ComfyUI Node
🎲 LLM Sampling Options
A ComfyUI node in prompt/wildcards with 10 inputs and 2 outputs.
🎲 LLM Sampling Options
- server
- server
- options
◄use_model_default_samplingfalse►
◄temperature-1.00►
◄top_k0►
◄top_p0.00►
◄min_p0.00►
◄repeat_penalty1.00►
◄context_size0►
◄system_prompt►
◄show_everything_in_consolefalse►
Categoryprompt/wildcards
Inputs (10)
| Name | Type | Default | Description |
|---|---|---|---|
| server | LLM_SERVER | — | |
| use_model_default_samplingopt | BOOLEAN | false | If ON, drop all sampling overrides and let the model use its built-in defaults. Temperature from the base server is still applied. |
| temperatureopt | FLOAT | -1.00-1–2 | Override the server's temperature (-1 = keep server value). 0.0 = deterministic, 2.0 = very random. |
| top_kopt | INT | 00–200 | Sample from top K tokens (0 = disabled / not sent). |
| top_popt | FLOAT | 0.000–1 | Nucleus sampling threshold (0.0 = not sent). |
| min_popt | FLOAT | 0.000–1 | Minimum probability relative to the top token (0.0 = not sent). |
| repeat_penaltyopt | FLOAT | 1.001–2 | Repetition penalty (1.0 = no penalty / not sent). |
| context_sizeopt | INT | 00–131072 | Context window size in tokens (0 = use server default / not sent). |
| system_promptopt | STRING | Custom system prompt. Used only when wired into a node that reads it (e.g. Prompt Generator). Leave empty to use the consumer's defaults. | |
| show_everything_in_consoleopt | BOOLEAN | false | Print system prompt, user prompt, sampling options, and raw replies to the ComfyUI console. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| server | LLM_SERVER | — |
| options | OPTIONS | — |