aprender-core 0.29.2

Next-generation machine learning library in pure Rust
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

use crate::serialization::safetensors::{save_safetensors, MappedSafeTensors, SafeTensorsDType};
use std::collections::BTreeMap;
use std::fs;

#[test]
fn test_get_tensor_raw_tensor_not_found() {
    let path = "/tmp/test_get_tensor_raw_not_found.safetensors";

    let mut tensors = BTreeMap::new();
    tensors.insert("weight".to_string(), (vec![1.0f32], vec![1]));
    save_safetensors(path, &tensors).expect("save");

    let mapped = MappedSafeTensors::open(path).expect("open");
    let result = mapped.get_tensor_raw("nonexistent");

    assert!(result.is_err());
    let err = result.unwrap_err();
    assert!(
        err.contains("not found"),
        "Error should mention 'not found': {err}"
    );
    assert!(
        err.contains("nonexistent"),
        "Error should mention tensor name: {err}"
    );

    fs::remove_file(path).ok();
}

#[test]
fn test_get_tensor_raw_f16_dtype() {
    // Build a SafeTensors file with F16 data manually
    let path = "/tmp/test_get_tensor_raw_f16.safetensors";

    // F16 representation of 1.0 = 0x3C00
    let f16_data: Vec<u8> = vec![0x00, 0x3C, 0x00, 0x40]; // two f16 values: 1.0, 2.0

    let metadata_json = serde_json::json!({
        "tensor_f16": {
            "dtype": "F16",
            "shape": [2],
            "data_offsets": [0, 4]
        }
    });
    let metadata_str = serde_json::to_string(&metadata_json).expect("serialize");
    let metadata_bytes = metadata_str.as_bytes();

    let mut file_bytes = Vec::new();
    file_bytes.extend_from_slice(&(metadata_bytes.len() as u64).to_le_bytes());
    file_bytes.extend_from_slice(metadata_bytes);
    file_bytes.extend_from_slice(&f16_data);

    fs::write(path, &file_bytes).expect("write");

    let mapped = MappedSafeTensors::open(path).expect("open");
    let raw = mapped.get_tensor_raw("tensor_f16").expect("get raw");

    assert!(matches!(raw.dtype, SafeTensorsDType::F16));
    assert_eq!(raw.shape, vec![2]);
    assert_eq!(raw.bytes.len(), 4);

    fs::remove_file(path).ok();
}

#[test]
fn test_get_tensor_raw_bf16_dtype() {
    let path = "/tmp/test_get_tensor_raw_bf16.safetensors";

    // BF16 representation of 1.0 = 0x3F80
    let bf16_data: Vec<u8> = vec![0x80, 0x3F]; // one BF16 value

    let metadata_json = serde_json::json!({
        "tensor_bf16": {
            "dtype": "BF16",
            "shape": [1],
            "data_offsets": [0, 2]
        }
    });
    let metadata_str = serde_json::to_string(&metadata_json).expect("serialize");
    let metadata_bytes = metadata_str.as_bytes();

    let mut file_bytes = Vec::new();
    file_bytes.extend_from_slice(&(metadata_bytes.len() as u64).to_le_bytes());
    file_bytes.extend_from_slice(metadata_bytes);
    file_bytes.extend_from_slice(&bf16_data);

    fs::write(path, &file_bytes).expect("write");

    let mapped = MappedSafeTensors::open(path).expect("open");
    let raw = mapped.get_tensor_raw("tensor_bf16").expect("get raw");

    assert!(matches!(raw.dtype, SafeTensorsDType::BF16));
    assert_eq!(raw.shape, vec![1]);
    assert_eq!(raw.bytes.len(), 2);

    fs::remove_file(path).ok();
}

#[test]
fn test_get_tensor_raw_unsupported_dtype() {
    let path = "/tmp/test_get_tensor_raw_unsupported.safetensors";

    // Create a SafeTensors file with an unsupported dtype
    let data = vec![0u8; 8];

    let metadata_json = serde_json::json!({
        "tensor_i32": {
            "dtype": "I32",
            "shape": [2],
            "data_offsets": [0, 8]
        }
    });
    let metadata_str = serde_json::to_string(&metadata_json).expect("serialize");
    let metadata_bytes = metadata_str.as_bytes();

    let mut file_bytes = Vec::new();
    file_bytes.extend_from_slice(&(metadata_bytes.len() as u64).to_le_bytes());
    file_bytes.extend_from_slice(metadata_bytes);
    file_bytes.extend_from_slice(&data);

    fs::write(path, &file_bytes).expect("write");

    let mapped = MappedSafeTensors::open(path).expect("open");
    let result = mapped.get_tensor_raw("tensor_i32");

    assert!(result.is_err());
    let err = result.unwrap_err();
    assert!(
        err.contains("Unsupported dtype"),
        "Error should mention unsupported: {err}"
    );

    fs::remove_file(path).ok();
}

#[test]
fn test_get_tensor_raw_multiple_tensors() {
    let path = "/tmp/test_get_tensor_raw_multi.safetensors";

    let mut tensors = BTreeMap::new();
    tensors.insert("alpha".to_string(), (vec![1.0f32, 2.0], vec![2]));
    tensors.insert("beta".to_string(), (vec![3.0f32, 4.0, 5.0], vec![3]));
    save_safetensors(path, &tensors).expect("save");

    let mapped = MappedSafeTensors::open(path).expect("open");

    let alpha_raw = mapped.get_tensor_raw("alpha").expect("get alpha");
    assert!(matches!(alpha_raw.dtype, SafeTensorsDType::F32));
    assert_eq!(alpha_raw.shape, vec![2]);
    let alpha_f32 = alpha_raw.to_f32().expect("convert");
    assert_eq!(alpha_f32, vec![1.0, 2.0]);

    let beta_raw = mapped.get_tensor_raw("beta").expect("get beta");
    assert!(matches!(beta_raw.dtype, SafeTensorsDType::F32));
    assert_eq!(beta_raw.shape, vec![3]);
    let beta_f32 = beta_raw.to_f32().expect("convert");
    assert_eq!(beta_f32, vec![3.0, 4.0, 5.0]);

    fs::remove_file(path).ok();
}