tf explicit variables

import tensorflow as tf

class EvoNetwork:
    def __init__(self, input_size, weights, use_biases=False):
        with tf.Graph().as_default():
            self.inputs = tf.placeholder(tf.float32, [None, input_size])

            hidden_layer = self.inputs
            for i, weight in enumerate(weights):
                with tf.variable_scope('hidden_layer_{}'.format(i)):
                    W = tf.Variable(
                        initial_value=weight.astype(np.float32),
                        trainable=False,
                        name='W',
                    )

                    b = tf.Variable(
                        initial_value=np.zeros((weight.shape[1])).astype(np.float32),
                        trainable=False,
                        name='b',
                    )

                    hidden_layer = tf.add(tf.matmul(hidden_layer, W), b)
                    
                    if i+1 < len(weights):
                        hidden_layer = tf.nn.tanh(hidden_layer)

            self.output_layer = hidden_layer

            self.session = tf.Session()
            self.session.run(tf.global_variables_initializer())

    def infer(self, inputs):
        [infer] = self.session.run([self.output_layer], {self.inputs: inputs})
        return infer