Mohammed Innat
Mohammed Innat
@james77777778 Thanks for the info. That sounds great. Nowadays, `tf.data` is widely adopted across many projects, **largely inspired** by Keras 2. It would be really helpful if `keras.ops` could work...
I’ve replaced the built-in keras training API with pure torch for a keras model running on the torch backend, so I don’t have any runnable script to share at the...
I see. I will create another ticket for that after preparing some 🔁 code.
@sachinprasadhs With all due respect, I think there may have been a misunderstanding. I’d kindly ask you to revisit my original message. Just to clarify — my suggestion isn't about...
Nice work! Just a small suggestion, since this is under ops, would it be possible to unify the API for both 2D and 3D cases? For example, similar to how...
Not really, the review take ages. However, if this is a high priority feature, I can at least provide guidance to the contributor by breaking it down into sub task.
https://github.com/google-deepmind/videoprism
@ashpakshaikh26732 Interesting. I've been developing MONAI version in Keras 3, [MEDIC-AI](https://github.com/innat/medic-ai). It would be nice if you ping me there with some performance level effect of this augmentation layer -...
[collab](https://colab.research.google.com/drive/1J5V4tH4M9ruBBwJoW1i7ebVa7jZ8aL0K?usp=sharing)