Batch_dataset for cylinder_flow of MGN?

[ziweiyang]

Hi,

I am trying to reproduce the MGN for my own simulation data. So, I need to understand the batch mechanism in the original code. I have noticed in the dataset.py file, the function batch_dataset was never called in the executable file (run_model.py) in the main function for cylinder_flow case. I am curious that does this indicate each individual timestep is a batch for its own? And all learnable variables got updated for each training_step for each batch?

Thanks, ZY