keras-team/keras · error · ValueError

`strides > 1` not supported in conjunction with `dilation_ra

Error message

`strides > 1` not supported in conjunction with `dilation_rate > 1`. Received: strides={self.strides} and dilation_rate={self.dilation_rate}

What it means

Keras transpose-convolution layers (Conv2DTranspose etc.) reject any configuration where both strides and dilation_rate exceed 1. The backend algorithms for transposed convolutions cannot combine strided upsampling with dilated kernels, so the constructor fails fast at layer creation.

Source

Thrown at keras/src/layers/convolutional/base_conv_transpose.py:159

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

        if self.output_padding is not None:
            for i, (op, s) in enumerate(zip(self.output_padding, self.strides)):
                if op >= s:
                    raise ValueError(
                        "`output_padding` must be strictly less than "
                        f"`strides` for all dimensions. At dimension {i}, "
                        f"`output_padding` is {op} but `strides` is {s}. "
                        f"Received: output_padding={self.output_padding}, "
                        f"strides={self.strides}"
                    )

        if max(self.strides) > 1 and max(self.dilation_rate) > 1:
            raise ValueError(
                "`strides > 1` not supported in conjunction with "
                f"`dilation_rate > 1`. Received: strides={self.strides} and "
                f"dilation_rate={self.dilation_rate}"
            )

        if self.output_padding is not None:
            for i, (op, s) in enumerate(zip(self.output_padding, self.strides)):
                if op >= s:
                    raise ValueError(
                        "Invalid `output_padding` argument. "
                        "Each value in `output_padding` must be strictly "
                        "less than the corresponding `strides` value.\n"
                        f"At index {i}, `output_padding` is {op} and `strides` "
                        f"is {s}.\n"
                        f"Received: output_padding={self.output_padding}, "
                        f"strides={self.strides}."
                    )

View on GitHub (pinned to 7a34a03db6)

Solutions

1. Set dilation_rate back to 1 and keep the stride, growing receptive field via kernel_size instead.
2. Replace the transpose conv with UpSampling2D followed by a dilated Conv2D, which supports the combination.
3. Assert the combo in your model factory so invalid configs fail at config time.

Example fix

# before
layer = keras.layers.Conv2DTranspose(64, 3, strides=2, dilation_rate=2)

# after
layer = keras.layers.Conv2DTranspose(64, 3, strides=2)
# or: UpSampling2D(size=2) + Conv2D(64, 3, dilation_rate=2)

assert valid_transpose_cfg(strides, dilation_rate), 'strides>1 with dilation>1 unsupported'

def valid_transpose_cfg(strides, dilation_rate) -> bool:
    s = strides if isinstance(strides, (tuple, list)) else (strides,)
    d = dilation_rate if isinstance(dilation_rate, (tuple, list)) else (dilation_rate,)
    return not (max(s) > 1 and max(d) > 1)

When it happens

Trigger: Constructing keras.layers.Conv2DTranspose (or Conv1DTranspose/Conv3DTranspose) with e.g. strides=2 and dilation_rate=2: max of any stride dim >1 AND max of any dilation dim >1 triggers it.

Common situations: Copying a regular Conv2D config into a Conv*DTranspose; attempting dilated upsampling architectures (generator networks) without knowing transposed conv limits.

Related errors

• Invalid `output_padding` argument. Each value in `output_pad
• Invalid value for argument `depth_multiplier`. Expected a st
• The argument `kernel_size` cannot contain 0. Received kernel
• The argument `strides` cannot contains 0. Received strides={
• 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/a46b3e91da27125f. Report an issue: GitHub.