Skip to main content
The CLIPAttentionMultiply node allows you to adjust the attention mechanism in CLIP models by applying multiplication factors to different components of the self-attention layers. It works by modifying the query, key, value, and output projection weights and biases in the CLIP model’s attention mechanism. This experimental node creates a modified copy of the input CLIP model with the specified scaling factors applied.

​
Inputs

ParameterDescriptionData TypeRequiredRange
clipThe CLIP model to modifyCLIPYes-
qMultiplication factor for query projection weights and biases (default: 1.0)FLOATYes0.0 - 10.0
kMultiplication factor for key projection weights and biases (default: 1.0)FLOATYes0.0 - 10.0
vMultiplication factor for value projection weights and biases (default: 1.0)FLOATYes0.0 - 10.0
outMultiplication factor for output projection weights and biases (default: 1.0)FLOATYes0.0 - 10.0

​
Outputs

Output NameDescriptionData Type
CLIPReturns a modified CLIP model with the specified attention scaling factors appliedCLIP
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): 1f5c437cb4a372dfd69ad3c10e2b8cfc68821e0fe82dcf37df5e6fd64aa0b436