spine.io.sample

Used to define which dataset entries to load at each iteration

Classes

AbstractBatchSampler(dataset, batch_size[, ...])	Abstract sampler class.
AbstractJointBatchSampler(dataset, batch_size)	Pair primary sampler indices with independently sampled secondary indices.
BootstrapBatchSampler(dataset, batch_size[, ...])	Sampler used for bootstrap sampling of the entire dataset.
DistributedProxySampler(sampler, ...)	Sampler that restricts data loading to a subset of input sampler indices.
DistributedSampler(*args, **kwargs)	Import-safe stand-in used when PyTorch is unavailable.
JointBootstrapBatchSampler(dataset, batch_size)	Bootstrap primary sampling with paired bootstrap secondary sampling.
JointRandomSequenceBatchSampler(dataset, ...)	Random-sequence primary sampling with paired random secondary sampling.
JointSequentialBatchSampler(dataset, batch_size)	Sequential primary sampling with paired sequential secondary sampling.
RandomSequenceBatchSampler(dataset, batch_size)	Samples sequential batches randomly within the dataset.
Sampler(*args, **kwargs)	Import-safe stand-in used when PyTorch is unavailable.
SequentialBatchSampler(dataset, batch_size)	Samples batches sequentially within the dataset.

class spine.io.sample.SequentialBatchSampler(dataset: Sized, batch_size: int, seed: int | None = None, drop_last: bool = True)

Samples batches sequentially within the dataset.

name = 'sequential'

class spine.io.sample.RandomSequenceBatchSampler(dataset: Sized, batch_size: int, seed: int | None = None, drop_last: bool = True)

Samples sequential batches randomly within the dataset.

name = 'random_sequence'

class spine.io.sample.BootstrapBatchSampler(dataset: Sized, batch_size: int, seed: int | None = None, drop_last: bool = True)

Sampler used for bootstrap sampling of the entire dataset.

This is particularly useful for training an ensemble of networks (bagging).

name = 'bootstrap'

class spine.io.sample.JointSequentialBatchSampler(dataset: Sized, batch_size: int, seed: int | None = None, drop_last: bool = True, pair_probability: float = 1.0)

Sequential primary sampling with paired sequential secondary sampling.

Methods

sampler_cls

name = 'joint_sequential'

sampler_cls

alias of SequentialBatchSampler

class spine.io.sample.JointRandomSequenceBatchSampler(dataset: Sized, batch_size: int, seed: int | None = None, drop_last: bool = True, pair_probability: float = 1.0)

Random-sequence primary sampling with paired random secondary sampling.

Methods

sampler_cls

name = 'joint_random_sequence'

sampler_cls

alias of RandomSequenceBatchSampler

class spine.io.sample.JointBootstrapBatchSampler(dataset: Sized, batch_size: int, seed: int | None = None, drop_last: bool = True, pair_probability: float = 1.0)

Bootstrap primary sampling with paired bootstrap secondary sampling.

Methods

sampler_cls

name = 'joint_bootstrap'

sampler_cls

alias of BootstrapBatchSampler