blazefl.utils.RandomState

class blazefl.utils.RandomState(random: tuple, environ: str, numpy: dict, torch_seed: int, torch_rng_state: Tensor, cuda: CUDARandomState | None)

Bases: object

A dataclass representing the random state for Python, NumPy, and PyTorch.

random

The state of Python’s random module.

Type:

tuple

environ

The PYTHONHASHSEED environment variable.

Type:

str

numpy

The state of NumPy’s RNG.

Type:

dict

torch_seed

The initial seed for PyTorch.

Type:

int

torch_rng_state

The RNG state for PyTorch.

Type:

torch.Tensor

cuda

The CUDA-specific random state, if available.

Type:

CUDARandomState | None

__init__(random: tuple, environ: str, numpy: dict, torch_seed: int, torch_rng_state: Tensor, cuda: CUDARandomState | None) → None

Methods

__init__(random, environ, numpy, torch_seed, ...)
get_random_state(device)	Capture the current random state.
set_random_state(random_state)	Restore the random state from a RandomState object.

Attributes

random
environ
numpy
torch_seed
torch_rng_state
cuda

cuda: CUDARandomState | None

environ: str

classmethod get_random_state(device: str) → RandomState

Capture the current random state.

Parameters:

device (str) – The device type (‘cpu’ or ‘cuda’).

Returns:

The captured random state.

Return type:

RandomState

numpy: dict

random: tuple

static set_random_state(random_state: RandomState) → None

Restore the random state from a RandomState object.

Parameters:

random_state (RandomState) – The random state to restore.

Returns:

None

torch_rng_state: Tensor

torch_seed: int