process+tokenize string and return string,label,and stringlen

gets the index'th string from the dataset

read file and splice strings based on string ending array `self.ends`

Initialize n-best list of hypotheses.

Constructs BertConfig. Args: vocab_size_or_config_json_file: Vocabulary size of `inputs_ids` in `BertModel`. hidden_s

Construct a layernorm module in the TF style (epsilon inside the square root).

Create a new InputExample. :param guid: a unique textual identifier :param text_a: the sequence of text :param text_

Create a new PVP. :param args: the args :param tokenizer: the tokenizer :param label_list: the list of labels