Source code for nlp_shap.runtime.metrics
"""Performance summaries for explain pipeline runs."""
from dataclasses import dataclass
[docs]
@dataclass(frozen=True, slots=True)
class PerfSummary:
"""Wall-clock breakdown for one explain or reanalyze run."""
total_ms: float
"""Total wall-clock time for the run."""
generation_ms: float
"""Time spent in base and coalition generation."""
scoring_ms: float
"""Time spent scoring coalition outputs with the value function."""
aggregation_ms: float
"""Time spent aggregating coalition payoffs into attributions."""