ComfyUI-DareMerge
Merge two checkpoint models by dare ties (https://github.com/yule-BUAA/MergeLM), sort of.
Node List
DareModelMerger
| category | node name | input type | output type | desc. |
|---|---|---|---|---|
| merge | DareModelMerger | MODEL, MODEL |
MODEL |
Performs a DARE block merge |
- input, middle, and out are the block region weights.
- sparsity is the level to sample the peak differences between the two models. Zero applies nothing, 1 applies the full normal block weights for the second model.
- threshold_type is the way we determine where our threshold lies in our distribution since we use chunks, quantile will err towards the sparsity, median will use the chunk median.
- invert is whether we invert the threshold, so we keep the weights that are below the threshold instead of above.