Metric

minnt.Metric

Bases: Module, ABC

An abstract metric interface.

update abstractmethod

update(
    y: TensorOrTensors,
    y_true: TensorOrTensors | None = None,
    sample_weights: TensorOrTensors | None = None,
) -> Self

Update the internal state of the metric with new predictions and possibly gold targets.

Optional sample weights might be provided if supported by the metric.

Parameters:

• y (TensorOrTensors) –

  The predicted outputs.

• y_true (TensorOrTensors | None, default: None ) –

  Optional ground-truth targets.

• sample_weights (TensorOrTensors | None, default: None ) –

  Optional sample weights.

Returns:

• Self –

  self

compute abstractmethod

compute() -> Tensor

Compute the accumulated metric value.

Returns:

• Tensor –

  A (usually scalar) tensor representing the accumulated metric value.

reset abstractmethod

reset() -> Self

Reset the internal state of the metric.

Returns:

• Self –

  self