Desc: return the size of input x in DIM.

Creates a `truncated_normal_initializer` with the given range.

Returns a list of the shape of tensor, preferring static dimensions. Args: tensor: A tf.Tensor object to find the shape of. expected_

Creates an `input_fn` closure to be passed to TPUEstimator.

Desc: the same functionality of tf.boolean_mask. The tf.boolean_mask operation is not available on the cloud TPU.

Desc: the same functionality of tf.boolean_mask. The tf.boolean_mask operation is not available on the cloud TPU.

Desc: expected cluster ids is : [[[21, 25], [18, 18]], [[63, 65], [46, 48], [27, 29]], [[88, 88], [89, 89]]] Args:

Gets final hidden layer of encoder. Returns: float Tensor of shape [batch_size, seq_length, hidden_size] corresponding to

Desc: hidden_size should be equal to embeddding_size. Args: doc_overlap_window_embs: (num_window, window_siz

Converts a sequence of [tokens|ids] using the vocab.

Perform dropout. Args: input_tensor: float Tensor. dropout_prob: Python float. The probability of dropping out a value (NOT of

Desc: fully-connected neural network. transform non-linearly the [input] tensor with [hidden_size] to a fix [output_

Constructs a `BertConfig` from a json file of parameters.

Desc: forward function for training mention proposal module. Args: instance: a tuple of train/dev/test data

Desc: transform the mention start/end position index tf.int32 Tensor to a tf.int32 Tensor with 1/0 labels for the input subtoken

Desc: compute the unifrom start/end loss and probabilities. Args: pred_sequence_logits: (input_shape, 1)

Desc: compute the unifrom start/end loss and probabilities. Args: pred_sequence_logits: (input_shape, 1)

Converts `text` to Unicode (if it's not already), assuming utf-8 input.

Desc: fully-connected neural network. transform non-linearly the [input] tensor with [hidden_size] to a fix [output_

Desc: the original gold labels is padded to the fixed length and only contains the position index of gold mentions.

Desc: construct question based on the selected mention.

Run layer normalization on the last dimension of the tensor.

Reshapes a >= rank 2 tensor to a rank 2 tensor (i.e., a matrix).

Runs basic whitespace cleaning and splitting on a piece of text.

Get the n-best logits from a list.

Loads a data file into a list of `InputBatch`s.

Creates an `input_fn` closure to be passed to TPUEstimator.

Read a SQuAD json file into a list of SquadExample.

Reshapes a rank 2 tensor back to its original rank >= 2 tensor.

Desc: transform the mention start/end position index tf.int32 Tensor to a tf.int32 Tensor with 1/0 labels for the input subtoken

Desc: transform the mention (start, end) position pairs to a span matrix gold_span_sequence_labels. matrix[i][j]

Check if this is the 'max context' doc span for the token.

Performs invalid character removal and whitespace cleanup on text.

Compute softmax probability over raw logits.

Decodes a record to a TensorFlow example.

Decodes a record to a TensorFlow example.

Whether to use L2 weight decay for `param_name`.

Get the variable name from the tensor name.

Returns tokenized answer spans that better match the annotated answer.

Checks whether CP is the codepoint of a CJK character.

Checks whether `chars` is a control character.

Checks whether `chars` is a punctuation character.

Checks whether `chars` is a whitespace character.

Splits punctuation on a piece of text.

Strips accents from a piece of text.

Adds whitespace around any CJK character.

See base class.

Raises an exception if the tensor rank is not of the expected rank. Args: tensor: A tf.Tensor to check the rank of. expected_rank: Py

Performs multi-headed attention from `from_tensor` to `to_tensor`. This is an implementation of multi-headed attention based on "Attention is

construct sliding windows for BERT processing :param sequence_length: e.g. 9 :param sliding_window_size: e.g. 4 :return: [(0, 4, [1,

construct sliding windows, allocate tokens and masks into each window :param tokenized_document: :param sliding_window_size: :return:

convert coreference information to clusters :param annotations: :return:

Create 3D attention mask from a 2D tensor mask. Args: from_tensor: 2D or 3D Tensor of shape [batch_size, from_seq_length, ...]. to_ma

Creates a classification model.

Looks up words embeddings for id tensor. Args: input_ids: int32 Tensor of shape [batch_size, seq_length] containing word ids.

Performs various post-processing on a word embedding tensor. Args: input_tensor: float Tensor of shape [batch_size, seq_length, emb

add speaker name information :param tokenized_document: tokenized document information :return:

Constructs a `BertConfig` from a Python dictionary of parameters.

Maps a string to a Python function, e.g., "relu" => `tf.nn.relu`. Args: activation_string: String name of the activation function. Ret

Project the tokenized prediction back to the original text.

Runs layer normalization followed by dropout.