huggingface/transformers · error · ValueError

max_size = {max_size} must be strictly greater than the requ

Error message

max_size = {max_size} must be strictly greater than the requested size for the smaller edge size = {size}

What it means

When resizing so the shortest edge becomes `size` while capping the longest at `max_size`, the cap only makes sense if max_size > size; otherwise every image would violate the cap immediately, so the helper raises ValueError. This mirrors torchvision Resize semantics.

Source

Thrown at src/transformers/image_transforms.py:303

            return tuple(size)
        elif len(size) == 1:
            # Perform same logic as if size was an int
            size = size[0]
        else:
            raise ValueError("size must have 1 or 2 elements if it is a list or tuple")

    if default_to_square:
        return (size, size)

    height, width = get_image_size(input_image, input_data_format)
    short, long = (width, height) if width <= height else (height, width)
    requested_new_short = size

    new_short, new_long = requested_new_short, int(requested_new_short * long / short)

    if max_size is not None:
        if max_size <= requested_new_short:
            raise ValueError(
                f"max_size = {max_size} must be strictly greater than the requested "
                f"size for the smaller edge size = {size}"
            )
        if new_long > max_size:
            new_short, new_long = int(max_size * new_short / new_long), max_size

    return (new_long, new_short) if width <= height else (new_short, new_long)


def resize(
    image: np.ndarray,
    size: tuple[int, int],
    resample: Optional["PILImageResampling"] = None,
    reducing_gap: int | None = None,
    data_format: ChannelDimension | None = None,
    return_numpy: bool = True,
    input_data_format: str | ChannelDimension | None = None,
) -> np.ndarray:

View on GitHub (pinned to a597f97485)

Solutions

1. Ensure max_size > size (e.g. size=800, max_size=1333).
2. If you want a hard bound on both dims, use explicit (height, width) size instead of max_size.
3. Add an assert in config code: assert max_size is None or max_size > size.

Example fix

# before
size_dict = get_size_dict(800, max_size=800, default_to_square=False)

# after
size_dict = get_size_dict(800, max_size=1333, default_to_square=False)

assert max_size is None or max_size > size, "max_size must be strictly greater than size"

When it happens

Trigger: get_resize_output_image_dims(img, size=800, max_size=800, default_to_square=False), or max_size < size such as size=1024/max_size=512. Note the comparison is strict: equal values also raise.

Common situations: Configs copied from detection models (800/1333) and then edited so max_size ends up <= size; programmatic size searches that set both from one variable; forgetting that equality is invalid.

Related errors

• Cannot specify both size as an int, with default_to_square=T
• Cannot specify both default_to_square=True and max_size
• Could not convert size input to size dict: {size}
• {param_name} must have one of the following set of keys: {VA
• size must have 1 or 2 elements if it is a list or tuple

AI-assisted analysis of huggingface/transformers@a597f97485 (2026-08-14). Data as JSON: /api/errors/ac69afcb239168da. Report an issue: GitHub.