Bert系列（二）——源码解读之模型主体

本篇文章主要是解读模型主体代码modeling.py。在阅读这篇文章之前希望读者们对bert的相关理论有一定的了解，尤其是transformer的结构原理，网上的资料很多，本文内容对原理部分就不做过多的介绍了。

我自己写出来其中一个目的也是帮助自己学习整理、当你输出的时候才也会明白哪里懂了哪里不懂。因为水平有限，很多地方理解不到位的，还请各位批评指正。

1、配置

class BertConfig(object):
  """Configuration for `BertModel`.""" def __init__(self, vocab_size, hidden_size=768, num_hidden_layers=12, num_attention_heads=12, intermediate_size=3072, hidden_act="gelu", hidden_dropout_prob=0.1, attention_probs_dropout_prob=0.1, max_position_embeddings=512, type_vocab_size=16, initializer_range=0.02): self.vocab_size = vocab_size self.hidden_size = hidden_size self.num_hidden_layers = num_hidden_layers self.num_attention_heads = num_attention_heads self.hidden_act = hidden_act self.intermediate_size = intermediate_size self.hidden_dropout_prob = hidden_dropout_prob self.attention_probs_dropout_prob = attention_probs_dropout_prob self.max_position_embeddings = max_position_embeddings self.type_vocab_size = type_vocab_size self.initializer_range = initializer_range 

模型配置，比较简单，依次是：词典大小、隐层神经元个数、transformer的层数、attention的头数、激活函数、中间层神经元个数、隐层dropout比例、attention里面dropout比例、sequence最大长度、token_type_ids的词典大小、truncated_normal_initializer的stdev。

2、word embedding

def embedding_lookup(input_ids,
                     vocab_size,
                     embedding_size=128, initializer_range=0.02, word_embedding_name="word_embeddings", use_one_hot_embeddings=False): if input_ids.shape.ndims == 2: input_ids = tf.expand_dims(input_ids, axis=[-1]) embedding_table = tf.get_variable( name=word_embedding_name, shape=[vocab_size, embedding_size], initializer=create_initializer(initializer_range)) if use_one_hot_embeddings: flat_input_ids = tf.reshape(input_ids, [-1]) one_hot_input_ids = tf.one_hot(flat_input_ids, depth=vocab_size) output = tf.matmul(one_hot_input_ids, embedding_table) else: output = tf.nn.embedding_lookup(embedding_table, input_ids) input_shape = get_shape_list(input_ids) output = tf.reshape(output, input_shape[0:-1] + [input_shape[-1] * embedding_size]) return (output, embedding_table) 

构造embedding_table，进行word embedding，可选one_hot的方式，返回embedding的结果和embedding_table

3、词向量的后续处理

def embedding_postprocessor(input_tensor,
                            use_token_type=False,
                            token_type_ids=None,
                            token_type_vocab_size=16, token_type_embedding_name="token_type_embeddings", use_position_embeddings=True, position_embedding_name="position_embeddings", initializer_range=0.02, max_position_embeddings=512, dropout_prob=0.1): input_shape = get_shape_list(input_tensor, expected_rank=3) batch_size = input_shape[0] seq_length = input_shape[1] width = input_shape[2] output = input_tensor if use_token_type: if token_type_ids is None: raise ValueError("`token_type_ids` must be specified if" "`use_token_type` is True.") token_type_table = tf.get_variable( name=token_type_embedding_name, shape=[token_type_vocab_size, width], initializer=create_initializer(initializer_range)) flat_token_type_ids = tf.reshape(token_type_ids, [-1]) one_hot_ids = tf.one_hot(flat_token_type_ids, depth=token_type_vocab_size) token_type_embeddings = tf.matmul(one_hot_ids, token_type_table) token_type_embeddings = tf.reshape(token_type_embeddings, [batch_size, seq_length, width]) output += token_type_embeddings if use_position_embeddings: assert_op = tf.assert_less_equal(seq_length, max_position_embeddings) with tf.control_dependencies([assert_op]): full_position_embeddings = tf.get_variable( name=position_embedding_name, shape=[max_position_embeddings, width], initializer=create_initializer(initializer_range)) position_embeddings = tf.slice(full_position_embeddings, [0, 0], [seq_length, -1]) num_dims = len(output.shape.as_list()) position_broadcast_shape = [] for _ in range(num_dims - 2): position_broadcast_shape.append(1) position_broadcast_shape.extend([seq_length, width]) position_embeddings = tf.reshape(position_embeddings, position_broadcast_shape) output += position_embeddings output = layer_norm_and_dropout(output, dropout_prob) return output 

主要是信息添加，可以将word的位置和word对应的token type等信息添加到词向量里面，并且layer正则化和dropout之后返回

4、构造attention mask

def create_attention_mask_from_input_mask(from_tensor, to_mask):
  from_shape = get_shape_list(from_tensor, expected_rank=[2, 3]) batch_size = from_shape[0] from_seq_length = from_shape[1] to_shape = get_shape_list(to_mask, expected_rank=2) to_seq_length = to_shape[1] to_mask = tf.cast( tf.reshape(to_mask, [batch_size, 1, to_seq_length]), tf.float32) broadcast_ones = tf.ones( shape=[batch_size, from_seq_length, 1], dtype=tf.float32) mask = broadcast_ones * to_mask return mask 

将shape为[batch_size, to_seq_length]的2D mask转换为一个shape 为[batch_size, from_seq_length, to_seq_length] 的3D mask用于attention当中。

5、attention layer

def attention_layer(from_tensor,
                    to_tensor,
                    attention_mask=None,
                    num_attention_heads=1, size_per_head=512, query_act=None, key_act=None, value_act=None, attention_probs_dropout_prob=0.0, initializer_range=0.02, do_return_2d_tensor=False, batch_size=None, from_seq_length=None, to_seq_length=None): def transpose_for_scores(input_tensor, batch_size, num_attention_heads, seq_length, width): output_tensor = tf.reshape( input_tensor, [batch_size, seq_length, num_attention_heads, width]) output_tensor = tf.transpose(output_tensor, [0, 2, 1, 3]) return output_tensor from_shape = get_shape_list(from_tensor, expected_rank=[2, 3]) to_shape = get_shape_list(to_tensor, expected_rank=[2, 3]) if len(from_shape) != len(to_shape): raise ValueError( "The rank of `from_tensor` must match the rank of `to_tensor`.") if len(from_shape) == 3: batch_size = from_shape[0] from_seq_length = from_shape[1] to_seq_length = to_shape[1] elif len(from_shape) == 2: if (batch_size is None or from_seq_length is None or to_seq_length is None): raise ValueError( "When passing in rank 2 tensors to attention_layer, the values " "for `batch_size`, `from_seq_length`, and `to_seq_length` " "must all be specified.") # Scalar dimensions referenced here: # B = batch size (number of sequences) # F = `from_tensor` sequence length # T = `to_tensor` sequence length # N = `num_attention_heads` # H = `size_per_head` from_tensor_2d = reshape_to_matrix(from_tensor) to_tensor_2d = reshape_to_matrix(to_tensor) # `query_layer` = [B*F, N*H] query_layer = tf.layers.dense( from_tensor_2d, num_attention_heads * size_per_head, activation=query_act, name="query", kernel_initializer=create_initializer(initializer_range)) # `key_layer` = [B*T, N*H] key_layer = tf.layers.dense( to_tensor_2d, num_attention_heads * size_per_head, activation=key_act, name="key", kernel_initializer=create_initializer(initializer_range)) # `value_layer` = [B*T, N*H] value_layer = tf.layers.dense( to_tensor_2d, num_attention_heads * size_per_head, activation=value_act, name="value", kernel_initializer=create_initializer(initializer_range)) # `query_layer` = [B, N, F, H] query_layer = transpose_for_scores(query_layer, batch_size, num_attention_heads, from_seq_length, size_per_head) # `key_layer` = [B, N, T, H] key_layer = transpose_for_scores(key_layer, batch_size, num_attention_heads, to_seq_length, size_per_head) attention_scores = tf.matmul(query_layer, key_layer, transpose_b=True) attention_scores = tf.multiply(attention_scores, 1.0 / math.sqrt(float(size_per_head))) if attention_mask is not None: # `attention_mask` = [B, 1, F, T] attention_mask = tf.expand_dims(attention_mask, axis=[1]) adder = (1.0 - tf.cast(attention_mask, tf.float32)) * -10000.0 attention_scores += adder attention_probs = tf.nn.softmax(attention_scores) attention_probs = dropout(attention_probs, attention_probs_dropout_prob) # `value_layer` = [B, T, N, H] value_layer = tf.reshape( value_layer, [batch_size, to_seq_length, num_attention_heads, size_per_head]) # `value_layer` = [B, N, T, H] value_layer = tf.transpose(value_layer, [0, 2, 1, 3]) # `context_layer` = [B, N, F, H] context_layer = tf.matmul(attention_probs, value_layer) # `context_layer` = [B, F, N, H] context_layer = tf.transpose(context_layer, [0, 2, 1, 3]) if do_return_2d_tensor: # `context_layer` = [B*F, N*V] context_layer = tf.reshape( context_layer, [batch_size * from_seq_length, num_attention_heads * size_per_head]) else: # `context_layer` = [B, F, N*V] context_layer = tf.reshape( context_layer, [batch_size, from_seq_length, num_attention_heads * size_per_head]) return context_layer 

整个网络的重头戏来了！tansformer的主要内容都在这里面，输入的from_tensor当作query，to_tensor当作key和value。当self attention的时候from_tensor和to_tensor是同一个值。

（1）函数一开始对输入的shape进行校验，获取batch_size、from_seq_length 、to_seq_length 。输入如果是3D张量则转化成2D矩阵(以输入为word_embedding为例[batch_size, seq_lenth, hidden_size] -> [batch_size*seq_lenth, hidden_size])

（2）通过全连接线性投影生成query_layer、key_layer 、value_layer，输出的第二个维度变成num_attention_heads * size_per_head（整个模型默认hidden_size=num_attention_heads * size_per_head）。然后通过transpose_for_scores转换成多头。

（3）根据公式计算attention_probs（attention score）：

 

Attention Score计算公式

如果attention_mask is not None，对mask的部分加上一个很大的负数，这样softmax之后相应的概率值接近为0，再dropout。

（4）最后再将value和attention_probs相乘，返回3D张量或者2D矩阵

总结：

同学们可以将这段代码与网络结构图对照起来看：

 

Attention Layer

该函数相比其他版本的的 transformer很多地方都有简化，有以下四点：

（1）缺少scale的操作；

（2）没有Causality mask，个人猜测主要是bert没有decoder的操作，所以对角矩阵mask是不需要的，从另一方面来说正好体现了双向transformer的特点；

（3）没有query mask。跟（2）理由类似，encoder都是self attention，query和key相同所以只需要一次key mask就够了

（4）没有query的Residual层和normalize

6、Transformer

def transformer_model(input_tensor,
                      attention_mask=None,
                      hidden_size=768, num_hidden_layers=12, num_attention_heads=12, intermediate_size=3072, intermediate_act_fn=gelu, hidden_dropout_prob=0.1, attention_probs_dropout_prob=0.1, initializer_range=0.02, do_return_all_layers=False): if hidden_size % num_attention_heads != 0: raise ValueError( "The hidden size (%d) is not a multiple of the number of attention " "heads (%d)" % (hidden_size, num_attention_heads)) attention_head_size = int(hidden_size / num_attention_heads) input_shape = get_shape_list(input_tensor, expected_rank=3) batch_size = input_shape[0] seq_length = input_shape[1] input_width = input_shape[2] if input_width != hidden_size: raise ValueError("The width of the input tensor (%d) != hidden size (%d)" % (input_width, hidden_size)) prev_output = reshape_to_matrix(input_tensor) all_layer_outputs = [] for layer_idx in range(num_hidden_layers): with tf.variable_scope("layer_%d" % layer_idx): layer_input = prev_output with tf.variable_scope("attention"): attention_heads = [] with tf.variable_scope("self"): attention_head = attention_layer( from_tensor=layer_input, to_tensor=layer_input, attention_mask=attention_mask, num_attention_heads=num_attention_heads, size_per_head=attention_head_size, attention_probs_dropout_prob=attention_probs_dropout_prob, initializer_range=initializer_range, do_return_2d_tensor=True, batch_size=batch_size, from_seq_length=seq_length, to_seq_length=seq_length) attention_heads.append(attention_head) attention_output = None if len(attention_heads) == 1: attention_output = attention_heads[0] else: attention_output = tf.concat(attention_heads, axis=-1) with tf.variable_scope("output"): attention_output = tf.layers.dense( attention_output, hidden_size, kernel_initializer=create_initializer(initializer_range)) attention_output = dropout(attention_output, hidden_dropout_prob) attention_output = layer_norm(attention_output + layer_input) with tf.variable_scope("intermediate"): intermediate_output = tf.layers.dense( attention_output, intermediate_size, activation=intermediate_act_fn, kernel_initializer=create_initializer(initializer_range)) with tf.variable_scope("output"): layer_output = tf.layers.dense( intermediate_output, hidden_size, kernel_initializer=create_initializer(initializer_range)) layer_output = dropout(layer_output, hidden_dropout_prob) layer_output = layer_norm(layer_output + attention_output) prev_output = layer_output all_layer_outputs.append(layer_output) if do_return_all_layers: final_outputs = [] for layer_output in all_layer_outputs: final_output = reshape_from_matrix(layer_output, input_shape) final_outputs.append(final_output) return final_outputs else: final_output = reshape_from_matrix(prev_output, input_shape) return final_output 

transformer是对attention的利用，分以下几步：

（1）计算attention_head_size，attention_head_size = int(hidden_size / num_attention_heads)即将隐层的输出等分给各个attention头。然后将input_tensor转换成2D矩阵；

（2）对input_tensor进行多头attention操作，再做：线性投影——dropout——layer norm——intermediate线性投影——线性投影——dropout——attention_output的residual——layer norm

其中intermediate线性投影的hidden_size可以自行指定，其他层的线性投影hidden_size需要统一，目的是为了对齐。

（3）如此循环计算若干次，且保存每一次的输出，最后返回所有层的输出或者最后一层的输出。

总结：

进一步证实该函数transformer只存在encoder，而不存在decoder操作，所以所有层的多头attention操作都是基于self encoder的。对应论文红框的部分：

 

The Transformer - model architecture

7、BertModel

class BertModel(object):
  def __init__(self, config, is_training, input_ids, input_mask=None, token_type_ids=None, use_one_hot_embeddings=True, scope=None): config = copy.deepcopy(config) if not is_training: config.hidden_dropout_prob = 0.0 config.attention_probs_dropout_prob = 0.0 input_shape = get_shape_list(input_ids, expected_rank=2) batch_size = input_shape[0] seq_length = input_shape[1] if input_mask is None: input_mask = tf.ones(shape=[batch_size, seq_length], dtype=tf.int32) if token_type_ids is None: token_type_ids = tf.zeros(shape=[batch_size, seq_length], dtype=tf.int32) with tf.variable_scope(scope, default_name="bert"): with tf.variable_scope("embeddings"): (self.embedding_output, self.embedding_table) = embedding_lookup( input_ids=input_ids, vocab_size=config.vocab_size, embedding_size=config.hidden_size, initializer_range=config.initializer_range, word_embedding_name="word_embeddings", use_one_hot_embeddings=use_one_hot_embeddings) self.embedding_output = embedding_postprocessor( input_tensor=self.embedding_output, use_token_type=True, token_type_ids=token_type_ids, token_type_vocab_size=config.type_vocab_size, token_type_embedding_name="token_type_embeddings", use_position_embeddings=True, position_embedding_name="position_embeddings", initializer_range=config.initializer_range, max_position_embeddings=config.max_position_embeddings, dropout_prob=config.hidden_dropout_prob) with tf.variable_scope("encoder"): attention_mask = create_attention_mask_from_input_mask( input_ids, input_mask) self.all_encoder_layers = transformer_model( input_tensor=self.embedding_output, attention_mask=attention_mask, hidden_size=config.hidden_size, num_hidden_layers=config.num_hidden_layers, num_attention_heads=config.num_attention_heads, intermediate_size=config.intermediate_size, intermediate_act_fn=get_activation(config.hidden_act), hidden_dropout_prob=config.hidden_dropout_prob, attention_probs_dropout_prob=config.attention_probs_dropout_prob, initializer_range=config.initializer_range, do_return_all_layers=True) self.sequence_output = self.all_encoder_layers[-1] with tf.variable_scope("pooler"): first_token_tensor = tf.squeeze(self.sequence_output[:, 0:1, :], axis=1) self.pooled_output = tf.layers.dense( first_token_tensor, config.hidden_size, activation=tf.tanh, kernel_initializer=create_initializer(config.initializer_range)) 

终于到模型入口了。

(1）设置各种参数，如果input_mask为None的话，就指定所有input_mask值为1，即不进行过滤；如果token_type_ids是None的话，就指定所有token_type_ids值为0；

(2）对输入的input_ids进行embedding操作，再embedding_postprocessor操作，前面我们说了。主要是加入位置和token_type信息到词向量里面；

(3）转换attention_mask 后，通过调用transformer_model进行encoder操作；

(4）获取最后一层的输出sequence_output和pooled_output，pooled_output是取sequence_output的第一个切片然后线性投影获得（可以用于分类问题）

8、总结：

（1）bert主要流程是先embedding（包括位置和token_type的embedding），然后调用transformer得到输出结果，其中embedding、embedding_table、所有transformer层输出、最后transformer层输出以及pooled_output都可以获得，用于迁移学习的fine-tune和预测任务；

（2）bert对于transformer的使用仅限于encoder，没有decoder的过程。这是因为模型存粹是为了预训练服务，而预训练是通过语言模型，不同于NLP其他特定任务。在做迁移学习时可以自行添加；

（3）正因为没有decoder的操作，所以在attention函数里面也相应地减少了很多不必要的功能。

其他非主要函数这里不做过多介绍，感兴趣的同学可以去看源码。

下一篇文章我们将继续学习bert源码的其他模块，包括训练、预测以及输入输出等相关功能。

本文上一篇系列

Bert系列（一）——demo运行
Bert系列（三）——源码解读之Pre-train
Bert系列（四）——源码解读之Fine-tune
Bert系列（五）——中文分词实践 F1 97.8%(附代码)

Reference

1.https://github.com/google-research/bert/blob/master/modeling.py

2.https://github.com/Kyubyong/transformer

3.Attention Is All You Need

4.BERT: Pre-training of Deep Bidirectional Transformers for
Language Understanding

作者：西溪雷神
链接：https://www.jianshu.com/p/d7ce41b58801
来源：简书
简书著作权归作者所有，任何形式的转载都请联系作者获得授权并注明出处。