boostr 0.1.0

ML framework built on numr - attention, quantization, model architectures
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

use super::*;
use crate::test_utils::cpu_setup;
use numr::runtime::cpu::CpuRuntime;
use std::io::Write;
use tempfile::NamedTempFile;

fn create_test_file() -> NamedTempFile {
    let mut file = NamedTempFile::new().unwrap();

    let header = serde_json::json!({
        "__metadata__": { "format": "pt" },
        "weight": {
            "dtype": "F32",
            "shape": [2, 3],
            "data_offsets": [0, 24]
        }
    });
    let header_str = header.to_string();
    let header_bytes = header_str.as_bytes();

    file.write_all(&(header_bytes.len() as u64).to_le_bytes())
        .unwrap();
    file.write_all(header_bytes).unwrap();

    for f in [1.0f32, 2.0, 3.0, 4.0, 5.0, 6.0] {
        file.write_all(&f.to_le_bytes()).unwrap();
    }
    file.flush().unwrap();
    file
}

#[test]
fn test_open_and_metadata() {
    let f = create_test_file();
    let st = SafeTensors::open(f.path()).unwrap();
    assert_eq!(st.len(), 1);
    assert_eq!(st.metadata().get("format"), Some(&"pt".to_string()));
}

#[test]
fn test_tensor_info() {
    let f = create_test_file();
    let st = SafeTensors::open(f.path()).unwrap();
    let info = st.tensor_info("weight").unwrap();
    assert_eq!(info.dtype, DType::F32);
    assert_eq!(info.shape, vec![2, 3]);
    assert_eq!(info.numel(), 6);
    assert_eq!(info.size_bytes(), 24);
}

#[test]
fn test_load_tensor_f32() {
    let (_, device) = cpu_setup();
    let f = create_test_file();
    let mut st = SafeTensors::open(f.path()).unwrap();
    let tensor = st.load_tensor::<CpuRuntime>("weight", &device).unwrap();
    assert_eq!(tensor.shape(), &[2, 3]);
    let data = tensor.to_vec::<f32>();
    assert!((data[0] - 1.0).abs() < 1e-6);
    assert!((data[5] - 6.0).abs() < 1e-6);
}

fn create_test_file_bf16() -> NamedTempFile {
    let mut file = NamedTempFile::new().unwrap();

    let header = serde_json::json!({
        "__metadata__": { "format": "pt" },
        "weight": {
            "dtype": "BF16",
            "shape": [2, 3],
            "data_offsets": [0, 12]
        }
    });
    let header_str = header.to_string();
    let header_bytes = header_str.as_bytes();

    file.write_all(&(header_bytes.len() as u64).to_le_bytes())
        .unwrap();
    file.write_all(header_bytes).unwrap();

    for f in [1.0f32, 2.0, 3.0, 4.0, 5.0, 6.0] {
        file.write_all(&half::bf16::from_f32(f).to_le_bytes())
            .unwrap();
    }
    file.flush().unwrap();
    file
}

#[test]
fn test_load_tensor_bf16() {
    let (_, device) = cpu_setup();
    let f = create_test_file_bf16();
    let mut st = SafeTensors::open(f.path()).unwrap();
    let tensor = st.load_tensor::<CpuRuntime>("weight", &device).unwrap();
    assert_eq!(tensor.shape(), &[2, 3]);
    assert_eq!(tensor.dtype(), DType::BF16);
    let data: Vec<half::bf16> = tensor.to_vec();
    assert!((data[0].to_f32() - 1.0).abs() < 1e-2);
    assert!((data[5].to_f32() - 6.0).abs() < 1e-2);
}

#[test]
fn test_tensor_not_found() {
    let f = create_test_file();
    let st = SafeTensors::open(f.path()).unwrap();
    assert!(st.tensor_info("nonexistent").is_err());
}

#[test]
fn test_save_and_load_roundtrip() {
    let (_, device) = cpu_setup();
    let tmp = NamedTempFile::new().unwrap();

    let mut tensors = HashMap::new();
    tensors.insert(
        "w1".to_string(),
        Tensor::<CpuRuntime>::from_slice(&[1.0f32, 2.0, 3.0, 4.0], &[2, 2], &device),
    );

    save_safetensors(tmp.path(), &tensors, None).unwrap();

    let mut loaded = SafeTensors::open(tmp.path()).unwrap();
    assert_eq!(loaded.len(), 1);
    let t = loaded.load_tensor::<CpuRuntime>("w1", &device).unwrap();
    assert_eq!(t.shape(), &[2, 2]);
    let data = t.to_vec::<f32>();
    assert!((data[0] - 1.0).abs() < 1e-6);
    assert!((data[3] - 4.0).abs() < 1e-6);
}