kl_l.append(kl_loss)
bow_l.append(bow_loss)
if global_step % test_step == 0:
time_now = strftime("%m-%d %H:%M:%S", gmtime())
p.put_step(epoch, global_step)
p.put_list([np.mean(recon_l), np.mean(kl_l), np.mean(bow_l)])
recon_l = []
kl_l = []
bow_l = []
if global_step % (test_step * 10) == 0:
""" EVAL and INFER """
# TEST
(test_recon_loss, test_kl_loss, test_bow_loss, test_ppl,
test_bleu_score, precisions,
_) = cvae.infer_and_eval(test_batches)
p.put_example(cvae)
p.put_step(epoch, global_step)
put_eval(test_recon_loss, test_kl_loss, test_bow_loss,
test_ppl, test_bleu_score, precisions, "TEST", log_f)
if kl_weight >= 0.35:
path = join(output_dir,
"breakpoints/at_step_%d.ckpt" % global_step)
save_path = saver.save(sess, path)
global_step += 1
"""GENERATE"""
# TRAIN SET
train_batches = batch_generator(train_data,
start_i,
saver1.restore(sess1, "classifier/07-16_14-33-58/breakpoints/best_test_loss.ckpt")
graph1.finalize()
"""build data"""
test_data = build_data(test_ori_f, test_rep_f, word2index)
test_batches = batch_generator(test_data, start_i, end_i, batch_size, permutate=False)
train_data = build_data(train_ori_f, train_rep_f, word2index)
global_step = best_step = 1
start_epoch = best_epoch = 1
total_step = (8 * len(train_data[0]) / batch_size)
(test_recon_loss, test_kl_loss, test_bow_loss,
perplexity, test_bleu_score, precisions, _) = seq2seq.infer_and_eval(test_batches)
p.put_example(seq2seq)
p.put_bleu(
test_recon_loss, test_kl_loss, test_bow_loss,
perplexity, test_bleu_score, precisions, "TEST")
lengths, ac, ac5 = seq2seq.policy_gen_eval(test_batches, classifier)
p.put_list([lengths, ac, ac5])
s = 1.
e = 1.
step = 500
for epoch in range(start_epoch, num_epoch + 1):
train_batches = batch_generator(train_data, start_i, end_i, batch_size, permutate=True)