hezar.metrics.seqeval module

class hezar.metrics.seqeval.Seqeval(config: SeqevalConfig, **kwargs)

Bases: Metric

Seqeval metric for sequence labeling tasks using seqeval.

Parameters:

• config (SeqevalConfig) – Metric configuration object.

• **kwargs – Extra configuration parameters passed as kwargs to update the config.

compute(predictions=None, targets=None, suffix: bool = None, mode: str | None = None, sample_weight: List[int] | None = None, zero_division: str | int = None, n_decimals: int = None, output_keys=None, **kwargs)

Computes the Seqeval scores for the given predictions against targets.

Parameters:

• predictions – Predicted labels.

• targets – Ground truth labels.

• suffix (bool) – Flag to indicate whether the labels have suffixes.

• mode (Optional[str]) – Evaluation mode for seqeval.

• sample_weight (Optional[List[int]]) – Sample weights for the seqeval metrics.

• zero_division (str | int) – Strategy for zero-division, default is 0.

• n_decimals (int) – Number of decimals for the final score.

• output_keys (tuple) – Filter the output keys.

Returns:

A dictionary of the metric results, with keys specified by output_keys.

Return type:

dict

required_backends: List[str | Backends] = [Backends.SEQEVAL]

class hezar.metrics.seqeval.SeqevalConfig(objective: str = 'maximize', output_keys: tuple = ('accuracy', 'recall', 'precision', 'f1'), n_decimals: int = 4, suffix: bool = False, mode: str | None = None, sample_weight: List[int] | None = None, zero_division: str | int = 0)

Bases: MetricConfig

Configuration class for Seqeval metric.

Parameters:

• name (MetricType) – The type of metric, Seqeval in this case.

• output_keys (tuple) – Keys to filter the metric results for output.

• suffix (bool) – Flag to indicate whether the labels have suffixes.

• mode (Optional[str]) – Evaluation mode for seqeval.

• sample_weight (Optional[List[int]]) – Sample weights for the seqeval metrics.

• zero_division (str | int) – Strategy for zero-division, default is 0.

mode: str | None = None

name: str = 'seqeval'

objective: str = 'maximize'

output_keys: tuple = ('accuracy', 'recall', 'precision', 'f1')

sample_weight: List[int] | None = None

suffix: bool = False

zero_division: str | int = 0