hidden_dim = 100
ae = AutoEncoder(max_sequence_len, vocab.NumIds(), embedding_dim,
hidden_dim)
ae.build_models()
reducelr_cb = ReduceLROnPlateau(monitor='val_loss',
factor=0.5,
patience=10,
verbose=1,
mode='auto',
epsilon=0.0001,
cooldown=0,
min_lr=1e-20)
checkpoint_cb = ModelCheckpoint(model_weights, period=1)
earlystopping_cb = EarlyStopping(min_delta=0.0001, patience=10)
callbacks_list = [reducelr_cb, checkpoint_cb, earlystopping_cb]
x = p.get_data()[:5000]
print(len(x))
print('-' * 30, 'Loaded data', '-' * 30)
ae.train(x, batch_size, num_epochs=30, callbacks_list=callbacks_list)
# AutoEncoder.save(ae, model_params, model_file, model_weights)
# ae = AutoEncoder.load(model_params, model_file, model_weights)
data = p.process_sentences([
'malignant melanoma is now the fifth most common cancer .',
'the babies are conjoined from the chest to the navel.',
'experts believe the rapidly rising rate in older people.'
])
print(data)
prediction = ae.predict(data)
sentences = predict_sentences(prediction, vocab)
for s in sentences:
