jax-ml/jax · error · ValueError

Arguments to jax.numpy.gcd must be integers.

Error message

Arguments to jax.numpy.gcd must be integers.

What it means

jax.numpy.gcd computes the greatest common divisor elementwise and only accepts integer-dtyped inputs. After converting arguments to arrays and promoting dtypes, it checks that the common dtype is a subclass of np.integer and raises ValueError otherwise. This mirrors NumPy's requirement that gcd operands be integers.

Source

Thrown at jax/_src/numpy/lax_numpy.py:8861

    Array inputs:

    >>> x1 = jnp.array([12, 18, 24])
    >>> x2 = jnp.array([5, 10, 15])
    >>> jnp.gcd(x1, x2)
    Array([1, 2, 3], dtype=int32)

    Broadcasting:

    >>> x1 = jnp.array([12])
    >>> x2 = jnp.array([6, 9, 12])
    >>> jnp.gcd(x1, x2)
    Array([ 6,  3, 12], dtype=int32)
  """
  x1, x2 = util.ensure_arraylike("gcd", x1, x2)
  x1, x2 = util.promote_dtypes(x1, x2)
  if not issubdtype(x1.dtype, np.integer):
    raise ValueError("Arguments to jax.numpy.gcd must be integers.")
  x1, x2 = broadcast_arrays(x1, x2)
  gcd, _ = control_flow.while_loop(_gcd_cond_fn, _gcd_body_fn, (ufuncs.abs(x1), ufuncs.abs(x2)))
  return gcd


@export
@api.jit
def lcm(x1: ArrayLike, x2: ArrayLike) -> Array:
  """Compute the least common multiple of two arrays.

  JAX implementation of :func:`numpy.lcm`.

  Args:
    x1: First input array. The elements must have integer dtype.
    x2: Second input array. The elements must have integer dtype.

  Returns:
    An array containing the least common multiple of the corresponding

View on GitHub (pinned to 1e1c6a8fc0)

Solutions

1. Cast inputs to an integer dtype before calling: jnp.gcd(x1.astype(jnp.int32), x2.astype(jnp.int32))
2. Ensure upstream computations don't produce floats (e.g. replace / with // where integer results are expected)
3. Check dtypes beforehand with jnp.issubdtype(x.dtype, jnp.integer)

Example fix

// before
jnp.gcd(6.0, 4.0)  # ValueError
// after
jnp.gcd(jnp.asarray(6.0, dtype=jnp.int32), jnp.asarray(4.0, dtype=jnp.int32))

def as_int_pair(x1, x2):
    x1, x2 = jnp.asarray(x1), jnp.asarray(x2)
    if not jnp.issubdtype(jnp.result_type(x1, x2), jnp.integer):
        x1, x2 = x1.astype(jnp.int32), x2.astype(jnp.int32)
    return x1, x2
x1, x2 = as_int_pair(x1, x2)
jnp.gcd(x1, x2)

def is_integer_array(x) -> bool:
    return jnp.issubdtype(jnp.asarray(x).dtype, jnp.integer)

When it happens

Trigger: Calling jnp.gcd(x1, x2) where either argument is float/complex/bool-promoted-to-float, e.g. jnp.gcd(6.0, 4.0) or arrays with dtype float32. Promotion of mixed int/float inputs yields a float dtype, failing the check.

Common situations: Passing Python floats or float arrays computed from division/mean operations; loading data with np.loadtxt (float64 default) then calling gcd; mixing an int array with a Python float scalar.

Related errors

• Arguments to jax.numpy.lcm must be integers.
• dtype argument to jnp.std must be inexact; got {dtype}
• len() of unsized object
• numpy masked arrays are not supported as direct inputs to JA
• Unsupported scalar attribute type: {type(val)}

AI-assisted analysis of jax-ml/jax@1e1c6a8fc0 (2026-08-27). Data as JSON: /api/errors/6a94fc2631d4978b. Report an issue: GitHub.