def __init__(self,
cuts: CutSet,
max_frames: int = None,
max_samples: int = None,
max_duration: Seconds = None,
max_cuts: Optional[int] = None,
**kwargs):
"""
SingleCutSampler's constructor.
:param cuts: the ``CutSet`` to sample data from.
:param max_frames: The maximum total number of feature frames from ``cuts``.
:param max_samples: The maximum total number of audio samples from ``cuts``.
:param max_duration: The maximum total recording duration from ``cuts``.
:param max_cuts: The maximum number of cuts sampled to form a mini-batch.
By default, this constraint is off.
:param kwargs: Arguments to be passed into ``CutSampler``.
"""
super().__init__(cuts.ids, provide_len=not cuts.is_lazy, **kwargs)
self.cuts = cuts
self.time_constraint = TimeConstraint(max_duration=max_duration,
max_frames=max_frames,
max_samples=max_samples)
self.max_cuts = max_cuts
assert self.time_constraint.is_active() or \
not (self.time_constraint.is_active() and self.max_cuts is not None)
# Constraints
assert is_none_or_gt(self.max_cuts, 0)
def __init__(
self,
cuts: CutSet,
max_frames: int = None,
max_samples: int = None,
max_duration: Seconds = None,
max_cuts: Optional[int] = None,
shuffle: bool = False,
drop_last: bool = False,
world_size: Optional[int] = None,
rank: Optional[int] = None,
seed: int = 0,
):
"""
SingleCutSampler's constructor.
:param cuts: the ``CutSet`` to sample data from.
:param max_frames: The maximum total number of feature frames from ``cuts``.
:param max_samples: The maximum total number of audio samples from ``cuts``.
:param max_duration: The maximum total recording duration from ``cuts``.
:param max_cuts: The maximum number of cuts sampled to form a mini-batch.
By default, this constraint is off.
:param shuffle: When ``True``, the cuts will be shuffled at the start of iteration.
Convenient when mini-batch loop is inside an outer epoch-level loop, e.g.:
`for epoch in range(10): for batch in dataset: ...` as every epoch will see a
different cuts order.
:param drop_last: When ``True``, the last batch is dropped if it's incomplete.
:param world_size: Total number of distributed nodes. We will try to infer it by default.
:param rank: Index of distributed node. We will try to infer it by default.
:param seed: Random seed used to consistently shuffle the dataset across different processes.
"""
super().__init__(
shuffle=shuffle,
world_size=world_size,
rank=rank,
seed=seed,
)
self.data_source = DataSource(cuts)
self.time_constraint = TimeConstraint(max_duration=max_duration,
max_frames=max_frames,
max_samples=max_samples)
self.drop_last = drop_last
self.max_cuts = max_cuts
assert self.time_constraint.is_active() or not (
self.time_constraint.is_active() and self.max_cuts is not None)
# Constraints
assert is_none_or_gt(self.max_cuts, 0)
def __post_init__(self) -> None:
assert exactly_one_not_null(*self._constraints) or all(
x is None for x in self._constraints)
for c in self._constraints:
assert is_none_or_gt(c, 0)