Skip to main content
The HyperTile node applies a tiling technique to the attention mechanism in diffusion models to optimize memory usage during image generation. It divides the latent space into smaller tiles and processes them separately, then reassembles the results. This allows for working with larger image sizes without running out of memory.

Inputs

ParameterDescriptionData TypeRequiredRange
modelThe diffusion model to apply the HyperTile optimization toMODELYes-
tile_sizeThe target tile size for processing (default: 256). The effective tile size is rounded down to a multiple of 8, with a minimum of 32.INTNo1 - 2048
swap_sizeControls how the tiles are rearranged during processing to improve efficiency (default: 2)INTNo1 - 128
max_depthThe maximum depth level (resolution scale) to apply tiling. A value of 0 applies tiling only at the highest resolution (default: 0)INTNo0 - 10
scale_depthWhen enabled, the tile size is scaled proportionally at deeper depth levels. This can help maintain quality at lower resolutions (default: False)BOOLEANNoTrue / False

Outputs

Output NameDescriptionData Type
modelThe modified model with HyperTile optimization appliedMODEL
This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub

Source fingerprint (SHA-256): ea770eea3c628ffb166c2a15a6f75757cebd99c3fdd6225f041b3ef01a0e73c7