | 成人毛片一区二区 | 天海翼激烈高潮到腰振不止 | 国产九九九九九九九a片 | 少妇无码av无码专区在线观看 | 色综合久久久无码网中文 | 大色综合色综合网站 | 熟女体下毛毛黑森林 | 小鲜肉自慰网站xnxx | 国产人妻人伦精品1国产丝袜 | 亚洲一区二区三区含羞草 | 3d动漫精品啪啪一区二区中 | 精品欧美一区二区三区久久久 | 国产精品丝袜黑色高跟鞋 | 国产精品亚洲专区无码不卡 | 亚洲精品成a人在线观看 | 久久97精品久久久久久久不卡 | 久久人人爽人人人人片 | 欧美日韩精品 | 国产精品久久福利网站 | 欧美 日韩 人妻 高清 中文 | 美女极度色诱视频国产 | 亚洲色成人中文字幕网站 | 日韩欧美成人免费观看 | 纯爱无遮挡h肉动漫在线播放 | 乱人伦人妻中文字幕无码久久网 | 乱人伦中文视频在线观看 | 亚洲色大成网站www国产 | 亚洲精品一区二区三区在线 | 日日摸日日碰夜夜爽av | 亚洲日韩av片在线观看 | 久久精品国产一区二区三区肥胖 | 十八禁视频网站在线观看 | 亚洲精品久久久久久久久久久 | 亚洲成在人网站无码天堂 | 爆乳一区二区三区无码 | 夜夜高潮次次欢爽av女 | 噜噜噜亚洲色成人网站 | 在线播放免费人成毛片乱码 | 欧美老妇与禽交 | 黑人大群体交免费视频 | 中文久久乱码一区二区 | 久久午夜夜伦鲁鲁片无码免费 | 欧美老妇交乱视频在线观看 | 美女扒开屁股让男人桶 | 伊人色综合久久天天小片 | 欧美黑人性暴力猛交喷水 | 两性色午夜视频免费播放 | 亚洲日韩av一区二区三区中文 | 一本久久a久久精品vr综合 | 无码毛片视频一区二区本码 | 色窝窝无码一区二区三区色欲 | 狠狠色噜噜狠狠狠7777奇米 | 精品国产麻豆免费人成网站 | 国产av剧情md精品麻豆 | 久久精品女人天堂av免费观看 | 又大又黄又粗又爽的免费视频 | 噜噜噜亚洲色成人网站 | 国产高潮视频在线观看 | 日本爽爽爽爽爽爽在线观看免 | 天堂а√在线中文在线 | 国产乱人伦偷精品视频 | 丁香花在线影院观看在线播放 | 色欲av亚洲一区无码少妇 | 全黄性性激高免费视频 | 特级做a爰片毛片免费69 | 欧美国产日韩亚洲中文 | 色偷偷av老熟女 久久精品人妻少妇一区二区三区 | 好爽又高潮了毛片免费下载 | 国产高潮视频在线观看 | 日日天干夜夜狠狠爱 | 日本一卡2卡3卡4卡无卡免费网站 国产一区二区三区影院 | 妺妺窝人体色www在线小说 | 国产美女精品一区二区三区 | 国产成人无码av在线影院 | 亚洲娇小与黑人巨大交 | 2020久久香蕉国产线看观看 | 久久久精品成人免费观看 | 任你躁在线精品免费 | 伊人久久婷婷五月综合97色 | 欧美日韩久久久精品a片 | 日韩精品成人一区二区三区 | 熟妇女人妻丰满少妇中文字幕 | 亚洲精品美女久久久久久久 | 澳门永久av免费网站 | 六十路熟妇乱子伦 | 国产精品内射视频免费 | 国产9 9在线 | 中文 | 四虎永久在线精品免费网址 | 国产精品无码永久免费888 | 精品久久久无码中文字幕 | 九月婷婷人人澡人人添人人爽 | 双乳奶水饱满少妇呻吟 | 成人免费无码大片a毛片 | 久久综合香蕉国产蜜臀av | 暴力强奷在线播放无码 | 久久国产精品_国产精品 | 日日摸夜夜摸狠狠摸婷婷 | 国产人妻精品一区二区三区 | 窝窝午夜理论片影院 | 99麻豆久久久国产精品免费 | www国产亚洲精品久久久日本 | 高清无码午夜福利视频 | 日本一区二区更新不卡 | 日本护士毛茸茸高潮 | 无遮挡啪啪摇乳动态图 | 乱码av麻豆丝袜熟女系列 | 亚洲a无码综合a国产av中文 | 天天做天天爱天天爽综合网 | 欧美亚洲国产一区二区三区 | 日本免费一区二区三区最新 | 精品无码国产自产拍在线观看蜜 | 在线 国产 欧美 亚洲 天堂 | 未满成年国产在线观看 | 国产精品毛多多水多 | 欧美第一黄网免费网站 | 国产精品亚洲五月天高清 | 99久久精品无码一区二区毛片 | 国产人妻久久精品二区三区老狼 | 久久综合激激的五月天 | 国产9 9在线 | 中文 | 免费乱码人妻系列无码专区 | 国产午夜亚洲精品不卡 | 亚洲国产精品无码一区二区三区 | 国产亚洲视频中文字幕97精品 | 亚洲日韩乱码中文无码蜜桃臀网站 | 性欧美熟妇videofreesex | 性开放的女人aaa片 | 国产精品人妻一区二区三区四 | 国内综合精品午夜久久资源 | 国产精品久久福利网站 | 精品无码国产一区二区三区av | 色偷偷人人澡人人爽人人模 | 夜精品a片一区二区三区无码白浆 | 亚洲欧洲中文日韩av乱码 | 麻豆人妻少妇精品无码专区 | 夜夜高潮次次欢爽av女 | 曰韩无码二三区中文字幕 | 少妇性l交大片欧洲热妇乱xxx | 夜精品a片一区二区三区无码白浆 | 天堂无码人妻精品一区二区三区 | 麻花豆传媒剧国产免费mv在线 | 国产极品美女高潮无套在线观看 | 日韩欧美成人免费观看 | 荫蒂添的好舒服视频囗交 | 丰满肥臀大屁股熟妇激情视频 | 曰本女人与公拘交酡免费视频 | 国产精品igao视频网 | 国精品人妻无码一区二区三区蜜柚 | 国产一区二区三区日韩精品 | 人妻夜夜爽天天爽三区 | 国产高潮视频在线观看 | 久久国产精品萌白酱免费 | 精品国产福利一区二区 | 欧美老熟妇乱xxxxx | 狂野欧美性猛xxxx乱大交 | 成人无码精品1区2区3区免费看 | 久久久久se色偷偷亚洲精品av | 老熟妇仑乱视频一区二区 | 中文字幕乱码亚洲无线三区 | 精品厕所偷拍各类美女tp嘘嘘 | 在线视频网站www色 | 国产无遮挡又黄又爽免费视频 | 狠狠cao日日穞夜夜穞av | 性做久久久久久久免费看 | 色老头在线一区二区三区 | 久久五月精品中文字幕 | 一二三四在线观看免费视频 | 亚洲成a人片在线观看日本 | 少妇人妻偷人精品无码视频 | 欧美自拍另类欧美综合图片区 | 正在播放东北夫妻内射 | 亚洲精品国产品国语在线观看 | 爆乳一区二区三区无码 | 久久精品国产99久久6动漫 | 国产 浪潮av性色四虎 | aⅴ在线视频男人的天堂 | 国产亚洲精品久久久久久国模美 | 玩弄少妇高潮ⅹxxxyw | 亚洲成熟女人毛毛耸耸多 | 99久久婷婷国产综合精品青草免费 | 色综合天天综合狠狠爱 | 欧美性黑人极品hd | 色五月丁香五月综合五月 | 久久久久久国产精品无码下载 | 久久国内精品自在自线 | 免费无码午夜福利片69 | 夜精品a片一区二区三区无码白浆 | 99久久无码一区人妻 | 亚洲欧美日韩国产精品一区二区 | 日韩少妇白浆无码系列 | 亚洲啪av永久无码精品放毛片 | 黑人巨大精品欧美黑寡妇 | 激情国产av做激情国产爱 | 97精品国产97久久久久久免费 | 国产激情无码一区二区app | 国产绳艺sm调教室论坛 | 老子影院午夜精品无码 | 婷婷综合久久中文字幕蜜桃三电影 | 国产网红无码精品视频 | 久久久久亚洲精品男人的天堂 | 国产偷自视频区视频 | 亚洲狠狠婷婷综合久久 | 久久精品人人做人人综合 | 秋霞成人午夜鲁丝一区二区三区 | 国产午夜视频在线观看 | 18禁止看的免费污网站 | 午夜熟女插插xx免费视频 | 综合人妻久久一区二区精品 | 少妇被粗大的猛进出69影院 | 国内精品一区二区三区不卡 | 最近的中文字幕在线看视频 | 久久精品中文闷骚内射 | 激情内射亚州一区二区三区爱妻 | 国产人妖乱国产精品人妖 | 国产午夜视频在线观看 | 亚洲欧洲无卡二区视頻 | 精品国精品国产自在久国产87 | 丰满人妻翻云覆雨呻吟视频 | 国产成人综合美国十次 | 欧美放荡的少妇 | 国产人妻人伦精品1国产丝袜 | 久久99国产综合精品 | 一本大道久久东京热无码av | 波多野结衣乳巨码无在线观看 | 丰满人妻翻云覆雨呻吟视频 | 撕开奶罩揉吮奶头视频 | 人人妻人人澡人人爽欧美精品 | 中文字幕人妻无码一区二区三区 | 欧美丰满老熟妇xxxxx性 | 永久黄网站色视频免费直播 | 国产精品高潮呻吟av久久4虎 | 男女下面进入的视频免费午夜 | 国产疯狂伦交大片 | 亚洲性无码av中文字幕 | 国产精品亚洲一区二区三区喷水 | 九月婷婷人人澡人人添人人爽 | 国产精品爱久久久久久久 | 97久久精品无码一区二区 | 国产麻豆精品一区二区三区v视界 | 77777熟女视频在线观看 а天堂中文在线官网 | 永久免费观看国产裸体美女 | 欧美高清在线精品一区 | 波多野结衣 黑人 | 2019nv天堂香蕉在线观看 | 国产成人综合美国十次 | 无码人妻出轨黑人中文字幕 | 又紧又大又爽精品一区二区 | 男女爱爱好爽视频免费看 | 欧美怡红院免费全部视频 | 欧洲欧美人成视频在线 | 国产九九九九九九九a片 | 久久亚洲国产成人精品性色 | 人妻少妇精品无码专区动漫 | 国产极品美女高潮无套在线观看 | 亚洲欧美日韩国产精品一区二区 | 国产精品-区区久久久狼 | 国产真实乱对白精彩久久 | 国产9 9在线 | 中文 | 荫蒂被男人添的好舒服爽免费视频 | 久久综合网欧美色妞网 | 久久国产精品二国产精品 | 国产疯狂伦交大片 | 欧美亚洲日韩国产人成在线播放 | 精品少妇爆乳无码av无码专区 | 欧美丰满熟妇xxxx性ppx人交 | 国产免费无码一区二区视频 | 蜜桃视频插满18在线观看 | 无码成人精品区在线观看 | 国产亚洲人成在线播放 | 国产特级毛片aaaaaa高潮流水 | 亚洲码国产精品高潮在线 | 久青草影院在线观看国产 | 国产综合久久久久鬼色 | 精品无码成人片一区二区98 | 国产人妻精品一区二区三区不卡 | 欧美熟妇另类久久久久久多毛 | 婷婷丁香五月天综合东京热 | 国产精品无码成人午夜电影 | 久久久久av无码免费网 | 天天做天天爱天天爽综合网 | 亚洲国精产品一二二线 | 日韩av无码一区二区三区不卡 | 久热国产vs视频在线观看 | 亚洲国产精品美女久久久久 | 久久久久成人片免费观看蜜芽 | 色欲人妻aaaaaaa无码 | 成在人线av无码免观看麻豆 | 55夜色66夜色国产精品视频 | 久久亚洲日韩精品一区二区三区 | 国产又爽又黄又刺激的视频 | 精品人妻人人做人人爽 | 自拍偷自拍亚洲精品10p | 亚洲午夜福利在线观看 | 亚洲另类伦春色综合小说 | 久久www免费人成人片 | 精品国产乱码久久久久乱码 | 激情人妻另类人妻伦 | 99re在线播放 | 无码中文字幕色专区 | 国产成人无码a区在线观看视频app | 欧洲vodafone精品性 | 成年女人永久免费看片 | 77777熟女视频在线观看 а天堂中文在线官网 | 免费看少妇作爱视频 | 一本久久a久久精品亚洲 | 久9re热视频这里只有精品 | 人妻天天爽夜夜爽一区二区 | 全球成人中文在线 | 国产亚洲精品久久久久久久 | 国产免费久久精品国产传媒 | 老熟妇仑乱视频一区二区 | 最近免费中文字幕中文高清百度 | 亚洲精品午夜国产va久久成人 | 成人精品视频一区二区 | 动漫av网站免费观看 | 一二三四在线观看免费视频 | 天干天干啦夜天干天2017 | 黑人玩弄人妻中文在线 | 久久99精品国产麻豆蜜芽 | 理论片87福利理论电影 | 亚洲一区二区观看播放 | 国产特级毛片aaaaaaa高清 | 久久久久久a亚洲欧洲av冫 | 色综合天天综合狠狠爱 | 精品亚洲韩国一区二区三区 | 日本一卡二卡不卡视频查询 | 久久亚洲精品中文字幕无男同 | 性啪啪chinese东北女人 | 色欲人妻aaaaaaa无码 | 天天拍夜夜添久久精品大 | 狠狠色噜噜狠狠狠7777奇米 | 国产精品久久久久久久影院 | 色综合久久网 | a在线亚洲男人的天堂 | 亚洲理论电影在线观看 | 欧美人与物videos另类 | 亚洲乱亚洲乱妇50p | 夜精品a片一区二区三区无码白浆 | 久久亚洲a片com人成 | 无码人中文字幕 | 国产性猛交╳xxx乱大交 国产精品久久久久久无码 欧洲欧美人成视频在线 | 国产69精品久久久久app下载 | 美女毛片一区二区三区四区 | 男人扒开女人内裤强吻桶进去 | 日韩精品成人一区二区三区 | 成人免费无码大片a毛片 | 久久久亚洲欧洲日产国码αv | 亚洲欧洲日本无在线码 | 狠狠噜狠狠狠狠丁香五月 | 麻豆果冻传媒2021精品传媒一区下载 | 清纯唯美经典一区二区 | 天堂久久天堂av色综合 | 久久精品人人做人人综合试看 | 中国大陆精品视频xxxx | 扒开双腿吃奶呻吟做受视频 | 成人亚洲精品久久久久软件 | 国产97色在线 | 免 | 熟妇人妻无乱码中文字幕 | 国产精品理论片在线观看 | 久久久精品456亚洲影院 | 久久久中文字幕日本无吗 | 色婷婷香蕉在线一区二区 | 丝袜 中出 制服 人妻 美腿 | 一本色道久久综合狠狠躁 | 久久99国产综合精品 | 久久国产精品精品国产色婷婷 | 国产亚洲精品久久久久久国模美 | 国产成人无码一二三区视频 | 精品国产成人一区二区三区 | 人人爽人人澡人人高潮 | 欧美 丝袜 自拍 制服 另类 | 日韩 欧美 动漫 国产 制服 | 丝袜 中出 制服 人妻 美腿 | 欧美成人午夜精品久久久 | 国内揄拍国内精品人妻 | 一本久道久久综合狠狠爱 | 亚洲区欧美区综合区自拍区 | 国内揄拍国内精品人妻 | 亚洲 激情 小说 另类 欧美 | 久久国产精品精品国产色婷婷 | 最近中文2019字幕第二页 | 国产日产欧产精品精品app | 国产网红无码精品视频 | 一个人看的视频www在线 | 日本熟妇大屁股人妻 | www国产精品内射老师 | 国产精品igao视频网 | 久久国产精品二国产精品 | 一区二区三区高清视频一 | 亚洲日韩中文字幕在线播放 | 亚洲中文字幕无码中文字在线 | 久久亚洲精品成人无码 | 丁香花在线影院观看在线播放 | 久久久久久国产精品无码下载 | 久在线观看福利视频 | 国产性猛交╳xxx乱大交 国产精品久久久久久无码 欧洲欧美人成视频在线 | 久精品国产欧美亚洲色aⅴ大片 | 欧美人与牲动交xxxx | 日本一区二区三区免费高清 | 十八禁真人啪啪免费网站 | yw尤物av无码国产在线观看 | 人妻少妇精品无码专区二区 | 初尝人妻少妇中文字幕 | 99久久无码一区人妻 | 中文毛片无遮挡高清免费 | 亚洲成av人片天堂网无码】 | 国产成人精品无码播放 | 午夜免费福利小电影 | 欧美变态另类xxxx | 网友自拍区视频精品 | 国产精品第一国产精品 | 一本久道高清无码视频 | 十八禁真人啪啪免费网站 | 国内精品久久毛片一区二区 | 特级做a爰片毛片免费69 | 999久久久国产精品消防器材 | 久久久久成人精品免费播放动漫 | 无遮挡啪啪摇乳动态图 | 亚洲 激情 小说 另类 欧美 | 国内少妇偷人精品视频 | 丰满人妻翻云覆雨呻吟视频 | 免费人成在线观看网站 | 亚拍精品一区二区三区探花 | 午夜无码人妻av大片色欲 | 国内精品久久毛片一区二区 | 国产福利视频一区二区 | 麻花豆传媒剧国产免费mv在线 | 天堂а√在线中文在线 | 中文字幕乱码人妻无码久久 | 午夜精品久久久久久久久 | 在线精品国产一区二区三区 | 婷婷五月综合缴情在线视频 | 日本乱人伦片中文三区 | 亚洲国产欧美日韩精品一区二区三区 | 岛国片人妻三上悠亚 | 蜜臀av在线播放 久久综合激激的五月天 | 久久精品国产日本波多野结衣 | 国产av一区二区三区最新精品 | 国产在线aaa片一区二区99 | 亚洲精品国产品国语在线观看 | 中文字幕精品av一区二区五区 | 中文字幕中文有码在线 | 日本大香伊一区二区三区 | 波多野结衣 黑人 | 人妻少妇精品无码专区二区 | 无码人妻丰满熟妇区五十路百度 | 小sao货水好多真紧h无码视频 | 精品人人妻人人澡人人爽人人 | 丰满岳乱妇在线观看中字无码 | 天天躁夜夜躁狠狠是什么心态 | 亚洲欧美国产精品久久 | 青青久在线视频免费观看 | 日本熟妇大屁股人妻 | 理论片87福利理论电影 | 无码一区二区三区在线观看 | 国产熟妇高潮叫床视频播放 | 国产熟女一区二区三区四区五区 | 女人被男人躁得好爽免费视频 | 国内少妇偷人精品视频免费 | 中文字幕乱码人妻无码久久 | 2019午夜福利不卡片在线 | 国产国产精品人在线视 | 女人和拘做爰正片视频 | 精品无码国产自产拍在线观看蜜 | 国产精品久免费的黄网站 | 欧美 日韩 亚洲 在线 | 丰满少妇高潮惨叫视频 | 成人片黄网站色大片免费观看 | 中文字幕av无码一区二区三区电影 | 又紧又大又爽精品一区二区 | 色一情一乱一伦一区二区三欧美 | 中文字幕日韩精品一区二区三区 | 欧美三级a做爰在线观看 | 欧美日韩精品 | 国产高清不卡无码视频 | 久久国产精品偷任你爽任你 | 最近中文2019字幕第二页 | 免费中文字幕日韩欧美 | 国产精品久久久久9999小说 | 日韩精品无码一区二区中文字幕 | 青青草原综合久久大伊人精品 | 黑人巨大精品欧美一区二区 | 中文字幕无码免费久久9一区9 | 成人亚洲精品久久久久 | 激情亚洲一区国产精品 | 日韩人妻系列无码专区 | 亚洲一区二区三区无码久久 | 精品国产乱码久久久久乱码 | 欧美性生交活xxxxxdddd | 强奷人妻日本中文字幕 | 综合人妻久久一区二区精品 | 毛片内射-百度 | 国产美女精品一区二区三区 | 亚洲精品美女久久久久久久 | 亚洲一区二区观看播放 | 国产片av国语在线观看 | 日产国产精品亚洲系列 | 蜜臀av无码人妻精品 | 性做久久久久久久免费看 | 兔费看少妇性l交大片免费 | 少妇无码av无码专区在线观看 | 牲交欧美兽交欧美 | 丝袜美腿亚洲一区二区 | 九九久久精品国产免费看小说 | 97资源共享在线视频 | 国产精品久久久久7777 | 欧美性黑人极品hd | 无码av岛国片在线播放 | 亚洲色成人中文字幕网站 | 亚洲va中文字幕无码久久不卡 | 国产成人无码av在线影院 | ass日本丰满熟妇pics | 亚洲色欲久久久综合网东京热 | 色婷婷欧美在线播放内射 | 国内揄拍国内精品少妇国语 | 极品尤物被啪到呻吟喷水 | 乌克兰少妇性做爰 | 在教室伦流澡到高潮hnp视频 | 国产在线aaa片一区二区99 | 成人无码精品1区2区3区免费看 | 精品国产一区二区三区四区 | 欧美一区二区三区视频在线观看 | 永久免费观看国产裸体美女 | 久久99热只有频精品8 | 亚洲色无码一区二区三区 | 国产69精品久久久久app下载 | 亚洲男女内射在线播放 | 欧美丰满熟妇xxxx性ppx人交 | 中文字幕日产无线码一区 | 青青草原综合久久大伊人精品 | 欧美猛少妇色xxxxx | 亚拍精品一区二区三区探花 | 小sao货水好多真紧h无码视频 | 国产激情精品一区二区三区 | 麻豆蜜桃av蜜臀av色欲av | 成人无码视频在线观看网站 | 亚洲欧洲日本综合aⅴ在线 | 51国偷自产一区二区三区 | 久久久av男人的天堂 | 色狠狠av一区二区三区 | 精品午夜福利在线观看 | 青青草原综合久久大伊人精品 | 欧美日韩在线亚洲综合国产人 | 国产av一区二区三区最新精品 | 国语自产偷拍精品视频偷 | 国产手机在线αⅴ片无码观看 | 中文字幕无码免费久久9一区9 | 色 综合 欧美 亚洲 国产 | 久久成人a毛片免费观看网站 | 成人影院yy111111在线观看 | 特黄特色大片免费播放器图片 | 水蜜桃色314在线观看 | 久久 国产 尿 小便 嘘嘘 | 色婷婷av一区二区三区之红樱桃 | 人人爽人人澡人人人妻 | 夜夜高潮次次欢爽av女 | 成人性做爰aaa片免费看不忠 | 国产黑色丝袜在线播放 | 中文无码伦av中文字幕 | 色综合久久久无码网中文 | 国产精品理论片在线观看 | 欧美性猛交内射兽交老熟妇 | 国色天香社区在线视频 | 黑人粗大猛烈进出高潮视频 | 国产午夜福利100集发布 | 欧洲vodafone精品性 | 精品久久8x国产免费观看 | 狠狠色色综合网站 | 国产av久久久久精东av | 国产两女互慰高潮视频在线观看 | 超碰97人人射妻 | 麻豆国产丝袜白领秘书在线观看 | 无码人妻精品一区二区三区下载 | 中文字幕中文有码在线 | 久久久久久久久888 | 天天躁日日躁狠狠躁免费麻豆 | 色偷偷av老熟女 久久精品人妻少妇一区二区三区 | 国产精品欧美成人 | 熟妇女人妻丰满少妇中文字幕 | 性开放的女人aaa片 | 四虎国产精品免费久久 | 国内精品久久久久久中文字幕 | 欧美性生交xxxxx久久久 | 九九热爱视频精品 | 性生交大片免费看女人按摩摩 | 一本色道久久综合亚洲精品不卡 | 国产超碰人人爽人人做人人添 | 国产亚洲精品久久久久久久久动漫 | 波多野结衣乳巨码无在线观看 | 亚洲综合无码久久精品综合 | 亚洲欧洲日本综合aⅴ在线 | a国产一区二区免费入口 | 夜夜夜高潮夜夜爽夜夜爰爰 | 一本一道久久综合久久 | 女人被爽到呻吟gif动态图视看 | 亚洲精品国偷拍自产在线观看蜜桃 | 国产三级久久久精品麻豆三级 | 国产suv精品一区二区五 | 国产va免费精品观看 | 国产精品久久精品三级 | 丰满少妇人妻久久久久久 | 亚洲第一无码av无码专区 | 2020最新国产自产精品 | 无码av岛国片在线播放 | 欧美性生交活xxxxxdddd | 精品亚洲成av人在线观看 | 性生交大片免费看l | 人妻尝试又大又粗久久 | 日本精品久久久久中文字幕 | 国产av剧情md精品麻豆 | 在线a亚洲视频播放在线观看 | 成人女人看片免费视频放人 | 性色欲网站人妻丰满中文久久不卡 | 亚洲精品午夜无码电影网 | 波多野结衣一区二区三区av免费 | 精品偷自拍另类在线观看 | 清纯唯美经典一区二区 | 国产极品视觉盛宴 | 丰满少妇弄高潮了www | 国产成人无码av一区二区 | 丰满人妻翻云覆雨呻吟视频 | 久久久久久久人妻无码中文字幕爆 | 色综合久久网 | 亚洲日韩av一区二区三区四区 | 影音先锋中文字幕无码 | 亚洲狠狠婷婷综合久久 | 98国产精品综合一区二区三区 | 亚洲综合另类小说色区 | 伦伦影院午夜理论片 | 人人妻人人澡人人爽人人精品浪潮 | 欧美成人高清在线播放 | 亚洲色偷偷男人的天堂 | 国产亚洲精品久久久久久久 | 亚洲区欧美区综合区自拍区 | 丰满少妇弄高潮了www | 色综合久久网 | 黑人玩弄人妻中文在线 | 最近中文2019字幕第二页 | 扒开双腿疯狂进出爽爽爽视频 | 3d动漫精品啪啪一区二区中 | 中文字幕亚洲情99在线 | 欧美大屁股xxxxhd黑色 | 国产成人av免费观看 | 国产精品自产拍在线观看 | 欧美成人免费全部网站 | 国产又爽又黄又刺激的视频 | 国产在线无码精品电影网 | 骚片av蜜桃精品一区 | 欧美激情一区二区三区成人 | 人人妻人人澡人人爽人人精品浪潮 | 黑人巨大精品欧美一区二区 | 蜜桃av蜜臀av色欲av麻 999久久久国产精品消防器材 | 无码纯肉视频在线观看 | 久久国产36精品色熟妇 | 无码免费一区二区三区 | 久久精品99久久香蕉国产色戒 | 中文字幕乱码人妻无码久久 | 狠狠cao日日穞夜夜穞av | 亚洲国产精华液网站w | 日产精品99久久久久久 | 国产凸凹视频一区二区 | 人妻尝试又大又粗久久 | 奇米影视888欧美在线观看 | 欧洲熟妇色 欧美 | 两性色午夜免费视频 | 中文字幕无码av波多野吉衣 | 女高中生第一次破苞av | 国产av无码专区亚洲a∨毛片 | 成人无码精品1区2区3区免费看 | 国产亚洲精品久久久久久久久动漫 | 丝袜美腿亚洲一区二区 | 一本久道高清无码视频 | 久久久久久久久蜜桃 | 成人一在线视频日韩国产 | 国产suv精品一区二区五 | www国产亚洲精品久久网站 | 真人与拘做受免费视频一 | a片免费视频在线观看 | 无码一区二区三区在线 |