

Summarizer

class Summarizer

__init__

def __init__()

init_model

def init_model(preprocessor, vectorizer, embedding_weights_encoder, embedding_weights_decoder)

Initializes the model and provides necessary information for compilation.

Args

• preprocessor: Preprocessor object that preprocesses text for training and prediction.

• vectorizer: Vectorizer object that performs vectorization of the text.

• embedding_weights_encoder (optional): Matrix to initialize the encoder embedding.

• embedding_weights_decoder (optional): Matrix to initialize the decoder embedding.

predict

def predict(text)

Predicts summary of an input text.

predict_vectors

def predict_vectors(input_text, target_text)

Predicts summary of an input text and outputs information needed for evaluation: output logits, input tokens, output tokens, predicted tokens, preprocessed text, attention alignment.

Args

• input_text: Text used as input for prediction.

• target_text: Text used for evaluation.

Returns

new_train_step

def new_train_step(loss_function, batch_size, apply_gradients)

Initializes the train_step function to train the model on batches of data.

Args

• loss_function: Loss function to perform backprop on.

• batch_size: Batch size to use for training.

• apply_gradients: Whether to apply the gradients, i.e. False if you want to validate the model on test data.

save

def save(out_path)

Saves the model to a file.

Args

• out_path: Path to directory for saving the model.

load

def load(in_path)

Loads the model from a file.

Args

• in_path: Path to the model directory.