pandas-dev/pandas · error · TypeError

dtype {data.dtype} cannot be converted to timedelta64[ns]

Error message

dtype {data.dtype} cannot be converted to timedelta64[ns]

What it means

Raised by sequence_to_td64ns when data.dtype is not integer, float, object, or timedelta64. The else-branch explicitly lists datetime64 as a known trigger (GH#23539, GH#29794): you cannot convert a datetime array into a timedelta array directly.

Source

Thrown at pandas/core/arrays/timedeltas.py:1308

        data = data.astype(np.float64, copy=False)
        try:
            data = cast_from_unit_vectorized(data, unit or "ns")
        except OutOfBoundsDatetime as err:
            raise OutOfBoundsTimedelta(*err.args) from err
        data[mask] = iNaT
        data = data.view("m8[ns]")
        copy = False

    elif lib.is_np_dtype(data.dtype, "m"):
        if not is_supported_dtype(data.dtype):
            # cast to closest supported unit, i.e. s or ns
            new_dtype = get_supported_dtype(data.dtype)
            data = astype_overflowsafe(data, dtype=new_dtype, copy=False)
            copy = False

    else:
        # This includes datetime64-dtype, see GH#23539, GH#29794
        raise TypeError(f"dtype {data.dtype} cannot be converted to timedelta64[ns]")

    if not copy:
        data = np.asarray(data)
    else:
        data = np.array(data, copy=copy)

    assert data.dtype.kind == "m"
    assert data.dtype != "m8"  # i.e. not unit-less

    return data


def _ints_to_td64ns(data, unit: str = "ns") -> tuple[np.ndarray, bool]:
    """
    Convert an ndarray with integer-dtype to timedelta64[ns] dtype, treating
    the integers as multiples of the given timedelta unit.

    Parameters

View on GitHub (pinned to 71959b8cb9)

Solutions

1. If you have datetimes, compute differences: `dt_a - dt_b` already yields timedelta.
2. Cast object columns of strings first via pd.to_timedelta on the raw strings.
3. Check data.dtype before calling to_timedelta and branch accordingly.

Example fix

// before
s = pd.to_timedelta(df['timestamp'])  # datetime64

// after
s = df['timestamp'] - df['timestamp'].min()  # timedelta

import numpy as np
if np.asarray(data).dtype.kind == 'M':
    raise TypeError('input is datetime64; compute differences to get timedelta')

def is_timedelta_convertible(data) -> bool:
    import numpy as np
    k = np.asarray(data).dtype.kind
    return k in 'iufOM' or k == 'm'

try:
    s = pd.to_timedelta(data)
except TypeError as e:
    if 'cannot be converted to timedelta64' in str(e):
        s = data - np.min(data)  # convert datetime to timedelta via diff
    else:
        raise

When it happens

Trigger: `pd.to_timedelta(pd.to_datetime(['2020-01-01']))`, or passing a datetime64 ndarray / complex-dtype array to to_timedelta / TimedeltaIndex constructor.

Common situations: Confusing duration vs timestamp; subtracting two datetimes but forgetting to wrap with subtraction that yields timedelta; passing the wrong column from a pipeline.

Related errors

• overflow in timedelta operation
• cannot add {type(self).__name__} and {type(other).__name__}
• cannot subtract a datelike from a {type(self).__name__}
• Supported units are 's', 'ms', 'us', 'ns'
• {dtype=} does not have a resolution.

AI-assisted analysis of pandas-dev/pandas@71959b8cb9 (2026-08-07). Data as JSON: /api/errors/49ada9b06b661b76. Report an issue: GitHub.