class BidirectionalOneShotIterator(dataloader_head: torch.utils.data.DataLoader, dataloader_tail: torch.utils.data.DataLoader)

Bases: object

Iterator over the data

static one_shot_iterator(dataloader: torch.utils.data.DataLoader) → Generator

Transform a PyTorch Dataloader into python iterator

Parameters:

dataloader – Dataloader

Returns:

Generator of data

class TestDataset(*args: Any, **kwargs: Any)

Bases: Dataset

Dataset with test data

static collate_fn(data: List[Any]) → Tuple[torch.Tensor, torch.Tensor, str]

Collate

Parameters:

data – data to collate

Returns:

(Tuple[Tensor, Tensor, Tensor, str]) positive_sample, negative_sample, mode

class TrainDataset(*args: Any, **kwargs: Any)

Bases: Dataset

Dataset with training data

static collate_fn(data: List[Any]) → Tuple[torch.Tensor, torch.Tensor, torch.Tensor, str]

Collate

Parameters:

data – data to collate

Returns:

(Tuple[Tensor, Tensor, Tensor, str]) positive_sample, negative_sample, subsample_weight, mode

get_info(triples: Tuple[Dict[str, Any], Dict[str, Any], Dict[str, Any]], dataset: TrainDataset | None = None, do_count: bool = False) → Tuple[int, int, int, int, int, str]

Get dataset info

Parameters:

• dataset – Dataset

• triples – (Tuple[Dict[str, Any], Dict[str, Any], Dict[str, Any]]) Dicts with triples splitted by train test and validation

Returns:

Tuple[number of entities, numbers of relations, volume train, volume validdation, volume_test, info_log]