Functions — github.com/darkautism/sensevoice-rs

CodeHub MCPcopy Create free account

hub / github.com/darkautism/sensevoice-rs / functions

Functions46 in github.com/darkautism/sensevoice-rs

⨍ Functions 46 ◇ Types & classes 22

↓ 8 callersMethodforward

(&self, xs: &Tensor)

src/asr_candle_pt.rs:147

↓ 5 callersMethodforward

( &self, input: &Tensor, cache: &Tensor, )

src/silero_vad.rs:36

↓ 4 callersMethodreset

src/silero_vad.rs:349

↓ 3 callersMethoddestroy

Destroys the `SenseVoiceSmall` instance, releasing associated resources. This method ensures that the RKNN model resources are properly cleaned up.

↓ 2 callersMethodextract_features

Extracts mel-frequency features from a waveform. Processes the input waveform to produce fbank features, applies LFR, and optionally applies CMVN. #

src/wavfrontend.rs:346

↓ 2 callersMethodfinalize_segment

(&mut self, trim_tail: bool)

src/silero_vad.rs:316

↓ 2 callersMethodfinish

src/silero_vad.rs:359

↓ 2 callersMethodids_to_text

Helper to convert token IDs to text

↓ 2 callersMethodinfer_file

Performs speech recognition on an audio file.

↓ 2 callersFunctionparse_line

Parses a string line into a `VoiceText` instance based on a specific format. The expected format is: `<|language|><|emotion|><|event|><|punctuation|>

↓ 2 callersMethodrecognition

(&self, segment: &[i16])

↓ 2 callersMethodrun

( &self, audio_feats: &Array2<f32>, _speech_length: i64, language: i64,

src/asr_candle_pt.rs:267

↓ 1 callersMethodapply_cmvn

Applies cepstral mean and variance normalization (CMVN) to features. Normalizes the features using precomputed mean and variance values if available.

src/wavfrontend.rs:307

↓ 1 callersMethodapply_lfr

Applies low frame rate (LFR) processing to fbank features. Stacks and subsamples frames to reduce the frame rate, padding as necessary. # Arguments

src/wavfrontend.rs:255

↓ 1 callersMethodcompute_fbank_features

Computes mel-frequency filterbank (fbank) features from a waveform. Uses the `kaldi_fbank_rust` library to extract fbank features based on the config

src/wavfrontend.rs:194

↓ 1 callersMethoddecode_asr_output

(&self, output: &[f32])

↓ 1 callersMethoddecode_onnx_output

Decodes ONNX output.

↓ 1 callersMethodfill_buffer

examples/stream.rs:36

↓ 1 callersMethodforward_simple

(&self, xs: &Tensor)

src/asr_candle_pt.rs:74

↓ 1 callersMethodinfer_stream

( &'a mut self, input_stream: S, )

↓ 1 callersMethodinfer_vec

Performs speech recognition on a vector of audio samples.

↓ 1 callersMethodpositional_encoding

( &self, seq_len: usize, dim: usize, )

src/asr_candle_pt.rs:209

↓ 1 callersMethodpredict_probability

( &mut self, chunk: &[i16; CHUNK_SIZE], )

src/silero_vad.rs:124

↓ 1 callersMethodprepare_rknn_input_advanced

( &self, feats: &Array2<f32>, language: usize, use_itn: bool, )

↓ 1 callersMethodprocess_chunk

(&mut self, chunk: &[i16; CHUNK_SIZE])

src/silero_vad.rs:258

↓ 1 callersMethodread

(&mut self, buf: &mut [u8])

examples/stream.rs:51

↓ 1 callersFunctionreader_to_stream

(mut reader: DelayedReader)

examples/stream.rs:71

↓ 1 callersMethodset_notify_silence_after_ms

更新通知靜音的設定

src/silero_vad.rs:251

src/silero_vad.rs:195

Creates a `WavFrontendConfig` instance with default values. # Returns A `WavFrontendConfig` instance with commonly used default settings.

src/wavfrontend.rs:58

(&self, f: &mut fmt::Formatter<'_>)

(&self, f: &mut std::fmt::Formatter<'_>)

src/silero_vad.rs:68

Converts a string to a `SenseVoiceLanguage` variant. This method parses a string (case-insensitive) and returns the corresponding language variant.

Initializes a new `SenseVoiceSmall` instance. If the `rknpu` feature is enabled, it initializes the RKNN backend using the default RKNN model. Otherw

Methodinit_official_model_pt

Initializes using official `FunAudioLLM/SenseVoiceSmall` assets (`model.pt` path). This uses the native Candle PT backend.

Methodinit_with_config

Initializes a new `SenseVoiceSmall` instance with custom configuration. # Arguments `config` - The configuration containing file paths. `vadconfig`

Methodload_cmvn

Loads CMVN statistics from a file. Parses a CMVN file to extract mean and variance vectors for normalization. # Arguments `path` - Path to the CMVN

src/wavfrontend.rs:141

examples/official_model_pt.rs:4

examples/stream.rs:95

examples/basic.rs:4

( vb: VarBuilder, n_head: usize, in_dim: usize, hidden_dim: usize, )

src/asr_candle_pt.rs:37

Creates a new `SenseVoiceSmallError` instance with the given message. # Arguments `message` - The error message to encapsulate. # Returns A new `S

(vb: VarBuilder)

src/silero_vad.rs:22

Creates a new `WavFrontend` instance with the specified configuration. If a CMVN file is provided in the config, it loads the mean and variance value

src/wavfrontend.rs:112

(mut file: File, chunk_size: usize, delay: Duration)

examples/stream.rs:24

Methodset_vad_silence_notification

(&mut self, ms: Option<u32>)