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

use crate::{
    compute::{Kernel, StaticKernel},
    element::WgpuElement,
    kernel::{
        self, elemwise_workgroup,
        pool::{build_output_and_info_pool2d, build_pool2d_info},
        KernelSettings, StaticKernelSource, WORKGROUP_DEFAULT,
    },
    kernel_wgsl,
    ops::numeric::empty_device,
    tensor::WgpuTensor,
};

kernel_wgsl!(AvgPool2dRaw, "../../template/pool/avg_pool2d.wgsl");
kernel_wgsl!(
    AvgPool2dBackwardRaw,
    "../../template/pool/avg_pool2d_backward.wgsl"
);

struct AvgPool2dBackward<const COUNT_INCLUDE_PAD: bool>;
struct AvgPool2d<const COUNT_INCLUDE_PAD: bool>;

impl<const COUNT_INCLUDE_PAD: bool> StaticKernelSource for AvgPool2dBackward<COUNT_INCLUDE_PAD> {
    fn source() -> kernel::SourceTemplate {
        AvgPool2dBackwardRaw::source().register("count_include_pad", format!("{COUNT_INCLUDE_PAD}"))
    }
}

impl<const COUNT_INCLUDE_PAD: bool> StaticKernelSource for AvgPool2d<COUNT_INCLUDE_PAD> {
    fn source() -> kernel::SourceTemplate {
        AvgPool2dRaw::source().register("count_include_pad", format!("{COUNT_INCLUDE_PAD}"))
    }
}

pub(crate) fn avg_pool2d<E: WgpuElement>(
    x: WgpuTensor<E, 4>,
    kernel_size: [usize; 2],
    stride: [usize; 2],
    padding: [usize; 2],
    count_include_pad: bool,
) -> WgpuTensor<E, 4> {
    let (info_handle, output) =
        build_output_and_info_pool2d(&x, kernel_size, stride, padding, [1, 1]);

    let workgroup = elemwise_workgroup(output.shape.num_elements(), WORKGROUP_DEFAULT);
    let kernel: Box<dyn Kernel> = match count_include_pad {
        true => Box::new(StaticKernel::<
            KernelSettings<AvgPool2d<true>, E, i32, WORKGROUP_DEFAULT, WORKGROUP_DEFAULT, 1>,
        >::new(workgroup)),
        false => Box::new(StaticKernel::<
            KernelSettings<AvgPool2d<false>, E, i32, WORKGROUP_DEFAULT, WORKGROUP_DEFAULT, 1>,
        >::new(workgroup)),
    };

    x.client
        .execute(kernel, &[&x.handle, &output.handle, &info_handle]);

    output
}

pub(crate) fn avg_pool2d_backward<E: WgpuElement>(
    x: WgpuTensor<E, 4>,
    grad: WgpuTensor<E, 4>,
    kernel_size: [usize; 2],
    stride: [usize; 2],
    padding: [usize; 2],
    count_include_pad: bool,
) -> WgpuTensor<E, 4> {
    let grad = kernel::into_contiguous(grad);
    let output = empty_device(x.client.clone(), x.device.clone(), x.shape.clone());
    let info_handle = build_pool2d_info(&x, &grad, kernel_size, stride, padding, [1, 1]);
    let workgroup = elemwise_workgroup(output.shape.num_elements(), WORKGROUP_DEFAULT);

    let kernel: Box<dyn Kernel> = match count_include_pad {
        true => Box::new(StaticKernel::<
            KernelSettings<
                AvgPool2dBackward<true>,
                E,
                i32,
                WORKGROUP_DEFAULT,
                WORKGROUP_DEFAULT,
                1,
            >,
        >::new(workgroup)),
        false => Box::new(StaticKernel::<
            KernelSettings<
                AvgPool2dBackward<false>,
                E,
                i32,
                WORKGROUP_DEFAULT,
                WORKGROUP_DEFAULT,
                1,
            >,
        >::new(workgroup)),
    };

    x.client
        .execute(kernel, &[&grad.handle, &output.handle, &info_handle]);

    output
}

#[cfg(test)]
mod tests {
    use crate::tests::{ReferenceBackend, TestBackend};
    use burn_tensor::{backend::Backend, module, ops::ModuleOps, Distribution, Tensor};

    #[test]
    fn avg_pool2d_should_work_with_multiple_invocations() {
        let tensor = Tensor::<TestBackend, 4>::random([32, 32, 32, 32], Distribution::Default);
        let tensor_ref = Tensor::<ReferenceBackend, 4>::from_data(tensor.to_data());
        let kernel_size = [3, 4];
        let stride = [1, 2];
        let padding = [1, 2];
        let count_include_pad = true;

        let pooled = module::avg_pool2d(tensor, kernel_size, stride, padding, count_include_pad);
        let pooled_ref =
            module::avg_pool2d(tensor_ref, kernel_size, stride, padding, count_include_pad);

        pooled
            .into_data()
            .assert_approx_eq(&pooled_ref.into_data(), 3);
    }

    #[test]
    fn avg_pool2d_backward_should_work_with_multiple_invocations() {
        TestBackend::seed(0);
        ReferenceBackend::seed(0);
        let tensor = Tensor::<TestBackend, 4>::random([32, 32, 32, 32], Distribution::Default);
        let tensor_ref = Tensor::<ReferenceBackend, 4>::from_data(tensor.to_data());
        let kernel_size = [3, 3];
        let stride = [1, 1];
        let padding = [1, 1];
        let count_include_pad = true;

        let shape_out = module::avg_pool2d(
            tensor.clone(),
            kernel_size,
            stride,
            padding,
            count_include_pad,
        )
        .shape();
        let grad_output = Tensor::<TestBackend, 4>::random(shape_out, Distribution::Default);
        let grad_output_ref = Tensor::<ReferenceBackend, 4>::from_data(grad_output.to_data());

        let grad: Tensor<TestBackend, 4> =
            Tensor::from_primitive(TestBackend::avg_pool2d_backward(
                tensor.into_primitive(),
                grad_output.into_primitive(),
                kernel_size,
                stride,
                padding,
                count_include_pad,
            ));
        let grad_ref: Tensor<ReferenceBackend, 4> =
            Tensor::from_primitive(ReferenceBackend::avg_pool2d_backward(
                tensor_ref.into_primitive(),
                grad_output_ref.into_primitive(),
                kernel_size,
                stride,
                padding,
                count_include_pad,
            ));

        grad.into_data().assert_approx_eq(&grad_ref.into_data(), 3);
    }
}