rustorch 0.6.29

Production-ready PyTorch-compatible deep learning library in Rust with special mathematical functions (gamma, Bessel, error functions), statistical distributions, Fourier transforms (FFT/RFFT), matrix decomposition (SVD/QR/LU/eigenvalue), automatic differentiation, neural networks, computer vision transforms, complete GPU acceleration (CUDA/Metal/OpenCL), SIMD optimizations, parallel processing, WebAssembly browser support, comprehensive distributed learning support, and performance validation
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
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
//! Numeric safety and overflow protection
//! 数値安全性とオーバーフロー保護

use crate::error::{RusTorchError, RusTorchResult};
use num_traits::Float;
use std::fmt::Debug;

/// Safe numeric operations with overflow detection
/// オーバーフロー検出付きの安全な数値演算
pub trait SafeNumeric: Float + Debug + Clone {
    /// Safe addition that detects overflow
    /// オーバーフローを検出する安全な加算
    fn safe_add(&self, other: Self) -> RusTorchResult<Self>;

    /// Safe multiplication that detects overflow
    /// オーバーフローを検出する安全な乗算
    fn safe_mul(&self, other: Self) -> RusTorchResult<Self>;

    /// Safe division with zero check
    /// ゼロチェック付きの安全な除算
    fn safe_div(&self, other: Self) -> RusTorchResult<Self>;

    /// Safe exponential with overflow check
    /// オーバーフローチェック付きの安全な指数関数
    fn safe_exp(&self) -> RusTorchResult<Self>;

    /// Check if value is finite (not NaN or infinite)
    /// 値が有限か確認(NaNや無限大でない)
    fn is_safe(&self) -> bool;

    /// Clamp value to safe range
    /// 値を安全な範囲にクランプ
    fn clamp_safe(&self) -> Self;
}

impl SafeNumeric for f32 {
    fn safe_add(&self, other: Self) -> RusTorchResult<Self> {
        let result = self + other;
        if result.is_finite() {
            Ok(result)
        } else {
            Err(RusTorchError::numeric("Addition overflow"))
        }
    }

    fn safe_mul(&self, other: Self) -> RusTorchResult<Self> {
        let result = self * other;
        if result.is_finite() {
            Ok(result)
        } else {
            Err(RusTorchError::numeric("Multiplication overflow"))
        }
    }

    fn safe_div(&self, other: Self) -> RusTorchResult<Self> {
        if other.abs() < Self::epsilon() {
            return Err(RusTorchError::numeric("Division by zero"));
        }

        let result = self / other;
        if result.is_finite() {
            Ok(result)
        } else {
            Err(RusTorchError::numeric("Division overflow"))
        }
    }

    fn safe_exp(&self) -> RusTorchResult<Self> {
        // Check for potential overflow before computing
        if *self > 88.0 {
            // exp(88) is close to f32::MAX
            return Err(RusTorchError::numeric("Numeric overflow"));
        }

        let result = self.exp();
        if result.is_finite() {
            Ok(result)
        } else {
            Err(RusTorchError::numeric("Exponential overflow"))
        }
    }

    fn is_safe(&self) -> bool {
        self.is_finite()
    }

    fn clamp_safe(&self) -> Self {
        if self.is_nan() {
            0.0
        } else if *self == Self::infinity() {
            <Self as num_traits::Bounded>::max_value()
        } else if *self == Self::neg_infinity() {
            <Self as num_traits::Bounded>::min_value()
        } else {
            *self
        }
    }
}

impl SafeNumeric for f64 {
    fn safe_add(&self, other: Self) -> RusTorchResult<Self> {
        let result = self + other;
        if result.is_finite() {
            Ok(result)
        } else {
            Err(RusTorchError::numeric("Addition overflow"))
        }
    }

    fn safe_mul(&self, other: Self) -> RusTorchResult<Self> {
        let result = self * other;
        if result.is_finite() {
            Ok(result)
        } else {
            Err(RusTorchError::numeric("Multiplication overflow"))
        }
    }

    fn safe_div(&self, other: Self) -> RusTorchResult<Self> {
        if other.abs() < Self::epsilon() {
            return Err(RusTorchError::numeric("Division by zero"));
        }

        let result = self / other;
        if result.is_finite() {
            Ok(result)
        } else {
            Err(RusTorchError::numeric("Division overflow"))
        }
    }

    fn safe_exp(&self) -> RusTorchResult<Self> {
        // Check for potential overflow before computing
        if *self > 709.0 {
            // exp(709) is close to f64::MAX
            return Err(RusTorchError::numeric("Exponential overflow"));
        }

        let result = self.exp();
        if result.is_finite() {
            Ok(result)
        } else {
            Err(RusTorchError::numeric("Exponential overflow"))
        }
    }

    fn is_safe(&self) -> bool {
        self.is_finite()
    }

    fn clamp_safe(&self) -> Self {
        if self.is_nan() {
            0.0
        } else if *self == Self::infinity() {
            <Self as num_traits::Bounded>::max_value()
        } else if *self == Self::neg_infinity() {
            <Self as num_traits::Bounded>::min_value()
        } else {
            *self
        }
    }
}

/// Safe tensor with numeric overflow protection
/// 数値オーバーフロー保護付きの安全テンソル
#[derive(Debug, Clone, PartialEq)]
pub struct SafeTensor<T: SafeNumeric> {
    data: Vec<T>,
    shape: Vec<usize>,
}

impl<T: SafeNumeric> SafeTensor<T> {
    /// Create new safe tensor with validation
    /// 検証付きで新しい安全テンソルを作成
    pub fn new(data: Vec<T>, shape: Vec<usize>) -> RusTorchResult<Self> {
        if data.len() != shape.iter().product::<usize>() {
            return Err(RusTorchError::shape_mismatch(
                &[data.len()],
                &[shape.iter().product::<usize>()],
            ));
        }

        // Validate all values are safe
        if !data.iter().all(|x| x.is_safe()) {
            return Err(RusTorchError::numeric(
                "Tensor contains unsafe values (NaN or infinity)",
            ));
        }

        Ok(Self { data, shape })
    }

    /// Get tensor shape
    /// テンソル形状を取得
    pub fn shape(&self) -> &[usize] {
        &self.shape
    }

    /// Get tensor data
    /// テンソルデータを取得
    pub fn data(&self) -> &[T] {
        &self.data
    }

    /// Safe element-wise addition
    /// 安全な要素ごとの加算
    pub fn safe_add(&self, other: &Self) -> RusTorchResult<Self> {
        if self.shape != other.shape {
            return Err(RusTorchError::shape_mismatch(&self.shape, &other.shape));
        }

        let mut result_data = Vec::with_capacity(self.data.len());
        for (a, b) in self.data.iter().zip(other.data.iter()) {
            result_data.push(a.safe_add(*b)?);
        }

        Ok(Self {
            data: result_data,
            shape: self.shape.clone(),
        })
    }

    /// Safe element-wise multiplication
    /// 安全な要素ごとの乗算
    pub fn safe_mul(&self, other: &Self) -> RusTorchResult<Self> {
        if self.shape != other.shape {
            return Err(RusTorchError::shape_mismatch(&self.shape, &other.shape));
        }

        let mut result_data = Vec::with_capacity(self.data.len());
        for (a, b) in self.data.iter().zip(other.data.iter()) {
            result_data.push(a.safe_mul(*b)?);
        }

        Ok(Self {
            data: result_data,
            shape: self.shape.clone(),
        })
    }

    /// Safe scalar multiplication
    /// 安全なスカラー乗算
    pub fn safe_mul_scalar(&self, scalar: T) -> RusTorchResult<Self> {
        let mut result_data = Vec::with_capacity(self.data.len());
        for value in self.data.iter() {
            result_data.push(value.safe_mul(scalar)?);
        }

        Ok(Self {
            data: result_data,
            shape: self.shape.clone(),
        })
    }

    /// Safe exponential function
    /// 安全な指数関数
    pub fn safe_exp(&self) -> RusTorchResult<Self> {
        let mut result_data = Vec::with_capacity(self.data.len());
        for value in self.data.iter() {
            result_data.push(value.safe_exp()?);
        }

        Ok(Self {
            data: result_data,
            shape: self.shape.clone(),
        })
    }

    /// Clamp all values to safe range
    /// すべての値を安全な範囲にクランプ
    pub fn clamp_safe(&self) -> Self {
        let clamped_data: Vec<T> = self.data.iter().map(|x| x.clamp_safe()).collect();

        Self {
            data: clamped_data,
            shape: self.shape.clone(),
        }
    }

    /// Check if all values are safe
    /// すべての値が安全か確認
    pub fn is_all_safe(&self) -> bool {
        self.data.iter().all(|x| x.is_safe())
    }

    /// Get statistics about unsafe values
    /// 安全でない値の統計を取得
    pub fn unsafe_stats(&self) -> UnsafeStats {
        let mut nan_count = 0;
        let mut inf_count = 0;
        let mut neg_inf_count = 0;

        for value in &self.data {
            if value.is_nan() {
                nan_count += 1;
            } else if value.is_infinite() {
                if value.is_sign_positive() {
                    inf_count += 1;
                } else {
                    neg_inf_count += 1;
                }
            }
        }

        UnsafeStats {
            total_elements: self.data.len(),
            nan_count,
            inf_count,
            neg_inf_count,
        }
    }
}

/// Statistics about unsafe numeric values
/// 安全でない数値の統計
#[derive(Debug, Clone, PartialEq)]
pub struct UnsafeStats {
    /// Total number of elements in the tensor
    /// テンソル内の要素の総数
    pub total_elements: usize,
    /// Number of NaN (Not-a-Number) values
    /// NaN(非数)値の数
    pub nan_count: usize,
    /// Number of positive infinity values
    /// 正の無限大値の数
    pub inf_count: usize,
    /// Number of negative infinity values
    /// 負の無限大値の数
    pub neg_inf_count: usize,
}

impl UnsafeStats {
    /// Check if tensor has any unsafe values
    /// テンソルに安全でない値があるか確認
    pub fn has_unsafe(&self) -> bool {
        self.nan_count > 0 || self.inf_count > 0 || self.neg_inf_count > 0
    }

    /// Get total unsafe value count
    /// 安全でない値の総数を取得
    pub fn unsafe_total(&self) -> usize {
        self.nan_count + self.inf_count + self.neg_inf_count
    }

    /// Get percentage of unsafe values
    /// 安全でない値の割合を取得
    pub fn unsafe_percentage(&self) -> f64 {
        if self.total_elements == 0 {
            0.0
        } else {
            (self.unsafe_total() as f64 / self.total_elements as f64) * 100.0
        }
    }
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn test_safe_operations_f32() {
        let a = 5.0f32;
        let b = 3.0f32;

        assert_eq!(a.safe_add(b).unwrap(), 8.0);
        assert_eq!(a.safe_mul(b).unwrap(), 15.0);
        assert_eq!(a.safe_div(b).unwrap(), 5.0 / 3.0);
    }

    #[test]
    fn test_overflow_detection() {
        let large = f32::MAX;
        let result = large.safe_add(large);
        assert!(result.is_err());

        if result.is_err() {
            // Expected error
        } else {
            panic!("Expected overflow error");
        }
    }

    #[test]
    fn test_division_by_zero() {
        let a = 5.0f32;
        let zero = 0.0f32;

        let result = a.safe_div(zero);
        assert!(result.is_err());

        if result.is_err() {
            // Expected error
        } else {
            panic!("Expected division by zero error");
        }
    }

    #[test]
    fn test_safe_tensor_creation() {
        let data = vec![1.0f32, 2.0, 3.0, 4.0];
        let shape = vec![2, 2];

        let tensor = SafeTensor::new(data, shape).unwrap();
        assert_eq!(tensor.shape(), &[2, 2]);
        assert_eq!(tensor.data(), &[1.0f32, 2.0, 3.0, 4.0]);
    }

    #[test]
    fn test_safe_tensor_operations() {
        let data1 = vec![1.0f32, 2.0, 3.0, 4.0];
        let data2 = vec![2.0f32, 3.0, 4.0, 5.0];
        let shape = vec![2, 2];

        let tensor1 = SafeTensor::new(data1, shape.clone()).unwrap();
        let tensor2 = SafeTensor::new(data2, shape).unwrap();

        let sum = tensor1.safe_add(&tensor2).unwrap();
        assert_eq!(sum.data(), &[3.0f32, 5.0, 7.0, 9.0]);

        let product = tensor1.safe_mul(&tensor2).unwrap();
        assert_eq!(product.data(), &[2.0f32, 6.0, 12.0, 20.0]);
    }

    #[test]
    fn test_unsafe_value_detection() {
        let data = vec![1.0f32, f32::NAN, f32::INFINITY, -f32::INFINITY, 2.0];
        let shape = vec![5];

        let tensor = SafeTensor::new(data.clone(), shape);
        assert!(tensor.is_err()); // Should fail due to NaN and infinity

        // Create tensor without validation for testing
        let unsafe_tensor = SafeTensor {
            data,
            shape: vec![5],
        };
        let stats = unsafe_tensor.unsafe_stats();

        assert_eq!(stats.nan_count, 1);
        assert_eq!(stats.inf_count, 1);
        assert_eq!(stats.neg_inf_count, 1);
        assert!(stats.has_unsafe());
        assert_eq!(stats.unsafe_percentage(), 60.0); // 3 out of 5
    }

    #[test]
    fn test_clamp_safe() {
        let nan = f32::NAN;
        let inf = f32::INFINITY;
        let neg_inf = f32::NEG_INFINITY;
        let normal = 5.0f32;

        assert_eq!(nan.clamp_safe(), 0.0);
        assert_eq!(inf.clamp_safe(), f32::MAX);
        assert_eq!(neg_inf.clamp_safe(), f32::MIN);
        assert_eq!(normal.clamp_safe(), 5.0);
    }

    #[test]
    fn test_exp_overflow_protection() {
        let large_value = 100.0f32; // exp(100) would overflow
        let result = large_value.safe_exp();
        assert!(result.is_err());

        let safe_value = 1.0f32;
        let result = safe_value.safe_exp();
        assert!(result.is_ok());
        assert!((result.unwrap() - 1.0f32.exp()).abs() < 1e-6);
    }
}