Does resume_from_checkpoint also make the trainer not go through the same data?

Would resume_from_checkpoint={path_to_checkpoint} just make the trainer go through the data again, regardless of whether it should be a new epoch? Or would it start from the point in the data at which it previously stopped?

I want to change the save_steps parameter in TrainingArguments in the middle of training and so I decided to stop the trainer and then let it run with the new save_steps. But I want to know whether the trainer also saves the information about which part of the training data it has already trained on.