jax-ml/jax · error · TypeError

Cannot interpret ' ' as a data type.\n\nDid you…

Error message

Cannot interpret '{dtype}' as a data type.\n\nDid you accidentally write `jax.numpy.ones({shape}, {dtype})` when you meant `jax.numpy.ones(({shape}, {dtype}))`, i.e. with a single tuple argument for the shape?

What it means

jnp.ones (via the shared shape/dtype checking helper) received a dtype argument that cannot be interpreted as a dtype — commonly because shape and dtype were passed as two positional arguments where a single tuple was intended. The hint shows the likely parenthesization fix.

Solutions

Pass shape and dtype as separate arguments: jnp.ones((2, 3), jnp.float32)

Example fix

# before
a = jnp.ones((2, 3, jnp.float32))
# after
a = jnp.ones((2, 3), jnp.float32)

Defensive patterns

Strategy: validation

Validate before calling

assert all(isinstance(d, int) for d in shape), 'dtype accidentally inside shape tuple?'

Type guard

null

Try / catch

null

Prevention

Separate dtype from shape arguments
Use a linter rule flagging dtype objects inside shape tuples

When it happens

Trigger: jnp.ones((2, 3, dtype)) — one tuple argument mixing shape dims with a dtype.

Common situations: Editing or templating shape/dtype on one line and misplacing the parenthesis; porting torch-style size/dtype calls.

Related errors

AI-assisted analysis of jax-ml/jax@1e1c6a8fc0 (2026-08-27). Data as JSON: /api/errors/2d1e60fe97195d8f. Report an issue: GitHub.

Appendix: source

Thrown at jax/_src/numpy/array_creation.py:138

  See also:
    - :func:`jax.numpy.ones_like`
    - :func:`jax.numpy.empty`
    - :func:`jax.numpy.zeros`
    - :func:`jax.numpy.full`

  Examples:
    >>> jnp.ones(4)
    Array([1., 1., 1., 1.], dtype=float32)
    >>> jnp.ones((2, 3), dtype=bool)
    Array([[ True,  True,  True],
           [ True,  True,  True]], dtype=bool)

  .. _explicit sharding: https://docs.jax.dev/en/latest/parallel.html
  """
  if isinstance(shape, types.GeneratorType):
    raise TypeError("expected sequence object with len >= 0 or a single integer")
  if (m := _check_forgot_shape_tuple("ones", shape, dtype)): raise TypeError(m)
  shape = canonicalize_shape(shape)
  dtype = dtypes.check_and_canonicalize_user_dtype(
      float if dtype is None else dtype, "ones")
  sharding = util.choose_device_or_out_sharding(
      device, out_sharding, 'jnp.ones')
  return lax.full(shape, 1, dtype, sharding=sharding)


@export
def empty(shape: Any, dtype: DTypeLike | None = None, *,
          device: xc.Device | Sharding | None = None,
          out_sharding: NamedSharding | P | None = None) -> Array:
  """Create an empty array.

  JAX implementation of :func:`numpy.empty`. Starting in JAX v0.11.0, this
  returns an uninitialized array on platforms that support doing so. Prior to
  v0.11.0, this function returned an array filled with zeros on all platforms.

View on GitHub (pinned to 1e1c6a8fc0)