Skip to main content
This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub
The UNetCrossAttentionMultiply node applies multiplication factors to the cross-attention mechanism in a UNet model. It allows you to scale the query, key, value, and output components of the cross-attention layers to experiment with different attention behaviors and effects.

Inputs

ParameterData TypeRequiredRangeDescription
modelMODELYes-The UNet model to modify with attention scaling factors
qFLOATNo0.0 - 10.0Scaling factor for query components in cross-attention (default: 1.0)
kFLOATNo0.0 - 10.0Scaling factor for key components in cross-attention (default: 1.0)
vFLOATNo0.0 - 10.0Scaling factor for value components in cross-attention (default: 1.0)
outFLOATNo0.0 - 10.0Scaling factor for output components in cross-attention (default: 1.0)

Outputs

Output NameData TypeDescription
modelMODELThe modified UNet model with scaled cross-attention components