ComfyUI Node
Ciru AMD Halo Qwen 2.1 Turbo
Qwen Image 2.1 only. Twelve full denoiser evaluations by default. For image editing, set full evaluations equal to sampler steps, or opt in to experimental edit prediction. Auto keeps native attention at 1024 and uses the gfx1151 path at 2048 square.
Ciru AMD Halo Qwen 2.1 Turbo
- model
- MODEL
◄enabledtrue►
◄full_evaluations12►
◄attentionauto►
◄allow_edit_predictionfalse►
CategoryCiru/Image Acceleration
Inputs (5)
| Name | Type | Default | Description |
|---|---|---|---|
| model | MODEL | — | |
| enabled | BOOLEAN | true | — |
| full_evaluations | INT | 126–10000 | — |
| attention | COMBO | auto | 3 options: auto, native, strix |
| allow_edit_predictionopt | BOOLEAN | false | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| MODEL | MODEL | — |