import tensorflow as tf

tf.compat.v1.disable_eager_execution()

#Parameters
W = tf.Variable([.3],tf.float32)
b = tf.Variable([-.3],tf.float32)

#Input and output
x = tf.compat.v1.placeholder(tf.float32)

linear_model = W*x+b

y = tf.compat.v1.placeholder(tf.float32)

#Loss 
square_delta = tf.square(linear_model-y)
loss = tf.reduce_sum(square_delta)

#Optimize
optimizer = tf.compat.v1.train.GradientDescentOptimizer(0.01)
train = optimizer.minimize(loss)

init = tf.compat.v1.global_variables_initializer()

sess = tf.compat.v1.Session()

sess.run(init)

for i in range (100):
    sess.run(train,{x:[1,2,3,4],y:[0,-1,-2,-3]})
    print(sess.run(loss,{x:[1,2,3,4],y:[0,-1,-2,-3]}))


print(sess.run([W,b]))