Skip to content

Type alias: LlamaModelOptions

ts
type LlamaModelOptions: {
  gpuLayers: number;
  modelPath: string;
  useMlock: boolean;
  useMmap: boolean;
  vocabOnly: boolean;
};
type LlamaModelOptions: {
  gpuLayers: number;
  modelPath: string;
  useMlock: boolean;
  useMmap: boolean;
  vocabOnly: boolean;
};

Type declaration

gpuLayers

ts
gpuLayers?: number;
gpuLayers?: number;

number of layers to store in VRAM

modelPath

ts
modelPath: string;
modelPath: string;

path to the model on the filesystem

useMlock

ts
useMlock?: boolean;
useMlock?: boolean;

force system to keep model in RAM

useMmap

ts
useMmap?: boolean;
useMmap?: boolean;

use mmap if possible

vocabOnly

ts
vocabOnly?: boolean;
vocabOnly?: boolean;

only load the vocabulary, no weights

Source

llamaEvaluator/LlamaModel.ts:7