"""Training KGE models based on the LCWA."""
import logging
from collections.abc import Callable
from math import ceil
from typing import ClassVar
from torch.nn import functional
from torch.utils.data import DataLoader, TensorDataset
from torch_max_mem.api import is_oom_error
from .training_loop import TrainingLoop
from ..constants import get_target_column
from ..losses import Loss
from ..models import Model
from ..triples import CoreTriplesFactory, LCWAInstances
from ..triples.instances import LCWABatch
from ..typing import FloatTensor, InductiveMode, MappedTriples, TargetHint
__all__ = [
"LCWATrainingLoop",
"SymmetricLCWATrainingLoop",
]
logger = logging.getLogger(__name__)
[docs]
class LCWATrainingLoop(TrainingLoop[LCWABatch]):
r"""A training loop that is based upon the local closed world assumption (LCWA).
Under the LCWA, for a given true training triple $(h, r, t) \in \mathcal{T}_{train}$, all triples
$(h, r, t') \notin \mathcal{T}_{train}$ are assumed to be false. The training approach thus uses a 1-n scoring,
where it efficiently computes scores for all triples $(h, r, t')$ for $t' \in \mathcal{E}$, i.e., sharing the
same (head, relation)-pair.
This implementation slightly generalizes the original LCWA, and allows to make the same assumption for relation, or
head entity. In particular the second, i.e., predicting the relation, is commonly encountered in visual relation
prediction.
[ruffinelli2020]_ call the LCWA ``KvsAll`` in their work.
"""
supports_slicing: ClassVar[bool] = True
num_targets: int
def __init__(self, *, target: TargetHint = None, **kwargs) -> None:
"""
Initialize the training loop.
:param target:
The target column. Defaults to tail prediction.
:param kwargs:
Additional keyword-based parameters passed to TrainingLoop.__init__
:raises ValueError:
If an invalid target column is given
"""
super().__init__(**kwargs)
# normalize target column
self.target = get_target_column(target)
# The type inference is so confusing between the function switching
# and polymorphism introduced by slicability that these need to be ignored
if self.target == 0:
self.score_method = self.model.score_h
elif self.target == 1:
self.score_method = self.model.score_r # type: ignore[assignment]
elif self.target == 2:
self.score_method = self.model.score_t # type: ignore[assignment]
else:
raise ValueError(f"Invalid target column: {self.target}. Must be from {{0, 1, 2}}.")
# Explicit mentioning of num_transductive_entities since in the evaluation there will be a different number
# of total entities from another inductive inference factory
self.num_targets = self.model.num_relations if self.target == 1 else self.model._get_entity_len(mode=self.mode)
# docstr-coverage: inherited
def _create_training_data_loader(
self, triples_factory: CoreTriplesFactory, sampler: str | None, **kwargs
) -> DataLoader[LCWABatch]: # noqa: D102
if sampler:
raise NotImplementedError(
f"LCWA training does not support non-default batch sampling. Expected sampler=None, but got "
f"sampler='{sampler}'.",
)
dataset = LCWAInstances.from_triples_factory(
triples_factory,
target=self.target,
loss_weighter=self.loss_weighter,
loss_weighter_kwargs=self.loss_weighter_kwargs,
)
return DataLoader(dataset=dataset, **kwargs)
@staticmethod
# docstr-coverage: inherited
def _get_batch_size(batch: LCWABatch) -> int: # noqa: D102
return batch["pairs"].shape[0]
@staticmethod
def _process_batch_static(
model: Model,
score_method: Callable,
loss: Loss,
num_targets: int,
mode: InductiveMode | None,
batch: LCWABatch,
start: int | None,
stop: int | None,
label_smoothing: float = 0.0,
slice_size: int | None = None,
) -> FloatTensor:
# Split batch components
batch_pairs = batch["pairs"]
batch_labels_full = batch["target"]
batch_weights = batch.get("weights")
# Send batch to device
batch_pairs = batch_pairs[start:stop].to(device=model.device)
batch_labels_full = batch_labels_full[start:stop].to(device=model.device)
if batch_weights is not None:
batch_weights = batch_weights[start:stop].to(device=model.device)
predictions = score_method(batch_pairs, slice_size=slice_size, mode=mode)
return (
loss.process_lcwa_scores(
predictions=predictions,
labels=batch_labels_full,
label_smoothing=label_smoothing,
num_entities=num_targets,
weights=batch_weights,
)
+ model.collect_regularization_term()
)
# docstr-coverage: inherited
def _process_batch(
self,
batch: LCWABatch,
start: int,
stop: int,
label_smoothing: float = 0.0,
slice_size: int | None = None,
) -> FloatTensor: # noqa: D102
return self._process_batch_static(
model=self.model,
score_method=self.score_method,
loss=self.loss,
num_targets=self.num_targets,
mode=self.mode,
batch=batch,
start=start,
stop=stop,
label_smoothing=label_smoothing,
slice_size=slice_size,
)
# docstr-coverage: inherited
def _slice_size_search(
self,
*,
triples_factory: CoreTriplesFactory,
batch_size: int,
sub_batch_size: int,
supports_sub_batching: bool,
) -> int: # noqa: D102
self._check_slicing_availability(supports_sub_batching)
reached_max = False
evaluated_once = False
logger.info("Trying slicing now.")
# Since the batch_size search with size 1, i.e. one tuple ((h, r) or (r, t)) scored on all entities,
# must have failed to start slice_size search, we start with trying half the entities.
slice_size = ceil(self.model.num_entities / 2)
while True:
try:
logger.debug(f"Trying {slice_size=:_} now.")
self._train(
triples_factory=triples_factory,
num_epochs=1,
batch_size=batch_size,
sub_batch_size=sub_batch_size,
slice_size=slice_size,
only_size_probing=True,
)
except RuntimeError as runtime_error: # noqa: PERF203
self._free_graph_and_cache()
if not is_oom_error(runtime_error):
raise runtime_error
if evaluated_once:
slice_size //= 2
logger.info(f"Concluded search with {slice_size=:_}.")
break
if slice_size == 1:
raise MemoryError(
f"Even {slice_size=:_} doesn't fit into your memory with these parameters."
) from runtime_error
logger.debug(f"The {slice_size=:_} was too big, trying less now.")
slice_size //= 2
reached_max = True
else:
self._free_graph_and_cache()
if reached_max:
logger.info(f"Concluded search with {slice_size=:_}.")
break
slice_size *= 2
evaluated_once = True
return slice_size
def _check_slicing_availability(self, supports_sub_batching: bool):
if self.target == 0:
return
if self.target == 1:
return
if self.target == 2:
return
if supports_sub_batching:
report = (
"This model supports sub-batching, but it also requires slicing,"
" which is not implemented for this model yet."
)
else:
report = "This model doesn't support sub-batching and slicing is not implemented for this model yet."
logger.warning(report)
raise MemoryError("The current model can't be trained on this hardware with these parameters.")
# note: we use Tuple[Tensor] here, so we can re-use TensorDataset instead of having to create a custom one
[docs]
class SymmetricLCWATrainingLoop(TrainingLoop[tuple[MappedTriples]]):
r"""A "symmetric" LCWA scoring heads *and* tails at once.
This objective was introduced by [lacroix2018]_ as
.. math ::
l_{i,j,k}(X) = - X_{i,j,k} + \log \left(
\sum_{k'} \exp(X_{i,j,k′})
\right) - X_{k,j+P,i} + \log \left(
\sum_{i'} \exp (X_{k, j+P, i'})
\right)
which can be seen as a "symmetric LCWA", where for one batch of triples, we score both, heads *and* tails, given
the remainder of the triple.
.. note ::
at the same time, there is a also a difference to the :class:`LCWATrainingLoop`: we do not group by e.g.,
head+relation pairs. Thus, the name might be suboptimal and change in the future.
"""
# docstr-coverage: inherited
def _create_training_data_loader(
self, triples_factory: CoreTriplesFactory, sampler: str | None, **kwargs
) -> DataLoader[tuple[MappedTriples]]: # noqa: D102
assert sampler is None
return DataLoader(dataset=TensorDataset(triples_factory.mapped_triples), **kwargs)
# docstr-coverage: inherited
def _process_batch(
self,
batch: tuple[MappedTriples],
start: int,
stop: int,
label_smoothing: float = 0,
slice_size: int | None = None,
) -> FloatTensor: # noqa: D102
# unpack
hrt_batch = batch[0]
# Send batch to device
hrt_batch = hrt_batch[start:stop].to(device=self.model.device)
return (
# head prediction
self.loss.process_lcwa_scores(
predictions=self.model.score_h(rt_batch=hrt_batch[:, 1:], slice_size=slice_size, mode=self.mode),
# TODO: exploit sparsity
# note: this is different to what we do for LCWA, where we collect *all* training entities
# for which the combination is true
labels=functional.one_hot(hrt_batch[:, 0], num_classes=self.model.num_entities).float(),
label_smoothing=label_smoothing,
num_entities=self.model.num_entities,
)
# tail prediction
+ self.loss.process_lcwa_scores(
predictions=self.model.score_t(hr_batch=hrt_batch[:, :-1], slice_size=slice_size, mode=self.mode),
# TODO: exploit sparsity
labels=functional.one_hot(hrt_batch[:, 2], num_classes=self.model.num_entities).float(),
label_smoothing=label_smoothing,
num_entities=self.model.num_entities,
)
# regularization
+ self.model.collect_regularization_term()
)
@staticmethod
# docstr-coverage: inherited
def _get_batch_size(batch: tuple[MappedTriples]) -> int: # noqa: D102
assert len(batch) == 1
return batch[0].shape[0]
def _slice_size_search(self, **kwargs) -> int:
# TODO?
raise MemoryError("The current model can't be trained on this hardware with these parameters.")