concision_neural/model/
mod.rs

1/*
2    Appellation: model <module>
3    Contrib: @FL03
4*/
5//! This module provides the scaffolding for creating models and layers in a neural network.
6
7#[doc(inline)]
8pub use self::{config::StandardModelConfig, store::ModelParams};
9
10pub mod config;
11pub mod store;
12
13pub(crate) mod prelude {
14    pub use super::store::*;
15}
16
17use crate::ModelFeatures;
18use cnc::data::Dataset;
19
20/// This trait defines the base interface for all models, providing access to the models
21/// configuration, layout, and learned parameters.
22pub trait Model<T = f32> {
23    /// returns an immutable reference to the models configuration; this is typically used to
24    /// access the models hyperparameters (i.e. learning rate, momentum, etc.) and other
25    /// related control parameters.
26    fn config(&self) -> &StandardModelConfig<T>;
27    /// returns a mutable reference to the models configuration; useful for setting hyperparams
28    fn config_mut(&mut self) -> &mut StandardModelConfig<T>;
29    /// returns a copy of the models features (or layout); this is used to define the structure
30    /// of the model and its consituents.
31    fn features(&self) -> ModelFeatures;
32    /// returns an immutable reference to the model parameters
33    fn params(&self) -> &ModelParams<T>;
34    /// returns a mutable reference to the model's parameters
35    fn params_mut(&mut self) -> &mut ModelParams<T>;
36    /// propagates the input through the model; each layer is applied in sequence meaning that
37    /// the output of each previous layer is the input to the next layer. This pattern
38    /// repeats until the output layer returns the final result.
39    ///
40    /// By default, the trait simply passes each output from one layer to the next, however,
41    /// custom models will likely override this method to inject activation methods and other
42    /// related logic
43    fn forward<U, V>(&self, inputs: &U) -> cnc::Result<V>
44    where
45        Self: cnc::Forward<U, Output = V>,
46    {
47        <Self as cnc::Forward<U>>::forward(self, inputs)
48    }
49    /// returns a model trainer prepared to train the model; this is a convenience method
50    /// that creates a new trainer instance and returns it. Trainers are lazily evaluated
51    /// meaning that the training process won't begin until the user calls the `begin` method.
52    fn train<U, V>(
53        &mut self,
54        dataset: Dataset<U, V>,
55    ) -> crate::train::trainer::Trainer<'_, Self, T, Dataset<U, V>>
56    where
57        Self: Sized,
58        T: Default,
59    {
60        crate::train::trainer::Trainer::new(self, dataset)
61    }
62}
concision_neural/model/mod.rs

concision_neural/model/
mod.rs
