for train, test in get_10_fold_data(data_1, num_labels):
flag += 1
data_1_train = data_1[train]
data_2_train = data_2[train]
y_train = labels_all[train]
data_1_test = data_1[test]
data_2_test = data_2[test]
y_test = labels_all[test]
end2end_model = E2E_MCTN_Model(configs, data_1_train, data_2_train)
for i in range(10):
print('now it is epoch %d' % i)
x2_train = end2end_model.embeding_model.predict(data_2_train).mean(1)
x2_test = end2end_model.embeding_model.predict(data_2_test).mean(1)
end2end_model.model.fit(
x=[data_1_train],
y=[x2_train, y_train],
epochs=1,
validation_data=[data_1_test, [x2_test, y_test]],
# self.input_test,
batch_size=256,
verbose=2,
)
predictions = end2end_model.model.predict(data_1_test)[-1]
get_preds_statistics(predictions, y_test)
print(
'--------------------------finish %d cross validation------------------------------------------------'
% flag)
ModelCheckpoint(weights_path,
monitor='val_loss',
save_best_only=True,
verbose=1),
]
try:
end2end_model.model.load_weights(weights_path)
print("\nWeights loaded from {}\n".format(weights_path))
except:
print("\nCannot load weight. Training from scratch\n")
#
print("TRAINING NOW...")
train = 1
if train == 1:
history = end2end_model.train(weights_path=weights_path,
n_epochs=args.train_epoch,
val_split=args.val_split,
batch_size=args.batch_size,
callbacks=callbacks)
with open('history_params.sav', 'wb') as f:
pickle.dump(history.history, f, -1)
print("PREDICTING...")
predictions = end2end_model.predict()
# predictions = predictions.reshape(-1, )
get_preds_statistics(predictions, feats_dict['test_labels'])
