dataeval.scope.Prioritize.Config¶

class dataeval.scope.Prioritize.Config¶

Configuration for Prioritize evaluator.

extractor¶

Feature extractor used to compute embeddings from a dataset. Optional only when pre-computed embeddings are passed to evaluate().

Type:¶

FeatureExtractor or None

batch_size¶

Batch size for embedding computation. When None, uses the global batch size from get_batch_size().

Type:¶

int or None, default None

method¶

Ranking method to use.

Type:¶

{“knn”, “kmeans_distance”, “kmeans_complexity”, “hdbscan_distance”, “hdbscan_complexity”}, default “knn”

k¶

Number of nearest neighbors for “knn” method.

Type:¶

int or None, default None

c¶

Number of clusters for clustering methods.

Type:¶

int or None, default None

n_init¶

Number of K-means initializations (kmeans methods only).

Type:¶

int or “auto”, default “auto”

max_cluster_size¶

Maximum cluster size for HDBSCAN methods.

Type:¶

int or None, default None

order¶

Sort direction for output indices.

Type:¶

{“easy_first”, “hard_first”}, default “easy_first”

policy¶

Selection policy to apply after ranking.

Type:¶

{“difficulty”, “stratified”, “class_balanced”}, default “difficulty”

num_bins¶

Number of bins for “stratified” policy.

Type:¶

int, default 50