Like re.split but skips matches outside of configured range.

An NQL encoding of an appropriate Tensorflow constant. The constant should be a Tensorflow expression corresponding to a minibatch of weighte

Run model on given query and entity. Args: query: String. Returns: logits: Numpy array of logits for each output entity.

Find a variable `var_to_find` the arguments and replace it by `replace_by_var`.

Main entry point for this class.

Add an element to the sketch. Args: sketch: sketch to add x to x: name or id in string

Load the model, its tokenizer, and weights from the checkpoint. Args: model_checkpoint: Checkpoint to restore from, from .save() mode

Makes a cell name from coordinates.

Follow a relation, or relation group, in the knowledge graph. Specifically, if x evaluates to a set of knowledge graph entities, then x.follo

(DEPRECATED) An NQL encoding of a Tensorflow placeholder. The placehold will be configured to hold weighted sets of entities of the appropria

Encode the example using the input TF example. Args: claim: Claim text evidence: Evidence text claim_label: Claim verif

Returns the legal action space after a sequence of actions.

Add an entry to the chart.

A universal set containing all entities of some type. Args: type_name: the string type name Returns: A NeuralQueryExpression for

Return the shape of the matrix defining this relation. Args: rel_name: a string naming a KG relation. Returns: a tuple of two in

Test the model on this dataset over the examples in a dataset. Args: dset: a tf.data.Dataset Returns: a dictionary of results fr

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

Train the model on this dataset over the examples in a dataset. Args: dset: a tf.data.Dataset

Eval one tf_prediction. Args: tf_prediction: a dictionary of predicted values

Return True if can be parsed given target CFG.

Evaluate the Tensorflow expression associated with this NeuralQueryExpression. Args: session: (DEPRECATED) tf.Session used to evaluate if n

An NQL encoding of a one-hot set containing just this entity. Args: entity_name: a string naming an entity type_name: the string type

Tokenize text. Args: text: text in natural language Returns: a list of tokens, segment_ids, and input mask

Declare the domain and range types for a relation. Args: rel_name: string naming a relation domain_type: string naming the type of su

Reads a tab separated value file.

Convert a Tensorflow expression to a NeuralQueryExpression. Args: expr: a Tensorflow expression which evaluates to a dense matrix of size

Add an entry to the chart.

Sample a source and target pair. Args: score_fn: A scoring function that takes parent_rule (QCFGRule), nt_idx, child_rule (QCFGRule

Persist the encoder and the model to disk.

An integer upper bound on symbol ids. Returns: one plus largest assigned symbol id.

Turn an english natural language string into tokens. Args: text: to transform Returns: tokens: list of string tokens

Reads a tab separated value file.

Builds model by instantiating flax module associated with task.

Try to insert a symbol into the table. No action will be taken after the table has been frozen. Args: symbol: a string Raises:

Given examples in JSON format, predict evidence relevance. Args: examples: List of claim/evidence pairs to rank Returns: Scalar

The range for this relation. Args: rel_name: a string naming a declared relation. Returns: a string naming the type of the objec

SqlSpan constructor where only one argument should be non-None.

The domain for this relation. Args: rel_name: a string naming a declared relation. Returns: a string naming the type of the subj

Produces function to preprocess batches. The difference compared to `preprocess_fn` is that `collater_fn_` does not receive individual exampl

Save generated memories into files.

Wrapper of follow operation inside EmQL.

Wrapper to encode a single relation into its EmQL representation.

Convert word pieces to text.

Like re.finditer but skips matches outside of configured range.

Retrieve the maximum index of any entity with given type. Args: type_name: string name of a declared type Returns: Maximum index

Produces model-specific dummy input batch. Tasks are responsible for producing a dummy input in the form of a py-tree with the same shape as

Returns the string serialization of the logical form.

Reads a tab separated value file.

Reads a tab separated value file.

See base class.

Declare an entity type. Declared entity types can be used as domain and ranges for relations. There are three flavors of entity types. By d

Return the unknown symbol id or None if nonexistent.

Intersect two sketches. Args: sk1: first sketch sk2: second sketch Returns: a countmin sketch for intersection

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

Retrieve the string name the entity with given index and type. Args: entity_id: integer id for entity type_name: string name of decla

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

Return the answer score for documents_list.

(DEPRECATED) Construct and return a Model. Args: feature_ph_dict: maps feature names to placeholders that will hold the correspondi