Builds the input pipeline configuration and adds it to the output_spec. Args: pipeline_name: the name of the pipeline to add to the

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

:returns next idx, (agg_op id, col_id)

:returns next idx, (agg_op id, col_id)

:returns next idx, (agg_op id, col_id)

Initializer

Resets Tensorflow's FLAG values

Creates a configurable class from a dictionary. The dictionary must have "class" and "params" properties. The class can be either fully qualified, o

Recursively merges dict_y into dict_x.

Flattens a nested dictionary. Namedtuples within the dictionary are converted to dicts. Args: dict_: The dictionary to flatten. parent_ke

Runs a multi-bleu test.

Helper function to test the full model pipeline.

inplace break value nodes with a string separaed by spaces

canonicalize the query, replace strings to a special place holder

combine a list of json to a single json, which is the most consistent on the given keys :param json_list: a list of similar json with same ke

Computes the sequence loss for this model. seq_loss is the cross entropy loss for the output sequence. Returns a tuple `(losses, loss)

Computes the decoder outputs.

If key is not in spec, requests a value from the user and modifies spec to include that key:value pair. Args: spec: A dictionary sp

Applies final transformation to the decoder output once decoding is finished.

Get similar samples from train set based on text embedding :param train: the train set :param text: the text to embed, which can be the

Creates features and labels.

redo *break_value_nodes*

Converts a dictionary of batch items to a batch/list of dictionary items.

Initializer. Args: initialize_fn: callable that returns `(finished, next_inputs)` for the first iteration. sample_fn: callabl