BiMPM (Bilateral Multi-Perspective Matching) model implementation.

class allennlp.models.bimpm.BiMpm(vocab:, text_field_embedder: allennlp.modules.text_field_embedders.text_field_embedder.TextFieldEmbedder, matcher_word: allennlp.modules.bimpm_matching.BiMpmMatching, encoder1: allennlp.modules.seq2seq_encoders.seq2seq_encoder.Seq2SeqEncoder, matcher_forward1: allennlp.modules.bimpm_matching.BiMpmMatching, matcher_backward1: allennlp.modules.bimpm_matching.BiMpmMatching, encoder2: allennlp.modules.seq2seq_encoders.seq2seq_encoder.Seq2SeqEncoder, matcher_forward2: allennlp.modules.bimpm_matching.BiMpmMatching, matcher_backward2: allennlp.modules.bimpm_matching.BiMpmMatching, aggregator: allennlp.modules.seq2vec_encoders.seq2vec_encoder.Seq2VecEncoder, classifier_feedforward: allennlp.modules.feedforward.FeedForward, dropout: float = 0.1, initializer: allennlp.nn.initializers.InitializerApplicator = <allennlp.nn.initializers.InitializerApplicator object>, regularizer: Optional[allennlp.nn.regularizers.regularizer_applicator.RegularizerApplicator] = None)[source]

Bases: allennlp.models.model.Model

This Model implements BiMPM model described in Bilateral Multi-Perspective Matching for Natural Language Sentences by Zhiguo Wang et al., 2017. Also please refer to the TensorFlow implementation and PyTorch implementation.


Used to embed the premise and hypothesis TextFields we get as input to the model.


BiMPM matching on the output of word embeddings of premise and hypothesis.


First encoder layer for the premise and hypothesis


BiMPM matching for the forward output of first encoder layer


BiMPM matching for the backward output of first encoder layer


Second encoder layer for the premise and hypothesis


BiMPM matching for the forward output of second encoder layer


BiMPM matching for the backward output of second encoder layer


Aggregator of all BiMPM matching vectors


Fully connected layers for classification.

dropoutfloat, optional (default=0.1)

Dropout percentage to use.

initializerInitializerApplicator, optional (default=``InitializerApplicator()``)

If provided, will be used to initialize the model parameters.

regularizerRegularizerApplicator, optional (default=``None``)

If provided, will be used to calculate the regularization penalty during training.

decode(self, output_dict: Dict[str, torch.Tensor]) → Dict[str, torch.Tensor][source]

Converts indices to string labels, and adds a "label" key to the result.

forward(self, premise: Dict[str, torch.LongTensor], hypothesis: Dict[str, torch.LongTensor], label: torch.LongTensor = None, metadata: List[Dict[str, Any]] = None) → Dict[str, torch.Tensor][source]
premiseDict[str, torch.LongTensor]

The premise from a TextField

hypothesisDict[str, torch.LongTensor]

The hypothesis from a TextField

labeltorch.LongTensor, optional (default = None)

The label for the pair of the premise and the hypothesis

metadataList[Dict[str, Any]], optional, (default = None)

Additional information about the pair

An output dictionary consisting of:

A tensor of shape (batch_size, num_labels) representing unnormalised log probabilities of the entailment label.

losstorch.FloatTensor, optional

A scalar loss to be optimised.

get_metrics(self, reset: bool = False) → Dict[str, float][source]

Returns a dictionary of metrics. This method will be called by in order to compute and use model metrics for early stopping and model serialization. We return an empty dictionary here rather than raising as it is not required to implement metrics for a new model. A boolean reset parameter is passed, as frequently a metric accumulator will have some state which should be reset between epochs. This is also compatible with Metrics should be populated during the call to ``forward`, with the Metric handling the accumulation of the metric until this method is called.