simian 0.1.0

A command-line tool for exploring and implementing Machine Learning algorithms in Rust.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164

#![doc = include_str!("README.md")]
use crate::Device;
use crate::data::Dataset;
use candle_core::{Result, Tensor};

pub mod algo;
pub mod plot;

pub trait Model {
  /// Preprocesses a dataset before training or evaluation.
  fn preprocess(&self, dataset: &Dataset) -> Result<Dataset> {
    Ok(dataset.clone())
  }

  /// Forward pass (Inference). Expects preprocessed input.
  fn forward(&self, x: &Tensor) -> Result<Tensor>;

  /// Calculates gradients manually based on the loss derivative.
  fn backward(&self, x: &Tensor, d_loss_d_y: &Tensor) -> Result<Vec<Tensor>>;

  /// Exposes mutable references to the model's parameters.
  fn params_mut(&mut self) -> Vec<&mut Tensor>;

  /// Convenience method for inference on raw data.
  fn predict(&self, raw_dataset: &Dataset, device: &Device) -> Result<Tensor> {
    let preprocessed = self.preprocess(raw_dataset)?;
    let features = preprocessed
      .to_tensor(device)
      .map_err(|e| candle_core::Error::Msg(e.to_string()))?;
    self.forward(&features)
  }

  type Optim: crate::optim::Optimizer;
  type LossFn: crate::loss::Loss;

  fn optimizer(&self, learning_rate: f64) -> Self::Optim;
  fn loss(&self) -> Self::LossFn;

  fn fit<'a>(
    &'a mut self,
    train_data: &'a Dataset,
    val_data: &'a Dataset,
    target_col: &'a str,
    device: &'a Device,
  ) -> FitJob<'a, Self>
  where
    Self: Sized,
  {
    FitJob {
      model: self,
      train_data,
      val_data,
      target_col,
      device,
      learning_rate: 0.0001,
      epochs: 1000,
      batch_size: 32,
      patience: 10,
      tolerance: 1e-4,
      log_path: None,
    }
  }

  #[allow(clippy::too_many_arguments)]
  fn fit_with<O: crate::optim::Optimizer, L: crate::loss::Loss>(
    &mut self,
    optimizer: &mut O,
    loss_fn: &L,
    train_data: &Dataset,
    val_data: &Dataset,
    target_col: &str,
    device: &Device,
    epochs: usize,
    batch_size: usize,
    patience: usize,
    tolerance: f64,
    log_path: Option<String>,
  ) -> Result<()>
  where
    Self: Sized,
  {
    crate::train::fit(
      self, optimizer, loss_fn, train_data, val_data, target_col, device, epochs, batch_size,
      patience, tolerance, log_path,
    )
  }

  fn plot<'a>(
    &'a self,
    dataset: &'a Dataset,
    x_cols: &[&'a str],
    y_col: &'a str,
    device: &'a Device,
  ) -> crate::model::plot::ModelPlotJob<'a, Self>
  where
    Self: Sized,
  {
    crate::model::plot::ModelPlotJob::new(dataset, self, x_cols, y_col, device)
  }
}

pub struct FitJob<'a, M: Model> {
  model: &'a mut M,
  train_data: &'a Dataset,
  val_data: &'a Dataset,
  target_col: &'a str,
  device: &'a Device,
  learning_rate: f64,
  epochs: usize,
  batch_size: usize,
  patience: usize,
  tolerance: f64,
  log_path: Option<String>,
}

impl<'a, M: Model> FitJob<'a, M> {
  pub fn learning_rate(mut self, lr: f64) -> Self {
    self.learning_rate = lr;
    self
  }

  pub fn epochs(mut self, e: usize) -> Self {
    self.epochs = e;
    self
  }

  pub fn batch_size(mut self, b: usize) -> Self {
    self.batch_size = b;
    self
  }

  pub fn patience(mut self, p: usize) -> Self {
    self.patience = p;
    self
  }

  pub fn tolerance(mut self, t: f64) -> Self {
    self.tolerance = t;
    self
  }

  pub fn log_path(mut self, lp: Option<String>) -> Self {
    self.log_path = lp;
    self
  }

  pub fn run(self) -> Result<()> {
    let mut opt = self.model.optimizer(self.learning_rate);
    let loss = self.model.loss();
    self.model.fit_with(
      &mut opt,
      &loss,
      self.train_data,
      self.val_data,
      self.target_col,
      self.device,
      self.epochs,
      self.batch_size,
      self.patience,
      self.tolerance,
      self.log_path,
    )
  }
}