ComfyUI Node
Llama.cpp Hardware Runtime Profile
Bundles hardware-dependent batch, offload, CPU, attention, and mmap settings. n_ubatch=0 uses the backend default.
Llama.cpp Hardware Runtime Profile
- hardware profile
◄profileGPU Full Offload►
◄n_batch512►
◄n_ubatch0►
◄gpu_layersall►
◄main_gpu0►
◄n_threads0►
◄flash_attentionauto►
◄use_mmaptrue►
CategoryOllama/llama_cpp/compact
Inputs (8)
| Name | Type | Default | Description |
|---|---|---|---|
| profile | COMBO | GPU Full Offload | 5 options: GPU Full Offload, GPU Vision 512, Automatic Offload, CPU, Custom |
| n_batch | INT | 5121–65536 | — |
| n_ubatch | INT | 00–65536 | 0 uses the llama.cpp backend default. |
| gpu_layers | COMBO | all | 3 options: all, auto, cpu |
| main_gpu | INT | 00–31 | — |
| n_threads | INT | 00–1024 | — |
| flash_attention | COMBO | auto | 3 options: auto, enabled, disabled |
| use_mmap | BOOLEAN | true | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| hardware profile | OLLAMA_IMAGE_LIST_LLAMA_CPP_HARDWARE_RUNTIME_PROFILE | — |