Loss function returns 0 after first epoch for training set only when using validation data in training

[fabiodimarco]

Loss function returns 0 after first epoch for training set only when using validation data in training

[fabiodimarco]

Hi @sabre-sr, thank you for reporting the issue. I'll look into it later today. What are the input output dimensions of the data you are using?

[sabre-sr]

Thank you for your swift reply.
My data has dimensions of 48 floats as inputs and 24 floats as outputs, with 365 samples - 48x365 and 24x365 respectively.
Modifications to sample size (48x760/24x760), as well as modifying validation split ratio or splitting data with train_test_split() from sklearn library didn't resolve the issue.

I'm also attaching library versions below:

Tensorflow/Keras: 2.10.0
pandas: 1.5.3
numpy: 1.24.2
sklearn: 1.2.1

Thank you in advance for your help.

[fabiodimarco]

I committed a fix for that. Let me know if it does work on your side.

[sabre-sr]

The fix works, thank you very much once again. The issue can be now closed.