Resume training at specific batch number

I know how to resume training from saved weights. However, looking at the code of train.py in example/bin, I see there is a line :

self.optimizer.step(self.batch) # Needed when resuming, harmless with batch=0

so I am wondering if there is a way to resume training from saved weights AND from a specific batch iteration number ? I assume this is useful for producing meaningful loss evolution with respect to number of epochs ?

But I could not understand from reading the code how to use such a possibility ? Furthermore, in my case, I had to modify

self.optimizer.step(self.batch) to self.optimizer.step()

because it triggered an error