src/flowfm/training_utils.py

"""Shared training and evaluation utilities."""

import math

import numpy as np

try:
    from src.metric import pearsonr_score
except ImportError:
    from metric import pearsonr_score


def validate_finite_loss(loss_value: float, batch_idx: int) -> None:
    """Fail fast when training diverges."""
    if math.isnan(loss_value) or math.isinf(loss_value):
        raise RuntimeError(f"NaN/Inf loss encountered on step {batch_idx + 1}; exiting")


def compute_subject_metrics(
    samples: np.ndarray,
    outputs: np.ndarray,
    subjects: list[int],
) -> tuple[float, dict[str, np.ndarray | float]]:
    """Compute per-subject and global Pearson metrics."""
    metrics: dict[str, np.ndarray | float] = {}

    dim = samples.shape[-1]
    acc = 0.0
    acc_map = np.zeros(dim)

    for ii, sub in enumerate(subjects):
        y_true = samples[ii].reshape(-1, dim)
        y_pred = outputs[ii].reshape(-1, dim)

        acc_map_i = pearsonr_score(y_true, y_pred)
        acc_i = float(np.mean(acc_map_i))

        metrics[f"accmap_sub-{sub}"] = acc_map_i
        metrics[f"acc_sub-{sub}"] = acc_i

        acc_map += acc_map_i / len(subjects)
        acc += acc_i / len(subjects)

    metrics["accmap_avg"] = acc_map
    metrics["acc_avg"] = acc

    return acc, metrics


def format_subject_accuracies(metrics: dict[str, np.ndarray | float]) -> str:
    """Render a stable, compact per-subject accuracy string for logs."""
    values = [
        f"{val:.3f}"
        for key, val in metrics.items()
        if key.startswith("acc_sub-") and isinstance(val, (float, np.floating))
    ]
    return ",".join(values)