This step creates a checkpoint and shuffles the order of the batches before the execution of the next step. The batch_randomizer step is renamed and was former called the randomizer step (3.1.0 and older).
parameters
checkpoint_processing(optional): Boolean which indicates if a checkpoint gets created in a non-training execution.
example
{
"step": "batch_randomizer",
"type": "batch_randomizer",
"checkpoint_processing": true
}