tensorflow 2 api regression tensorflow.python.framework.ops.EagerTensor' object is not callable

Question

I am trying to implement a multivariate regression using tensorflow 2 api.

import tensorflow as tf
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

df = pd.DataFrame({'A': np.array([100, 105.4, 108.3, 111.1, 113, 114.7]),
                   'B': np.array([11, 11.8, 12.3, 12.8, 13.1,13.6]),
                   'C': np.array([55, 56.3, 57, 58, 59.5, 60.4]),
                   'Target': np.array([4000, 4200.34, 4700, 5300, 5800, 6400])})

X = df.iloc[:, :3].values
Y = df.iloc[:, 3].values

plt.scatter(X[:, 0], Y)
plt.show()

X = tf.convert_to_tensor(X, dtype=tf.float32)
Y = tf.convert_to_tensor(Y, dtype=tf.float32)

def poly_model(X, w, b):
    mult = tf.matmul(X, w)
    pred = tf.add(tf.matmul(X, w), b)
    return pred

w = tf.cast(tf.Variable(np.random.randn(3, 1), name='weight'), tf.float32)
b = tf.Variable(np.random.randn(), name='bias')

model = poly_model(X, w, b)

cost = tf.reduce_sum(tf.square(Y - model))

train_op = tf.optimizers.SGD(0.001)

train_op.minimize(cost, var_list=[w])

At the last line it throws me:

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

Also, I am a bit confused:

1) How to proceed without using Session. Just do something like : output = train_op(X)?

2) Do I need to use tf.GradientTape() as tape or this is only for graphs?

-- error trace --

TypeError                                 Traceback (most recent call last)
<ipython-input-1-ffbbbe1a3709> in <module>()
     32 train_op = tf.optimizers.SGD(0.001)
     33 
---> 34 train_op.minimize(cost, var_list=[w])

~/anaconda3/envs/dpl/lib/python3.6/site-packages/tensorflow/python/keras/optimizer_v2/optimizer_v2.py in minimize(self, loss, var_list, grad_loss, name)
    294     """
    295     grads_and_vars = self._compute_gradients(
--> 296         loss, var_list=var_list, grad_loss=grad_loss)
    297 
    298     return self.apply_gradients(grads_and_vars, name=name)

~/anaconda3/envs/dpl/lib/python3.6/site-packages/tensorflow/python/keras/optimizer_v2/optimizer_v2.py in _compute_gradients(self, loss, var_list, grad_loss)
    326     with backprop.GradientTape() as tape:
    327       tape.watch(var_list)
--> 328       loss_value = loss()
    329     grads = tape.gradient(loss_value, var_list, grad_loss)
    330 

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

Have you tried enabling eager execution? tf.enable_eager_execution() # requires r1.7 tensorflow.org/tutorials/eager/eager_basics — Scott Skiles
– Scott Skiles, Commented Mar 12, 2019 at 18:46
@ScottSkiles:I am using tensorflow 2.0 as I said, in which the eager is enabled by default.If I check if it is enab;ed , it is true. — George
– George, Commented Mar 12, 2019 at 22:20
Can you add more of the stacktrace leading up to the error message? — Scott Skiles
– Scott Skiles, Commented Mar 13, 2019 at 19:09

DecentGradient · Accepted Answer · 2019-03-14 05:13:39Z

8

+50

2) You definitely need to use GradientTape.

Check out the Effective TF2 Guide

1)Something like this:

import tensorflow as tf
import numpy as np

print("TensorFlow version: {}".format(tf.__version__))
print("Eager execution: {}".format(tf.executing_eagerly()))

x = np.array([
    [100, 105.4, 108.3, 111.1, 113, 114.7],
    [11, 11.8, 12.3, 12.8, 13.1, 13.6],
    [55, 56.3, 57, 58, 59.5, 60.4]
])

y = np.array([4000, 4200.34, 4700, 5300, 5800, 6400])


class Model(object):
    def __init__(self, x, y):
        # Initialize variable to (5.0, 0.0)
        # In practice, these should be initialized to random values.
        self.W = tf.Variable(tf.random.normal((len(x), len(x[0]))))
        self.b = tf.Variable(tf.random.normal((len(y),)))

    def __call__(self, x):
        return self.W * x + self.b


def loss(predicted_y, desired_y):
    return tf.reduce_sum(tf.square(predicted_y - desired_y))

optimizer = tf.optimizers.Adam(0.1)
# noinspection PyPep8Naming
def train(model, inputs, outputs):
    with tf.GradientTape() as t:
        current_loss = loss(model(inputs), outputs)
    grads = t.gradient(current_loss, [model.W, model.b])
    optimizer.apply_gradients(zip(grads,[model.W, model.b]))
    print(current_loss)


model = Model(x, y)

for i in range(10000):
    # print(model.b.numpy())
    train(model,x,y)

edited Mar 14, 2019 at 5:13

answered Mar 13, 2019 at 20:38

DecentGradient

7317 silver badges17 bronze badges

Sign up to request clarification or add additional context in comments.

7 Comments

George Over a year ago

I am not sure..It doesn't state anywhere that is mandatory.Only recommended.

DecentGradient Over a year ago

From what I understand, that's how gradients are calculated and applied with eager execution

DecentGradient Over a year ago

This page may help tensorflow.org/alpha/tutorials/eager/…

Scott Skiles Over a year ago

@DecentGradient: Nice one!for a great answer it would be really helpful if you explained how to fix the problem using GradientTape!

Scott Skiles Over a year ago

Also, I don't see anything in your link? Can you link to something more specific than the entire 2.0 change notes?

|

Collectives™ on Stack Overflow

tensorflow 2 api regression tensorflow.python.framework.ops.EagerTensor' object is not callable

1 Answer 1

7 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

7 Comments

Your Answer

Sign up or log in

Post as a guest

Related