Loss Accumulation over several training steps (epochs)

[shaimove]

Hi,
I have limited GPU memory, and in order to increase my batch size, I want to perform loss accumulation over several training steps.
How do I change the tuner object to do that? In a simple loop, it's very clear, but I using Bayesian Optimization for my project.
thanks,
Sharon

[NikitaShubin]

It's not related to Keras Tuner.
KT using model.fit:
IMHO, you shuld modify your model methods instead.