tracel-ai/burn · error

Scale factor for height is too large

Error message

Scale factor for height is too large

What it means

This panic comes from Interpolate2d::calculate_output_size in burn-nn. When the module is configured with a scale_factor, the output height is computed as input_height * scale_factor[0] in f64; if the product exceeds usize::MAX (or is infinite/NaN-adjacent), the resulting usize cannot represent it, so the library panics instead of wrapping or saturating. It is a guard against silent numeric overflow when upsampling.

Source

Thrown at crates/burn-nn/src/modules/interpolate/interpolate2d.rs:148

/// or if the scale factor results in dimensions exceeding usize::MAX.
fn calculate_output_size(
    input_dims: [usize; 4],
    output_size: Option<[usize; 2]>,
    scale_factor: Option<[f32; 2]>,
) -> [usize; 2] {
    match (output_size, scale_factor) {
        (Some(output_size), None) => {
            // Use provided
            output_size
        }
        (None, Some(scale_factor)) => {
            // Calculate output size based on scale factor
            let [_, _, h, w] = input_dims;

            let new_dim_h = (h as f64) * (scale_factor[0] as f64);

            if new_dim_h > usize::MAX as f64 {
                panic!("Scale factor for height is too large");
            }

            let new_dim_w = (w as f64) * (scale_factor[1] as f64);

            if new_dim_w > usize::MAX as f64 {
                panic!("Scale factor for width is too large");
            }

            [new_dim_h as usize, new_dim_w as usize]
        }
        _ => panic!("Either output_size or scale_factor must be provided"),
    }
}

impl ModuleDisplay for Interpolate2d {
    fn custom_settings(&self) -> Option<DisplaySettings> {
        DisplaySettings::new()
            .with_new_line_after_attribute(false)

View on GitHub (pinned to d16f7ba2ed)

Solutions

1. Reduce scale_factor[0] so that input_height * scale_factor[0] stays within usize range (keep it a small multiple like 2.0 or 4.0).
2. Specify the exact target with with_output_size([h, w]) instead of a scale factor when you know the desired dimensions.
3. Validate/parse the scale from external config: reject non-finite or > ~1e6 values before constructing Interpolate2dConfig.
4. If you legitimately need extreme upsampling, downscale the input first or process it in tiles.

Example fix

// before
let cfg = Interpolate2dConfig::new().with_scale_factor([f64::INFINITY, 2.0]);
// after
let scale_h = if scale_h.is_finite() { scale_h } else { 1.0 };
let cfg = Interpolate2dConfig::new().with_scale_factor([scale_h, 2.0]);
// or, when the target size is known:
let cfg = Interpolate2dConfig::new().with_output_size([224, 224]);

let new_dim_h = (h as f64) * (scale_factor[0] as f64);
if !new_dim_h.is_finite() || new_dim_h > usize::MAX as f64 {
    // reject or clamp before calling forward
    return Err("height scale out of range");
}

fn valid_scale(s: f64) -> bool {
    s.is_finite() && s > 0.0 && s <= 1e6
}

let result = std::panic::catch_unwind(std::panic::AssertUnwindSafe(|| module.forward(input)));
match result {
    Ok(out) => out,
    Err(_) => fallback_to_output_size_config(),
}

When it happens

Trigger: Calling Interpolate2d::forward (directly or via a model) where the config used with_scale_factor([s_h, s_w]) and (input_height as f64) * s_h > usize::MAX as f64. Typical concrete triggers: scale_factor[0] = f64::INFINITY, an extremely large finite scale (e.g. 1e19), or a huge input height combined with a moderate scale.

Common situations: Typo in scale_factor (e.g. writing a target size like 4096 as a scale instead of an output_size), reading the scale from user input or a config file without bounds checks, accidentally passing f64::INFINITY or f64::MAX as a sentinel, or running on 32-bit targets where usize::MAX is much smaller.

Related errors

• Scale factor for width is too large
• Either output_size or scale_factor must be provided
• Standard deviation is required to be non-negative, but got {
• Affine is set to true, but gamma or beta is None
• capture tensor operations must run inside CaptureDevice::cap

AI-assisted analysis of tracel-ai/burn@d16f7ba2ed (2026-09-05). Data as JSON: /api/errors/0d4c41613a50bc15. Report an issue: GitHub.