Нейросеть на tensorflow

import math
import numpy as np
import tensorflow as tf
import matplotlib.pyplot as plt
 
iterations = 10000 # количество итераций обучения
learn_rate = 0.1 # коэффициент переобучения
 
hiddenSize = 4 # размер скрытого слоя
inputs = 2
outputs = 1
 
def read_from_file(name):
train_h2 = []
train_ch4 = []
tr_h2 = [0] * inputs
tr_ch4 = [0] * outputs
with open(name, "r") as file_values:
value = file_values.readlines()
for elem in value:
tr_h2 = [0] * inputs
tr_ch4 = [0] * outputs
j = 0
s = elem.split()
for i in range(len(s)):
if i < inputs:
tr_h2[i] = float(s[i])
else: 
tr_ch4[j] = float(s[i])
j += 1
train_h2.append(tr_h2)
train_ch4.append(tr_ch4)
 
def_train_dataset = np.array(train_h2)
def_train_values = np.array(train_ch4)
return def_train_dataset, def_train_values
 
h2 = tf.Variable(initial_value=50.0, dtype=tf.float32)
ch4 = tf.Variable(initial_value=80.0, dtype=tf.float32)
 
f =tf.divide(ch4,h2)
 
# выходной слой
model = tf.layers.dense(inputs, outputs,
activation = None,
kernel_initializer = 'random_uniform',
bias_initializer='zeros',
name = "output")
 
#функция подсчёта ошибки
cost = tf.losses.mean_squared_error(y, model)
 
train = tf.train.GradientDescentOptimizer(learn_rate).minimize(cost)
 
init = tf.initializers.global_variables()
 
with tf.Session() as session:
session.run(init)
 
name = "IO1.txt"
train_dataset, train_values = read_from_file(name)
for _ in range(iterations):
_, cost_now=session.run([train, cost], feed_dict = {h2: train_dataset, ch4: train_values})
print("cost = " + str(cost_now))
 
name = "test_IO1.txt"
train_dataset, train_values = read_from_file(name)
train_values1 = session.run(model, feed_dict = {h2: train_dataset})
 
'''print("input:")
print(train_dataset)
print("output(real):")
print(train_values)'''
 
# прежде всего нужно инициализировать все глобальные переменные
# в нашем случае это только x
tf.global_variables_initializer().run()
 
result_f, result_ch4, result_h2 = session.run([f, ch4, h2])
print("f = %.1f / %.1f= %.1f" % (result_ch4, result_h2, result_f))