Provide a list of additional instances of a type. If the type doesn't previously exist it will be defined. Args: type_name: string nam

Get the numeric id of a symbol. This assumes that either the symbol has been inserted in the past - ie that self.has_id(symbol) succeeds - or

Construct RNN cells. Args: mode: train or eval. Keys from tf.estimator.ModeKeys. hps: Hyperparameters. input_dim: input size. num_u

Produces function to preprocess samples. Tasks should be aware of what type of input they require and are responsible for creating these inpu

Run prediction.

Read a json file into a list of Example.

Wrapper to encode a single entity into its EmQL representation.

Return a string based on x.id or x.id[x.name].

Prevent further additions to the symbol table. After a table is frozen, subsequent calls to insert() will not change the table, and calls to

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

Load knowledge graph data from a files or lines of strings. Each line in the file is tab-separated with fields rel_name head tail [confide

Convert text to word pieces.

Check if the sketch contains x. Args: sketch: sketch to add xs to x: name or id in string Returns: True or False

Creates a `truncated_normal_initializer` with the given range.

Gets a single sequence representation for classification.

Test if the relation has been declared as 'dense'. Args: rel_name: a string naming a declared relation. Returns: a boolean

Creates task loss function. See BaseTask. EaE is pre-trained with a combination of 1) MLM loss, 2) entity-linking loss comparing mention

Checks the results of running `CreateTFExampleFn.process`. Args: entry: The output of `preproc.create_entry_from_json`. creator_fn: T

Takes the retrieval results of an example and yields seq2seq examples. Args: example: a RawExample. Yields: AugmentedExample, on

Creates the TPUEstimator object.

Load vocabulary.

Produces function to postprocess task samples (input and output). The method is occasionally called during training or evaluation to save mod

A one-hot 1-by-N matrix encoding this entity. Args: entity_name: a string naming an entity. type_name: the string type name of the na

Resplit a dataset of jsonlines with QA pairs.

Weight elements in set by confidence in some relationship holding. Specifically, weight all elements x in a set by the confidence given to th

Reads a tab separated value file.

Sums values corresponding to repeated indices. Returns the unique indices and their summed values. Args: indices: [num_nnz, rank] Tensor.

Adds base hparams for LM.

Create a new NeuralQueryContext and configure it. Args: context_factory: factory to construct the NQL context gpus: number of gpus av

Convert symbol strings to CFGRule. Args: lhs: String symbol for LHS. rhs: List of string symbols for RHS. rule_idx: Integer ind

Constructs a ConverterConfig from the given dict.

All declared relations. Returns: an list of string relation names.

Test if the type has been previously declared. Args: type_name: a string naming a declared type. Returns: boolean

Lower text and remove punctuation, articles and extra whitespace.

Render an iterable of renderable objects. Arguments: renderables: An iterable of renderable objects. sep: The seperator. Returns: Th

Save numpy array into multiple shards. Input array will be divided and saved into files with names "prefix-X-of-num_shards", where `X` is in [off

Return the score for each candidate, order does not change. Args: claim: The claim to match candidates: The candidates to rank

Gets the most specific dimension size(s) of the given tensor. This is a wrapper around the two ways to get the shape of a tensor: (1) t.get_shape

Tokenizes a string (in raw bytes) return a list of string tokens.

Return set of non-terminal symbols in `rhs`.

Test restrict behavior on types written and read through tempfile. Args: input_types: A list of the types to start with. restrict_wri

Declare a group of relations. This must be called after all the triples for all the relations in the group have already been defined. A

Returns the final documents for the current step. Returns: List of documents at the end of an episode.

Gets a sequence representation, one position per character.

Detects whether an XML element has the given tag.

Return only rules having the to-be-expanded non-terminal as lhs.

Simple linear transform of x. Args: x: <float>[batch_size, length, input_size] output_size: Integer specifying output size. scope: Stri

Filter invalid nodes.

Runs a prediction on a set of features. Calling multiple times does *not* regenerate the graph which makes predict much faster. Args:

Converts TyDi entries into serialized tf examples. Args: entry: "TyDi entries", dicts as returned by `create_entry_from_json`. errors

Adds a value action to the output. Args: item: A string value present in the SQL query. example: The NLToSQLExample being constructed.

Convert a tensorflow dataset of unbatched, text examples to TF batches. Args: dataset: TF Dataset to transform train: Whether to enco