quantile

QuantileMetric

Bases: CatMetric

Computes quantiles over a set of observations.

__init__

__init__(q: Tensor | list[Tensor] | tuple[Tensor, ...], interpolation: Literal['linear', 'lower', 'higher', 'nearest', 'midpoint'] | str = 'linear', nan_strategy: str | float = 'warn', **kwargs: Any) -> None

Construct a QuantileMetric.

Parameters:

Name	Type	Description	Default
q	Tensor | list[Tensor] | tuple[Tensor, ...]	The quantiles to compute. Values should be in the range [0, 1]. Should be a tensor, or a list or tuple of hstack-able tensors. The result is flattened and quantile values are computed for each element of q.	required
interpolation	Literal['linear', 'lower', 'higher', 'nearest', 'midpoint'] | str	One of: * 'linear': a + (b - a) * fraction, where fraction is the fractional part of the computed quantile index. * 'lower': a. * 'higher': b. * 'nearest': a or b, whichever's index is closer to the computed quantile index (rounding down for .5 fractions). * 'midpoint': (a + b) / 2.	'linear'
nan_strategy	str | float	One of: * 'error': if any nan values are encountered will give a RuntimeError * 'warn': if any nan values are encountered will give a warning and continue * 'ignore': all nan values are silently removed * a float: if a float is provided will impute any nan values with this value	'warn'
**kwargs	Any	Additional arguments to pass to the base metric class.	required

compute

compute() -> dict[float, torch.Tensor]

Compute the configured quantiles over the observations.

Returns:

Type	Description
dict[float, torch.Tensor]	A mapping of quantiles to their corresponding value.