pandas-dev/pandas · error · TypeError

FloatingArray does not support np.float16 dtype.

Error message

FloatingArray does not support np.float16 dtype.

What it means

Raised by NumericArray.__init__ specifically when values.dtype == np.float16. Although float16 is a floating dtype and passes the _checker, FloatingArray maps numpy float types to Float32/Float64 only; the dtype-mapping lookup in NumericArray.dtype would otherwise fail later, so pandas fails fast here with an explicit, actionable message. float16 has limited precision and no Float16Dtype.

Source

Thrown at pandas/core/arrays/numeric.py:275

    _dtype_cls: type[NumericDtype]

    def __init__(
        self, values: np.ndarray, mask: npt.NDArray[np.bool_], copy: bool = False
    ) -> None:
        checker = self._dtype_cls._checker
        if not (isinstance(values, np.ndarray) and checker(values.dtype)):
            descr = (
                "floating"
                if self._dtype_cls.kind == "f"  # type: ignore[comparison-overlap]
                else "integer"
            )
            raise TypeError(
                f"values should be {descr} numpy array. Use "
                "the 'pd.array' function instead"
            )
        if values.dtype == np.float16:
            # If we don't raise here, then accessing self.dtype would raise
            raise TypeError("FloatingArray does not support np.float16 dtype.")

        # NB: if is_nan_na() is True
        #  then caller is responsible for ensuring
        #  assert mask[np.isnan(values)].all()

        super().__init__(values, mask, copy=copy)

    @cache_readonly
    def dtype(self) -> NumericDtype:
        mapping = self._dtype_cls._get_dtype_mapping()
        return mapping[self._data.dtype]

    @classmethod
    def _coerce_to_array(
        cls, value, *, dtype: DtypeObj, copy: bool = False
    ) -> tuple[np.ndarray, np.ndarray]:
        dtype_cls = cls._dtype_cls
        values, mask = _coerce_to_data_and_mask(value, dtype, copy, dtype_cls)

View on GitHub (pinned to 3b7651241d)

Solutions

1. Upcast to float32 or float64 before wrapping: values.astype(np.float64).
2. Use pd.array(values, dtype='Float32') which coerces through the standard path.
3. Keep float16 data in a plain numpy array or a numpy-backed Series instead of a FloatingArray.

Example fix

# before
pd.arrays.FloatingArray(np.array([1.0], dtype=np.float16), mask=np.zeros(1, bool))  # raises

# after
pd.arrays.FloatingArray(np.array([1.0], dtype=np.float16).astype(np.float32), mask=np.zeros(1, bool))

import numpy as np

def to_float_array(values):
    arr = np.asarray(values)
    if arr.dtype == np.float16:
        arr = arr.astype(np.float32)
    return arr

def is_supported_float_dtype(values) -> bool:
    import numpy as np
    dt = np.asarray(values).dtype
    return dt.kind == 'f' and dt != np.float16

try:
    pd.arrays.FloatingArray(values, mask)
except TypeError as e:
    if 'float16' in str(e):
        values = np.asarray(values, dtype=np.float32)
        pd.arrays.FloatingArray(values, mask)
    else:
        raise

When it happens

Trigger: pd.arrays.FloatingArray(np.array([1.0], dtype=np.float16), mask=...) ; or pd.array(values, dtype='Float64') where the underlying numpy array was upcast incorrectly and float16 leaked through an internal path. Also when a user manually builds a float16 ndarray and calls the constructor.

Common situations: Memory-constrained pipelines that downcast to float16 upstream and then try to wrap in a nullable floating extension array; Arrow import paths that materialize float16.

Related errors

• values must be a 1D list-like
• values should be {descr} numpy array. Use the 'pd.array' fun
• Invalid value '{value!s}' for dtype '{self.dtype}'
• cannot convert to '{dtype}'-dtype NumPy array with missing v
• interpolate is not implemented for dtype={self.dtype}

AI-assisted analysis of pandas-dev/pandas@3b7651241d (2026-08-11). Data as JSON: /api/errors/c6d1c3406bb03286. Report an issue: GitHub.