realizar 0.8.4

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


// ============================================================================
// SafetensorsModel::get_tensor_f32 boundary cases
// ============================================================================

#[test]
fn test_get_tensor_f32_exact_boundary() {
    // Tensor data exactly fills the data region
    let json = r#"{"w":{"dtype":"F32","shape":[3],"data_offsets":[0,12]}}"#;
    let json_bytes = json.as_bytes();
    let mut data = Vec::new();
    data.extend_from_slice(&(json_bytes.len() as u64).to_le_bytes());
    data.extend_from_slice(json_bytes);
    data.extend_from_slice(&1.0f32.to_le_bytes());
    data.extend_from_slice(&2.0f32.to_le_bytes());
    data.extend_from_slice(&3.0f32.to_le_bytes());

    let model = SafetensorsModel::from_bytes(&data).expect("parse");
    let values = model.get_tensor_f32("w").expect("get");
    assert_eq!(values, vec![1.0, 2.0, 3.0]);
}

#[test]
fn test_get_tensor_f32_special_values() {
    let json = r#"{"w":{"dtype":"F32","shape":[4],"data_offsets":[0,16]}}"#;
    let json_bytes = json.as_bytes();
    let mut data = Vec::new();
    data.extend_from_slice(&(json_bytes.len() as u64).to_le_bytes());
    data.extend_from_slice(json_bytes);
    data.extend_from_slice(&f32::INFINITY.to_le_bytes());
    data.extend_from_slice(&f32::NEG_INFINITY.to_le_bytes());
    data.extend_from_slice(&f32::NAN.to_le_bytes());
    data.extend_from_slice(&0.0f32.to_le_bytes());

    let model = SafetensorsModel::from_bytes(&data).expect("parse");
    let values = model.get_tensor_f32("w").expect("get");
    assert!(values[0].is_infinite() && values[0] > 0.0);
    assert!(values[1].is_infinite() && values[1] < 0.0);
    assert!(values[2].is_nan());
    assert_eq!(values[3], 0.0);
}

// ============================================================================
// SafetensorsConfig: deserialization from raw JSON
// ============================================================================

#[test]
fn test_config_deserialize_minimal() {
    let config: SafetensorsConfig = serde_json::from_str("{}").expect("parse");
    assert!(config.hidden_size.is_none());
    assert!(config.num_hidden_layers.is_none());
    assert!(config.vocab_size.is_none());
    assert_eq!(config.num_kv_heads(), 1); // all None -> default 1
    assert_eq!(config.architecture(), "unknown");
}

#[test]
fn test_config_deserialize_extra_fields_ignored() {
    let json = r#"{
        "hidden_size": 768,
        "unknown_field": "some_value",
        "another_unknown": 42
    }"#;
    let config: SafetensorsConfig = serde_json::from_str(json).expect("parse");
    assert_eq!(config.hidden_size, Some(768));
}

// ============================================================================
// MappedSafeTensorsModel: GH-213 truncated file detection
// ============================================================================

#[cfg(not(target_arch = "wasm32"))]
mod mapped_tests_part_02 {
    use super::*;
    use std::io::Write;

    #[test]
    fn test_mapped_truncated_tensor_data_detected_at_load() {
        // GH-213: File has valid header+metadata but tensor data is truncated
        let mut file = tempfile::NamedTempFile::new().expect("temp file");
        let json = r#"{"weight":{"dtype":"F32","shape":[1000],"data_offsets":[0,4000]}}"#;
        file.write_all(&(json.len() as u64).to_le_bytes())
            .expect("header");
        file.write_all(json.as_bytes()).expect("metadata");
        // Only write 100 bytes of tensor data instead of 4000
        file.write_all(&[0u8; 100]).expect("partial data");
        file.flush().expect("flush");

        let result = MappedSafeTensorsModel::load(file.path());
        assert!(result.is_err());
        let err = format!("{:?}", result.unwrap_err());
        assert!(
            err.contains("truncated"),
            "Expected truncation error, got: {}",
            err
        );
    }

    #[test]
    fn test_mapped_empty_file() {
        let mut file = tempfile::NamedTempFile::new().expect("temp file");
        file.write_all(&[]).expect("write nothing");
        file.flush().expect("flush");

        let result = MappedSafeTensorsModel::load(file.path());
        assert!(result.is_err());
        let err = format!("{:?}", result.unwrap_err());
        assert!(err.contains("too small") || err.contains("File too small"));
    }

    #[test]
    fn test_mapped_exactly_8_bytes_zero_metadata() {
        let mut file = tempfile::NamedTempFile::new().expect("temp file");
        file.write_all(&0u64.to_le_bytes()).expect("header");
        file.flush().expect("flush");

        let result = MappedSafeTensorsModel::load(file.path());
        // metadata_len=0 means json_bytes is empty -> parse error
        assert!(result.is_err());
    }

    #[test]
    fn test_mapped_valid_no_tensors() {
        let mut file = tempfile::NamedTempFile::new().expect("temp file");
        let json = b"{}";
        file.write_all(&(json.len() as u64).to_le_bytes())
            .expect("header");
        file.write_all(json).expect("metadata");
        file.flush().expect("flush");

        let model = MappedSafeTensorsModel::load(file.path()).expect("load");
        assert_eq!(model.tensor_count(), 0);
        assert!(model.tensor_names().is_empty());
        assert_eq!(model.file_size(), 8 + 2); // 8-byte header + 2-byte JSON
    }

    #[test]
    fn test_mapped_debug_format() {
        let mut file = tempfile::NamedTempFile::new().expect("temp file");
        let json = r#"{"w":{"dtype":"F32","shape":[1],"data_offsets":[0,4]}}"#;
        file.write_all(&(json.len() as u64).to_le_bytes())
            .expect("header");
        file.write_all(json.as_bytes()).expect("metadata");
        file.write_all(&1.0f32.to_le_bytes()).expect("data");
        file.flush().expect("flush");

        let model = MappedSafeTensorsModel::load(file.path()).expect("load");
        let debug = format!("{:?}", model);
        assert!(debug.contains("MappedSafeTensorsModel"));
    }

    #[test]
    fn test_mapped_has_tensor_and_get_info_consistency() {
        let mut file = tempfile::NamedTempFile::new().expect("temp file");
        let json = r#"{
            "a":{"dtype":"F32","shape":[2],"data_offsets":[0,8]},
            "b":{"dtype":"F16","shape":[4],"data_offsets":[8,16]},
            "c":{"dtype":"BF16","shape":[3],"data_offsets":[16,22]}
        }"#;
        file.write_all(&(json.len() as u64).to_le_bytes())
            .expect("header");
        file.write_all(json.as_bytes()).expect("metadata");
        file.write_all(&[0u8; 22]).expect("data");
        file.flush().expect("flush");

        let model = MappedSafeTensorsModel::load(file.path()).expect("load");

        // has_tensor and get_tensor_info should be consistent
        for name in ["a", "b", "c"] {
            assert!(model.has_tensor(name));
            let info = model.get_tensor_info(name).expect("info");
            assert_eq!(info.name, name);
        }
        assert!(!model.has_tensor("nonexistent"));
        assert!(model.get_tensor_info("nonexistent").is_none());
    }
}