ComfyUI Node
LoRA Block Filter (Apply)
LoraLoaderModelOnly with a per-block filter. Tensors belonging to no numbered block (embedders, heads) are always applied -- they are not blocks, and dropping them would change the adapter in a way the block spec never asked for. Nothing is written to disk.
LoRA Block Filter (Apply)
- model
- model
- applied
◄lora_name▾►
◄strength_model1.00►
◄blocks►
◄mode▾►
CategoryLoRA Block Surgeon
Inputs (5)
| Name | Type | Default | Description |
|---|---|---|---|
| model | MODEL | — | |
| lora_name | COMBO | Opened read-only. Never modified. | |
| strength_model | FLOAT | 1.00-100–100 | — |
| blocks | STRING | e.g. 31-35 or 0-2,31,35. Empty + keep mode applies no blocks at all. | |
| mode | COMBO | keep = apply only these blocks. drop = apply everything except these. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| model | MODEL | — |
| applied | STRING | — |