Nodes/Skoogeer-Noise/Model (RotorQuant Attention)
ComfyUI Node

Model (RotorQuant Attention)

A ComfyUI node in model/patch with 9 inputs and 1 output.

By ttulttul·Created 8 months ago·Updated 2 months ago· 14
Model (RotorQuant Attention)
  • model
  • MODEL
keep_components3
min_token_product65536
attention_scope
layer_start-1
layer_end-1
rotation_seed0
max_head_dim256
force_fp32
Categorymodel/patch

Inputs (9)

NameTypeDefaultDescription
modelMODELModel to clone and patch with RotorQuant-style attention.
keep_componentsINT31–3How many rotated coordinates to keep from each 3D rotor group. Values below 3 are currently forced back to 3 because the lossy modes degrade image quality too much.
min_token_productINT655360–1073741824Only patch attention calls where query_tokens * key_tokens meets this threshold.
attention_scopeCOMBOWhich attention calls to patch. 'self' is usually the most useful for diffusion latents.
layer_startINT-1-1–4096First transformer block index to patch. -1 disables the lower bound.
layer_endINT-1-1–4096Last transformer block index to patch. -1 disables the upper bound.
rotation_seedINT00–18446744073709550000Seed used to generate the per-triplet rotor rotations.
max_head_dimINT2563–4096Skip heads larger than this to avoid excessive projection overhead.
force_fp32COMBOCast q/k/v to fp32 inside the override for extra numerical stability.

Outputs (1)

NameTypeDescription
MODELMODEL