Struct WhisperOptions 

#[repr(C)]
pub struct WhisperOptions {

    pub beam_size: usize,
    pub patience: f32,
    pub length_penalty: f32,
    pub repetition_penalty: f32,
    pub no_repeat_ngram_size: usize,
    pub max_length: usize,
    pub sampling_topk: usize,
    pub sampling_temperature: f32,
    pub num_hypotheses: usize,
    pub return_scores: bool,
    pub return_logits_vocab: bool,
    pub return_no_speech_prob: bool,
    pub max_initial_timestamp_index: usize,
    pub suppress_blank: bool,
    pub suppress_tokens: Vec<i32>,
}

Available on crate feature whisper only.

Options for whisper generation.

Examples

Example of creating a default WhisperOptions:

use ct2rs::sys::WhisperOptions;

let options = WhisperOptions::default();

Fields

beam_size: usize

Beam size to use for beam search (set 1 to run greedy search). (default: 5)

patience: f32

Beam search patience factor, as described in https://arxiv.org/abs/2204.05424. The decoding will continue until beam_size*patience hypotheses are finished. (default: 1.0)

length_penalty: f32

Exponential penalty applied to the length during beam search. (default: 1.0)

repetition_penalty: f32

Penalty applied to the score of previously generated tokens, as described in https://arxiv.org/abs/1909.05858 (set > 1 to penalize). (default: 1.0)

no_repeat_ngram_size: usize

Prevent repetitions of ngrams with this size (set 0 to disable). (default: 0)

max_length: usize

Maximum generation length. (default: 448)

sampling_topk: usize

Randomly sample from the top K candidates (set 0 to sample from the full distribution). (default: 1)

sampling_temperature: f32

High temperatures increase randomness. (default: 1.0)

num_hypotheses: usize

Number of hypotheses to include in the result. (default: 1)

return_scores: bool

Include scores in the result. (default: false)

return_logits_vocab: bool

Include log probs of each token in the result. (default: false)

return_no_speech_prob: bool

Include the probability of the no speech token in the result. (default: false)

max_initial_timestamp_index: usize

Maximum index of the first predicted timestamp. (default: 50)

suppress_blank: bool

Suppress blank outputs at the beginning of the sampling. (default: true)

suppress_tokens: Vec<i32>

List of token IDs to suppress. -1 will suppress a default set of symbols as defined in the model config.json file. (default: [-1])

Trait Implementations

impl Clone for WhisperOptions

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for WhisperOptions

fn fmt(&self, formatter: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for WhisperOptions

fn default() -> Self

Returns the “default value” for a type.

impl ExternType for WhisperOptions

type Kind = Trivial

Either cxx::kind::Opaque or cxx::kind::Trivial.

type Id

A type-level representation of the type’s C++ namespace and type name.