jax-ml/jax · error · TypeError

Invalid scalar value {x}

Error message

Invalid scalar value {x}

What it means

jax.dtypes.scalar_type_of maps a Python/numpy scalar to int/float/complex/bool; anything else (strings, None, arbitrary objects) is not a scalar and raises TypeError. It is used by lax._const when coercing Python values into typed constants.

Source

Thrown at jax/_src/dtypes.py:436

@export
def scalar_type_of(x: Any) -> type:
  """Return the scalar type associated with a JAX value."""
  typ = dtype(x)
  if typ in _custom_float_dtypes:
    return float
  elif typ in _intn_dtypes:
    return int
  elif np.issubdtype(typ, np.bool_):
    return bool
  elif np.issubdtype(typ, np.integer):
    return int
  elif np.issubdtype(typ, np.floating):
    return float
  elif np.issubdtype(typ, np.complexfloating):
    return complex
  else:
    raise TypeError(f"Invalid scalar value {x}")


def scalar_type_to_dtype(typ: type, value: Any = None) -> DType:
  """Return the numpy dtype for the given scalar type.

  Raises
  ------
  OverflowError: if `typ` is `int` and the value is too large for int64.

  Examples
  --------
  >>> scalar_type_to_dtype(int)
  dtype('int32')
  >>> scalar_type_to_dtype(float)
  dtype('float32')
  >>> scalar_type_to_dtype(complex)
  dtype('complex64')
  >>> scalar_type_to_dtype(int)

View on GitHub (pinned to 1e1c6a8fc0)

Solutions

1. Convert inputs to scalars before passing to JAX ops: int(v)/float(v)
2. Validate with isinstance(x, (bool, int, float, complex)) (plus numpy scalars) at your API boundary
3. Fix data pipeline producing strings/None for numeric fields

Example fix

# before
jnp.float32(config['scale'])  # scale is '2.0'

# after
jnp.float32(float(config['scale']))

assert isinstance(x, (bool, int, float, complex)), f'not a scalar: {x!r}'

def is_python_scalar(x) -> bool:
    return isinstance(x, (bool, int, float, complex)) and not isinstance(x, bool) or isinstance(x, bool)

try:
    fn(x)
except TypeError:
    fn(float(x))  # when a numeric string was intended

When it happens

Trigger: Passing a string, bytes, None, or custom object where a Python scalar is expected, e.g. lax ops building constants from user-supplied values (lax._const -> scalar_type_of).

Common situations: Trace-time constants from config (strings meant to be numbers); None defaults leaking into math; unhashable/non-scalar objects in place of scalars.

Related errors

• iteration over a 0-d array
• Argument to to_dlpack must be a jax.Array, got {type(x)}
• The array passed to from_dlpack must have __dlpack__ and __d
• dtype {dtype!r} not understood
• cond_fun must return a boolean scalar, but got output type(s

AI-assisted analysis of jax-ml/jax@1e1c6a8fc0 (2026-08-27). Data as JSON: /api/errors/cb53b611fee790a7. Report an issue: GitHub.