keras-team/keras · error · ValueError

Invalid `output_size`. Expected length 3 (D, H, W). Got: out

Error message

Invalid `output_size`. Expected length 3 (D, H, W). Got: output_size={output_size}

What it means

The 3D branch of keras.ops.image.reconstruct_patches requires output_size as a 3-element sequence (D, H, W). It raises when len(output_size) != 3, before touching tensors, because the target volume shape is undefined.

Source

Thrown at keras/src/ops/image.py:1328

                f"size * grid. Got output_size=({H},{W}), "
                f"grid=({gH},{gW}), size=({pH},{pW})."
            )

    if _unbatched:
        x = backend.numpy.squeeze(x, axis=0)
    return x


def _reconstruct_patches_3d(
    patches,
    size,
    output_size,
    strides=None,
    padding="valid",
    data_format=None,
):
    if len(output_size) != 3:
        raise ValueError(
            "Invalid `output_size`. Expected length 3 (D, H, W). "
            f"Got: output_size={output_size}"
        )
    if padding not in ("same", "valid"):
        raise ValueError(
            f"Invalid `padding`. Expected 'same' or 'valid'. Got: {padding}"
        )
    _validate_reconstruct_strides(size, strides, "reconstruct_patches")
    data_format = backend.standardize_data_format(data_format)
    if data_format == "channels_first":
        # Reconstruct in channels_last layout, then move channels back.
        # Patches are (flat, gD, gH, gW) unbatched or (B, flat, gD, gH, gW).
        if len(patches.shape) == 4:
            patches = backend.numpy.transpose(patches, axes=(1, 2, 3, 0))
        elif len(patches.shape) == 5:
            patches = backend.numpy.transpose(patches, axes=(0, 2, 3, 4, 1))
        else:
            raise ValueError(

View on GitHub (pinned to 7a34a03db6)

Solutions

1. Pass exactly three spatial ints: output_size=(D,H,W), no batch/channel axes
2. Confirm the patches tensor really is the 3D kind (rank 5 batched / rank 4 unbatched)
3. Add an assert len(output_size)==3 guard in the data pipeline

Example fix

# before
reconstruct_patches(patches, size=(4,8,8), output_size=(28,28))

# after
reconstruct_patches(patches, size=(4,8,8), output_size=(16,28,28))

if len(output_size) != 3:
    raise ValueError('output_size must be (D,H,W)')

def is_3d_output_size(output_size) -> bool:
    return hasattr(output_size, '__len__') and len(output_size) == 3

When it happens

Trigger: reconstruct_patches on rank-5 (batched 3D) or rank-4 (unbatched 3D) patches with output_size of length 2 or 4, e.g. (28,28) or (D,H,W,C).

Common situations: Porting 2D reconstruction code to 3D volumes (medical CT/MRI, video) and not extending output_size; including a channel or batch entry in output_size; passing a numpy array where a typo drops an element.

Understand the failure class

Background: "Must be a positive integer", "Invalid value", "Unsupported": the invalid-argument-value error family, when a library rejects the value you pass — this error's family across 35 libraries.

Related errors

• For `padding='same'`, `output_size` width ({W}) must be in t
• `padding='valid'` requires output_size to equal size * grid.
• `patches` has unexpected rank for 3D reconstruction. Expecte
• For `padding='same'`, `output_size` depth ({D}) must be in t
• `padding='valid'` requires output_size to equal size * grid.

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/e264e7a20097cd59. Report an issue: GitHub.