pub fn create_fallback_image(
config: &ImagePreprocessConfig,
device: Device,
) -> Result<Tensor>Expand description
Create a minimal fallback image (1x1 random pixel expanded to full size)
This is used when no image is provided but the model requires vision input. The image is essentially a solid color with a random RGB value.
§Arguments
config- Preprocessing configuration (determines output size)device- Device to create tensor on
§Returns
Tensor of shape [1, 3, H, W] with random solid color