ComfyUI Node
ClybGuidance
A ComfyUI node in sampling/custom_sampling with 13 inputs and 1 output.
ClybGuidance
- model
- MODEL
◄eta1.00►
◄norm_threshold0.0►
◄momentum0.00►
◄momentum_beta0.75►
◄momentum_renorm1.00►
◄scalar_projectionfalse►
◄scalar_logsumexpfalse►
◄rescale_phi0.00►
◄var_rescalefalse►
◄scale_up_ratio0.00►
◄scale_up_shift1.00►
◄atan2sin_ratio0.00►
Categorysampling/custom_sampling
Inputs (13)
| Name | Type | Default | Description |
|---|---|---|---|
| model | MODEL | — | |
| eta | FLOAT | 1.00-50–50 | Controls the scale of the parallel guidance vector. Default CFG behavior at a setting of 1. |
| norm_threshold | FLOAT | 0.00–50 | Normalize guidance vector to this value, normalization disable at a setting of 0. |
| momentum | FLOAT | 0.00-10–10 | Controls the amount of momentum applied to the latent, disabled at a setting of 0. |
| momentum_beta | FLOAT | 0.750–0.999 | Controls a running average of guidance during diffusion. |
| momentum_renorm | FLOAT | 1.000–1 | Re-normalizes your latent after applying momentum, back to its norm before momentum. |
| scalar_projection | BOOLEAN | false | Applies scalar projection of cond -> uncond onto the uncond. |
| scalar_logsumexp | BOOLEAN | false | Whether we use torch.logsumexp (true) or torch.sum (false) for scalar projection. |
| rescale_phi | FLOAT | 0.000–1 | Applies standard deviation renormalization of CFG to cond at this rate. |
| var_rescale | BOOLEAN | false | Whether we use torch.var (true) or torch.std (false) for rescaling. |
| scale_up_ratio | FLOAT | 0.000–1 | Weight of: Initiating CFG at guidance scale 1, increasing to your guidance scale in the middle of diffusion, and lower back to 1. |
| scale_up_shift | FLOAT | 1.000.1–10 | Whether to shift to your CFG scale later (lower than 1.0) or earlier (higher than 1.0) in the schedule. |
| atan2sin_ratio | FLOAT | 0.00-100–100 | Applies standard deviation renormalization of CFG to cond at this rate. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| MODEL | MODEL | — |