entrenar 0.7.13

Training & Optimization library with autograd, LoRA, quantization, and model merging
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
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270

//! Tests for evaluation metrics

use crate::Tensor;

use super::{Accuracy, F1Score, Metric, Precision, R2Score, Recall, MAE, RMSE};

#[test]
fn test_accuracy_perfect() {
    let metric = Accuracy::default();
    let pred = Tensor::from_vec(vec![0.9, 0.1, 0.8, 0.2], false);
    let target = Tensor::from_vec(vec![1.0, 0.0, 1.0, 0.0], false);

    let acc = metric.compute(&pred, &target);
    assert!((acc - 1.0).abs() < 1e-5);
}

#[test]
fn test_accuracy_half() {
    let metric = Accuracy::default();
    let pred = Tensor::from_vec(vec![0.9, 0.9, 0.1, 0.1], false);
    let target = Tensor::from_vec(vec![1.0, 0.0, 1.0, 0.0], false);

    let acc = metric.compute(&pred, &target);
    assert!((acc - 0.5).abs() < 1e-5);
}

#[test]
fn test_precision() {
    let metric = Precision::default();
    // 2 predicted positives (0.9, 0.8), 1 TP (0.9 -> 1.0)
    let pred = Tensor::from_vec(vec![0.9, 0.8, 0.2], false);
    let target = Tensor::from_vec(vec![1.0, 0.0, 0.0], false);

    let prec = metric.compute(&pred, &target);
    assert!((prec - 0.5).abs() < 1e-5);
}

#[test]
fn test_recall() {
    let metric = Recall::default();
    // 2 actual positives, 1 correctly predicted
    let pred = Tensor::from_vec(vec![0.9, 0.2, 0.8], false);
    let target = Tensor::from_vec(vec![1.0, 1.0, 0.0], false);

    let rec = metric.compute(&pred, &target);
    assert!((rec - 0.5).abs() < 1e-5);
}

#[test]
fn test_f1_score() {
    let metric = F1Score::default();
    let pred = Tensor::from_vec(vec![0.9, 0.8, 0.2, 0.1], false);
    let target = Tensor::from_vec(vec![1.0, 0.0, 1.0, 0.0], false);

    let f1 = metric.compute(&pred, &target);
    // Precision = 0.5 (1/2), Recall = 0.5 (1/2)
    // F1 = 2 * 0.5 * 0.5 / (0.5 + 0.5) = 0.5
    assert!((f1 - 0.5).abs() < 1e-5);
}

#[test]
fn test_f1_perfect() {
    let metric = F1Score::default();
    let pred = Tensor::from_vec(vec![0.9, 0.1], false);
    let target = Tensor::from_vec(vec![1.0, 0.0], false);

    let f1 = metric.compute(&pred, &target);
    assert!((f1 - 1.0).abs() < 1e-5);
}

#[test]
fn test_r2_perfect() {
    let metric = R2Score;
    let pred = Tensor::from_vec(vec![1.0, 2.0, 3.0], false);
    let target = Tensor::from_vec(vec![1.0, 2.0, 3.0], false);

    let r2 = metric.compute(&pred, &target);
    assert!((r2 - 1.0).abs() < 1e-5);
}

#[test]
fn test_r2_mean_prediction() {
    let metric = R2Score;
    // Predicting mean of targets
    let pred = Tensor::from_vec(vec![2.0, 2.0, 2.0], false);
    let target = Tensor::from_vec(vec![1.0, 2.0, 3.0], false);

    let r2 = metric.compute(&pred, &target);
    assert!(r2.abs() < 1e-5); // R² ≈ 0
}

#[test]
fn test_mae() {
    let metric = MAE;
    let pred = Tensor::from_vec(vec![1.0, 2.0, 3.0], false);
    let target = Tensor::from_vec(vec![1.5, 2.5, 3.5], false);

    let mae = metric.compute(&pred, &target);
    assert!((mae - 0.5).abs() < 1e-5);
}

#[test]
fn test_mae_perfect() {
    let metric = MAE;
    let pred = Tensor::from_vec(vec![1.0, 2.0, 3.0], false);
    let target = Tensor::from_vec(vec![1.0, 2.0, 3.0], false);

    let mae = metric.compute(&pred, &target);
    assert!(mae < 1e-5);
}

#[test]
fn test_rmse() {
    let metric = RMSE;
    let pred = Tensor::from_vec(vec![1.0, 2.0, 3.0], false);
    let target = Tensor::from_vec(vec![2.0, 3.0, 4.0], false);

    let rmse = metric.compute(&pred, &target);
    // MSE = mean([1, 1, 1]) = 1, RMSE = 1
    assert!((rmse - 1.0).abs() < 1e-5);
}

#[test]
fn test_rmse_perfect() {
    let metric = RMSE;
    let pred = Tensor::from_vec(vec![1.0, 2.0, 3.0], false);
    let target = Tensor::from_vec(vec![1.0, 2.0, 3.0], false);

    let rmse = metric.compute(&pred, &target);
    assert!(rmse < 1e-5);
}

#[test]
fn test_higher_is_better() {
    assert!(Accuracy::default().higher_is_better());
    assert!(Precision::default().higher_is_better());
    assert!(Recall::default().higher_is_better());
    assert!(F1Score::default().higher_is_better());
    assert!(R2Score.higher_is_better());
    assert!(!MAE.higher_is_better());
    assert!(!RMSE.higher_is_better());
}

#[test]
fn test_metric_names() {
    assert_eq!(Accuracy::default().name(), "Accuracy");
    assert_eq!(Precision::default().name(), "Precision");
    assert_eq!(Recall::default().name(), "Recall");
    assert_eq!(F1Score::default().name(), "F1");
    assert_eq!(R2Score.name(), "R²");
    assert_eq!(MAE.name(), "MAE");
    assert_eq!(RMSE.name(), "RMSE");
}

#[test]
fn test_empty_input() {
    let metric = Accuracy::default();
    let pred = Tensor::from_vec(vec![], false);
    let target = Tensor::from_vec(vec![], false);

    let acc = metric.compute(&pred, &target);
    assert_eq!(acc, 0.0);
}

#[test]
fn test_precision_no_predictions() {
    let metric = Precision::default();
    let pred = Tensor::from_vec(vec![0.1, 0.2, 0.3], false);
    let target = Tensor::from_vec(vec![1.0, 1.0, 1.0], false);

    let prec = metric.compute(&pred, &target);
    assert_eq!(prec, 0.0); // No positive predictions
}

#[test]
fn test_recall_no_positives() {
    let metric = Recall::default();
    let pred = Tensor::from_vec(vec![0.9, 0.8, 0.7], false);
    let target = Tensor::from_vec(vec![0.0, 0.0, 0.0], false);

    let rec = metric.compute(&pred, &target);
    assert_eq!(rec, 0.0); // No actual positives
}

#[test]
fn test_f1_zero_precision_and_recall() {
    let metric = F1Score::default();
    // Neither precision nor recall will be > 0
    let pred = Tensor::from_vec(vec![0.1, 0.2], false);
    let target = Tensor::from_vec(vec![0.0, 0.0], false);

    let f1 = metric.compute(&pred, &target);
    assert_eq!(f1, 0.0);
}

#[test]
fn test_metric_with_custom_threshold() {
    let metric = Accuracy::new(0.3);
    // Predictions > 0.3 are positive
    let pred = Tensor::from_vec(vec![0.4, 0.2, 0.35], false);
    let target = Tensor::from_vec(vec![1.0, 0.0, 1.0], false);

    let acc = metric.compute(&pred, &target);
    assert!((acc - 1.0).abs() < 1e-5); // 3/3 correct
}

#[test]
fn test_precision_with_custom_threshold() {
    let metric = Precision::new(0.3);
    let pred = Tensor::from_vec(vec![0.4, 0.2, 0.35], false);
    let target = Tensor::from_vec(vec![1.0, 0.0, 1.0], false);

    let prec = metric.compute(&pred, &target);
    assert!((prec - 1.0).abs() < 1e-5); // 2 true positives, 0 false positives
}

#[test]
fn test_recall_with_custom_threshold() {
    let metric = Recall::new(0.3);
    let pred = Tensor::from_vec(vec![0.4, 0.2, 0.35], false);
    let target = Tensor::from_vec(vec![1.0, 0.0, 1.0], false);

    let rec = metric.compute(&pred, &target);
    assert!((rec - 1.0).abs() < 1e-5); // 2 true positives out of 2 actual positives
}

#[test]
fn test_f1_with_custom_threshold() {
    let metric = F1Score::new(0.3);
    let pred = Tensor::from_vec(vec![0.4, 0.2, 0.35], false);
    let target = Tensor::from_vec(vec![1.0, 0.0, 1.0], false);

    let f1 = metric.compute(&pred, &target);
    assert!((f1 - 1.0).abs() < 1e-5); // Perfect precision and recall
}

#[test]
fn test_accuracy_default_threshold() {
    let metric = Accuracy::default_threshold();
    assert!((metric.threshold - 0.5).abs() < 1e-5);
}

#[test]
fn test_accuracy_clone() {
    let metric = Accuracy::new(0.5);
    let cloned = metric.clone();
    assert!((metric.threshold - cloned.threshold).abs() < 1e-5);
}

#[test]
fn test_precision_clone() {
    let metric = Precision::new(0.5);
    let cloned = metric.clone();
    assert!((metric.threshold - cloned.threshold).abs() < 1e-5);
}

#[test]
fn test_recall_clone() {
    let metric = Recall::new(0.5);
    let cloned = metric.clone();
    assert!((metric.threshold - cloned.threshold).abs() < 1e-5);
}

#[test]
fn test_f1_clone() {
    let metric = F1Score::new(0.5);
    let cloned = metric.clone();
    // F1 stores precision and recall internally
    assert_eq!(metric.name(), cloned.name());
}