AttributeError: 'Tensor' object has no attribute 'numpy' in custom loss function (Tensorflow 2.1.0)

Question

I want to train a model with a custom loss function, in order to do that, I need to convert the tensor to numpy array inside the method below:

def median_loss_estimation(y_true, y_predicted):
    a = y_predicted.numpy()

but I have this error:

AttributeError: 'Tensor' object has no attribute 'numpy'

Why? How can I convert the tensor to a numpy array?

score 5 · Accepted Answer · answered Apr 07 '20 at 14:06

5

The answer is: put run_eagerly=True in model.compile!

answered Apr 07 '20 at 14:06

Francesco Scala

201
3
15

score 1 · Answer 2 · answered Apr 06 '20 at 17:26

1

You're doing the right thing, only Tensorflow 2.1 is currently broken in that aspect. This would normally happen if you run the code without eager mode enabled. However, Tensorflow 2 by default runs in eager mode... or at least it should. The issue is tracked here.

There are at least two solutions to this:

Install the latest nightly build.
Set model.run_eagerly = True.

answered Apr 06 '20 at 17:26

Lukasz Tracewski

10,794
3
34
53

using `model.run_eagerly = True` the error remains! – Francesco Scala Apr 07 '20 at 10:25
@FrancescoScala Try then the first option, – Lukasz Tracewski Apr 07 '20 at 12:10
tried: tf `2.0.1`, `2.1.0`, `2.2.0rc2`, `tb-nightly (2.3.0something)` and I got every time the same error! – Francesco Scala Apr 07 '20 at 12:53

AttributeError: 'Tensor' object has no attribute 'numpy' in custom loss function (Tensorflow 2.1.0)

2 Answers2