realizar 0.8.6

Pure Rust ML inference engine built from scratch - model serving for GGUF and safetensors
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


#[cfg(test)]
#[cfg(feature = "cuda")]
mod tests {
    use super::*;

    fn create_executor() -> Option<CudaExecutor> {
        CudaExecutor::new(0).ok()
    }

    // ========================================================================
    // GEMM Tests (gemm, gemm_optimized, gemm_fused)
    // ========================================================================

    #[test]
    fn test_gemm_identity_matrix() {
        let Some(mut exec) = create_executor() else {
            return;
        };
        // 2x2 identity @ 2x2 vector
        let a = vec![1.0, 0.0, 0.0, 1.0];
        let b = vec![3.0, 4.0, 5.0, 6.0];
        let mut c = vec![0.0; 4];
        exec.gemm(&a, &b, &mut c, 2, 2, 2).unwrap();
        assert!((c[0] - 3.0).abs() < 0.1);
        assert!((c[3] - 6.0).abs() < 0.1);
    }

    #[test]
    fn test_gemm_small_matrix() {
        let Some(mut exec) = create_executor() else {
            return;
        };
        // A = [[1,2],[3,4]], B = [[5,6],[7,8]]
        // C = A@B = [[19,22],[43,50]]
        let a = vec![1.0, 2.0, 3.0, 4.0];
        let b = vec![5.0, 6.0, 7.0, 8.0];
        let mut c = vec![0.0; 4];
        exec.gemm(&a, &b, &mut c, 2, 2, 2).unwrap();
        // Tiled GEMM may have some numerical error
        assert!((c[0] - 19.0).abs() < 1.0);
        assert!((c[3] - 50.0).abs() < 1.0);
    }

    #[test]
    fn test_gemm_m1_uses_gemv() {
        let Some(mut exec) = create_executor() else {
            return;
        };
        // M=1 should use GEMV kernel path
        let a = vec![1.0, 2.0, 3.0, 4.0]; // 1x4
        let b = vec![1.0, 0.5, 0.25, 0.125, 2.0, 1.0, 0.5, 0.25]; // 4x2
        let mut c = vec![0.0; 2]; // 1x2
        exec.gemm(&a, &b, &mut c, 1, 2, 4).unwrap();
        // dot(a, b[:, 0]) = 1*1 + 2*0.5 + 3*0.25 + 4*0.125 = 1+1+0.75+0.5 = 3.25
        assert!(c[0] > 0.0);
    }

    #[test]
    fn test_gemm_size_mismatch_error() {
        let Some(mut exec) = create_executor() else {
            return;
        };
        let a = vec![1.0, 2.0]; // Wrong size
        let b = vec![1.0, 2.0, 3.0, 4.0];
        let mut c = vec![0.0; 4];
        let result = exec.gemm(&a, &b, &mut c, 2, 2, 2);
        assert!(result.is_err());
    }

    #[test]
    fn test_gemm_optimized_tile32() {
        let Some(mut exec) = create_executor() else {
            return;
        };
        let a = vec![1.0; 64]; // 8x8
        let b = vec![1.0; 64]; // 8x8
        let mut c = vec![0.0; 64]; // 8x8
        exec.gemm_optimized(&a, &b, &mut c, 8, 8, 8, 32).unwrap();
        // Each element should be sum of 8 ones = 8
        assert!(c[0] > 0.0);
    }

    #[test]
    fn test_gemm_fused_no_activation() {
        let Some(mut exec) = create_executor() else {
            return;
        };
        let a = vec![1.0; 4];
        let b = vec![1.0; 4];
        let mut c = vec![0.0; 4];
        exec.gemm_fused(&a, &b, None, &mut c, 2, 2, 2, 0).unwrap();
        assert!(c[0] > 0.0);
    }

    #[test]
    fn test_gemm_fused_with_relu() {
        let Some(mut exec) = create_executor() else {
            return;
        };
        let a = vec![1.0, -1.0, 1.0, -1.0]; // Some negative
        let b = vec![1.0; 4];
        let mut c = vec![0.0; 4];
        // activation=1 is ReLU
        exec.gemm_fused(&a, &b, None, &mut c, 2, 2, 2, 1).unwrap();
        // ReLU should clamp negatives to 0
        for val in &c {
            assert!(*val >= 0.0);
        }
    }

    #[test]
    fn test_gemm_fused_with_bias() {
        let Some(mut exec) = create_executor() else {
            return;
        };
        let a = vec![1.0; 4];
        let b = vec![1.0; 4];
        let bias = vec![10.0, 20.0];
        let mut c = vec![0.0; 4];
        exec.gemm_fused(&a, &b, Some(&bias), &mut c, 2, 2, 2, 0)
            .unwrap();
        // Bias should be added
        assert!(c[0] > 10.0);
    }

    #[test]
    fn test_gemm_fused_bias_size_mismatch() {
        let Some(mut exec) = create_executor() else {
            return;
        };
        let a = vec![1.0; 4];
        let b = vec![1.0; 4];
        let bias = vec![10.0]; // Wrong size, should be 2
        let mut c = vec![0.0; 4];
        let result = exec.gemm_fused(&a, &b, Some(&bias), &mut c, 2, 2, 2, 0);
        assert!(result.is_err());
    }

    // ========================================================================
    // Softmax Tests
    // ========================================================================

    #[test]
    fn test_softmax_uniform() {
        let Some(mut exec) = create_executor() else {
            return;
        };
        let mut data = vec![1.0, 1.0, 1.0, 1.0];
        exec.softmax(&mut data).unwrap();
        // Uniform input -> uniform output
        let expected = 0.25;
        for val in &data {
            assert!((*val - expected).abs() < 0.01);
        }
    }

    #[test]
    fn test_softmax_single_max() {
        let Some(mut exec) = create_executor() else {
            return;
        };
        let mut data = vec![10.0, 0.0, 0.0, 0.0];
        exec.softmax(&mut data).unwrap();
        // First element should dominate
        assert!(data[0] > 0.9);
        let sum: f32 = data.iter().sum();
        assert!((sum - 1.0).abs() < 0.01);
    }

    #[test]
    fn test_softmax_sum_to_one() {
        let Some(mut exec) = create_executor() else {
            return;
        };
        let mut data = vec![1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0];
        exec.softmax(&mut data).unwrap();
        let sum: f32 = data.iter().sum();
        assert!((sum - 1.0).abs() < 0.01);
    }

    #[test]
    fn test_softmax_large_values() {
        let Some(mut exec) = create_executor() else {
            return;
        };
        let mut data = vec![100.0, 101.0, 102.0, 103.0];
        exec.softmax(&mut data).unwrap();
        // Should still sum to 1 (numerically stable)
        let sum: f32 = data.iter().sum();
        assert!((sum - 1.0).abs() < 0.01);
    }

    // ========================================================================
    // Allocate Buffer Tests
    // ========================================================================

    #[test]
    fn test_allocate_buffer_basic() {
        let Some(exec) = create_executor() else {
            return;
        };
        let buf = exec.allocate_buffer(1024).unwrap();
        assert_eq!(buf.len(), 1024);
    }

    #[test]
    fn test_allocate_buffer_small() {
        let Some(exec) = create_executor() else {
            return;
        };
        let buf = exec.allocate_buffer(1).unwrap();
        assert_eq!(buf.len(), 1);
    }

    #[test]
    fn test_allocate_buffer_large() {
        let Some(exec) = create_executor() else {
            return;
        };
        let buf = exec.allocate_buffer(1024 * 1024).unwrap();
        assert_eq!(buf.len(), 1024 * 1024);
    }

    // ========================================================================
    // Synchronize Tests
    // ========================================================================

    #[test]
    fn test_synchronize_compute() {
        let Some(exec) = create_executor() else {
            return;
        };
        assert!(exec.synchronize_compute().is_ok());
    }

    #[test]
    fn test_synchronize_transfer() {
        let Some(exec) = create_executor() else {
            return;
        };
        assert!(exec.synchronize_transfer().is_ok());
    }

    #[test]
    fn test_synchronize_all() {
        let Some(exec) = create_executor() else {
            return;
        };
        assert!(exec.synchronize_all().is_ok());
    }

    // ========================================================================
    // Cached GEMM Tests
    // ========================================================================

    #[test]
    fn test_gemm_cached_weight_not_found() {
        let Some(mut exec) = create_executor() else {
            return;
        };
        let b = vec![1.0; 4];
        let mut c = vec![0.0; 4];
        let result = exec.gemm_cached("nonexistent_weight", &b, &mut c, 2, 2, 2);
        assert!(result.is_err());
    }

    #[test]
    fn test_gemm_b_cached_weight_not_found() {
        let Some(mut exec) = create_executor() else {
            return;
        };
        let a = vec![1.0; 4];
        let mut c = vec![0.0; 4];
        let result = exec.gemm_b_cached("nonexistent_weight", &a, &mut c, 2, 2, 2);
        assert!(result.is_err());
    }

    #[test]
    fn test_gemv_cached_weight_not_found() {
        let Some(mut exec) = create_executor() else {
            return;
        };
        let x = vec![1.0; 4];
        let mut y = vec![0.0; 4];
        let result = exec.gemv_cached("nonexistent_weight", &x, &mut y, 4, 4);
        assert!(result.is_err());
    }

    #[test]
    fn test_gemv_cached_input_size_mismatch() {
        let Some(mut exec) = create_executor() else {
            return;
        };
        let x = vec![1.0; 2]; // Wrong size
        let mut y = vec![0.0; 4];
        let result = exec.gemv_cached("test", &x, &mut y, 4, 4);
        assert!(result.is_err());
    }

    #[test]
    fn test_gemv_cached_output_size_mismatch() {
        let Some(mut exec) = create_executor() else {
            return;
        };
        let x = vec![1.0; 4];
        let mut y = vec![0.0; 2]; // Wrong size
        let result = exec.gemv_cached("test", &x, &mut y, 4, 4);
        assert!(result.is_err());
    }

    // ========================================================================
    // Async Copy Tests
    // ========================================================================

    #[test]
    fn test_async_copy_roundtrip() {
        let Some(exec) = create_executor() else {
            return;
        };
        let data = vec![1.0f32, 2.0, 3.0, 4.0];
        let mut buf = exec.allocate_buffer(4).unwrap();

        // Copy to GPU async
        // SAFETY: Preconditions verified by caller or enclosing context
        unsafe {
            exec.copy_to_gpu_async(&mut buf, &data).unwrap();
        }
        exec.synchronize_transfer().unwrap();

        // Copy back from GPU async
        let mut result = vec![0.0f32; 4];
        // SAFETY: Preconditions verified by caller or enclosing context
        unsafe {
            exec.copy_from_gpu_async(&buf, &mut result).unwrap();
        }
        exec.synchronize_transfer().unwrap();

        assert_eq!(result, data);
    }
}