latexsnipper-runtime 2.0.0

Runtime abstraction — Session, Provider, ModelHandle
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
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145

/// Platform detection and ONNX Runtime library selection.
/// Detected platform type.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum Platform {
    WindowsX64,
    WindowsArm64,
    LinuxX64,
    LinuxAarch64,
    MacOsArm64,
    Unknown,
}

/// Detected acceleration capability.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum Acceleration {
    CpuOnly,
    Cuda12,
    Cuda13,
    Directml,
    Tensorrt,
}

impl Platform {
    /// Detect current platform at compile time.
    pub fn detect() -> Self {
        #[cfg(all(target_os = "windows", target_arch = "x86_64"))]
        {
            Platform::WindowsX64
        }
        #[cfg(all(target_os = "windows", target_arch = "aarch64"))]
        {
            Platform::WindowsArm64
        }
        #[cfg(all(target_os = "linux", target_arch = "x86_64"))]
        {
            Platform::LinuxX64
        }
        #[cfg(all(target_os = "linux", target_arch = "aarch64"))]
        {
            Platform::LinuxAarch64
        }
        #[cfg(all(target_os = "macos", target_arch = "aarch64"))]
        {
            Platform::MacOsArm64
        }
        #[cfg(not(any(
            all(target_os = "windows", target_arch = "x86_64"),
            all(target_os = "windows", target_arch = "aarch64"),
            all(target_os = "linux", target_arch = "x86_64"),
            all(target_os = "linux", target_arch = "aarch64"),
            all(target_os = "macos", target_arch = "aarch64"),
        )))]
        {
            Platform::Unknown
        }
    }

    /// Get the ONNX Runtime download URL for this platform.
    pub fn onnx_url(&self, version: &str, accel: Acceleration) -> String {
        let base = format!(
            "https://github.com/microsoft/onnxruntime/releases/download/v{}",
            version
        );
        match (self, accel) {
            (Platform::WindowsX64, Acceleration::CpuOnly) => {
                format!("{}/onnxruntime-win-x64-{}.zip", base, version)
            }
            (Platform::WindowsX64, Acceleration::Cuda12) => {
                format!("{}/onnxruntime-win-x64-gpu_cuda12-{}.zip", base, version)
            }
            (Platform::WindowsX64, Acceleration::Cuda13) => {
                format!("{}/onnxruntime-win-x64-gpu_cuda13-{}.zip", base, version)
            }
            (Platform::WindowsX64, Acceleration::Directml) => {
                format!("{}/onnxruntime-win-x64-{}.zip", base, version)
            }
            (Platform::WindowsX64, Acceleration::Tensorrt) => {
                format!("{}/onnxruntime-win-x64-gpu_cuda12-{}.zip", base, version)
            } // TensorRT uses CUDA backend
            (Platform::WindowsArm64, _) => {
                format!("{}/onnxruntime-win-arm64-{}.zip", base, version)
            }
            (Platform::LinuxX64, Acceleration::CpuOnly) => {
                format!("{}/onnxruntime-linux-x64-{}.tgz", base, version)
            }
            (Platform::LinuxX64, Acceleration::Cuda12) => {
                format!("{}/onnxruntime-linux-x64-gpu_cuda12-{}.tgz", base, version)
            }
            (Platform::LinuxX64, Acceleration::Cuda13) => {
                format!("{}/onnxruntime-linux-x64-gpu_cuda13-{}.tgz", base, version)
            }
            (Platform::LinuxX64, Acceleration::Directml) => {
                format!("{}/onnxruntime-linux-x64-{}.tgz", base, version)
            }
            (Platform::LinuxX64, Acceleration::Tensorrt) => {
                format!("{}/onnxruntime-linux-x64-gpu_cuda12-{}.tgz", base, version)
            } // TensorRT uses CUDA backend
            (Platform::LinuxAarch64, _) => {
                format!("{}/onnxruntime-linux-aarch64-{}.tgz", base, version)
            }
            (Platform::MacOsArm64, _) => format!("{}/onnxruntime-osx-arm64-{}.tgz", base, version),
            (Platform::Unknown, _) => format!("{}/onnxruntime-win-x64-{}.zip", base, version),
        }
    }

    /// Check if GPU acceleration is available on this platform.
    pub fn detect_gpu() -> Acceleration {
        // Check for TensorRT (highest priority on NVIDIA)
        if (cfg!(target_os = "windows") || cfg!(target_os = "linux"))
            && (std::env::var("TENSORRT_PATH").is_ok()
                || std::path::Path::new("/usr/local/tensorrt").exists()
                || std::path::Path::new("C:\\Program Files\\NVIDIA TensorRT").exists())
        {
            return Acceleration::Tensorrt;
        }

        // Check for CUDA
        if (cfg!(target_os = "windows") || cfg!(target_os = "linux"))
            && (std::env::var("CUDA_PATH").is_ok()
                || std::path::Path::new("/usr/local/cuda").exists())
        {
            return Acceleration::Cuda12;
        }

        // Check for DirectML on Windows
        if cfg!(target_os = "windows") {
            return Acceleration::Directml;
        }

        Acceleration::CpuOnly
    }
}

impl std::fmt::Display for Platform {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        match self {
            Platform::WindowsX64 => write!(f, "win-x64"),
            Platform::WindowsArm64 => write!(f, "win-arm64"),
            Platform::LinuxX64 => write!(f, "linux-x64"),
            Platform::LinuxAarch64 => write!(f, "linux-aarch64"),
            Platform::MacOsArm64 => write!(f, "osx-arm64"),
            Platform::Unknown => write!(f, "unknown"),
        }
    }
}