ComfyUI Node
Load Z-Image Turbo (SDNQ)
A ComfyUI node in Z-Image (SDNQ) with 7 inputs and 1 output.
Load Z-Image Turbo (SDNQ)
- pipeline
◄model_idDisty0/Z-Image-Turbo-SDNQ-uint4-svd-r32►
◄deviceauto►
◄attention_backenddefault►
◄enable_compilationfalse►
◄cpu_offloadfalse►
◄low_cpu_mem_usagetrue►
◄vae_tilingfalse►
CategoryZ-Image (SDNQ)
Inputs (7)
| Name | Type | Default | Description |
|---|---|---|---|
| model_id | STRING | Disty0/Z-Image-Turbo-SDNQ-uint4-svd-r32 | Hugging Face model ID for the SDNQ-quantized Z-Image Turbo model |
| device | COMBO | auto | Device to load the model on. 'auto' selects CUDA if available, otherwise CPU |
| attention_backendopt | COMBO | default | Attention mechanism: 'default' (SDPA), 'flash' (Flash Attention 2), 'flash3' (Flash Attention 3), 'sage' (Sage Attention). Flash/Sage may improve speed on supported GPUs |
| enable_compilationopt | BOOLEAN | false | Compile transformer with torch.compile for faster inference. First run will be slower while compiling |
| cpu_offloadopt | BOOLEAN | false | Enable sequential CPU offloading to reduce VRAM usage. Models are moved between CPU and GPU as needed |
| low_cpu_mem_usageopt | BOOLEAN | true | Use memory-efficient loading method. Recommended to keep enabled |
| vae_tilingopt | BOOLEAN | false | Process VAE decode in tiles to reduce VRAM usage. Useful for high resolutions |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| pipeline | ZIMAGE_SDNQ_PIPELINE | — |