sequince_density.utils

Utility helpers shared by density implementations.

Functions

validate_training_data(samples, dims[, stage])	Validate training particles using their native array backend.
draw_seed(rng)	Consume one deterministic integer seed from an optional RNG.
torch_random_context([rng, device])	Temporarily seed Torch and restore its global RNG state on exit.

Module Contents

sequince_density.utils.validate_training_data(samples, dims, stage='Raw')

Validate training particles using their native array backend.

Parameters:

• samples (array, shape (n_samples, dims)) – Nonempty, finite particle coordinates. Validation does not modify them.

• dims (int) – Expected number of dimensions.

• stage (str, default="Raw") – Label for the data in error messages, for example “Transformed”.

Raises:

ValueError – If the shape is invalid or any coordinates are non-finite. Non-finite errors report the affected dimension indices (zero-based).

Return type:

None

sequince_density.utils.draw_seed(rng)

Consume one deterministic integer seed from an optional RNG.

Parameters:

rng (orng.RandomGenerator | int | None)

Return type:

int | None

sequince_density.utils.torch_random_context(rng=None, *, device=None)

Temporarily seed Torch and restore its global RNG state on exit.

Parameters:

• rng (orng.RandomGenerator | int | None)

• device (Any)

Return type:

collections.abc.Generator[None, None, None]