NotImplementedError when assigning validation split

TensorFlow General Discussion
, ,
1

I am trying to recreate this example of VAE:
Variational AutoEncoder

I want to assign a validation split at the fit method:
vae.fit(x=mnist_digits, epochs=30,shuffle=True, batch_size=32,validation_split=0.2)

However I get this error:
NotImplementedError: in user code:

File “c:\Users\31613\anaconda3\envs\research\lib\site-packages\keras\engine\training.py”, line 1525, in test_function *
return step_function(self, iterator)
…
Call arguments received:
• inputs=tf.Tensor(shape=(32, 28, 28, 1), dtype=float32)
• training=False
• mask=None

Anyone knows how to fix it?

3

Have you checked NotImplementedError for predict keras for vae - Stack Overflow ?

4

Thank you Bhack. The error is the same, but I dont see a solution on it. In my case I get the error when I assign validation data.

5

Do you have provided the test_step as described in the official Doc?

https://keras.io/guides/customizing_what_happens_in_fit/

6

No, I used the standard tutorial found here:

I cant find an example with a keras variational autoencoder using the validation step

7