Training data in a sequential order, without shuffle

[kyrie-maker]

I'm currently using the Swift framework for model training and have encountered a scenario where I need to process training data in a sequential order, without shuffling. I understand that the default behavior is to shuffle data during training, and the documentation suggests that disabling this requires modifying the source code.

Before proceeding with source code modifications, I wanted to inquire if there's an existing method or recommended approach within the Swift framework to disable data shuffling during training. Specifically, is there a configuration parameter, command-line argument, or a way to customize the data loader to achieve this behavior?

[Jintao-Huang]

Please try: --dataset_shuffle false --train_dataloader_shuffle false