def get_eval_dataloader()

in optimum/graphcore/trainer.py [0:0]

• 37 lines of code
• 7 McCabe index (conditional complexity)

    def get_eval_dataloader(self, eval_dataset: Optional[Dataset] = None) -> poptorch.DataLoader:
        """
        Returns the evaluation `poptorch.DataLoader`.

        Subclass and override this method if you want to inject some custom behavior.

        Args:
            eval_dataset (`torch.utils.data.Dataset`, *optional*):
                If provided, will override `self.eval_dataset`. If it is a [`~datasets.Dataset`] dataset, the columns not accepted
                by the `model.forward()` method are automatically removed. It must implement `__len__`.
        """
        poptorch_specific_kwargs = {
            "auto_distributed_partitioning": not isinstance(eval_dataset, torch.utils.data.IterableDataset),
            "mode": DataLoaderMode.Sync,
            "worker_init_fn": _WorkerInit(123),
        }

        if eval_dataset is None and self.eval_dataset is None:
            raise ValueError("Trainer: evaluation requires an eval_dataset.")
        eval_dataset = eval_dataset if eval_dataset is not None else self.eval_dataset
        data_collator = self.eval_data_collator

        if is_datasets_available() and isinstance(eval_dataset, datasets.Dataset):
            eval_dataset = self._remove_unused_columns(eval_dataset, description="evaluation")
        else:
            data_collator = self._get_collator_with_removed_columns(data_collator, description="evaluation")

        if isinstance(eval_dataset, torch.utils.data.IterableDataset):
            return poptorch.DataLoader(
                self.eval_opts,
                eval_dataset,
                batch_size=self.args.per_device_eval_batch_size,
                collate_fn=data_collator,
                num_workers=self.args.dataloader_num_workers,
                pin_memory=self.args.dataloader_pin_memory,
                **poptorch_specific_kwargs,
            )

        eval_sampler = self._get_eval_sampler(eval_dataset)

        self._check_dataset_can_fill_batch(eval_dataset, for_inference=True)

        return poptorch.DataLoader(
            self.eval_opts,
            eval_dataset,
            sampler=eval_sampler,
            batch_size=self.args.per_device_eval_batch_size,
            collate_fn=data_collator,
            drop_last=self.args.dataloader_drop_last,
            num_workers=self.args.dataloader_num_workers,
            pin_memory=self.args.dataloader_pin_memory,
            **poptorch_specific_kwargs,
        )