random_mask

bayesflow.utils.random_mask(shape: tuple[int, ...], drop_prob: float, seed_generator: SeedGenerator = None) → Tensor | float

Generate an element-wise random mask.

Each element is independently drawn as 1 (keep) with probability 1 - drop_prob and 0 (drop) with probability drop_prob.

Parameters:

shapeShape

Shape of the mask to generate.

drop_probfloat

Probability of dropping each element. Must be in [0, 1].

seed_generatorkeras.random.SeedGenerator, optional

Seed generator used for randomness.

Returns:

Tensor or float

A mask tensor of the given shape, or 1.0 when drop_prob <= 0.