def run():
"""Builds model, loads data, trains and evaluates"""
#df = dataloader.DataLoader().load_data(CFG)
#print(df[2])
model = MyModel(CFG)
model.load_data()
model.build()
metrics = pd.DataFrame(model.train())
print(metrics)
def main_leePincombeWelsh():
f = open(Configuration.get_LOG(), "a+")
f.write(
"\n\n......................Experiment Begin......................\n")
f.write("%s\n" % t.time())
f.close()
Configuration.write_configuration(logfile=Configuration.get_LOG())
data_x_a, data_x_b, data_y, embedding_matrix = generate_data_leePincombeWelsh(
)
print(data_x_a.shape)
print(data_x_a[1:3, :])
print(data_x_b.shape)
print(data_y.shape)
print(embedding_matrix.shape)
print(embedding_matrix[1:3, :])
# exit()
indices = np.arange(data_x_a.shape[0])
np.random.shuffle(indices)
data_x_a = data_x_a[indices]
data_x_b = data_x_b[indices]
data_y = data_y[indices]
nb_validation_samples = int(Configuration.get_VALIDATION_SPLIT() *
data_x_a.shape[0])
x_a_train = data_x_a[:-nb_validation_samples]
x_b_train = data_x_b[:-nb_validation_samples]
y_train = data_y[:-nb_validation_samples]
x_a_test = data_x_a[-nb_validation_samples:]
x_b_test = data_x_b[-nb_validation_samples:]
y_test = data_y[-nb_validation_samples:]
# model = MyModel_New(embedding_matrix)
model = MyModel(inputLenght=Configuration.get_MAX_SEQUENCE_LENGTH(),
weights=embedding_matrix)
model.train(data_x_a=x_a_train,
data_x_b=x_b_train,
data_y=y_train,
batch_size=Configuration.get_BATCH_SIZE(),
episode=Configuration.get_EPIDOSE())
# print(model.predict(x_a_test, x_b_test))
f = open(Configuration.get_LOG(), mode="a+")
cosTrain, pearsonTrain = model.evaluate(
data_x_a=x_a_train,
data_x_b=x_b_train,
data_y=y_train,
batch_size=Configuration.get_BATCH_SIZE())
print("cosine of train data = %f, pearson of train data = %f" %
(cosTrain, pearsonTrain))
f.write("cosine of prediction of train data = %f\n" % cosTrain)
cosTest, pearsonTest = model.evaluate(
data_x_a=x_a_test,
data_x_b=x_b_test,
data_y=y_test,
batch_size=Configuration.get_BATCH_SIZE())
f.write("cosine of prediction of test data = %f\n" % cosTest)
print("cosine of test data = %f, pearson of test data = %f" %
(cosTest, pearsonTest))
f.close()
f = open(Configuration.get_LOG(), mode="a+")
f.write("......................Experiment End......................\n\n")
f.close()
