keras-team/keras · error · ValueError

Argument `constant_values` can only be provided when `mode…

Error message

Argument `constant_values` can only be provided when `mode == 'constant'`. Received: mode={mode}

What it means

Guard inside the NumPy backend's `pad`: NumPy's `constant_values` fill argument only applies when padding with mode 'constant'. If a user supplies `constant_values` together with any other mode ('reflect', 'edge', 'minimum', etc.), the combination is rejected rather than silently ignoring the fill value. Referenced by the custom-train-step JAX guide's model code paths, so it commonly surfaces when padding is configured from generic helper code that always sets a fill value.

Solutions

1. Drop the `constant_values` argument when using a non-'constant' pad mode
2. Keep mode='constant' if a specific fill value is required
3. Only pass constant_values conditionally: kwargs['constant_values'] = v if mode == 'constant' else None

When it happens

Trigger: Thrown at keras/src/backend/numpy/numpy.py:1270 when the library encounters an invalid state.

Common situations: See trigger scenarios.

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/fca270aa43064a1e. Report an issue: GitHub.

Appendix: source

Thrown at keras/src/backend/numpy/numpy.py:1270

def ones_like(x, dtype=None):
    return np.ones_like(x, dtype=dtype)


def outer(x1, x2):
    x1 = convert_to_tensor(x1)
    x2 = convert_to_tensor(x2)
    dtype = dtypes.result_type(x1.dtype, x2.dtype)
    x1 = x1.astype(dtype)
    x2 = x2.astype(dtype)
    return np.outer(x1, x2)


def pad(x, pad_width, mode="constant", constant_values=None):
    kwargs = {}
    if constant_values is not None:
        if mode != "constant":
            raise ValueError(
                "Argument `constant_values` can only be "
                "provided when `mode == 'constant'`. "
                f"Received: mode={mode}"
            )
        kwargs["constant_values"] = constant_values
    return np.pad(x, pad_width, mode=mode, **kwargs)


def percentile(x, q, axis=None, method="linear", keepdims=False):
    x = convert_to_tensor(x)
    q = convert_to_tensor(q)
    ori_dtype = standardize_dtype(x.dtype)
    if ori_dtype == "bool":
        x = x.astype(config.floatx())
    dtype = dtypes.result_type(x.dtype, float)
    return np.percentile(
        x, q, axis=axis, method=method, keepdims=keepdims
    ).astype(dtype)

View on GitHub (pinned to 7a34a03db6)