mlflow log_metric is designed for simple unidimensional metrics. Perhaps you want to look at the metric across different lead times. You can use the step parameter for this, which is normally used for looking how a metric varies over training steps or epochs.
def log_array_metric(metric_name: str, metric_array: np.ndarray, start_step=0) -> None:
for step, value in enumerate(metric_array, start=start_step):
mlflow.log_metric(metric_name, value, step=step)