### BERT源码工作流解读 #### 数据读取模块 处理MRPC数据的类 ~~~python class MrpcProcessor(DataProcessor): """Processor for the MRPC data set (GLUE version).""" def get_train_examples(self, data_dir): """See base class.""" return self._create_examples( self._read_tsv(os.path.join(data_dir, "train.tsv")), "train") def get_dev_examples(self, data_dir): """See base class.""" return self._create_examples( self._read_tsv(os.path.join(data_dir, "dev.tsv")), "dev") def get_test_examples(self, data_dir): """See base class.""" return self._create_examples( self._read_tsv(os.path.join(data_dir, "test.tsv")), "test") def get_labels(self): """See base class.""" return ["0", "1"] # 是否是二分类 def _create_examples(self, lines, set_type): """Creates examples for the training and dev sets.""" examples = [] for (i, line) in enumerate(lines): if i == 0: continue guid = "%s-%s" % (set_type, i) text_a = tokenization.convert_to_unicode(line[3]) # 相关的test_a和b怎么切分 text_b = tokenization.convert_to_unicode(line[4]) if set_type == "test": label = "0" else: label = tokenization.convert_to_unicode(line[0]) examples.append( InputExample(guid=guid, text_a=text_a, text_b=text_b, label=label)) return examples ~~~ 读取训练数据代码： ~~~python if FLAGS.do_train: train_examples = processor.get_train_examples(FLAGS.data_dir) num_train_steps = int( len(train_examples) / FLAGS.train_batch_size * FLAGS.num_train_epochs) # 得到需要迭代的次数，len(train_examples)计算出多少数据量 除以 我们设置的train_batch_size，再乘上epochs次数。 num_warmup_steps = int(num_train_steps * FLAGS.warmup_proportion) # 在刚开始时，让学习率偏小，经过warmup的百分比后，再还原回原始的学习率 ~~~ #### 数据预处理模块 ~~~python # 衔接上一个 file_based_convert_examples_to_features( train_examples, label_list, FLAGS.max_seq_length, tokenizer, train_file) # ctrl点击file_based_xxx函数跳转 def file_based_convert_examples_to_features( examples, label_list, max_seq_length, tokenizer, output_file): """Convert a set of `InputExample`s to a TFRecord file.""" writer = tf.python_io.TFRecordWriter(output_file) # TFRecord读取数据块，在bert中要求数据是TFRecord的形式。 for (ex_index, example) in enumerate(examples): if ex_index % 10000 == 0: tf.logging.info("Writing example %d of %d" % (ex_index, len(examples))) # for循环变量取数据 feature = convert_single_example(ex_index, example, label_list, max_seq_length, tokenizer) # ctrl点击convert_xxx跳转 def convert_single_example(ex_index, example, label_list, max_seq_length, tokenizer): """Converts a single `InputExample` into a single `InputFeatures`.""" if isinstance(example, PaddingInputExample): return InputFeatures( input_ids=[0] * max_seq_length, input_mask=[0] * max_seq_length, segment_ids=[0] * max_seq_length, label_id=0, is_real_example=False) label_map = {} # 构建标签0, 1 for (i, label) in enumerate(label_list): label_map[label] = i tokens_a = tokenizer.tokenize(example.text_a) # ctrl点击tokenize，对第一句话分词 tokens_b = None if example.text_b: # 第二句话分词 tokens_b = tokenizer.tokenize(example.text_b) if tokens_b: # Modifies `tokens_a` and `tokens_b` in place so that the total # length is less than the specified length. # Account for [CLS], [SEP], [SEP] with "- 3" # 保留3个特殊字符 _truncate_seq_pair(tokens_a, tokens_b, max_seq_length - 3) # 如果太长就截断的操作 else: # 没有b的时候保留两个字符 # Account for [CLS] and [SEP] with "- 2" if len(tokens_a) > max_seq_length - 2: tokens_a = tokens_a[0:(max_seq_length - 2)] # The convention in BERT is: # (a) For sequence pairs: # 将下面一对话，CLS开始，SEP断点，变成type_ids的0/1形式，0表示前一句，1表示后一句 # tokens: [CLS] is this jack ##son ##ville ? [SEP] no it is not . [SEP] # type_ids: 0 0 0 0 0 0 0 0 1 1 1 1 1 1 def tokenize(self, text): split_tokens = [] for token in self.basic_tokenizer.tokenize(text): # 词切片，将一个词切片成多个小段，让表达的含义更丰富 for sub_token in self.wordpiece_tokenizer.tokenize(token): split_tokens.append(sub_token) return split_tokens ~~~ #### tfrecord制作 ~~~~python # 延续上面的convert_single_example模块 # 开始构建，创建两个列表来承接 tokens = [] segment_ids = [] tokens.append("[CLS]") # 第一个词是CLS segment_ids.append(0) # 第一个的编码也肯定是0 for token in tokens_a: tokens.append(token) segment_ids.append(0) # 遍历获取，a（第一句话）都是0 tokens.append("[SEP]") # 遍历完增加个SEP连接符/断电 segment_ids.append(0) # tokens添加完SEP后，ids也添加对应的0 if tokens_b: for token in tokens_b: tokens.append(token) segment_ids.append(1) # b和a一样，唯一不同的是添加的是1 tokens.append("[SEP]") segment_ids.append(1) input_ids = tokenizer.convert_tokens_to_ids(tokens) # 转成ID的映射，就是vocab语料库索引 # The mask has 1 for real tokens and 0 for padding tokens. Only real # tokens are attended to. input_mask = [1] * len(input_ids) # Zero-pad up to the sequence length. 保证输入的长度是一样的，多退少补 while len(input_ids) < max_seq_length: # PAD的长度取决于设置的最大长度，小于全补0 input_ids.append(0) input_mask.append(0) segment_ids.append(0) assert len(input_ids) == max_seq_length assert len(input_mask) == max_seq_length assert len(segment_ids) == max_seq_length label_id = label_map[example.label] if ex_index < 5: tf.logging.info("*** Example ***") # 打印结果，这时候预处理的部分大致完成 ... return feature ~~~~ > 将数据制作成tfcord的形式，以便除了速度更快 返回原先的convert_single_example ~~~python for (ex_index, example) in enumerate(examples): # 不断遍历处理数据 if ex_index % 10000 == 0: tf.logging.info("Writing example %d of %d" % (ex_index, len(examples))) feature = convert_single_example(ex_index, example, label_list, max_seq_length, tokenizer) # ctrl点击convert_xxx跳 def create_int_feature(values): f = tf.train.Feature(int64_list=tf.train.Int64List(value=list(values))) return f features = collections.OrderedDict() # 下面执行格式处理，处理成模型所需的格式 features["input_ids"] = create_int_feature(feature.input_ids) features["input_mask"] = create_int_feature(feature.input_mask) features["segment_ids"] = create_int_feature(feature.segment_ids) features["label_ids"] = create_int_feature([feature.label_id]) features["is_real_example"] = create_int_feature( [int(feature.is_real_example)]) tf_example = tf.train.Example(features=tf.train.Features(feature=features)) # 最后转换成tf的数据格式 writer.write(tf_example.SerializeToString()) writer.close() ~~~ #### Embedding层的作用 ~~~python def create_model(bert_config, is_training, input_ids, input_mask, segment_ids, labels, num_labels, use_one_hot_embeddings): """Creates a classification model.""" model = modeling.BertModel( # ctrl点击BertModel跳转 config=bert_config, # 配置 is_training=is_training, input_ids=input_ids, # 特征 input_mask=input_mask, # 特征0/1 token_type_ids=segment_ids, # 特征维度表示第一句话还是第二句 use_one_hot_embeddings=use_one_hot_embeddings) 。。。 class BertModel(object): """BERT model ("Bidirectional Encoder Representations from Transformers"). Example usage: ```python # Already been converted into WordPiece token ids input_ids = tf.constant([[31, 51, 99], [15, 5, 0]]) input_mask = tf.constant([[1, 1, 1], [1, 1, 0]]) token_type_ids = tf.constant([[0, 0, 1], [0, 2, 0]]) config = modeling.BertConfig(vocab_size=32000, hidden_size=512, num_hidden_layers=8, num_attention_heads=6, intermediate_size=1024) model = modeling.BertModel(config=config, is_training=True, input_ids=input_ids, input_mask=input_mask, token_type_ids=token_type_ids) label_embeddings = tf.get_variable(...) pooled_output = model.get_pooled_output() logits = tf.matmul(pooled_output, label_embeddings) ... ``` """ def __init__(self, config, is_training, input_ids, input_mask=None, token_type_ids=None, use_one_hot_embeddings=False, scope=None): """Constructor for BertModel. Args: config: `BertConfig` instance. is_training: bool. true for training model, false for eval model. Controls whether dropout will be applied. input_ids: int32 Tensor of shape [batch_size, seq_length]. input_mask: (optional) int32 Tensor of shape [batch_size, seq_length]. token_type_ids: (optional) int32 Tensor of shape [batch_size, seq_length]. use_one_hot_embeddings: (optional) bool. Whether to use one-hot word embeddings or tf.embedding_lookup() for the word embeddings. scope: (optional) variable scope. Defaults to "bert". Raises: ValueError: The config is invalid or one of the input tensor shapes is invalid. """ 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: # 如果没设置mask，默认都是1 input_mask = tf.ones(shape=[batch_size, seq_length], dtype=tf.int32) if token_type_ids is None: # 没设置就默认一句话 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"): # Perform embedding lookup on the word ids. 词的embeddings (self.embedding_output, self.embedding_table) = embedding_lookup( # ctrl点击embedding_lookup跳转 input_ids=input_ids, # 词 vocab_size=config.vocab_size, # 语料库 embedding_size=config.hidden_size, # 编码映射成多少维 initializer_range=config.initializer_range, # 初始化范围 word_embedding_name="word_embeddings", use_one_hot_embeddings=use_one_hot_embeddings) ~~~ ~~~python def embedding_lookup(input_ids, vocab_size, embedding_size=128, initializer_range=0.02, word_embedding_name="word_embeddings", use_one_hot_embeddings=False): """Looks up words embeddings for id tensor. Args: input_ids: int32 Tensor of shape [batch_size, seq_length] containing word ids. vocab_size: int. Size of the embedding vocabulary. embedding_size: int. Width of the word embeddings. initializer_range: float. Embedding initialization range. word_embedding_name: string. Name of the embedding table. use_one_hot_embeddings: bool. If True, use one-hot method for word embeddings. If False, use `tf.gather()`. Returns: float Tensor of shape [batch_size, seq_length, embedding_size]. """ # This function assumes that the input is of shape [batch_size, seq_length, # num_inputs]. # # If the input is a 2D tensor of shape [batch_size, seq_length], we # reshape to [batch_size, seq_length, 1]. if input_ids.shape.ndims == 2: input_ids = tf.expand_dims(input_ids, axis=[-1]) embedding_table = tf.get_variable( # 词映射矩阵 name=word_embedding_name, # 词向量 shape=[vocab_size, embedding_size], # 获取语料库大表vovab.txt initializer=create_initializer(initializer_range)) flat_input_ids = tf.reshape(input_ids, [-1]) if use_one_hot_embeddings: one_hot_input_ids = tf.one_hot(flat_input_ids, depth=vocab_size) # 查出所有词做one_hot output = tf.matmul(one_hot_input_ids, embedding_table) # 运算一个batch里所有的映射结果 else: output = tf.gather(embedding_table, flat_input_ids) input_shape = get_shape_list(input_ids) output = tf.reshape(output, input_shape[0:-1] + [input_shape[-1] * embedding_size]) # 制作返回结果 return (output, embedding_table) # 返回，词变成了向量 ~~~ > 给数据做Embedding，再加入位置编码 #### 位置编码 ~~~python class BertModel(object): """BERT model ("Bidirectional Encoder Representations from Transformers"). Example usage: ```python # Already been converted into WordPiece token ids input_ids = tf.constant([[31, 51, 99], [15, 5, 0]]) input_mask = tf.constant([[1, 1, 1], [1, 1, 0]]) token_type_ids = tf.constant([[0, 0, 1], [0, 2, 0]]) config = modeling.BertConfig(vocab_size=32000, hidden_size=512, num_hidden_layers=8, num_attention_heads=6, intermediate_size=1024) model = modeling.BertModel(config=config, is_training=True, input_ids=input_ids, input_mask=input_mask, token_type_ids=token_type_ids) label_embeddings = tf.get_variable(...) pooled_output = model.get_pooled_output() logits = tf.matmul(pooled_output, label_embeddings) ... ``` """ def __init__(self, config, is_training, input_ids, input_mask=None, token_type_ids=None, use_one_hot_embeddings=False, scope=None): """Constructor for BertModel. Args: config: `BertConfig` instance. is_training: bool. true for training model, false for eval model. Controls whether dropout will be applied. input_ids: int32 Tensor of shape [batch_size, seq_length]. input_mask: (optional) int32 Tensor of shape [batch_size, seq_length]. token_type_ids: (optional) int32 Tensor of shape [batch_size, seq_length]. use_one_hot_embeddings: (optional) bool. Whether to use one-hot word embeddings or tf.embedding_lookup() for the word embeddings. scope: (optional) variable scope. Defaults to "bert". Raises: ValueError: The config is invalid or one of the input tensor shapes is invalid. """ ... # Add positional embeddings and token type embeddings, then layer # normalize and perform dropout. self.embedding_output = embedding_postprocessor( # 制作位置编码，ctrl点击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) ~~~ ~~~python 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): """Performs various post-processing on a word embedding tensor. Args: input_tensor: float Tensor of shape [batch_size, seq_length, embedding_size]. use_token_type: bool. Whether to add embeddings for `token_type_ids`. token_type_ids: (optional) int32 Tensor of shape [batch_size, seq_length]. Must be specified if `use_token_type` is True. token_type_vocab_size: int. The vocabulary size of `token_type_ids`. token_type_embedding_name: string. The name of the embedding table variable for token type ids. use_position_embeddings: bool. Whether to add position embeddings for the position of each token in the sequence. position_embedding_name: string. The name of the embedding table variable for positional embeddings. initializer_range: float. Range of the weight initialization. max_position_embeddings: int. Maximum sequence length that might ever be used with this model. This can be longer than the sequence length of input_tensor, but cannot be shorter. dropout_prob: float. Dropout probability applied to the final output tensor. Returns: float tensor with same shape as `input_tensor`. Raises: ValueError: One of the tensor shapes or input values is invalid. """ 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: # 判断是第一句还是第二句，再做相应处理 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)) # This vocab will be small so we always do one-hot here, since it is always # faster for a small vocabulary. 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: # 判断是否要做位置编码信息 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)) # Since the position embedding table is a learned variable, we create it # using a (long) sequence length `max_position_embeddings`. The actual # sequence length might be shorter than this, for faster training of # tasks that do not have long sequences. # # So `full_position_embeddings` is effectively an embedding table # for position [0, 1, 2, ..., max_position_embeddings-1], and the current # sequence has positions [0, 1, 2, ... seq_length-1], so we can just # perform a slice. position_embeddings = tf.slice(full_position_embeddings, [0, 0], [seq_length, -1]) # 如果位置编码给的过大，为了加速只需取出部分 num_dims = len(output.shape.as_list()) # Only the last two dimensions are relevant (`seq_length` and `width`), so # we broadcast among the first dimensions, which is typically just # the batch size. 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 ~~~ > 给数据加入位置编码 #### mask机制 ~~~python class BertModel(object): """BERT model ("Bidirectional Encoder Representations from Transformers"). Example usage: ```python # Already been converted into WordPiece token ids input_ids = tf.constant([[31, 51, 99], [15, 5, 0]]) input_mask = tf.constant([[1, 1, 1], [1, 1, 0]]) token_type_ids = tf.constant([[0, 0, 1], [0, 2, 0]]) config = modeling.BertConfig(vocab_size=32000, hidden_size=512, num_hidden_layers=8, num_attention_heads=6, intermediate_size=1024) model = modeling.BertModel(config=config, is_training=True, input_ids=input_ids, input_mask=input_mask, token_type_ids=token_type_ids) label_embeddings = tf.get_variable(...) pooled_output = model.get_pooled_output() logits = tf.matmul(pooled_output, label_embeddings) ... ``` """ def __init__(self, config, is_training, input_ids, input_mask=None, token_type_ids=None, use_one_hot_embeddings=False, scope=None): """Constructor for BertModel. Args: config: `BertConfig` instance. is_training: bool. true for training model, false for eval model. Controls whether dropout will be applied. input_ids: int32 Tensor of shape [batch_size, seq_length]. input_mask: (optional) int32 Tensor of shape [batch_size, seq_length]. token_type_ids: (optional) int32 Tensor of shape [batch_size, seq_length]. use_one_hot_embeddings: (optional) bool. Whether to use one-hot word embeddings or tf.embedding_lookup() for the word embeddings. scope: (optional) variable scope. Defaults to "bert". Raises: ValueError: The config is invalid or one of the input tensor shapes is invalid. """ with tf.variable_scope("encoder"): # This converts a 2D mask of shape [batch_size, seq_length] to a 3D # mask of shape [batch_size, seq_length, seq_length] which is used # for the attention scores. attention_mask = create_attention_mask_from_input_mask( input_ids, input_mask) # 创建mask矩阵 # 比如一个矩阵:[45，54，85，...，0，0，0] # [12，31，11，...，0，0，0] # [91，51，18，...，12，21，0] # 后面长度不足的都补0，mask后，有信息的变1，无信息的变0 # [1，1，1，...，0，0，0] # [1，1，1，...，0，0，0] # [1，1，1，...，1，1，0] # 不管要知道二维的，还要知道三维的，如开头这句话This converts a 2D mask of shape [batch_size, seq_length] to a 3D # 把里面的维度再分一个维度，如左上角的45 # [1，1，1，...，0，0，0] ， 这里的1是指45能看到的信息是那些，有的则为1，并与其计算，为0则不与其进行计算 # Run the stacked transformer. # `sequence_output` shape = [batch_size, seq_length, hidden_size]. self.all_encoder_layers = transformer_model( # Ctrl点击跳转transformer_model input_tensor=self.embedding_output, # 3种embedding attention_mask=attention_mask, # 上面的需不需要计算的0，1，1则是要计算 hidden_size=config.hidden_size, # 特征结果 num_hidden_layers=config.num_hidden_layers, # Transformer中的隐层神经元个数 num_attention_heads=config.num_attention_heads, # 多头机制，在bert的图解中有讲解 intermediate_size=config.intermediate_size, # 全连接层神经元个数 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) ~~~ > 对数据进行mask，此时数据部分已加工完，开始做QKV计算 #### 构建QKV矩阵 ~~~python # 通过上面的点击函数跳转到transformer_model 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): """Multi-headed, multi-layer Transformer from "Attention is All You Need". This is almost an exact implementation of the original Transformer encoder. See the original paper: https://arxiv.org/abs/1706.03762 Also see: https://github.com/tensorflow/tensor2tensor/blob/master/tensor2tensor/models/transformer.py Args: input_tensor: float Tensor of shape [batch_size, seq_length, hidden_size]. attention_mask: (optional) int32 Tensor of shape [batch_size, seq_length, seq_length], with 1 for positions that can be attended to and 0 in positions that should not be. hidden_size: int. Hidden size of the Transformer. num_hidden_layers: int. Number of layers (blocks) in the Transformer. num_attention_heads: int. Number of attention heads in the Transformer. intermediate_size: int. The size of the "intermediate" (a.k.a., feed forward) layer. intermediate_act_fn: function. The non-linear activation function to apply to the output of the intermediate/feed-forward layer. hidden_dropout_prob: float. Dropout probability for the hidden layers. attention_probs_dropout_prob: float. Dropout probability of the attention probabilities. initializer_range: float. Range of the initializer (stddev of truncated normal). do_return_all_layers: Whether to also return all layers or just the final layer. Returns: float Tensor of shape [batch_size, seq_length, hidden_size], the final hidden layer of the Transformer. Raises: ValueError: A Tensor shape or parameter is invalid. """ if hidden_size % num_attention_heads != 0: # 判断是否能整除，否则后面会报错 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] # The Transformer performs sum residuals on all layers so the input needs # to be the same as the hidden size. if input_width != hidden_size: raise ValueError("The width of the input tensor (%d) != hidden size (%d)" % (input_width, hidden_size)) # We keep the representation as a 2D tensor to avoid re-shaping it back and # forth from a 3D tensor to a 2D tensor. Re-shapes are normally free on # the GPU/CPU but may not be free on the TPU, so we want to minimize them to # help the optimizer. prev_output = reshape_to_matrix(input_tensor) all_layer_outputs = [] for layer_idx in range(num_hidden_layers): # 遍历层数，这层结果是下一层的输入 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( # Ctrl点击attention_layer跳转 from_tensor=layer_input, to_tensor=layer_input, # from和to都是self_tensor，即自己和自己本句的关联 attention_mask=attention_mask, # 0/1 num_attention_heads=num_attention_heads, # 多头参数 size_per_head=attention_head_size, # 头大小 attention_probs_dropout_prob=attention_probs_dropout_prob, # 丢弃 initializer_range=initializer_range, # 初始化位置 do_return_2d_tensor=True, # 是否返回2D特征 batch_size=batch_size, from_seq_length=seq_length, to_seq_length=seq_length) attention_heads.append(attention_head) ... def create_attention_mask_from_input_mask(from_tensor, to_mask): """Create 3D attention mask from a 2D tensor mask. Args: from_tensor: 2D or 3D Tensor of shape [batch_size, from_seq_length, ...]. to_mask: int32 Tensor of shape [batch_size, to_seq_length]. Returns: float Tensor of shape [batch_size, from_seq_length, to_seq_length]. """ 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) # We don't assume that `from_tensor` is a mask (although it could be). We # don't actually care if we attend *from* padding tokens (only *to* padding) # tokens so we create a tensor of all ones. # # `broadcast_ones` = [batch_size, from_seq_length, 1] broadcast_ones = tf.ones( shape=[batch_size, from_seq_length, 1], dtype=tf.float32) # Here we broadcast along two dimensions to create the mask. mask = broadcast_ones * to_mask return mask 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): """Performs multi-headed attention from `from_tensor` to `to_tensor`. This is an implementation of multi-headed attention based on "Attention is all you Need". If `from_tensor` and `to_tensor` are the same, then this is self-attention. Each timestep in `from_tensor` attends to the corresponding sequence in `to_tensor`, and returns a fixed-with vector. This function first projects `from_tensor` into a "query" tensor and `to_tensor` into "key" and "value" tensors. These are (effectively) a list of tensors of length `num_attention_heads`, where each tensor is of shape [batch_size, seq_length, size_per_head]. Then, the query and key tensors are dot-producted and scaled. These are softmaxed to obtain attention probabilities. The value tensors are then interpolated by these probabilities, then concatenated back to a single tensor and returned. In practice, the multi-headed attention are done with transposes and reshapes rather than actual separate tensors. Args: from_tensor: float Tensor of shape [batch_size, from_seq_length, from_width]. to_tensor: float Tensor of shape [batch_size, to_seq_length, to_width]. attention_mask: (optional) int32 Tensor of shape [batch_size, from_seq_length, to_seq_length]. The values should be 1 or 0. The attention scores will effectively be set to -infinity for any positions in the mask that are 0, and will be unchanged for positions that are 1. num_attention_heads: int. Number of attention heads. size_per_head: int. Size of each attention head. query_act: (optional) Activation function for the query transform. key_act: (optional) Activation function for the key transform. value_act: (optional) Activation function for the value transform. attention_probs_dropout_prob: (optional) float. Dropout probability of the attention probabilities. initializer_range: float. Range of the weight initializer. do_return_2d_tensor: bool. If True, the output will be of shape [batch_size * from_seq_length, num_attention_heads * size_per_head]. If False, the output will be of shape [batch_size, from_seq_length, num_attention_heads * size_per_head]. batch_size: (Optional) int. If the input is 2D, this might be the batch size of the 3D version of the `from_tensor` and `to_tensor`. from_seq_length: (Optional) If the input is 2D, this might be the seq length of the 3D version of the `from_tensor`. to_seq_length: (Optional) If the input is 2D, this might be the seq length of the 3D version of the `to_tensor`. Returns: float Tensor of shape [batch_size, from_seq_length, num_attention_heads * size_per_head]. (If `do_return_2d_tensor` is true, this will be of shape [batch_size * from_seq_length, num_attention_heads * size_per_head]). Raises: ValueError: Any of the arguments or tensor shapes are invalid. """ 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] 加速内积计算 query_layer = transpose_for_scores(query_layer, batch_size, num_attention_heads, from_seq_length, size_per_head) # `key_layer` = [B, N, T, H] 加速内积计算 key_layer = transpose_for_scores(key_layer, batch_size, num_attention_heads, to_seq_length, size_per_head) ~~~ > 此时完成QKV的计算，接下来消除维度影响、softmax #### 完成Transformer模块构建 ~~~python # 衔接上面的 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): ... # Take the dot product between "query" and "key" to get the raw # attention scores. # `attention_scores` = [B, N, F, T] 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))) # 消除维度对结果的影响 if attention_mask is not None: # `attention_mask` = [B, 1, F, T] attention_mask = tf.expand_dims(attention_mask, axis=[1]) # Since attention_mask is 1.0 for positions we want to attend and 0.0 for # masked positions, this operation will create a tensor which is 0.0 for # positions we want to attend and -10000.0 for masked positions. adder = (1.0 - tf.cast(attention_mask, tf.float32)) * -10000.0 # mask为1时结果为0，mask为0时结果为非常大的负数 # Since we are adding it to the raw scores before the softmax, this is # effectively the same as removing these entirely. attention_scores += adder # 把上面的值加入原始得分里相当于mask为1不变，mask为0则变成很大的负数 # Normalize the attention scores to probabilities. # `attention_probs` = [B, N, F, T] attention_probs = tf.nn.softmax(attention_scores) # 再softmax时，非常大的负数则无限接近于0，就相当于不考虑 # This is actually dropping out entire tokens to attend to, which might # seem a bit unusual, but is taken from the original Transformer paper. 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: # 返回结果前判断维度是否一样，因为连接了很多层，会不断输入输出 # `context_layer` = [B*F, N*H] context_layer = tf.reshape( context_layer, [batch_size * from_seq_length, num_attention_heads * size_per_head]) else: # `context_layer` = [B, F, N*H] context_layer = tf.reshape( context_layer, [batch_size, from_seq_length, num_attention_heads * size_per_head]) return context_layer ~~~ > 上面处理完后，还有残差连接，防止训练结果比不训练的更差 #### 训练BERT模型 ~~~python # 回到transformer_model 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): # Run a linear projection of `hidden_size` then add a residual # with `layer_input`. with tf.variable_scope("output"): # 残差连接 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) # The activation is only applied to the "intermediate" hidden layer. with tf.variable_scope("intermediate"): intermediate_output = tf.layers.dense( attention_output, intermediate_size, activation=intermediate_act_fn, kernel_initializer=create_initializer(initializer_range)) # Down-project back to `hidden_size` then add the residual. with tf.variable_scope("output"): # 残差连接完，数据维度会增大，需要变回一直的维度 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 ~~~ > 最终，所有的结果已处理完成并输出向量，这样BertModel模块已经讲完，modeling.py的部分也完成了，我们再回到run_classifier.py ~~~python def create_model(bert_config, is_training, input_ids, input_mask, segment_ids, labels, num_labels, use_one_hot_embeddings): """Creates a classification model.""" ... # 前面的modeling.BertModel已经看过了，最终我们得到了QVK计算后的softemax层和残差连接后的结果 # In the demo, we are doing a simple classification task on the entire # segment. # # If you want to use the token-level output, use model.get_sequence_output() # instead. output_layer = model.get_pooled_output() hidden_size = output_layer.shape[-1].value # 获取向量 output_weights = tf.get_variable( # 构造全连接层，二分类的权重参数 "output_weights", [num_labels, hidden_size], initializer=tf.truncated_normal_initializer(stddev=0.02)) output_bias = tf.get_variable( # 构造偏值b "output_bias", [num_labels], initializer=tf.zeros_initializer()) with tf.variable_scope("loss"): # 常规的loss function if is_training: # I.e., 0.1 dropout output_layer = tf.nn.dropout(output_layer, keep_prob=0.9) logits = tf.matmul(output_layer, output_weights, transpose_b=True) # 结果乘上权重 logits = tf.nn.bias_add(logits, output_bias) # 再加上偏值项 probabilities = tf.nn.softmax(logits, axis=-1) # 加上softmax层 log_probs = tf.nn.log_softmax(logits, axis=-1) # 加上softmax层 one_hot_labels = tf.one_hot(labels, depth=num_labels, dtype=tf.float32) per_example_loss = -tf.reduce_sum(one_hot_labels * log_probs, axis=-1) # 计算得到损失 loss = tf.reduce_mean(per_example_loss) # 优化损失 return (loss, per_example_loss, logits, probabilities) # 返回结果 ~~~