# print('set 1')
# test_features = pickle.load(open('/mnt/shdstorage/for_classification/elmo/elmo_sent_test_v7.pkl', 'rb'))
# label = pd.read_csv('/mnt/shdstorage/for_classification/test_v7.csv')['label'].values.tolist()
# dtest = xgb.DMatrix(test_features)
# data = model.predict(dtest)
# print(data)
# prediction = np.array(data)
# print(prediction)
# prediction = prediction >= 0.5
# prediction = prediction.astype(int)
#
# test_1, test_0 = calculate_all_metrics(label, prediction, 'TEST')
#
# print('set 2')
print('Ver')
ver_features = pickle.load(
open('/mnt/shdstorage/for_classification/elmo/elmo_sent_ver.pkl', 'rb'))
label = pd.read_csv('/mnt/shdstorage/for_classification/new_test.csv'
)['negative'].values.tolist()
dver = xgb.DMatrix(ver_features)
data = model.predict(dver)
print(data)
prediction = np.array(data)
print(prediction)
prediction = prediction >= 0.5
prediction = prediction.astype(int)
verif_1, verif_0 = calculate_all_metrics(label, prediction, 'VERIFICATION')
# ================================= MODEL SAVE =========================================
# path_to_weights = "models_dir/model_%s.h5" % (timing)
path_to_weights = '/mnt/shdstorage/for_classification/models_dir/model_%s.h5' % timing
path_to_architecture = "/mnt/shdstorage/for_classification/models_dir/architecture/model_%s.h5"
model.save_weights(path_to_weights)
model.save(path_to_architecture)
print('Model is saved %s' % path_to_weights)
else:
timing = previous_weights.split('/')[-1].split('_')[-1].split('.')[0]
score, acc = model.evaluate(X_test, y_test, batch_size=batch_size)
print('Test score:', score)
print('Test accuracy:', acc)
train_res = model.predict_classes(X_train)
train_res = [i[0] for i in train_res]
train_1, train_0 = calculate_all_metrics(y_train, train_res, 'TRAIN')
test_res = model.predict_classes(X_test)
test_res = [i[0] for i in test_res]
test_1, test_0 = calculate_all_metrics(y_test, test_res, 'TEST')
ver_res = model.predict_classes(X_ver)
data = pd.read_csv('/mnt/shdstorage/for_classification/new_test.csv')
label = data['label'].tolist()
ver_res = [i[0] for i in ver_res]
verif_1, verif_0 = calculate_all_metrics(label, ver_res, 'VERIFICATION')
# ============================== PRINT METRICS =====================================
# train_res = model.predict_classes(X_train)
# train_res = [i[0] for i in train_res]
# train_1, train_0 = calculate_all_metrics(y_train, train_res, 'TRAIN')
#
# test_res = model.predict_classes(X_test)
# test_res = [i[0] for i in test_res]
# test_1, test_0 = calculate_all_metrics(y_test, test_res, 'TEST')
ver_res = model.predict_classes(verification)
path_to_verification = verification_name
data = pd.read_csv(path_to_verification)
label = data['negative'].tolist()
ver_res = [i[0] for i in ver_res]
verif_1, verif_0 = calculate_all_metrics(label, ver_res, 'VERIFICATION')
# ======================= LOGS =======================
logs_name = 'logs/qrnn/%s.txt' % timing
try:
stream = open('logs/qrnn/%s.txt' % timing, 'r')
stream.close()
except:
with open(logs_name, 'w') as f:
f.write('======= DATASETS =======\n')
f.write('Train set data: %s\n' % x_train_set_name)
f.write('Test set data: %s\n' % x_test_set_name)
f.write('Train labels: %s\n' % y_train_labels)