get NER of users' query, and classify the label of entity :param query: user input :return: llm output, consist of entity and

@param current: index of current step @param values: list of tuples (name, value_for_last_step). The progress bar

Runs decoding based on the encoder outputs.

Dumps the options to a file in the model directory. Args: model_dir: Path to the model directory. The options will be dumped into a f

Loads options from the given model directory. Args: model_dir: Path to the model directory.

Returns the proper variable scope for this module.

get the depth-first, left-to-right sequence of rule applications returns a list of production rules and a map to their parent rules

Encodes an input sequence. Args: inputs: The inputs to encode. A float32 tensor of shape [B, T, ...]. sequence_length: The lengt

Returns `(finished, next_inputs, next_state)`.

Tests a MetricSpec

Returns `sample_ids`.

Parses `item` only if it is a string. If `item` is a dictionary it is returned as-is.

Runs the cell in a session

parse a python code into a tree structure code -> AST tree -> AST tree to internal tree structure

Splits multiple sentences into words and flattens the result

Get fewshots according to different input parameters :param question: the question :param entities: the entities :param sql: the sql

Runs the bridge with the given arguments

Called per step of decoding (but only once for dynamic decoding). Args: time: Scalar `int32` tensor. inputs: Input (possibly nested t

Creates default parameters used by multiple RNN encoders.

Disables dropout during eval/inference mode

Creates the decoder module. This must be implemented by child classes and instantiate the appropriate decoder to be tested.

The computed learning rate decay function.