关于transformer模型总结（源码）

本文主要是对transfermer模型的源码进行解析：
transfermer主要结构是由encoder和decoder构成。其中，encoder是由embedding + positional_encoding作为输入，然后加一个dropout层，然后输入放到6个multihead_attention构成的结构中，每个multihead_attention后面跟一个feedforward。而decoder是由decoder embedding + positional_encoding作为输入，输入到dropout层，然后后面跟六个self multihead_attention+ multihead_attention，最后后面跟一个feedward。最后加一个liner projection。

1.主体模块介绍

1) positiion encoding

这个模块主要是对sequence中word的顺序进行编码，主要采用的是google论文中提出的公式。

def positional_encoding(inputs,
                        num_units,
                        zero_pad=True,
                        scale=True,
                        scope="positional_encoding",
                        reuse=None):

    N, T = inputs.get_shape().as_list()
    with tf.variable_scope(scope, reuse=reuse):
        position_ind = tf.tile(tf.expand_dims(tf.range(T), 0), [tf.shape(inputs)[0], 1])

        # First part of the PE function: sin and cos argument
        position_enc = np.array([
            [pos / np.power(10000, 2.*i/num_units) for i in range(num_units)]
            for pos in range(T)])

        # Second part, apply the cosine to even columns and sin to odds.
        position_enc[:, 0::2] = np.sin(position_enc[:, 0::2])  # dim 2i
        position_enc[:, 1::2] = np.cos(position_enc[:, 1::2])  # dim 2i+1

        # Convert to a tensor
        lookup_table = tf.convert_to_tensor(position_enc)
        lookup_table = tf.cast(lookup_table,tf.float32)

        if zero_pad:
            lookup_table = tf.concat((tf.zeros(shape=[1, num_units]),
                                      lookup_table[1:, :]), 0)
        outputs = tf.nn.embedding_lookup(lookup_table, position_ind)

        if scale:
            outputs = outputs * num_units**0.5

        return outputs

2）layer normalize

这个部分主要是对multi-head attention中的输出做layer normlaze，代码如下，与batch normalize相似，不过layer是横向的做normalize。

# layer normalize
def normalize(inputs, 
              epsilon = 1e-8,
              scope="ln",
              reuse=None):
    with tf.variable_scope(scope, reuse=reuse):
        inputs_shape = inputs.get_shape()
        params_shape = inputs_shape[-1:]

        mean, variance = tf.nn.moments(inputs, [-1], keep_dims=True)
        beta= tf.Variable(tf.zeros(params_shape))
        gamma = tf.Variable(tf.ones(params_shape))
        normalized = (inputs - mean) / ( (variance + epsilon) ** (.5) )
        outputs = gamma * normalized + beta

    return outputs

3）embedding

这个部分比较简单，是将语句中的每个词转化为向量

def embedding(inputs, 
              vocab_size, 
              num_units, 
              zero_pad=True, 
              scale=True,
              scope="embedding", 
              reuse=None):

    with tf.variable_scope(scope, reuse=reuse):
        lookup_table = tf.get_variable('lookup_table',
                                       dtype=tf.float32,
                                       shape=[vocab_size, num_units],
                                       initializer=tf.contrib.layers.xavier_initializer())
        if zero_pad:
            lookup_table = tf.concat((tf.zeros(shape=[1, num_units]),
                                      lookup_table[1:, :]), 0)
        outputs = tf.nn.embedding_lookup(lookup_table, inputs)

        if scale:
            outputs = outputs * (num_units ** 0.5) 

    return outputs

4） label smoothing

对输入数据做平滑处理，这个部分不是太理解。

def label_smoothing(inputs, epsilon=0.1):
    K = inputs.get_shape().as_list()[-1]    # number of channels
    return ((1-epsilon) * inputs) + (epsilon / K)

5）multihead attention

5）multihead attention
def multihead_attention(queries, 
                        keys, 
                        num_units=None, 
                        num_heads=8, 
                        dropout_rate=0,
                        is_training=True,
                        causality=False,
                        scope="multihead_attention", 
                        reuse=None):

    with tf.variable_scope(scope, reuse=reuse):
        # Set the fall back option for num_units
        if num_units is None:
            num_units = queries.get_shape().as_list[-1]

        # Linear projections
        Q = tf.layers.dense(queries, num_units, activation=tf.nn.relu) # (N, T_q, C)
        K = tf.layers.dense(keys, num_units, activation=tf.nn.relu) # (N, T_k, C)
        V = tf.layers.dense(keys, num_units, activation=tf.nn.relu) # (N, T_k, C)

        # Split and concat
        Q_ = tf.concat(tf.split(Q, num_heads, axis=2), axis=0) # (h*N, T_q, C/h) 
        K_ = tf.concat(tf.split(K, num_heads, axis=2), axis=0) # (h*N, T_k, C/h) 
        V_ = tf.concat(tf.split(V, num_heads, axis=2), axis=0) # (h*N, T_k, C/h) 

        # Multiplication
        outputs = tf.matmul(Q_, tf.transpose(K_, [0, 2, 1])) # (h*N, T_q, T_k)

        # Scale
        outputs = outputs / (K_.get_shape().as_list()[-1] ** 0.5)

        # Key Masking
        key_masks = tf.sign(tf.abs(tf.reduce_sum(keys, axis=-1))) # (N, T_k)
        key_masks = tf.tile(key_masks, [num_heads, 1]) # (h*N, T_k)
        key_masks = tf.tile(tf.expand_dims(key_masks, 1), [1, tf.shape(queries)[1], 1]) # (h*N, T_q, T_k)

        paddings = tf.ones_like(outputs)*(-2**32+1)
        outputs = tf.where(tf.equal(key_masks, 0), paddings, outputs) # (h*N, T_q, T_k)

        # Causality = Future blinding
        if causality:
            diag_vals = tf.ones_like(outputs[0, :, :]) # (T_q, T_k)
            tril = tf.contrib.linalg.LinearOperatorLowerTriangular(diag_vals).to_dense() # (T_q, T_k)
            masks = tf.tile(tf.expand_dims(tril, 0), [tf.shape(outputs)[0], 1, 1]) # (h*N, T_q, T_k)

            paddings = tf.ones_like(masks)*(-2**32+1)
            outputs = tf.where(tf.equal(masks, 0), paddings, outputs) # (h*N, T_q, T_k)

        # Activation
        outputs = tf.nn.softmax(outputs) # (h*N, T_q, T_k)

        # Query Masking
        query_masks = tf.sign(tf.abs(tf.reduce_sum(queries, axis=-1))) # (N, T_q)
        query_masks = tf.tile(query_masks, [num_heads, 1]) # (h*N, T_q)
        query_masks = tf.tile(tf.expand_dims(query_masks, -1), [1, 1, tf.shape(keys)[1]]) # (h*N, T_q, T_k)
        outputs *= query_masks # broadcasting. (N, T_q, C)

        # Dropouts
        outputs = tf.layers.dropout(outputs, rate=dropout_rate, training=tf.convert_to_tensor(is_training))

        # Weighted sum
        outputs = tf.matmul(outputs, V_) # ( h*N, T_q, C/h)

        # Restore shape
        outputs = tf.concat(tf.split(outputs, num_heads, axis=0), axis=2) # (N, T_q, C)

        # Residual connection
        outputs += queries

        # Normalize
        outputs = normalize(outputs) # (N, T_q, C)

    return outputs

6）feedforward

前向网络

def feedforward(inputs, 
                num_units=[2048, 512],
                scope="multihead_attention", 
                reuse=None):

    with tf.variable_scope(scope, reuse=reuse):
        # Inner layer
        params = {"inputs": inputs, "filters": num_units[0], "kernel_size": 1,
                  "activation": tf.nn.relu, "use_bias": True}
        outputs = tf.layers.conv1d(**params)

        # Readout layer
        params = {"inputs": outputs, "filters": num_units[1], "kernel_size": 1,
                  "activation": None, "use_bias": True}
        outputs = tf.layers.conv1d(**params)

        # Residual connection
        outputs += inputs
        # Normalize
        outputs = normalize(outputs)

    return outputs

二 模型训练部分

模型图的构建如下所示：

class Graph():
    def __init__(self, is_training=True):
        self.graph = tf.Graph()
        with self.graph.as_default():
            if is_training:
                self.x, self.y, self.num_batch = get_batch_data() # (N, T)
            else:  # inference
                self.x = tf.placeholder(tf.int32, shape=(None, hp.maxlen))
                self.y = tf.placeholder(tf.int32, shape=(None, hp.maxlen))

            # define decoder inputs(??)
            self.decoder_inputs = tf.concat((tf.ones_like(self.y[:, :1])*2, self.y[:, :-1]), -1) # 2:<S>

            # Load vocabulary    
            de2idx, idx2de = load_de_vocab()
            en2idx, idx2en = load_en_vocab()

            # Encoder
            with tf.variable_scope("encoder"):
                # Embedding
                self.enc = embedding(self.x,
                                     vocab_size=len(de2idx),
                                     num_units=hp.hidden_units,
                                     scale=True,
                                     scope="enc_embed")

                # Positional Encoding
                if hp.sinusoid:
                    self.enc += positional_encoding(self.x,
                                                    num_units=hp.hidden_units,
                                                    zero_pad=False,
                                                    scale=False,
                                                    scope="enc_pe")
                else:
                    self.enc += embedding(tf.tile(tf.expand_dims(tf.range(tf.shape(self.x)[1]), 0), [tf.shape(self.x)[0], 1]),
                                          vocab_size=hp.maxlen,
                                          num_units=hp.hidden_units,
                                          zero_pad=False,
                                          scale=False,
                                          scope="enc_pe")

                # Dropout
                self.enc = tf.layers.dropout(self.enc, 
                                             rate=hp.dropout_rate,
                                             training=tf.convert_to_tensor(is_training))

                # Blocks
                for i in range(hp.num_blocks):
                    with tf.variable_scope("num_blocks_{}".format(i)):
                        # Multihead Attention
                        self.enc = multihead_attention(queries=self.enc, 
                                                       keys=self.enc,
                                                       num_units=hp.hidden_units,
                                                       num_heads=hp.num_heads,
                                                       dropout_rate=hp.dropout_rate,
                                                       is_training=is_training,
                                                       causality=False)

                        # Feed Forward
                        self.enc = feedforward(self.enc, num_units=[4*hp.hidden_units, hp.hidden_units])

            # Decoder
            with tf.variable_scope("decoder"):
                # Embedding
                self.dec = embedding(self.decoder_inputs, 
                                      vocab_size=len(en2idx), 
                                      num_units=hp.hidden_units,
                                      scale=True, 
                                      scope="dec_embed")

                # Positional Encoding
                if hp.sinusoid:
                    self.dec += positional_encoding(self.decoder_inputs,
                                      num_units=hp.hidden_units,
                                      zero_pad=False, 
                                      scale=False,
                                      scope="dec_pe")
                else:
                    self.dec += embedding(tf.tile(tf.expand_dims(tf.range(tf.shape(self.decoder_inputs)[1]), 0), [tf.shape(self.decoder_inputs)[0], 1]),
                                          vocab_size=hp.maxlen,
                                          num_units=hp.hidden_units,
                                          zero_pad=False,
                                          scale=False,
                                          scope="dec_pe")

                # Dropout
                self.dec = tf.layers.dropout(self.dec, 
                                            rate=hp.dropout_rate, 
                                            training=tf.convert_to_tensor(is_training))

                # Blocks
                for i in range(hp.num_blocks):
                    with tf.variable_scope("num_blocks_{}".format(i)):
                        ## Multihead Attention (self-attention)
                        self.dec = multihead_attention(queries=self.dec, 
                                                        keys=self.dec, 
                                                        num_units=hp.hidden_units, 
                                                        num_heads=hp.num_heads, 
                                                        dropout_rate=hp.dropout_rate,
                                                        is_training=is_training,
                                                        causality=True, 
                                                        scope="self_attention")

                        # Multihead Attention (vanilla attention)
                        self.dec = multihead_attention(queries=self.dec, 
                                                       keys=self.enc,
                                                       num_units=hp.hidden_units,
                                                       num_heads=hp.num_heads,
                                                       dropout_rate=hp.dropout_rate,
                                                       is_training=is_training,
                                                       causality=False,
                                                       scope="vanilla_attention")

                        # Feed Forward
                        self.dec = feedforward(self.dec, num_units=[4*hp.hidden_units, hp.hidden_units])

            # Final linear projection
            self.logits = tf.layers.dense(self.dec, len(en2idx))
            self.preds = tf.to_int32(tf.argmax(self.logits, axis=-1))
            self.istarget = tf.to_float(tf.not_equal(self.y, 0))
            self.acc = tf.reduce_sum(tf.to_float(tf.equal(self.preds, self.y))*self.istarget) / (tf.reduce_sum(self.istarget))
            tf.summary.scalar('acc', self.acc)
            if is_training:
                # Loss
                self.y_smoothed = label_smoothing(tf.one_hot(self.y, depth=len(en2idx)))
                self.loss = tf.nn.softmax_cross_entropy_with_logits_v2(logits=self.logits, labels=self.y_smoothed)
                self.mean_loss = tf.reduce_sum(self.loss*self.istarget) / (tf.reduce_sum(self.istarget))

                # Training Scheme
                self.global_step = tf.Variable(0, name='global_step', trainable=False)
                self.optimizer = tf.train.AdamOptimizer(learning_rate=hp.lr, beta1=0.9, beta2=0.98, epsilon=1e-8)
                self.train_op = self.optimizer.minimize(self.mean_loss, global_step=self.global_step)

                # Summary 
                tf.summary.scalar('mean_loss', self.mean_loss)
                self.merged = tf.summary.merge_all()