jax-ml/jax · error · ValueError

attempt to get argmin of an empty sequence

Error message

attempt to get argmin of an empty sequence

What it means

An index of the minimum is undefined when the reduced axis contains no elements. The jitted _argmin helper fires when a.shape[axis] == 0 before delegating to the underlying reduction.

Solutions

1. Check size before calling and handle empty explicitly (return a default index or skip)
2. Ensure upstream filtering guarantees at least one element
3. Pad with +inf sentinel and post-process the sentinel index

Example fix

// before
best = jnp.argmin(losses)  # losses may be empty
// after
best = jnp.argmin(losses) if losses.size else 0

if a.size == 0: return 0  # or skip
return int(jnp.argmin(a))

def nonempty_for_argmin(a, axis=None):
    a = jnp.asarray(a)
    return a.size > 0 if axis is None else a.shape[axis] > 0

try:
    idx = jnp.argmin(a)
except ValueError as e:
    if 'empty sequence' in str(e):
        idx = 0
    else:
        raise

When it happens

Trigger: jnp.argmin(jnp.array([])); jnp.argmin(a, axis=k) with a.shape[k] == 0; empty result after boolean masking or slicing in jit-compiled code.

Common situations: Empty minibatches, all-elements-filtered arrays, boundary conditions in loops producing zero-length slices.

Related errors

• argmin and argmax require non-empty reduced dimension…
• attempt to get argmax of an empty sequence
• Cannot do a non-empty jnp.take() from an empty axis.
• expected non-empty vector for x
• fiedler_companion requires the last axis of 'a' to have…

AI-assisted analysis of jax-ml/jax@1e1c6a8fc0 (2026-08-27). Data as JSON: /api/errors/a049a6198992ff59. Report an issue: GitHub.

Appendix: source

Thrown at jax/_src/numpy/lax_numpy.py:8347

    Array([[0],
           [2]], dtype=int32)
  """
  arr = util.ensure_arraylike("argmin", a)
  if out is not None:
    raise NotImplementedError("The 'out' argument to jnp.argmin is not supported.")
  return _argmin(arr, None if axis is None else operator.index(axis),
                 keepdims=bool(keepdims))

@api.jit(static_argnames=('axis', 'keepdims'), inline=True)
def _argmin(a: Array, axis: int | None = None, keepdims: bool = False) -> Array:
  if axis is None:
    dims = list(range(np.ndim(a)))
    a = ravel(a)
    axis = 0
  else:
    dims = [axis]
  if a.shape[axis] == 0:
    raise ValueError("attempt to get argmin of an empty sequence")
  # TODO(phawkins): use an int64 index if the dimension is large enough.
  result = lax.argmin(a, _canonicalize_axis(axis, a.ndim), int)
  return expand_dims(result, dims) if keepdims else result


@export
def nanargmax(
    a: ArrayLike,
    axis: int | None = None,
    out: None = None,
    keepdims: bool | None = None,
) -> Array:
  """Return the index of the maximum value of an array, ignoring NaNs.

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

  Args:
    a: input array

View on GitHub (pinned to 1e1c6a8fc0)