Python "OverflowError" in CI

Python integers are unbounded, but floats and C-backed math functions are not. OverflowError is raised when a float result exceeds the largest representable double, commonly from math.exp, **, or float(huge_int).

What this error means

A numeric test fails with "OverflowError: math range error" or "OverflowError: (34, 'Result too large')" on an exponential or power expression.

python

    p = math.exp(x)
OverflowError: math range error

Common causes

An exponential or power on a large input

math.exp(x) or base ** big exceeds the float range when x or the exponent is large.

Converting a very large int to float

float(huge_int) overflows when the integer is larger than the maximum double.

How to fix it

Bound the input or work in log space

Clamp or validate inputs so they stay within a representable range.
For probability math, compute in log space (logsumexp) instead of raw exp.
Catch OverflowError and substitute a saturating value where appropriate.

Python

import math

def safe_exp(x: float) -> float:
    if x > 709:  # exp(710) overflows float64
        return math.inf
    return math.exp(x)

How to prevent it

Validate the numeric range of inputs before exp/pow.
Use log-space math for products of many small probabilities.
Test boundary inputs explicitly.