ModelMergeCosmosPredict2_2B - ComfyUI Built-in Node Documentation

The ModelMergeCosmosPredict2_2B node merges two diffusion models using a block-based approach with fine-grained control over different model components. It allows you to blend specific parts of two models by adjusting interpolation weights for position embedders, time embedders, transformer blocks, and final layers. This provides precise control over how different architectural components from each model contribute to the final merged result.

Inputs

Parameter	Description	Data Type	Required	Range
`model1`	The first model to merge	MODEL	Yes	-
`model2`	The second model to merge	MODEL	Yes	-
`pos_embedder.`	Position embedder interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`x_embedder.`	Input embedder interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`t_embedder.`	Time embedder interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`t_embedding_norm.`	Time embedding normalization interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.0.`	Transformer block 0 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.1.`	Transformer block 1 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.2.`	Transformer block 2 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.3.`	Transformer block 3 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.4.`	Transformer block 4 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.5.`	Transformer block 5 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.6.`	Transformer block 6 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.7.`	Transformer block 7 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.8.`	Transformer block 8 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.9.`	Transformer block 9 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.10.`	Transformer block 10 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.11.`	Transformer block 11 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.12.`	Transformer block 12 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.13.`	Transformer block 13 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.14.`	Transformer block 14 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.15.`	Transformer block 15 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.16.`	Transformer block 16 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.17.`	Transformer block 17 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.18.`	Transformer block 18 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.19.`	Transformer block 19 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.20.`	Transformer block 20 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.21.`	Transformer block 21 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.22.`	Transformer block 22 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.23.`	Transformer block 23 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.24.`	Transformer block 24 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.25.`	Transformer block 25 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.26.`	Transformer block 26 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`blocks.27.`	Transformer block 27 interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0
`final_layer.`	Final layer interpolation weight (default: 1.0)	FLOAT	Yes	0.0 - 1.0

Outputs

Output Name	Description	Data Type
`model`	The merged model combining features from both input models	MODEL

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): 53a8de66d6b731f5b29af326832f66cc973284bc8fdf09d779575f2346cc75a7

​Inputs

​Outputs

Inputs

Outputs