Method _run_ddp_forward

torch/nn/parallel/distributed.py:1357–1359

(self, *inputs, **kwargs)

            DistributedDataParallel._active_ddp_module = None

    def _run_ddp_forward(self, *inputs, **kwargs):
        with self._inside_ddp_forward():
            return self.module(*inputs, **kwargs)  # type: ignore[index]

    def _clear_grad_buffer(self):
        # Making param.grad points to the grad buffers before backward is based on the