k-Step Checkpoint Saving #318
Description
The training process now saves a checkpoint at every epoch, which is not efficient. Adding an option to save checkpoints only every k steps (for example, every 10 epochs) can improve efficiency and also reduce storage usage.