Describes a single generation action for a SQL query.

Nonterminal symbol.

Information on how a NeuralQueryExpression was constructed. Attributes: operation: A text description of the operation. inner: The nested o

A single training/test example for simple sequence classification.

This exception indicates an unexpected condition during parsing.

A single training/test example for simple sequence classification.

A single training/test example for simple sequence classification.

A single training/test example for simple sequence classification.

A copy action from the schema.

Rule of form Non-Terminal -> Terminal.

Logical form class supports object relation.

The relation is undefined or invalid. Attributes: relation_name: The relation that is not defined. message: An explanation of the error

Linear map: sum_i(args[i] * W[i]), where W[i] is a variable. Args: args: a 2D Tensor or a list of 2D, batch, n, Tensors. output_size: int,

For SQL queries that are not supported in Abstract SQL.

A namedtuple that groups together all BERT inputs.

A single training/test example for QA.

Writes InputFeature to TF example file.

Writes InputFeature to TF example file.

Helper class for generating / detecting delimiters within a fixed range. Attributes: constructor: An invertible delimiter constructor. star

Contains both inputs and outputs for a NL to SQL example.

A delimited sentence w/ entity id & evidence metadata.

Contains wordpiece information as a substring of an utterance.

A single set of features of data.

Creates / detects indexed renderables that delimit fields within text. Attributes: fstring: A format string with a `uid` field used to create t

Based on wrapper of tf.contrib.seq2seq.LuongAttention. Allows to use a customized mask.

NP Wrapper rule.

Based on tensorflow.contrib.rnn.OutputProjectionWrapper.

Term class that supports location in predicate.

The type is undefined. Attributes: type_name: The type that is not defined. message: An explanation of the error

Contains information about a table in a database.

Writes InputFeature to TF example file.

Represents an anchored rule application.

A single training/test example for simple sequence classification. For examples without an answer, the start and end position are -1.

Contains information about column in a database table.

Terminal symbol.

Transformer output layer. Attributes: hidden_size: <int32>[1]. hidden_dropout_rate: <int32>[1]. layer_norm_eps: <float32>[1]. kerne

Represents an aggregation over multiple nodes.

Represents an aggregation over multiple nodes.

Holds beam search state data.

See modeling.BertModel.

See modeling.BertModel.

DP rule.

The entity is undefined and cannot be mapped onto the unknown id. Attributes: entity_name: The entity that is not defined. type_name: The t

Class to prevent re-initialization of Estimator when predicting.

Writes InputFeature to TF example file.

Writes InputFeature to TF example file.

Writes InputFeature to TF example file.

Implements Luong-style (multiplicative) attention scoring. This attention has two forms. The first is standard Luong attention, as described in:

A single training/test example for simple sequence classification.

A single training/test example for simple sequence classification.

A single set of features of data.

A single set of features of data.

A single set of features of data.

A single set of features of data.

A single set of features of data.

Not enough GPU available as requested by the user. Attributes: requested_gpu: The number of GPUs requested by the user. available_gpu: The

Represents a QCFG rule with type annotations based on target CFG.

Hyperparameters for the QA model.

Wraps rule application scores.

(DEPRECATED) Help for building Estimator-friendly NQL models. To use this, subclass it, and implement config_* methods, so as to define a procedu

NP rule.

Fake example so the num input examples is a multiple of the batch size. When running eval/predict on the TPU, we need to pad the number of examples

Fake example so the num input examples is a multiple of the batch size. When running eval/predict on the TPU, we need to pad the number of examples

Fake example so the num input examples is a multiple of the batch size. When running eval/predict on the TPU, we need to pad the number of examples

Fake example so the num input examples is a multiple of the batch size. When running eval/predict on the TPU, we need to pad the number of examples

Fake example so the num input examples is a multiple of the batch size. When running eval/predict on the TPU, we need to pad the number of examples

Read and process a Wikipedia JSONL dump.

One single adjustment to the query.

Represents a string-formatted CFG rule.

Represents an anchored rule application.

Evaluation of a single example pair.

Encode a frame with slot-fillers, grouped by slot.

A simple tokenizer.

A single training instance (sentence pair).

Transformer interdediate layer after self-attention. Attributes: intermediate_size: <int32>[1]. hidden_act: <int32>[1]. kernel_init: In

Represents a node in a generic Trie data structure.

Represents a node in a generic Trie data structure.

Two types are not compatible for use in an operation. Attributes: type_name1: The name of type 1. type_name2: The name of type 2. opera

Object that maps words in string form to indices. Code adapted from https://github.com/LauraRuis/multimodal_seq2seq_gSCAN/blob/ master/seq2seq/

Intransitive VP rule.

Transitive VP rule.

Container for all the parts in WinoDict answer.

Contains information that we might want to print out in error messages.