keras-team/keras · error · ValueError

The argument `kernel_size` cannot contain 0. Received kernel

Error message

The argument `kernel_size` cannot contain 0. Received kernel_size={self.kernel_size}.

What it means

Depthwise convolution kernels must have non-zero size in every spatial dimension; a 0 in kernel_size (e.g. (3,0)) is a degenerate filter. The constructor uses all(self.kernel_size) to reject any zero component.

Source

Thrown at keras/src/layers/convolutional/base_depthwise_conv.py:139

        self.use_bias = use_bias
        self.depthwise_initializer = initializers.get(depthwise_initializer)
        self.bias_initializer = initializers.get(bias_initializer)
        self.depthwise_regularizer = regularizers.get(depthwise_regularizer)
        self.bias_regularizer = regularizers.get(bias_regularizer)
        self.depthwise_constraint = constraints.get(depthwise_constraint)
        self.bias_constraint = constraints.get(bias_constraint)
        self.input_spec = InputSpec(min_ndim=self.rank + 2)
        self.data_format = self.data_format

        if self.depth_multiplier is not None and self.depth_multiplier <= 0:
            raise ValueError(
                "Invalid value for argument `depth_multiplier`. Expected a "
                "strictly positive value. Received "
                f"depth_multiplier={self.depth_multiplier}."
            )

        if not all(self.kernel_size):
            raise ValueError(
                "The argument `kernel_size` cannot contain 0. Received "
                f"kernel_size={self.kernel_size}."
            )

        if not all(self.strides):
            raise ValueError(
                "The argument `strides` cannot contains 0. Received "
                f"strides={self.strides}"
            )

    def build(self, input_shape):
        if self.data_format == "channels_last":
            channel_axis = -1
            input_channel = input_shape[-1]
        else:
            channel_axis = 1
            input_channel = input_shape[1]
        self.input_spec = InputSpec(

View on GitHub (pinned to 7a34a03db6)

Solutions

1. Set kernel_size to positive integers in every spatial dimension.
2. Validate all(k > 0 for k in kernel_size) before constructing the layer.
3. Check tuple arity matches the layer rank.

Example fix

# before
keras.layers.DepthwiseConv2D(kernel_size=(3, 0))

# after
keras.layers.DepthwiseConv2D(kernel_size=(3, 3))

ks = kernel_size if isinstance(kernel_size, (tuple, list)) else (kernel_size,)
assert all(k > 0 for k in ks)

def valid_kernel_size(kernel_size, rank) -> bool:
    ks = kernel_size if isinstance(kernel_size, (tuple, list)) else (kernel_size,)
    return len(ks) == rank and all(k > 0 for k in ks)

When it happens

Trigger: Constructing DepthwiseConv1D/2D/3D with kernel_size containing a 0, typically a computed or config-driven kernel size.

Common situations: Kernel size derived from data shapes or config parsing that yields 0; rank/tuple arity mismatch producing a stray 0.

Related errors

• Invalid value for argument `depth_multiplier`. Expected a st
• The argument `strides` cannot contains 0. Received strides={
• The argument `kernel_size` cannot contain 0. Received: kerne
• `strides > 1` not supported in conjunction with `dilation_ra
• Invalid value for argument `depth_multiplier`. Expected a st

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/b644d0dc86a79598. Report an issue: GitHub.