SenseVoiceLargeOptions

Options & Configuration · AiDotNet.SpeechRecognition.AlibabaASR

Options for SenseVoice-Large: the autoregressive encoder-decoder SenseVoice (FunAudioLLM, 2024).

Inheritance: ModelOptions → SenseVoiceLargeOptions

For Beginners

Defaults are the Whisper-large geometry SenseVoice-Large follows: 128 mel bands over a 30-second window, a 32-layer encoder and a 32-layer decoder of width 1,280 with 20 heads. The report gives 1,587M parameters; its vocabulary is unpublished, so Whisper large-v3’s 51,866 tokens (1,550M parameters) is the default.

Constructors

Constructor Summary
SenseVoiceLargeOptions() Initializes a new instance with default values.
SenseVoiceLargeOptions(SenseVoiceLargeOptions other) Initializes a new instance by copying from another instance.

Properties

Property Summary
OnnxModelOptions OnnxOptions { get; set; } Gets or sets the ONNX Runtime settings used when the model is loaded from an ONNX file.
bool PredictEmotion { get; set; } Whether the prompt carries the SER task token, so the decoder predicts the speech emotion.
bool PredictEvents { get; set; } Whether the prompt carries the AED task token, so the decoder predicts audio events with timestamps.
bool PredictLanguage { get; set; } Whether the prompt carries the LID task token, so the decoder first predicts the language.
double LearningRate { get; set; } The AdamW learning rate used when no optimizer is injected (1e-4; the report states no optimizer).
double WeightDecay { get; set; } The AdamW weight decay used when no optimizer is injected (0.01).
int EncoderDim { get; set; } The encoder’s and the decoder’s width (1,280); the feed-forward is four times wider.
int MaxTextLength { get; set; } The longest decoded sequence, task outputs included (at most the decoder’s 448 positions).
int NumAttentionHeads { get; set; } Gets or sets the number of attention heads.
int NumDecoderLayers { get; set; } Decoder layers (32).
int NumEncoderLayers { get; set; } Encoder layers (32).
int NumMels { get; set; } Log-mel bands (128).
int SampleRate { get; set; } Gets or sets the sample rate, in Hz, that input audio is expected at.
int VocabSize { get; set; } The vocabulary (51,866).
string Language { get; set; } Gets or sets the language code used for speech processing.
string ModelPath { get; set; } Gets or sets the path to the model weights.
string[] Vocabulary { get; set; } Optional vocabulary mapping token IDs to text.