node-llama-cpp
Version:
Run AI models locally on your machine with node.js bindings for llama.cpp. Enforce a JSON schema on the model output on the generation level
38 lines (37 loc) • 1.01 kB
TypeScript
import { CommandModule } from "yargs";
import { BuildGpu } from "../../bindings/types.js";
type InfillCommand = {
modelPath?: string;
header?: string[];
gpu?: BuildGpu | "auto";
systemInfo: boolean;
prefix?: string;
prefixFile?: string;
suffix?: string;
suffixFile?: string;
contextSize?: number;
batchSize?: number;
flashAttention?: boolean;
threads?: number;
temperature: number;
minP: number;
topK: number;
topP: number;
seed?: number;
gpuLayers?: number;
repeatPenalty: number;
lastTokensRepeatPenalty: number;
penalizeRepeatingNewLine: boolean;
repeatFrequencyPenalty?: number;
repeatPresencePenalty?: number;
maxTokens: number;
tokenPredictionDraftModel?: string;
tokenPredictionModelContextSize?: number;
debug: boolean;
meter: boolean;
timing: boolean;
noMmap: boolean;
printTimings: boolean;
};
export declare const InfillCommand: CommandModule<object, InfillCommand>;
export {};