How can I solve this error in linear regression code with tensorflow 2.0?

Question

I was learning linear regression with tensorflow 2.0, and I intended to use the SGD optimizer in keras tool. Here's my code.

import tensorflow as tf
from tensorflow import keras

import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
%matplotlib inline

x_train = [1,2,3]
y_train = [1,2,3]

W = tf.Variable(np.random.normal([1]),name='weight')
b = tf.Variable(np.random.normal([1]),name='bias')

cost = tf.reduce_mean(tf.square(x_train*W + b-y_train))

opt = keras.optimizers.SGD(learning_rate=0.1)
fig=plt.grid()
plt.scatter(x_train,y_train)
plt.xlabel('x')
plt.ylabel('y')
for i in range(20):
    plt.title('hypothesis: epoch {}'.format(i+1))
    plt.plot(hypothesis, 'r.-',label='hypothesis')
    plt.legend(loc='best')
    opt.minimize(cost, var_list=[W,b])

I intended to print a plot for every single epoch, but I got this error at th last sentence of the loop.

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-8-be257fb20d71> in <module>
      8     plt.plot(hypothesis, 'r.-',label='hypothesis')
      9     plt.legend(loc='best')
---> 10     opt.minimize(cost, var_list=[W,b])

~\anaconda3\envs\tensorflow\lib\site-packages\tensorflow_core\python\keras\optimizer_v2\optimizer_v2.py in minimize(self, loss, var_list, grad_loss, name)
    315     """
    316     grads_and_vars = self._compute_gradients(
--> 317         loss, var_list=var_list, grad_loss=grad_loss)
    318 
    319     return self.apply_gradients(grads_and_vars, name=name)

~\anaconda3\envs\tensorflow\lib\site-packages\tensorflow_core\python\keras\optimizer_v2\optimizer_v2.py in _compute_gradients(self, loss, var_list, grad_loss)
    349       if not callable(var_list):
    350         tape.watch(var_list)
--> 351       loss_value = loss()
    352     if callable(var_list):
    353       var_list = var_list()

TypeError: 'tensorflow.python.framework.ops.EagerTensor' object is not callable

How can I solve this?

Answer 1

Your cost: cost = tf.reduce_mean(tf.square(x_train*W + b-y_train)) is actually a tensor, but the cost argument in opt.minimize(cost, var_list=[W,b]) requires a function.

So, you should create cost as a function instead of a tensor by calling lambda:

cost = lambda: tf.reduce_mean(tf.square(x_train*W + b-y_train))

And, you should train for longer epochs.