jax-ml/jax · error · TypeError

Cannot make an array with dtype {dtype} from an object with

Error message

Cannot make an array with dtype {dtype} from an object with dtype {object.dtype}.

What it means

When creating a string array via _make_string_array, JAX requires the requested dtype and the source NumPy array's dtype to agree in string-ness: you cannot request string dtype from a numeric array, or a numeric dtype from a string array.

Source

Thrown at jax/_src/numpy/array_constructors.py:93

  else:
    return True


def _make_string_array(
    object: np.ndarray,
    dtype: DTypeLike | None = None,
    ndmin: int = 0,
    device: xc.Device | Sharding | None = None,
) -> Array:
  if not isinstance(object, np.ndarray):
    raise TypeError(
        "Currently, string arrays can only be made from NumPy"
        f" arrays. Got:  {type(object)}."
    )
  if dtype is not None and (
      (object.dtype == dtypes.string_dtype) != (dtype == dtypes.string_dtype)
  ):
    raise TypeError(
        f"Cannot make an array with dtype {dtype} from an object with dtype"
        f" {object.dtype}."
    )
  if ndmin > object.ndim:
    raise TypeError(
        f"ndmin {ndmin} cannot be greater than object's ndims"
        f" {object.ndim} for string arrays."
    )

  # Just do a device_put since XLA does not support string as a data type.
  return api.device_put(x=object, device=device)


@export
def array(object: Any, dtype: DTypeLike | None = None, *args, copy: bool = True,
          order: str | None = "K", ndmin: int = 0,
          device: xc.Device | Sharding | None = None,
          out_sharding: NamedSharding | P | None = None) -> Array:

View on GitHub (pinned to 1e1c6a8fc0)

Solutions

1. Match dtypes: drop the dtype argument and let it be inferred, or convert in NumPy first (np.asarray(...).astype(...))
2. For numeric output from string input, parse/convert via NumPy before calling jnp.array

Example fix

# before
import numpy as np, jax.numpy as jnp
a = jnp.array(np.array(['1', '2']), dtype=jnp.float32)
# after
import numpy as np, jax.numpy as jnp
a = jnp.array(np.array(['1', '2']).astype(np.float32))

import jax.numpy as jnp, numpy as np
def string_or_none_dtype(np_arr, dtype):
    if dtype is None:
        return None
    same = (np_arr.dtype == object or np_arr.dtype.kind in 'US') == (dtype == jnp.string_dtype)
    return dtype if same else None  # None lets inference work

null

null

When it happens

Trigger: jnp.array(np.array(['a','b']), dtype=jnp.float32) or jnp.array(np.array([1,2]), dtype=jax.numpy.string_dtype) — a mismatch between object's string/numeric dtype and the requested dtype.

Common situations: Generic loading code that always passes dtype=float32, hitting a column of strings; or explicitly tagging numeric data with string_dtype by mistake.

Related errors

• Incorrect output dtype for return value #{i}: Expected: {out
• Value '{x}' with dtype {dt} is not a valid JAX array type. O
• {} function carry input and carry output must have equal typ
• lower and upper arguments to fori_loop must have equal types
• the first argument to pow must have an inexact dtype (float

AI-assisted analysis of jax-ml/jax@1e1c6a8fc0 (2026-08-27). Data as JSON: /api/errors/42efc08aeb4dcfec. Report an issue: GitHub.