DeepSpeech

class torchaudio.models.DeepSpeech(n_feature: int, n_hidden: int = 2048, n_class: int = 40, dropout: float = 0.0)

Deep Speech: Scaling up end-to-end speech recognition [Hannun et al., 2014] 中介紹的 DeepSpeech 架構。

參數:

• n_feature – 輸入特徵數量

• n_hidden – 內部隱藏單元大小。

• n_class – 輸出類別數量

方法

forward

DeepSpeech.forward(x: Tensor) → Tensor

參數:

x (torch.Tensor) – 維度為 (batch, channel, time, feature) 的張量。

返回:

維度為 (batch, time, class) 的預測張量。

返回類型:

Tensor