ComfyUI Node
LlamaServe-Doc Loader
配置原生 llama-server。后端在首次执行时自动下载，不需要安装 LM Studio。
LlamaServe-Doc Loader
- server_config
◄model▾►
◄mmproj▾►
◄context_length4096►
◄gpu_layers47►
◄flash_attentionon►
◄cache_type_kq8_0►
◄cache_type_vq8_0►
◄port8191►
CategoryLlamaServe-Doc
Inputs (8)
| Name | Type | Default | Description |
|---|---|---|---|
| model | COMBO | 1 options: No GGUF models found in models/LLM | |
| mmproj | COMBO | 1 options: None | |
| context_length | INT | 4096512–262144 | — |
| gpu_layers | INT | 47-1–999 | — |
| flash_attention | COMBO | on | 3 options: on, auto, off |
| cache_type_k | COMBO | q8_0 | 3 options: q8_0, f16, q4_0 |
| cache_type_v | COMBO | q8_0 | 3 options: q8_0, f16, q4_0 |
| port | INT | 81911024–65535 | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| server_config | LLAMA_SERVER_CONFIG | — |