jax-ml/jax · error · ValueError

y has more than 2 dimensions

Error message

y has more than 2 dimensions

What it means

jnp.cov computes covariance and, when an optional second variable array y is supplied, promotes m and y to inexact dtypes and requires y to have at most 2 dimensions. A y with ndim > 2 raises ValueError('y has more than 2 dimensions'), matching NumPy's error.

Source

Thrown at jax/_src/numpy/lax_numpy.py:9180

    Array([[ 1., -1.],
           [-1.,  1.]], dtype=float32)

    In general, the entries of the covariance matrix may be any positive
    or negative real value. For example, here is the covariance of 100
    points drawn from a 3-dimensional standard normal distribution:

    >>> key = jax.random.key(0)
    >>> x = jax.random.normal(key, shape=(3, 100))
    >>> with jnp.printoptions(precision=2):
    ...   print(jnp.cov(x))
    [[0.9  0.03 0.1 ]
     [0.03 1.   0.01]
     [0.1  0.01 0.85]]
  """
  if y is not None:
    m, y = util.promote_args_inexact("cov", m, y)
    if y.ndim > 2:
      raise ValueError("y has more than 2 dimensions")
  else:
    m, = util.promote_args_inexact("cov", m)

  if m.ndim > 2:
    raise ValueError("m has more than 2 dimensions")  # same as numpy error

  if dtype is not None and not dtypes.issubdtype(dtype, np.inexact):
    raise ValueError(f"cov: dtype must be a subclass of float or complex; got {dtype=}")

  X = atleast_2d(m)
  if not rowvar and m.ndim != 1:
    X = X.T
  if X.shape[0] == 0:
    return array([]).reshape(0, 0)

  if y is not None:
    y_arr = atleast_2d(y)
    if not rowvar and y_arr.shape[0] != 1:

View on GitHub (pinned to 1e1c6a8fc0)

Solutions

1. Reshape y to at most 2D before calling: y.reshape(-1, y.shape[-1])
2. If y is redundant, pass y=None
3. Compute covariance per-slice with vmap over the extra dimensions

Example fix

// before
jnp.cov(m, y_3d)  # ValueError
// after
jnp.cov(m, y_3d.reshape(-1, y_3d.shape[-1]))

if y is not None:
    y = jnp.asarray(y)
    assert y.ndim <= 2, 'y must be 1D or 2D'
jnp.cov(m, y)

def cov_ready_y(y) -> bool:
    return y is None or jnp.asarray(y).ndim <= 2

When it happens

Trigger: jnp.cov(m, y) where y is 3D or higher, e.g. an image batch of shape (N, H, W) passed as y.

Common situations: Feeding stacked/multidimensional tensors (e.g. batches of images or time-series windows) as the y argument without reshaping to 1D/2D.

Related errors

• m has more than 2 dimensions
• cannot handle multidimensional fweights
• cannot handle multidimensional aweights
• covariance does not have dimension {self.d}
• scan got `length` argument of {} which disagrees with leadin

AI-assisted analysis of jax-ml/jax@1e1c6a8fc0 (2026-08-27). Data as JSON: /api/errors/ef8fa628f263c794. Report an issue: GitHub.