File: deform_conv2d_kernel.cpp

package info (click to toggle)
pytorch-vision 0.21.0-3
  • links: PTS, VCS
  • area: main
  • in suites: forky, sid, trixie
  • size: 20,228 kB
  • sloc: python: 65,904; cpp: 11,406; ansic: 2,459; java: 550; sh: 265; xml: 79; objc: 56; makefile: 33
file content (55 lines) | stat: -rw-r--r-- 1,399 bytes parent folder | download | duplicates (2)
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
#include "../deform_conv2d.h"

#include <ATen/autocast_mode.h>
#include <torch/library.h>
#include <torch/types.h>

namespace vision {
namespace ops {

namespace {

at::Tensor deform_conv2d_autocast(
    const at::Tensor& input,
    const at::Tensor& weight,
    const at::Tensor& offset,
    const at::Tensor& mask,
    const at::Tensor& bias,
    int64_t stride_h,
    int64_t stride_w,
    int64_t pad_h,
    int64_t pad_w,
    int64_t dilation_h,
    int64_t dilation_w,
    int64_t groups,
    int64_t offset_groups,
    bool use_mask) {
  c10::impl::ExcludeDispatchKeyGuard no_autocast(c10::DispatchKey::Autocast);
  return deform_conv2d(
             at::autocast::cached_cast(at::kFloat, input),
             at::autocast::cached_cast(at::kFloat, weight),
             at::autocast::cached_cast(at::kFloat, offset),
             at::autocast::cached_cast(at::kFloat, mask),
             at::autocast::cached_cast(at::kFloat, bias),
             stride_h,
             stride_w,
             pad_h,
             pad_w,
             dilation_h,
             dilation_w,
             groups,
             offset_groups,
             use_mask)
      .to(input.scalar_type());
}

} // namespace

TORCH_LIBRARY_IMPL(torchvision, Autocast, m) {
  m.impl(
      TORCH_SELECTIVE_NAME("torchvision::deform_conv2d"),
      TORCH_FN(deform_conv2d_autocast));
}

} // namespace ops
} // namespace vision