aprender-serve 0.65.1

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


#[test]
fn test_generate_with_cache_with_temperature() {
    let config = create_test_config();
    let transformer = AprTransformer::new(config);
    let gen_config = GenerateConfig {
        max_tokens: 2,
        temperature: 1.0,
        ..Default::default()
    };

    let result = transformer.generate_with_cache(&[0], &gen_config);
    assert!(result.is_ok());
}

// ============================================================================
// Part 8: Utility Method Tests
// ============================================================================

#[test]
fn test_num_parameters() {
    let config = create_test_config();
    let transformer = AprTransformer::new(config);

    let params = transformer.num_parameters();
    assert!(params > 0);

    // Should include embedding, layers, and lm_head
    // Minimum: vocab_size * hidden_dim (embedding)
    assert!(params >= 100 * 64);
}

#[test]
fn test_memory_size() {
    let config = create_test_config();
    let transformer = AprTransformer::new(config);

    let size = transformer.memory_size();
    assert!(size > 0);

    // Memory size = num_parameters * 4 bytes
    assert_eq!(size, transformer.num_parameters() * 4);
}

#[test]
fn test_config_accessor() {
    let config = create_test_config();
    let transformer = AprTransformer::new(config.clone());

    let returned_config = transformer.config();
    assert_eq!(returned_config.hidden_dim, config.hidden_dim);
    assert_eq!(returned_config.num_layers, config.num_layers);
    assert_eq!(returned_config.vocab_size, config.vocab_size);
}

// ============================================================================
// Part 9: from_apr_bytes() Error Tests
// ============================================================================

#[test]
fn test_from_apr_bytes_too_small() {
    let data = vec![0u8; 32]; // Less than 64 bytes
    let result = AprTransformer::from_apr_bytes(&data);
    assert!(result.is_err());
}

#[test]
fn test_from_apr_bytes_invalid_magic() {
    let mut data = vec![0u8; 128];
    data[0..4].copy_from_slice(b"GGUF"); // Wrong magic

    let result = AprTransformer::from_apr_bytes(&data);
    assert!(result.is_err());
}

#[test]
fn test_from_apr_bytes_valid_magic_v1() {
    let mut data = vec![0u8; 128];
    data[0..4].copy_from_slice(b"APR\0"); // APR v1 magic

    // Set up minimal valid header
    // tensor_count at offset 8
    data[8..12].copy_from_slice(&0u32.to_le_bytes());
    // metadata_offset at offset 12
    data[12..20].copy_from_slice(&64u64.to_le_bytes());
    // metadata_size at offset 20
    data[20..24].copy_from_slice(&2u32.to_le_bytes());
    // tensor_index_offset at offset 24
    data[24..32].copy_from_slice(&66u64.to_le_bytes());
    // data_offset at offset 32
    data[32..40].copy_from_slice(&66u64.to_le_bytes());

    // Add minimal JSON metadata
    data[64] = b'{';
    data[65] = b'}';

    let result = AprTransformer::from_apr_bytes(&data);
    // Magic check passes (v1), but may fail on subsequent parsing (no tensors).
    // Key assertion: error is NOT about invalid magic.
    if let Err(ref e) = result {
        let msg = format!("{e}");
        assert!(
            !msg.contains("Invalid APR magic"),
            "APR v1 magic should be accepted, got: {msg}"
        );
    }
}

#[test]
fn test_from_apr_bytes_valid_magic_v2() {
    let mut data = vec![0u8; 128];
    data[0..4].copy_from_slice(b"APR2"); // APR v2 magic

    // Set up minimal header
    data[8..12].copy_from_slice(&0u32.to_le_bytes());
    data[12..20].copy_from_slice(&64u64.to_le_bytes());
    data[20..24].copy_from_slice(&2u32.to_le_bytes());
    data[24..32].copy_from_slice(&66u64.to_le_bytes());
    data[32..40].copy_from_slice(&66u64.to_le_bytes());

    // Minimal JSON
    data[64] = b'{';
    data[65] = b'}';

    let result = AprTransformer::from_apr_bytes(&data);
    // Magic check passes (v2), but may fail on subsequent parsing (no tensors).
    if let Err(ref e) = result {
        let msg = format!("{e}");
        assert!(
            !msg.contains("Invalid APR magic"),
            "APR v2 magic should be accepted, got: {msg}"
        );
    }
}

#[test]
fn test_from_apr_bytes_metadata_out_of_bounds() {
    let mut data = vec![0u8; 128];
    data[0..4].copy_from_slice(b"APR\0");

    // metadata_offset points beyond file
    data[12..20].copy_from_slice(&1000u64.to_le_bytes());
    data[20..24].copy_from_slice(&100u32.to_le_bytes());

    let result = AprTransformer::from_apr_bytes(&data);
    assert!(result.is_err());
}

// ============================================================================
// Part 10: Layer Tests
// ============================================================================

#[test]
fn test_layer_empty() {
    let layer = AprTransformerLayer::empty(64, 256);

    assert_eq!(layer.attn_norm_weight.len(), 64);
    assert_eq!(layer.qkv_weight.len(), 64 * 3 * 64);
    assert_eq!(layer.attn_output_weight.len(), 64 * 64);
    assert_eq!(layer.ffn_up_weight.len(), 64 * 256);
    assert_eq!(layer.ffn_down_weight.len(), 256 * 64);
}

#[test]
fn test_layer_empty_gqa() {
    let hidden_dim = 64;
    let num_heads = 8;
    let num_kv_heads = 2;
    let intermediate_dim = 256;

    let layer =
        AprTransformerLayer::empty_gqa(hidden_dim, num_heads, num_kv_heads, intermediate_dim);

    // QKV: Q (64) + K (16) + V (16) = 96 elements per hidden dim
    let head_dim = hidden_dim / num_heads; // 8
    let kv_dim = num_kv_heads * head_dim; // 16
    let qkv_out_dim = hidden_dim + 2 * kv_dim; // 64 + 32 = 96

    assert_eq!(layer.qkv_weight.len(), hidden_dim * qkv_out_dim);
}

#[test]
fn test_layer_num_parameters() {
    let layer = AprTransformerLayer::empty(64, 256);
    let params = layer.num_parameters();

    // Calculate expected
    let expected = 64  // attn_norm
        + 64 * 192  // qkv (3 * hidden)
        + 64 * 64   // attn_output
        + 64 * 256  // ffn_up
        + 256 * 64; // ffn_down

    assert_eq!(params, expected);
}

// ============================================================================
// Part 11: Q4K Layer Weights Tests
// ============================================================================

#[test]
fn test_q4k_layer_weights_default() {
    let weights = Q4KLayerWeights::default();

    assert!(weights.qkv_weight.is_none());
    assert!(weights.attn_q_weight.is_none());
    assert!(weights.attn_k_weight.is_none());
    assert!(weights.attn_v_weight.is_none());
    assert!(weights.attn_output_weight.is_none());
    assert!(weights.ffn_gate_weight.is_none());
    assert!(weights.ffn_up_weight.is_none());
    assert!(weights.ffn_down_weight.is_none());
}

#[test]
fn test_q4k_layer_weights_with_values() {
    let weights = Q4KLayerWeights {
        attn_q_weight: Some(vec![1, 2, 3]),
        attn_k_weight: Some(vec![4, 5, 6]),
        attn_v_weight: Some(vec![7, 8, 9]),
        ffn_down_weight: Some(vec![10, 11, 12]),
        ..Default::default()
    };

    assert!(weights.attn_q_weight.is_some());
    assert_eq!(weights.attn_q_weight.expect("test value should be present")[0], 1);
    assert!(weights.ffn_down_weight.is_some());
}

// ============================================================================
// Helper Functions
// ============================================================================

fn create_test_config() -> AprTransformerConfig {
    AprTransformerConfig {
        architecture: "test".to_string(),
        hidden_dim: 64,
        num_layers: 2,
        num_heads: 4,
        num_kv_heads: 4,
        vocab_size: 100,
        intermediate_dim: 256,
        context_length: 128,
        rope_theta: 10000.0,
        eps: 1e-5,
            eos_token_id: None,
    ..Default::default()
    }
}