Struct nnnoiseless::RnnModel

pub struct RnnModel { /* private fields */ }

An RnnModel contains all the model parameters for the denoising algorithm. nnnoiseless has a built-in model that should work for most purposes, but if you have specific needs then you might benefit from training a custom model. Scripts for model training are available as part of RNNoise; once the model is trained, you can load it here.

Implementations

impl RnnModel

pub fn from_bytes(bytes: &[u8]) -> Option<RnnModel>

Reads an RnnModel from an array of bytes, in the format produced by the nnnoiseless training scripts.

pub fn from_static_bytes(bytes: &'static [u8]) -> Option<RnnModel>

Reads an RnnModel from a static array of bytes, in the format produced by the nnnoiseless training scripts.

This differs from RnnModel::from_bytes in that the returned model doesn’t need to allocate its own byte buffers; it will just store references to the provided bytes array.

For example, if you have your neural network weights available at compile-time then the following code will embed them into your binary and initialize a model without allocation:

let weight_data: &'static [u8] = include_bytes!("/path/to/model/weights.rnn");
let model = RnnModel::from_static_bytes(weight_data).expect("Corrupted model file");

Trait Implementations

impl Clone for RnnModel

fn clone(&self) -> RnnModel

Returns a copy of the value.

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

Performs copy-assignment from source.

impl Default for RnnModel

fn default() -> RnnModel

Returns the “default value” for a type.