mlx-bun library API

Preparing search index...

@mlx-bun/inference
@mlx-bun/inference/execution/fit
DECODE_EFFICIENCY

Variable DECODE_EFFICIENCY`Const`

DECODE_EFFICIENCY: 0.82

Decode-efficiency vs theoretical bandwidth ceiling, measured on the reference machine (24.9 tok/s vs 30.3 ceiling @600 ctx).

Defined in packages/inference/src/execution/fit.ts:18

Settings

Member Visibility

Protected
Inherited
External

Theme

mlx-bun library API

Loading...

Generated using TypeDoc