def predict_valid(sess, model, logger):
"""Test on the valid data."""
time0 = time.time()
predict_labels_list = list() # 所有的预测结果
marked_labels_list = list()
for i in tqdm(range(int(n_va_batches))):
[X_batch, y_batch] = get_batch(data_valid_path, i)
marked_labels_list.extend(y_batch)
_batch_size = len(X_batch)
fetches = [model.y_pred]
feed_dict = {
model.X_inputs: X_batch,
model.batch_size: _batch_size,
model.tst: True,
model.keep_prob: 1.0
}
predict_labels = sess.run(fetches, feed_dict)[0]
predict_labels_list.extend(predict_labels)
f1_micro, f1_macro, score12 = cail_evaluator_least(predict_labels_list,
marked_labels_list)
print('precision_micro=%g, recall_micro=%g, score12=%g, time=%g s' %
(f1_micro, f1_macro, score12, time.time() - time0))
logger.info(
'\nValid predicting...\nEND:Global_step={}: f1_micro={}, f1_macro={}, score12={}, time=%g s'
.format(sess.run(model.global_step), f1_micro, f1_macro, score12,
time.time() - time0))
def predict(sess, model, logger):
"""Test on the test data."""
time0 = time.time()
te_batches = os.listdir(data_test_path)
n_te_batches = len(te_batches)
predict_labels_list = list() # 所有的预测结果
marked_labels_list = list()
for i in tqdm(range(n_te_batches)):
X_batch, y_batch = get_batch(data_test_path, i)
_batch_size = len(X_batch)
marked_labels_list.extend(y_batch)
fetches = [model.y_pred]
feed_dict = {
model.X_inputs: X_batch,
model.batch_size: _batch_size,
model.tst: True,
model.keep_prob: 1.0
}
predict_labels = sess.run(fetches, feed_dict)[0]
predict_labels_list.extend(predict_labels)
predict_scores_file = scores_path + model_name + '/' + 'predict.npy'
marked_scores_file = scores_path + model_name + '/' + 'origin.npy'
np.save(predict_scores_file, predict_labels_list)
np.save(marked_scores_file, marked_labels_list)
f1_micro, f1_macro, score12 = cail_evaluator_least(predict_labels_list,
marked_labels_list)
print('f1_micro=%g, f1_macro=%g, score12=%g, time=%g s' %
(f1_micro, f1_macro, score12, time.time() - time0))
logger.info(
'\nTest predicting...\nEND:Global_step={}: f1_micro={}, f1_macro={}, score12={}, time=%g s'
.format(sess.run(model.global_step), f1_micro, f1_macro, score12,
time.time() - time0))