pytext.models.semantic_parsers.rnng package

Submodules

pytext.models.semantic_parsers.rnng.rnng_constant module

pytext.models.semantic_parsers.rnng.rnng_data_structures module

class pytext.models.semantic_parsers.rnng.rnng_data_structures.CompositionalNN(lstm_dim: int)

Bases: torch.jit._script.ScriptModule

Combines a list / sequence of embeddings into one using a biLSTM

class pytext.models.semantic_parsers.rnng.rnng_data_structures.CompositionalSummationNN(lstm_dim: int)

Bases: torch.jit._script.ScriptModule

Simpler version of CompositionalNN

class pytext.models.semantic_parsers.rnng.rnng_data_structures.Element(node: Any)

Bases: object

Generic element representing a token / non-terminal / sub-tree on a stack. Used to compute valid actions in the RNNG parser.

class pytext.models.semantic_parsers.rnng.rnng_data_structures.ParserState(parser=None)

Bases: object

Maintains state of the Parser. Useful for beam search

copy()

finished()

class pytext.models.semantic_parsers.rnng.rnng_data_structures.StackLSTM(lstm: torch.nn.modules.rnn.LSTM)

Bases: collections.abc.Sized, typing.Generic

The Stack LSTM from Dyer et al: https://arxiv.org/abs/1505.08075

copy()

element_from_top(index: int) → pytext.models.semantic_parsers.rnng.rnng_data_structures.Element

embedding() → torch.Tensor

Shapes:

return value: (1, lstm_hidden_dim)

pop() → Tuple[torch.Tensor, pytext.models.semantic_parsers.rnng.rnng_data_structures.Element]

Pops and returns tuple of output embedding (1, lstm_hidden_dim) and element

push(expression: torch.Tensor, element: pytext.models.semantic_parsers.rnng.rnng_data_structures.Element) → None

Shapes:

expression: (1, lstm_input_dim)

pytext.models.semantic_parsers.rnng.rnng_parser module

class pytext.models.semantic_parsers.rnng.rnng_parser.RNNGParser(ablation: pytext.models.semantic_parsers.rnng.rnng_parser.RNNGParserBase.Config.AblationParams, constraints: pytext.models.semantic_parsers.rnng.rnng_parser.RNNGParserBase.Config.RNNGConstraints, lstm_num_layers: int, lstm_dim: int, max_open_NT: int, dropout: float, actions_vocab, shift_idx: int, reduce_idx: int, ignore_subNTs_roots: List[int], valid_NT_idxs: List[int], valid_IN_idxs: List[int], valid_SL_idxs: List[int], embedding: pytext.models.embeddings.embedding_list.EmbeddingList, p_compositional)

Bases: pytext.models.semantic_parsers.rnng.rnng_parser.RNNGParserBase

arrange_model_inputs(tensor_dict)

arrange_targets(tensor_dict)

get_export_input_names(tensorizers)

get_export_output_names(tensorizers)

vocab_to_export(tensorizers)

class pytext.models.semantic_parsers.rnng.rnng_parser.RNNGParserBase(ablation: pytext.models.semantic_parsers.rnng.rnng_parser.RNNGParserBase.Config.AblationParams, constraints: pytext.models.semantic_parsers.rnng.rnng_parser.RNNGParserBase.Config.RNNGConstraints, lstm_num_layers: int, lstm_dim: int, max_open_NT: int, dropout: float, actions_vocab, shift_idx: int, reduce_idx: int, ignore_subNTs_roots: List[int], valid_NT_idxs: List[int], valid_IN_idxs: List[int], valid_SL_idxs: List[int], embedding: pytext.models.embeddings.embedding_list.EmbeddingList, p_compositional)

Bases: pytext.models.model.BaseModel

The Recurrent Neural Network Grammar (RNNG) parser from Dyer et al.: https://arxiv.org/abs/1602.07776 and Gupta et al.: https://arxiv.org/abs/1810.07942d. RNNG is a neural constituency parsing algorithm that explicitly models compositional structure of a sentence. It is able to learn about hierarchical relationship among the words and phrases in a given sentence thereby learning the underlying tree structure. The paper proposes generative as well as discriminative approaches. In PyText we have implemented the discriminative approach for modeling intent slot models. It is a top-down shift-reduce parser than can output trees with non-terminals (intent and slot labels) and terminals (tokens)

contextualize(context)

Add additional context into model. context can be anything that helps maintaining/updating state. For example, it is used by DisjointMultitaskModel for changing the task that should be trained with a given iterator.

forward(tokens: torch.Tensor, seq_lens: torch.Tensor, dict_feat: Optional[Tuple[torch.Tensor, ...]] = None, actions: Optional[List[List[int]]] = None, contextual_token_embeddings: Optional[torch.Tensor] = None, beam_size=1, top_k=1) → List[Tuple[torch.Tensor, torch.Tensor]]

RNNG forward function.

Parameters:

• tokens (torch.Tensor) – list of tokens
• seq_lens (torch.Tensor) – list of sequence lengths
• dict_feat (Optional[Tuple[torch.Tensor, ..]]) – dictionary or gazetteer features for each token
• actions (Optional[List[List[int]]]) – Used only during training. Oracle actions for the instances.

Returns:

list of top k tuple of predicted actions tensor and corresponding scores tensor. Tensor shape: (batch_size, action_length) (batch_size, action_length, number_of_actions)

classmethod from_config(model_config, feature_config=None, metadata: pytext.data.data_handler.CommonMetadata = None, tensorizers: Dict[str, pytext.data.tensorizers.Tensorizer] = None)

get_loss(logits: List[Tuple[torch.Tensor, torch.Tensor]], target_actions: torch.Tensor, context: torch.Tensor)

Shapes:

logits[1]: action scores: (1, action_length, number_of_actions) target_actions: (1, action_length)

get_pred(logits: List[Tuple[torch.Tensor, torch.Tensor]], context=None, *args)

Return Shapes:

preds: batch (1) * topk * action_len scores: batch (1) * topk * (action_len * number_of_actions)

get_single_pred(logits: Tuple[torch.Tensor, torch.Tensor], *args)

push_action(state: pytext.models.semantic_parsers.rnng.rnng_data_structures.ParserState, target_action_idx: int) → None

Used for updating the state with a target next action

Parameters:

• state (ParserState) – The state of the stack, buffer and action
• target_action_idx (int) – Index of the action to process

save_modules(*args, **kwargs)

Save each sub-module in separate files for reusing later.

valid_actions(state: pytext.models.semantic_parsers.rnng.rnng_data_structures.ParserState) → List[int]

Used for restricting the set of possible action predictions

Parameters:state (ParserState) – The state of the stack, buffer and action Returns:indices of the valid actions Return type:List[int]

Module contents