def accumulated_iterations()

in opacus/optimizers/optimizer.py [0:0]

• 18 lines of code
• 6 McCabe index (conditional complexity)

    def accumulated_iterations(self) -> int:
        """
        Returns number of batches currently accumulated and not yet processed.

        In other words ``accumulated_iterations`` tracks the number of forward/backward
        passed done in between two optimizer steps. The value would typically be 1,
        but there are possible exceptions.

        Used by privacy accountants to calculate real sampling rate.
        """
        vals = []
        for p in self.params:
            if not hasattr(p, "grad_sample"):
                raise ValueError(
                    "Per sample gradient not found. Are you using GradSampleModule?"
                )
            if isinstance(p.grad_sample, torch.Tensor):
                vals.append(1)
            elif isinstance(p.grad_sample, list):
                vals.append(len(p.grad_sample))
            else:
                raise ValueError(f"Unexpected grad_sample type: {type(p.grad_sample)}")

        if len(set(vals)) > 1:
            raise ValueError(
                "Number of accumulated steps is inconsistent across parameters"
            )
        return vals[0]