torsh-backend 0.1.2

Backend abstraction layer for ToRSh
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
146
147
148
149

//! Rust bindings for neural network operation kernels

#![allow(unused_imports)]
#[allow(unused_imports)]
use crate::cuda::error::{CudaError, CudaResult};

/// Launch 2D convolution kernel
pub fn launch_conv2d_f32(
    input: *mut f32,
    weight: *mut f32,
    bias: *mut f32,
    output: *mut f32,
    batch_size: i32,
    in_channels: i32,
    out_channels: i32,
    input_height: i32,
    input_width: i32,
    kernel_height: i32,
    kernel_width: i32,
    pad_h: i32,
    pad_w: i32,
    stride_h: i32,
    stride_w: i32,
    dilation_h: i32,
    dilation_w: i32,
    stream: crate::cuda::CUstream,
) {
    unsafe {
        super::cuda_kernels::launch_conv2d_f32(
            input,
            weight,
            bias,
            output,
            batch_size,
            in_channels,
            out_channels,
            input_height,
            input_width,
            kernel_height,
            kernel_width,
            pad_h,
            pad_w,
            stride_h,
            stride_w,
            dilation_h,
            dilation_w,
            stream,
        );
    }
}

/// Launch 2D max pooling kernel
pub fn launch_maxpool2d_f32(
    input: *mut f32,
    output: *mut f32,
    batch_size: i32,
    channels: i32,
    input_height: i32,
    input_width: i32,
    output_height: i32,
    output_width: i32,
    kernel_height: i32,
    kernel_width: i32,
    pad_h: i32,
    pad_w: i32,
    stride_h: i32,
    stride_w: i32,
    stream: crate::cuda::CUstream,
) {
    unsafe {
        super::cuda_kernels::launch_maxpool2d_f32(
            input,
            output,
            batch_size,
            channels,
            input_height,
            input_width,
            output_height,
            output_width,
            kernel_height,
            kernel_width,
            pad_h,
            pad_w,
            stride_h,
            stride_w,
            stream,
        );
    }
}

/// Launch 2D batch normalization kernel
pub fn launch_batchnorm2d_f32(
    input: *mut f32,
    output: *mut f32,
    weight: *mut f32,
    bias: *mut f32,
    running_mean: *mut f32,
    running_var: *mut f32,
    batch_size: i32,
    channels: i32,
    height: i32,
    width: i32,
    eps: f32,
    momentum: f32,
    training: bool,
    stream: crate::cuda::CUstream,
) {
    unsafe {
        super::cuda_kernels::launch_batchnorm2d_f32(
            input,
            output,
            weight,
            bias,
            running_mean,
            running_var,
            batch_size,
            channels,
            height,
            width,
            eps,
            momentum,
            training,
            stream,
        );
    }
}

/// Launch softmax kernel
pub fn launch_softmax_f32(
    input: *mut f32,
    output: *mut f32,
    batch_size: i32,
    classes: i32,
    stream: crate::cuda::CUstream,
) {
    unsafe {
        super::cuda_kernels::launch_softmax_f32(input, output, batch_size, classes, stream);
    }
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn test_neural_kernel_bindings_exist() {
        // Verify the functions are properly linked
    }
}