SenseVoiceOptions

Options & Configuration · AiDotNet.SpeechRecognition.AlibabaASR

Options for SenseVoice: multi-task speech understanding model.

Inheritance: ModelOptions → SenseVoiceOptions

For Beginners

These options configure SenseVoice-Small. Supported defaults follow the released FunASR configuration: 512-wide SAN-M encoder, 50 + 20 blocks, 4 attention heads, 2048-wide feed-forward layers, FSMN kernel 11, low frame rate 7/6, 25,055 tokens, 80 mel bands, and AdamW at 2e-5.

Constructors

Constructor Summary
SenseVoiceOptions() Initializes a new instance with default values.
SenseVoiceOptions(SenseVoiceOptions 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 TargetsIncludeRichTokens { get; set; } Training targets carry the four rich tokens (language, emotion, event, text normalization) before the transcript, as the release’s do; their cross-entropy is then added to the CTC loss.
bool UseInverseTextNormalization { get; set; } The text-normalization query: with inverse text normalization (14) or without (15, the release’s default).
double LearningRate { get; set; } Gets or sets the optimizer learning rate.
double WeightDecay { get; set; } Gets or sets the weight decay of the default optimizer.
int EncoderDim { get; set; } Gets or sets the hidden width of the encoder.
int FeedForwardDim { get; set; } Gets or sets the hidden width of the feed-forward network.
int FsmnKernelSize { get; set; } The FSMN memory’s kernel (11).
int LfrM { get; set; } Filterbank frames stacked by the low frame rate front end (7).
int LfrN { get; set; } The stride between stacks (6, 60 ms).
int MaxAudioLengthSeconds { get; set; } Gets or sets the longest clip, in seconds, of the reference configuration.
int MaxTextLength { get; set; } Gets or sets the most tokens decoded for one utterance.
int NumAttentionHeads { get; set; } Gets or sets the number of heads in each attention layer.
int NumEncoderLayers { get; set; } Gets or sets the number of encoder layers.
int NumMels { get; set; } Gets or sets the number of mel filterbank channels in the input features.
int NumTpEncoderLayers { get; set; } SAN-M blocks after the first norm (tp_blocks, 20).
int SampleRate { get; set; } Gets or sets the sample rate, in Hz, that input audio is expected at.
int VocabSize { get; set; } Gets or sets the number of output tokens the model predicts over.
string Language { get; set; } The language query: auto (the release’s default), zh, en, yue, ja, ko or nospeech.
string ModelPath { get; set; } Gets or sets the path of the pretrained ONNX model.
string[] Vocabulary { get; set; } Token strings by id; ids past it decode as Unicode code points.