espnet3.systems.esp2_asr.inference.ASRInferenceRunner

source

class espnet3.systems.esp2_asr.inference.ASRInferenceRunner(provider: EnvironmentProvider, idx_key: str = 'utt_id', hyp_key: str | Sequence[str] = 'hyp', ref_key: str | Sequence[str] = 'ref', **kwargs)

Bases: InferenceRunner

Convert ESPnet2 ASR n-best results to the contracted text field.

Initialize the inference runner with output key settings.

• Parameters:
  • provider – Environment provider that supplies dataset/model/env.
  • idx_key – Output dict key used as the sample identifier written in the first column of each SCP line. This ties each inference result back to its dataset sample. Defaults to "utt_id".
  • hyp_key – Hypothesis key or keys expected in the output dict.
  • ref_key – Reference key or keys expected in the output dict.
  • **kwargs – Forwarded to BaseRunner.

static format_output(result, , batched)

Format best text hypotheses before the shared contract checks.

• Parameters:
  • result – The n-best results of Speech2Text (a list of them for a batch).
  • batched – Whether the call was a batch.
• Returns:{"text": ...} for one call, or a list of them for a batch.

####### Examples

>>> result = [("hello", ["h"], [1], None)]
>>> ASRInferenceRunner.format_output(result, batched=False)
{'text': 'hello'}

static prepare_model_inputs(inputs)

Pass all channels to ESPnet2 in its samples-first array layout.

• Parameters:inputs – The backend’s keyword arguments; speech holds an Audio (a list of them for a batch).
• Returns: A copy of inputs whose speech is a (samples, channels) array for stereo audio, or a 1-D array for mono.

####### Examples

>>> import numpy as np
>>> audio = Audio(np.zeros((2, 8000), dtype=np.float32), 16000)
>>> out = ASRInferenceRunner.prepare_model_inputs({"speech": audio})
>>> out["speech"].shape
(8000, 2)