beanmachine.ppl.inference.utils module

class beanmachine.ppl.inference.utils.VerboseLevel(value)

Bases: enum.Enum

Enum class which is used to set how much output is printed during inference. LOAD_BAR enables tqdm for full inference loop.

LOAD_BAR = 1

OFF = 0

beanmachine.ppl.inference.utils.detach_samples(samples: Dict[beanmachine.ppl.model.rv_identifier.RVIdentifier, torch.Tensor]) → Dict[beanmachine.ppl.model.rv_identifier.RVIdentifier, numpy.ndarray]

Detach pytorch tensors.

Parameters

samples (Dict[RVIdentifier, torch.Tensor]) – Dictionary of RVIdentifiers with original torch tensors.

Returns

Dictionary of RVIdentifiers with converted

NumPy arrays.

Return type

Dict[RVIdentifier, np.ndarray]

beanmachine.ppl.inference.utils.merge_dicts(dicts: List[Dict[beanmachine.ppl.model.rv_identifier.RVIdentifier, torch.Tensor]], dim: int = 0, stack_not_cat: bool = True) → Dict[beanmachine.ppl.model.rv_identifier.RVIdentifier, torch.Tensor]

A helper function that merge multiple dicts of samples into a single dictionary, stacking across a new dimension

beanmachine.ppl.inference.utils.safe_log_prob_sum(distrib, value: torch.Tensor) → torch.Tensor

Computes log_prob, converting out of support exceptions to -Infinity.

beanmachine.ppl.inference.utils.seed(seed: int) → None