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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290
|
#include <ATen/ATen.h>
#include <torch/library.h>
#include "../../cpu/roi_align_common.h"
namespace vision {
namespace ops {
namespace {
// BEGIN copy-pasted code from pytorch core
// https://github.com/pytorch/pytorch/blob/master/aten/src/ATen/native/quantized/affine_quantizer_base.cpp
// We're vendoring the quantize_val() and dequantize_val() functions here. The
// reason is that these functions belong in at::native, which is incompatible
// with android xplat support.
// FIXME: Remove this section once we can use at::native for android xplat
// builds, or when quantize_val() and dequantize_val() aren't in at::native
#ifdef USE_FBGEMM
template <typename T>
T quantize_val(double scale, int64_t zero_point, float value) {
// Internally, fbgemm::Quantize uses std::nearbyint.
// std::nearbyint results in nearest integer value according to the current
// rounding mode and the default rounding mode is rounds to even in half-way
// cases in most popular processor architectures like x86 and ARM. This is
// typically faster than an alternatives like std::round that rounds half-way
// cases away from zero, and can be consistent with SIMD implementations for
// example in x86 using _mm512_cvtps_epi32 or mm512_round_ps with
// _MM_FROUND_CUR_DIRECTION option that also follow the current rounding mode.
// NOLINTNEXTLINE(cppcoreguidelines-init-variables)
int32_t qvalue;
// NOLINTNEXTLINE(bugprone-signed-char-misuse)
qvalue = fbgemm::Quantize<typename T::underlying, false /*LEGACY*/>(
value,
static_cast<int32_t>(zero_point),
static_cast<float>(scale),
/*result_precision=*/CHAR_BIT * sizeof(typename T::underlying));
return static_cast<T>(qvalue);
}
template <typename T>
inline float dequantize_val(double scale, int64_t zero_point, T value) {
// NOLINTNEXTLINE(cppcoreguidelines-pro-type-member-init)
fbgemm::TensorQuantizationParams qparams;
qparams.scale = static_cast<float>(scale);
qparams.zero_point = static_cast<int32_t>(zero_point);
return fbgemm::Dequantize<typename T::underlying>(value.val_, qparams);
}
#else // USE_FBGEMM
#if defined(__ANDROID__) && !defined(__NDK_MAJOR__)
template <class T>
inline float Round(const float x) {
return ::nearbyintf(x);
}
inline double Round(const double x) {
return ::nearbyint(x);
}
#else
template <class T>
inline T Round(const T x) {
return std::nearbyint(x);
}
#endif
template <typename T>
T quantize_val(double scale, int64_t zero_point, float value) {
// std::nearbyint results in nearest integer value according to the current
// rounding mode and the default rounding mode is rounds to even in half-way
// cases in most popular processor architectures like x86 and ARM. This is
// typically faster than an alternatives like std::round that rounds half-way
// cases away from zero, and can be consistent with SIMD implementations for
// example in x86 using _mm512_cvtps_epi32 or mm512_round_ps with
// _MM_FROUND_CUR_DIRECTION option that also follow the current rounding mode.
int64_t qvalue;
constexpr int64_t qmin = std::numeric_limits<typename T::underlying>::min();
constexpr int64_t qmax = std::numeric_limits<typename T::underlying>::max();
float inv_scale = 1.0f / static_cast<float>(scale);
qvalue = static_cast<int64_t>(zero_point + Round(value * inv_scale));
qvalue = std::max<int64_t>(qvalue, qmin);
qvalue = std::min<int64_t>(qvalue, qmax);
return static_cast<T>(qvalue);
}
template <typename T>
float dequantize_val(double scale, int64_t zero_point, T value) {
// We need to convert the qint8 value to float to ensure the subtraction
// subexpression returns a float
return (static_cast<float>(value.val_) - zero_point) * scale;
}
#endif // USE_FBGEMM
// END copy-pasted code from pytorch core
template <typename T>
void qroi_align_forward_kernel_impl(
int n_rois,
const at::Tensor& t_input,
const float& spatial_scale,
int channels,
int height,
int width,
int pooled_height,
int pooled_width,
int sampling_ratio,
bool aligned,
const at::Tensor& t_rois,
T* output) {
// Don't delete these otherwise the .data_ptr() data might be undefined
auto t_input_cont = t_input.contiguous();
auto t_rois_cont = t_rois.contiguous();
const T* input = t_input_cont.data_ptr<T>();
int64_t input_zp = t_input.q_zero_point();
float input_scale = t_input.q_scale();
const T* rois = t_rois_cont.data_ptr<T>();
int64_t rois_zp = t_rois.q_zero_point();
float rois_scale = t_rois.q_scale();
for (int n = 0; n < n_rois; n++) {
int index_n = n * channels * pooled_width * pooled_height;
const T* offset_rois = rois + n * 5;
// FIXME: change this when batches of size > 1 are allowed
const int roi_batch_ind = 0;
// Do not using rounding; this implementation detail is critical
float offset = aligned ? 0.5 : 0.;
float roi_start_w =
dequantize_val(rois_scale, rois_zp, offset_rois[1]) * spatial_scale -
offset;
float roi_start_h =
dequantize_val(rois_scale, rois_zp, offset_rois[2]) * spatial_scale -
offset;
float roi_end_w =
dequantize_val(rois_scale, rois_zp, offset_rois[3]) * spatial_scale -
offset;
float roi_end_h =
dequantize_val(rois_scale, rois_zp, offset_rois[4]) * spatial_scale -
offset;
float roi_width = roi_end_w - roi_start_w;
float roi_height = roi_end_h - roi_start_h;
if (!aligned) {
// Force malformed ROIs to be 1x1
roi_width = std::max(roi_width, 1.f);
roi_height = std::max(roi_height, 1.f);
}
float bin_size_h = roi_height / pooled_height;
float bin_size_w = roi_width / pooled_width;
// We use roi_bin_grid to sample the grid and mimic integral
int roi_bin_grid_h = (sampling_ratio > 0)
? sampling_ratio
: ceil(roi_height / pooled_height); // e.g., = 2
int roi_bin_grid_w =
(sampling_ratio > 0) ? sampling_ratio : ceil(roi_width / pooled_width);
// We do average (integral) pooling inside a bin
// When the grid is empty, output zeros.
const float count =
std::max(roi_bin_grid_h * roi_bin_grid_w, 1); // e.g. = 4
// we want to precalculate indices and weights shared by all channels,
// this is the key point of optimization
std::vector<detail::PreCalc<float>> pre_calc(
roi_bin_grid_h * roi_bin_grid_w * pooled_width * pooled_height);
detail::pre_calc_for_bilinear_interpolate(
height,
width,
pooled_height,
pooled_width,
roi_start_h,
roi_start_w,
bin_size_h,
bin_size_w,
roi_bin_grid_h,
roi_bin_grid_w,
pre_calc);
for (int c = 0; c < channels; c++) {
int index_n_c = index_n + c * pooled_width * pooled_height;
const T* offset_input =
input + (roi_batch_ind * channels + c) * height * width;
int pre_calc_index = 0;
for (int ph = 0; ph < pooled_height; ph++) {
for (int pw = 0; pw < pooled_width; pw++) {
int index = index_n_c + ph * pooled_width + pw;
float output_val = 0.;
float sum_w = 0.;
for (int iy = 0; iy < roi_bin_grid_h; iy++) {
for (int ix = 0; ix < roi_bin_grid_w; ix++) {
detail::PreCalc<float> pc = pre_calc[pre_calc_index];
// Optimization: we use the raw values here and we'll dequantize
// later
output_val += pc.w1 * offset_input[pc.pos1].val_ +
pc.w2 * offset_input[pc.pos2].val_ +
pc.w3 * offset_input[pc.pos3].val_ +
pc.w4 * offset_input[pc.pos4].val_;
sum_w += pc.w1 + pc.w2 + pc.w3 + pc.w4;
pre_calc_index += 1;
}
}
// Dequantize here
output_val = input_scale * (output_val - (float)input_zp * sum_w);
output_val /= count; // Average pooling
output[index] = quantize_val<T>(input_scale, input_zp, output_val);
} // for pw
} // for ph
} // for c
} // for n
}
at::Tensor qroi_align_forward_kernel(
const at::Tensor& input,
const at::Tensor& rois,
double spatial_scale,
int64_t pooled_height,
int64_t pooled_width,
int64_t sampling_ratio,
bool aligned) {
TORCH_CHECK(input.device().is_cpu(), "input must be a CPU tensor");
TORCH_CHECK(rois.device().is_cpu(), "rois must be a CPU tensor");
TORCH_CHECK(rois.size(1) == 5, "rois must have shape as Tensor[K, 5]");
// The first column of the RoI tensor is an image index, but not all indices
// are representable depending on the quantization. For example 1, 3, 5...
// indices can't be represented when qscale is 2. To prevent any bug, we force
// a batch size of 1 and we ignore the first column
TORCH_CHECK(
input.size(0) == 1,
"Only one image per batch is allowed in roi_align when quantized tensors are passed.");
at::TensorArg input_t{input, "input", 1}, rois_t{rois, "rois", 2};
at::CheckedFrom c = "qroi_align_forward_kernel";
at::checkAllSameType(c, {input_t, rois_t});
auto num_rois = rois.size(0);
auto channels = input.size(1);
auto height = input.size(2);
auto width = input.size(3);
// FIXME: This is private, API might change:
// https://github.com/pytorch/pytorch/wiki/Introducing-Quantized-Tensor#quantized-tensor-apis
at::Tensor output = at::_empty_affine_quantized(
{num_rois, channels, pooled_height, pooled_width},
input.options(),
input.q_scale(),
input.q_zero_point());
if (output.numel() == 0)
return output;
AT_DISPATCH_QINT_TYPES(input.scalar_type(), "qroi_align_forward_kernel", [&] {
qroi_align_forward_kernel_impl<scalar_t>(
num_rois,
input,
spatial_scale,
channels,
height,
width,
pooled_height,
pooled_width,
sampling_ratio,
aligned,
rois,
output.data_ptr<scalar_t>());
});
return output;
}
} // namespace
TORCH_LIBRARY_IMPL(torchvision, QuantizedCPU, m) {
m.impl(
TORCH_SELECTIVE_NAME("torchvision::roi_align"),
TORCH_FN(qroi_align_forward_kernel));
}
} // namespace ops
} // namespace vision
|