Interface WhisperDecodeObserver

interface WhisperDecodeObserver {
    onAudioFeatures?(features: MlxArray): void;
    onStepLogits?(step: number, logits: MlxArray): void;
    onStepRows?(step: number, rows: number[][], sumLogprobs: number[]): void;
}

Index

Methods

onAudioFeatures? onStepLogits? onStepRows?

Methods

`Optional`onAudioFeatures

onAudioFeatures?(features: MlxArray): void
Parameters
- features: MlxArray
Returns void
- Defined in packages/inference/src/transcription/whisper/decode.ts:50

`Optional`onStepLogits

onStepLogits?(step: number, logits: MlxArray): void
Parameters
- step: number
- logits: MlxArray
Returns void
- Defined in packages/inference/src/transcription/whisper/decode.ts:51

`Optional`onStepRows

onStepRows?(step: number, rows: number[][], sumLogprobs: number[]): void
After each step: the live rows (token histories) and their cumulative log-probs.
Parameters
- step: number
- rows: number[][]
- sumLogprobs: number[]
Returns void
- Defined in packages/inference/src/transcription/whisper/decode.ts:53