aggregation.py

import math
from lm_eval.api.register import register_aggregation


def weighted_mean(items):
    a, b = zip(*items)
    return sum(a) / sum(b)


@register_aggregation("mean")
def mean(arr):
    return sum(arr) / len(arr)


@register_aggregation("median")
def median(arr):
    return arr[len(arr) // 2]


@register_aggregation("perplexity")
def perplexity(items):
    return math.exp(-mean(items))


@register_aggregation("weighted_perplexity")
def weighted_perplexity(items):
    return math.exp(-weighted_mean(items))


@register_aggregation("bits_per_byte")
def bits_per_byte(items):
    return -weighted_mean(items) / math.log(2)