jax-ml/jax · error · ValueError

varying and reduced cannot have common mesh axes. Got varyin

Error message

varying and reduced cannot have common mesh axes. Got varying={varying} and reduced={reduced}

What it means

_check_mat rejects a ManualAxisType where the same mesh axis is both 'varying' and 'reduced'. An axis is either replicated-varying or already reduced (result of a collective); it cannot be declared both.

Source

Thrown at jax/_src/core.py:2349

    reduced = frozenset(r for r in mat.reduced if mesh.shape[r] != 1)
    u_kind = mat.unreduced_kind if unreduced else None
    return mat.update(varying=varying, unreduced=unreduced, reduced=reduced,
                      unreduced_kind=u_kind)
  return mat


def get_memory_space(memory_space):
  assert memory_space is not None
  return memory_space


def _check_mat(varying, unreduced, reduced, unreduced_kind):
  if varying & unreduced:
    raise ValueError(
        "varying and unreduced cannot have common mesh axes. Got"
        f" varying={varying} and unreduced={unreduced}")
  if varying & reduced:
    raise ValueError(
        "varying and reduced cannot have common mesh axes. Got"
        f" varying={varying} and reduced={reduced}")
  assert not (varying & unreduced & reduced)

  if unreduced_kind is not None and not isinstance(unreduced_kind, UnreducedKind):
    raise TypeError(
        "Expected unreduced_kind to be of type `jax.sharding.UnreducedKind`"
        f" but got {type(unreduced_kind)}")
  if not unreduced and unreduced_kind is not None:
    raise ValueError(
        "`unreduced_kind` should be `None` when `unreduced` is an empty set."
        f" Got {unreduced_kind=} and {unreduced=}")

def _canonicalize_mat(name, val):
  if not isinstance(val, frozenset):
    if not isinstance(val, set):
      raise TypeError(
          f"{name} argument of ManualAxisType should "

View on GitHub (pinned to 1e1c6a8fc0)

Solutions

1. Remove the axis from 'reduced' if subsequent ops treat it as varying, or from 'varying' if the collective output is replicated
2. Derive mats automatically (JAX computes them for standard collectives) rather than setting them by hand

Example fix

// before
mat = ManualAxisType(varying={'x'}, reduced={'x'})

// after
mat = ManualAxisType(varying={'x'})

if varying & reduced:
    raise ValueError('axis in both varying and reduced')

def disjoint_mat(varying, unreduced, reduced): return not (varying & unreduced) and not (varying & reduced)

When it happens

Trigger: ManualAxisType(varying={'x'}, reduced={'x'}); typically from internal code or user-built avals after collectives like psum where the axis was already marked reduced.

Common situations: Composing collectives (psum then operations that assume varying) with manual mat bookkeeping; custom spmd primitives copying mat sets incorrectly.

Related errors

• Axes mentioned in `manual_axis_type` field of ShapedArray sh
• varying and unreduced cannot have common mesh axes. Got vary
• `unreduced_kind` should be `None` when `unreduced` is an emp
• Factor {factor} is not used in the rule, but size is provide
• {kind} contains duplicated factors

AI-assisted analysis of jax-ml/jax@1e1c6a8fc0 (2026-08-27). Data as JSON: /api/errors/a327e04225bb9fa8. Report an issue: GitHub.