LoRA Stack Apply (Anima Remap)

Apply a LORA_STACK to MODEL/CLIP with Anima cross-generation block remapping (28 / 40 / 52 layers).

By opaiopaio·Created 3 months ago·Updated 2 days ago· 4

◄bypassfalse►

◄enable_remaptrue►

◄target_modelAuto (detect)►

◄source_modelAuto (detect)►

◄remap_modeKeys only (recommended)►

◄extend_strength0.50►

◄weight_multiplier1.00►

◄apply_cliptrue►

CategoryLoRA Stack

Inputs (11)

Name	Type	Default	Description
model	MODEL		Model to apply the stack to. Its layer count is used to detect the target Anima generation.
bypass	BOOLEAN	false	Pass MODEL/CLIP through untouched (quick A/B toggle)
enable_remap	BOOLEAN	true	Anima only. Turn off to apply LoRAs untouched. For models that are not Anima this is a no-op either way.
target_model	COMBO	Auto (detect)	Generation of the connected model. Auto detects it from the layer count; override if Auto gets it wrong.
source_model	COMBO	Auto (detect)	Generation the LoRA was trained on. Auto infers it from the LoRA's highest block index -- which mis-reads adapters that only touch early layers, so override when in doubt.
remap_mode	COMBO	Keys only (recommended)	Keys only: rewrite indices (new layers get no delta). Extend: also project each source delta onto the layers inserted from it (experimental).
extend_strength	FLOAT	0.500–2	Only used by Extend mode. 1.0 copies the source delta as-is; scales linearly.
weight_multiplier	FLOAT	1.000–4	Multiplies every weight in the stack. 1.0 = unchanged; use it to dial the whole stack up or down at once.
apply_clip	BOOLEAN	true	OFF = never touch the text encoder: all CLIP weights are forced to 0 (model-only application).
clipopt	CLIP		Optional. Without it, CLIP weights are ignored.
lora_stackopt	LORA_STACK		Stack to apply (from LoRA Stack Auto Triggers or any other node that produces LORA_STACK).