Process-wide random seed management for reproducibility.
Provides a small global registry for a single "current" seed and applies it to Python's
random module and random in one call. Code that needs a source of randomness (train/test splitting, weight initialization, bootstrap resampling, ...) can call get_global_seed to discover whatever the caller last set with set_global_seed, instead of every call site inventing its own seeding convention.Notes
The seed is stored in a single module-level variable, so it is shared process-wide and is not thread-safe: concurrent calls to
set_global_seed from different threads can race.Functions
set_global_seed(seed: int | None) -> None
Set the process-wide random seed for random and numpy.random.
Parameters
seed
int or None
The integer seed to apply to Python's
random module and random. If None, the recorded global seed is cleared without reseeding either generator.
python
>>> from tuiml.utils.seed import set_global_seed, get_global_seed
>>> set_global_seed(42)
>>> get_global_seed()
42get_global_seed() -> int | None
Return the seed most recently set with set_global_seed.
Returns
seed
int or None
The seed currently in effect, or
None if it was never set (or was last cleared by passing None to set_global_seed).
python
>>> from tuiml.utils.seed import set_global_seed, get_global_seed
>>> set_global_seed(7)
>>> get_global_seed()
7
>>> set_global_seed(None)
>>> get_global_seed() is None
True