d = w2v.vector_size
similar_val = 0.9999
non_similar_val = 0.0001
steps_per_epoch = 500 # 196961 muestras
epochs = 200
validation_steps = 150
shann_obj = SHANN(max_len_doc_sents, max_len_doc_sent_words,
max_len_summ_sents, max_len_summ_sent_words, d, path_models,
name_models)
shann_obj._set_model()
train_file = "../../Corpora/CNNDM/dev.csv"
dev_file = "../../Corpora/CNNDM/dev.csv"
x_tr, y_tr = ut.load_csv_samples(train_file)
x_dv, y_dv = ut.load_csv_samples(dev_file)
generator_train = ut.generator_2(x_tr,
y_tr,
max_len_doc_sents,
max_len_summ_sents,
max_len_doc_sent_words,
max_len_summ_sent_words,
padding_val,
pos_pairs,
neg_pairs,
w2v,
d,
similar_val=similar_val,
non_similar_val=non_similar_val)
topk_sentences = 3
shann_obj = SHANN(max_len_doc_sents, max_len_doc_sent_words,
max_len_summ_sents, max_len_summ_sent_words,
d, path_models, name_models)
shann_obj._set_model()
shann_obj.load_weights(path_weights)
decoder = Decoder(max_len_doc_sents, max_len_doc_sent_words,
w2v, d, shann_obj.get_all_att_model(),
topk_sentences=topk_sentences)
test_file = "../../Corpora/CNNDM/test.csv"
x_ts, y_ts = ut.load_csv_samples(test_file)
"""
# Word Level #
print(len(x_ts))
summaries = decoder._word_decoder(x_ts[15])
print(summaries)
print(y_ts[15])
exit()
with open(output_file_words, "w", encoding="utf8") as fw:
for i in range(len(summaries)):
fw.write(summaries[i].strip() + "\t" + y_ts.iloc[i].strip() + "\n")
"""
# Sentence Level #
print(len(x_ts))
