neuralNetworkTrainer.py

from uwimg import *
import time
import sys

def softmax_model(inputs, outputs):
    l = [make_layer(inputs, outputs, SOFTMAX)]
    return make_model(l)

def neural_net(inputs, outputs):
    print (inputs)
    l = [make_layer(inputs, 32, RELU),
    		make_layer(32, 16,RELU),
		#make_layer(16, 8, RELU),
            make_layer(16, outputs, SOFTMAX)]
    return make_model(l)

def save_to_file(filepath, process_to_run):

	orig_stdout = sys.stdout
	with open('filepath', 'w') as f:
		sys.stdout = f

		process_to_run

		sys.stdout = orig_stdout
		f.close()

def neural_on_mnist_dataset():
	start_time = time.time()
	print("loading data...")
	train_file_path = "mnist.train"
	labels_path = "mnist/mnist.labels"
	test_file_path = "mnist.test"

	train = load_classification_data(c_char_p(train_file_path.encode('utf-8')), c_char_p(labels_path.encode('utf-8')), 1)
	test  = load_classification_data(c_char_p(test_file_path.encode('utf-8')),c_char_p(labels_path.encode('utf-8')) , 1)
	print("Loading Data done")
	print()

	print("training model...")
	batch = 128
	iters = 1000
	rate = .01
	momentum = .9
	decay = .0005

	m = neural_net(train.X.cols, train.y.cols)
	print(train_model(m, train, batch, iters, rate, momentum, decay))
	print("Training done")
	training_time = time.time()
	print('Training Ttime took {}'.format(training_time - start_time))

	testing_time_start = time.time()
	print("evaluating model...")
	print("training accuracy: %f", accuracy_model(m, train))
	print("test accuracy:     %f", accuracy_model(m, test))
	testing_time_ends = time.time()

	print('Testing took {}'.format(testing_time_ends - testing_time_start))

	print('Total Time Taken = {}'.format(testing_time_ends - start_time))

if __name__ == "__main__":
	save_to_file("results/mnist_Relu_Activation", neural_on_mnist_dataset())