For loop for creating neural net layers in tensorflow

NN-for-loop.py

'''
I've found this little setup (https://pythonprogramming.net/community/262/TensorFlow%20For%20loop%20to%20set%20weights%20and%20biases/) 
to create layers in a NN with a for loop. Unfortunately it doesn't really work - so here is a corrected version:

keep in mind, layer_config takes the form: [n_input, 200, 200, 200, 200, n_classes]
'''
def multilayer_perceptron(x, layer_config, name="neuralnet"):
    '''
    code from: https://pythonprogramming.net/community/262/TensorFlow%20For%20loop%20to%20set%20weights%20and%20biases/
    '''
    layers = {}
    layers_compute = {}

    with tf.name_scope(name):
        for i in range(1, len(layer_config)):
            new_layer = {'weights': tf.Variable(tf.random_normal([layer_config[i-1], layer_config[i]], 0, 0.1)),
                        'biases': tf.Variable(tf.random_normal([layer_config[i]], 0, 0.1))}
            layers[i-1] = new_layer

            with tf.name_scope("weights"):
                tf.summary.histogram("w_l"+str(i)+"_summary", new_layer['weights'])

            with tf.name_scope("biases"):
                tf.summary.histogram("b_l"+str(i)+"_summary", new_layer['biases'])

            l = tf.add(tf.matmul(x if i == 1 else layers_compute[i-2], layers[i-1]['weights']), layers[i-1]['biases'])
            
            with tf.name_scope(name):
                l = tf.nn.relu(l) if i != len(layer_config)-1 else l

            layers_compute[i-1] = l

    lastlayer = len(layers_compute)-1
    return layers_compute[lastlayer]