ComfyUI Node
GGUF Accelerator (patch_on_device)
A ComfyUI node in IAMCCS/Optimize with 8 inputs and 2 outputs.
GGUF Accelerator (patch_on_device)
- model
- model
- report
◄modeauto_oom_safe►
◄patch_on_devicetrue►
◄move_patches_nowtrue►
◄min_free_vram_mb1500►
◄oom_fallbacktrue►
◄move_policyall_or_nothing►
◄leave_free_vram_mb1024►
CategoryIAMCCS/Optimize
Inputs (8)
| Name | Type | Default | Description |
|---|---|---|---|
| model | MODEL | — | |
| mode | COMBO | auto_oom_safe | auto_oom_safe: tries patch_on_device+eager move, falls back to offload on OOM | manual: use toggles below |
| patch_on_device | BOOLEAN | true | — |
| move_patches_now | BOOLEAN | true | If enabled, attempts to pre-move patch tensors to the model load_device to reduce runtime transfers. Can increase VRAM usage. |
| min_free_vram_mb | INT | 15000–65536 | (auto_oom_safe) If free VRAM is below this, we disable patch_on_device to reduce OOM risk. 0 disables the check. |
| oom_fallback | BOOLEAN | true | If a CUDA OOM happens while moving patches, automatically switches to offload and continues. |
| move_policy | COMBO | all_or_nothing | When pre-moving patches: all_or_nothing moves everything only if it fits the VRAM budget; partial_* moves a subset within the budget. |
| leave_free_vram_mb | INT | 10240–65536 | When pre-moving patches to CUDA, try to keep at least this much VRAM free. 0 disables budget limiting. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| model | MODEL | — |
| report | STRING | — |