vllm.model_executor.models.funasr

Classes:

• FunASRAudioInputs –

  Dimensions:

FunASRAudioInputs

Bases: TensorSchema

Dimensions: - b: Batch size - nmb: Number of mel bins - t: Time frames (M)

Source code in vllm/model_executor/models/funasr.py

class FunASRAudioInputs(TensorSchema):
    """Dimensions:
    - b: Batch size
    - nmb: Number of mel bins
    - t: Time frames (M)
    """

    input_features: Annotated[
        list[torch.Tensor] | None,
        TensorShape("b", "nmb", "t"),
    ]
    speech_lengths: Annotated[
        list[torch.Tensor] | None,
        TensorShape("b"),
    ]
    fake_token_lengths: Annotated[
        list[torch.Tensor] | None,
        TensorShape("b"),
    ]