mirror of
https://github.com/Tencent/ncnn.git
synced 2026-10-17 04:41:47 +08:00
nihui
401 lines
15 KiB
C++
401 lines
15 KiB
C++
// Copyright 2019 Tencent
|
|
// SPDX-License-Identifier: BSD-3-Clause
|
|
|
|
#include "testutil.h"
|
|
|
|
#if NCNN_INT8
|
|
static int test_convolution_int8(int w, int h, int c, int outch, int kernel, int dilation, int stride, int pad, int bias, bool requant = false, int int8_scale_term = 0, bool sgemm = false, bool input_int8 = false, bool test_winograd43 = false, int winograd = 0)
|
|
{
|
|
ncnn::Mat a = RandomMat(w, h, c);
|
|
|
|
if (int8_scale_term == 0)
|
|
int8_scale_term = requant ? 101 : 1;
|
|
const bool use_requant = int8_scale_term > 100;
|
|
|
|
ncnn::ParamDict pd;
|
|
pd.set(0, outch);
|
|
pd.set(1, kernel);
|
|
pd.set(2, dilation);
|
|
pd.set(3, stride);
|
|
pd.set(4, pad);
|
|
pd.set(5, bias);
|
|
pd.set(6, outch * c * kernel * kernel);
|
|
pd.set(8, int8_scale_term); // int8_scale_term
|
|
|
|
int activation_type = RAND() % 7; // 0 1 2 3 4 5 6
|
|
ncnn::Mat activation_params(2);
|
|
activation_params[0] = (activation_type == 6) ? RandomFloat(0, 1) : RandomFloat(-1, 0); // alpha
|
|
activation_params[1] = RandomFloat(0, 1); // beta
|
|
pd.set(9, activation_type);
|
|
pd.set(10, activation_params);
|
|
|
|
std::vector<ncnn::Mat> weights(bias ? 5 : 4);
|
|
weights[0] = RandomMat(outch * c * kernel * kernel);
|
|
|
|
ncnn::Mat weight_scales = scales_mat(weights[0], outch, c * kernel * kernel, c * kernel * kernel);
|
|
ncnn::Mat input_scales = scales_mat(a, 1, w * h * c, a.cstep);
|
|
ncnn::Mat top_scales = use_requant ? scales_mat(a, 1, w * h * c, a.cstep) : ncnn::Mat();
|
|
|
|
ncnn::Mat a_int8 = a;
|
|
if (input_int8)
|
|
{
|
|
ncnn::Option opt;
|
|
opt.num_threads = 1;
|
|
opt.use_packing_layout = false;
|
|
ncnn::quantize_to_int8(a, a_int8, input_scales, opt);
|
|
}
|
|
|
|
if (kernel == 3 && dilation == 1 && stride == 1)
|
|
{
|
|
// test for 6bit quant
|
|
for (int i = 0; i < weight_scales.w; i++)
|
|
weight_scales[i] = weight_scales[i] / 4.f;
|
|
}
|
|
|
|
if (bias)
|
|
{
|
|
weights[1] = RandomMat(outch);
|
|
weights[2] = weight_scales;
|
|
weights[3] = input_scales;
|
|
weights[4] = top_scales;
|
|
}
|
|
else
|
|
{
|
|
weights[1] = weight_scales;
|
|
weights[2] = input_scales;
|
|
weights[3] = top_scales;
|
|
}
|
|
|
|
int flag = input_int8 ? TEST_LAYER_DISABLE_AUTO_INPUT_CASTING : 0;
|
|
if (input_int8 && (sgemm || winograd))
|
|
flag |= TEST_LAYER_ENABLE_THREADING;
|
|
|
|
if (winograd)
|
|
{
|
|
ncnn::Option opt;
|
|
opt.num_threads = 1;
|
|
opt.use_packing_layout = false;
|
|
opt.use_fp16_packed = false;
|
|
opt.use_fp16_storage = false;
|
|
opt.use_fp16_arithmetic = false;
|
|
opt.use_bf16_packed = false;
|
|
opt.use_bf16_storage = false;
|
|
opt.use_sgemm_convolution = false;
|
|
opt.use_winograd_convolution = true;
|
|
opt.use_winograd23_convolution = winograd == 23;
|
|
opt.use_winograd43_convolution = winograd == 43;
|
|
|
|
int ret = test_layer_opt("Convolution", pd, weights, opt, a_int8, use_requant ? 1.0f : 0.001f, flag);
|
|
if (ret != 0)
|
|
{
|
|
fprintf(stderr, "test_convolution_int8 failed w=%d h=%d c=%d outch=%d kernel=%d dilation=%d stride=%d pad=%d bias=%d int8_scale_term=%d input_int8=%d packing=%d winograd=%d act=%d actparams=[%f,%f]\n", w, h, c, outch, kernel, dilation, stride, pad, bias, int8_scale_term, input_int8, opt.use_packing_layout, winograd, activation_type, activation_params[0], activation_params[1]);
|
|
}
|
|
return ret;
|
|
}
|
|
|
|
int ret = 0;
|
|
if (input_int8)
|
|
{
|
|
ncnn::Option opt;
|
|
opt.num_threads = 1;
|
|
opt.use_fp16_packed = false;
|
|
opt.use_fp16_storage = false;
|
|
opt.use_fp16_arithmetic = false;
|
|
opt.use_bf16_packed = false;
|
|
opt.use_bf16_storage = false;
|
|
opt.use_sgemm_convolution = sgemm;
|
|
opt.use_winograd_convolution = false;
|
|
|
|
for (int packing = 0; packing < 2; packing++)
|
|
{
|
|
opt.use_packing_layout = packing != 0;
|
|
ret = test_layer_opt("Convolution", pd, weights, opt, a_int8, use_requant ? 1.0f : 0.001f, flag);
|
|
if (ret != 0)
|
|
{
|
|
fprintf(stderr, "test_convolution_int8 failed w=%d h=%d c=%d outch=%d kernel=%d dilation=%d stride=%d pad=%d bias=%d int8_scale_term=%d sgemm=%d input_int8=%d packing=%d act=%d actparams=[%f,%f]\n", w, h, c, outch, kernel, dilation, stride, pad, bias, int8_scale_term, sgemm, input_int8, opt.use_packing_layout, activation_type, activation_params[0], activation_params[1]);
|
|
return ret;
|
|
}
|
|
}
|
|
}
|
|
else
|
|
{
|
|
ret = test_layer("Convolution", pd, weights, a_int8, use_requant ? 1.0f : 0.001f, flag);
|
|
}
|
|
if (ret != 0)
|
|
{
|
|
fprintf(stderr, "test_convolution_int8 failed w=%d h=%d c=%d outch=%d kernel=%d dilation=%d stride=%d pad=%d bias=%d int8_scale_term=%d sgemm=%d input_int8=%d act=%d actparams=[%f,%f]\n", w, h, c, outch, kernel, dilation, stride, pad, bias, int8_scale_term, sgemm, input_int8, activation_type, activation_params[0], activation_params[1]);
|
|
return ret;
|
|
}
|
|
|
|
if (input_int8)
|
|
return ret;
|
|
|
|
if (kernel == 3 && dilation == 1 && stride == 1)
|
|
{
|
|
ncnn::Option opt;
|
|
opt.num_threads = 1;
|
|
opt.use_packing_layout = true;
|
|
opt.use_fp16_packed = false;
|
|
opt.use_fp16_storage = false;
|
|
opt.use_fp16_arithmetic = false;
|
|
opt.use_bf16_packed = false;
|
|
opt.use_bf16_storage = false;
|
|
opt.use_sgemm_convolution = false;
|
|
opt.use_winograd_convolution = true;
|
|
for (int i = 0; i < (test_winograd43 ? 2 : 1); i++)
|
|
{
|
|
opt.use_winograd23_convolution = i == 0;
|
|
opt.use_winograd43_convolution = i == 1;
|
|
|
|
ret = test_layer_opt("Convolution", pd, weights, opt, a, use_requant ? 1.0f : 0.001f, flag);
|
|
if (ret != 0)
|
|
{
|
|
fprintf(stderr, "test_convolution_int8 failed w=%d h=%d c=%d outch=%d kernel=%d dilation=%d stride=%d pad=%d bias=%d int8_scale_term=%d input_int8=%d packing=%d winograd=%d act=%d actparams=[%f,%f]\n", w, h, c, outch, kernel, dilation, stride, pad, bias, int8_scale_term, input_int8, opt.use_packing_layout, i == 0 ? 23 : 43, activation_type, activation_params[0], activation_params[1]);
|
|
return ret;
|
|
}
|
|
}
|
|
}
|
|
|
|
{
|
|
ncnn::Option opt;
|
|
opt.num_threads = 1;
|
|
opt.use_packing_layout = false;
|
|
opt.use_fp16_packed = false;
|
|
opt.use_fp16_storage = false;
|
|
opt.use_fp16_arithmetic = false;
|
|
opt.use_bf16_packed = false;
|
|
opt.use_bf16_storage = false;
|
|
opt.use_sgemm_convolution = false;
|
|
opt.use_winograd_convolution = false;
|
|
|
|
ret = test_layer_opt("Convolution", pd, weights, opt, a, use_requant ? 1.0f : 0.001f, flag);
|
|
if (ret != 0)
|
|
{
|
|
fprintf(stderr, "test_convolution_int8 failed w=%d h=%d c=%d outch=%d kernel=%d dilation=%d stride=%d pad=%d bias=%d int8_scale_term=%d act=%d actparams=[%f,%f]\n", w, h, c, outch, kernel, dilation, stride, pad, bias, int8_scale_term, activation_type, activation_params[0], activation_params[1]);
|
|
return ret;
|
|
}
|
|
}
|
|
|
|
{
|
|
ncnn::Option opt;
|
|
opt.num_threads = 1;
|
|
opt.use_packing_layout = true;
|
|
opt.use_fp16_packed = false;
|
|
opt.use_fp16_storage = false;
|
|
opt.use_fp16_arithmetic = false;
|
|
opt.use_bf16_packed = false;
|
|
opt.use_bf16_storage = false;
|
|
opt.use_sgemm_convolution = false;
|
|
opt.use_winograd_convolution = false;
|
|
|
|
ret = test_layer_opt("Convolution", pd, weights, opt, a, use_requant ? 1.0f : 0.001f, flag);
|
|
if (ret != 0)
|
|
{
|
|
fprintf(stderr, "test_convolution_int8 failed w=%d h=%d c=%d outch=%d kernel=%d dilation=%d stride=%d pad=%d bias=%d int8_scale_term=%d act=%d actparams=[%f,%f]\n", w, h, c, outch, kernel, dilation, stride, pad, bias, int8_scale_term, activation_type, activation_params[0], activation_params[1]);
|
|
return ret;
|
|
}
|
|
}
|
|
|
|
{
|
|
ncnn::Option opt;
|
|
opt.num_threads = 1;
|
|
opt.use_packing_layout = true;
|
|
opt.use_fp16_packed = true;
|
|
opt.use_fp16_storage = true;
|
|
opt.use_fp16_arithmetic = true;
|
|
opt.use_bf16_packed = true;
|
|
opt.use_bf16_storage = true;
|
|
opt.use_sgemm_convolution = false;
|
|
opt.use_winograd_convolution = false;
|
|
|
|
ret = test_layer_opt("Convolution", pd, weights, opt, a, use_requant ? 1.0f : 0.001f, flag);
|
|
if (ret != 0)
|
|
{
|
|
fprintf(stderr, "test_convolution_int8 failed w=%d h=%d c=%d outch=%d kernel=%d dilation=%d stride=%d pad=%d bias=%d int8_scale_term=%d act=%d actparams=[%f,%f]\n", w, h, c, outch, kernel, dilation, stride, pad, bias, int8_scale_term, activation_type, activation_params[0], activation_params[1]);
|
|
return ret;
|
|
}
|
|
}
|
|
|
|
if (sgemm)
|
|
{
|
|
ncnn::Option opt;
|
|
opt.num_threads = 1;
|
|
opt.use_packing_layout = true;
|
|
opt.use_fp16_packed = false;
|
|
opt.use_fp16_storage = false;
|
|
opt.use_fp16_arithmetic = false;
|
|
opt.use_bf16_packed = false;
|
|
opt.use_bf16_storage = false;
|
|
opt.use_sgemm_convolution = true;
|
|
opt.use_winograd_convolution = false;
|
|
|
|
ret = test_layer_opt("Convolution", pd, weights, opt, a, use_requant ? 1.0f : 0.001f, flag);
|
|
if (ret != 0)
|
|
{
|
|
fprintf(stderr, "test_convolution_int8 failed w=%d h=%d c=%d outch=%d kernel=%d dilation=%d stride=%d pad=%d bias=%d int8_scale_term=%d act=%d actparams=[%f,%f]\n", w, h, c, outch, kernel, dilation, stride, pad, bias, int8_scale_term, activation_type, activation_params[0], activation_params[1]);
|
|
return ret;
|
|
}
|
|
}
|
|
|
|
return ret;
|
|
}
|
|
|
|
static int test_convolution_1()
|
|
{
|
|
// exercise each packing direction on the pointwise and 3x3 paths
|
|
static const int kdsp_packing[2][4] = {
|
|
{1, 1, 1, 0},
|
|
{3, 1, 1, 1},
|
|
};
|
|
|
|
static const int channels_packing[10][2] = {
|
|
{1, 1},
|
|
{2, 2},
|
|
{3, 3},
|
|
{4, 4},
|
|
{7, 7},
|
|
{8, 8},
|
|
{15, 15},
|
|
{16, 15},
|
|
{15, 16},
|
|
{16, 16},
|
|
};
|
|
|
|
for (int q = 0; q < 2; q++)
|
|
{
|
|
for (int i = 0; i < 2; i++)
|
|
{
|
|
const int k = kdsp_packing[i][0];
|
|
const int d = kdsp_packing[i][1];
|
|
const int s = kdsp_packing[i][2];
|
|
const int p = kdsp_packing[i][3];
|
|
|
|
for (int j = 0; j < 10; j++)
|
|
{
|
|
const int c = channels_packing[j][0];
|
|
const int outch = channels_packing[j][1];
|
|
if (test_convolution_int8(9, 7, c, outch, k, d, s, p, 1, q != 0) != 0)
|
|
return -1;
|
|
}
|
|
}
|
|
}
|
|
|
|
// vary kernel geometry with packed, unpacked and mixed channel directions
|
|
static const int kdsp_geometry[5][4] = {
|
|
{2, 1, 1, 1},
|
|
{3, 2, 1, 1},
|
|
{4, 1, 1, 2},
|
|
{4, 2, 1, -234},
|
|
{5, 1, 1, -234},
|
|
};
|
|
|
|
static const int channels_geometry[6][2] = {
|
|
{1, 1},
|
|
{4, 4},
|
|
{8, 8},
|
|
{16, 15},
|
|
{15, 16},
|
|
{16, 16},
|
|
};
|
|
|
|
for (int q = 0; q < 2; q++)
|
|
{
|
|
for (int i = 0; i < 5; i++)
|
|
{
|
|
const int k = kdsp_geometry[i][0];
|
|
const int d = kdsp_geometry[i][1];
|
|
const int s = kdsp_geometry[i][2];
|
|
const int p = kdsp_geometry[i][3];
|
|
|
|
for (int j = 0; j < 6; j++)
|
|
{
|
|
const int c = channels_geometry[j][0];
|
|
const int outch = channels_geometry[j][1];
|
|
if (test_convolution_int8(9, 7, c, outch, k, d, s, p, 1, q != 0) != 0)
|
|
return -1;
|
|
}
|
|
}
|
|
}
|
|
|
|
// vary large kernels separately from channel packing transitions
|
|
static const int kdsp_7x7[2][4] = {
|
|
{7, 1, 1, 3},
|
|
{7, 2, 1, -233},
|
|
};
|
|
|
|
static const int channels_7x7[2][2] = {
|
|
{1, 1},
|
|
{8, 8},
|
|
};
|
|
|
|
for (int q = 0; q < 2; q++)
|
|
{
|
|
for (int i = 0; i < 2; i++)
|
|
{
|
|
const int k = kdsp_7x7[i][0];
|
|
const int d = kdsp_7x7[i][1];
|
|
const int s = kdsp_7x7[i][2];
|
|
const int p = kdsp_7x7[i][3];
|
|
|
|
for (int j = 0; j < 2; j++)
|
|
{
|
|
const int c = channels_7x7[j][0];
|
|
const int outch = channels_7x7[j][1];
|
|
if (test_convolution_int8(9, 7, c, outch, k, d, s, p, 1, q != 0) != 0)
|
|
return -1;
|
|
}
|
|
}
|
|
}
|
|
|
|
return 0
|
|
|| test_convolution_int8(11, 11, 8, 16, 3, 1, 1, 1, 1)
|
|
|| test_convolution_int8(13, 16, 16, 24, 3, 1, 1, 1, 1)
|
|
|| test_convolution_int8(8, 8, 16, 24, 3, 1, 1, 1, 0)
|
|
|| test_convolution_int8(4, 8, 16, 24, 3, 1, 1, 1, 1)
|
|
|| test_convolution_int8(4, 20, 16, 24, 3, 1, 1, 1, 0)
|
|
|| test_convolution_int8(25, 33, 16, 15, 3, 1, 1, 1, 0)
|
|
|| test_convolution_int8(25, 33, 31, 31, 3, 1, 1, 1, 0)
|
|
|| test_convolution_int8(9, 7, 32, 16, 3, 1, 1, 1, 1, false, 0, false, false, true)
|
|
|| test_convolution_int8(17, 13, 64, 64, 3, 1, 1, 1, 1, false, 0, false, false, true)
|
|
|| test_convolution_int8(13, 11, 65, 97, 3, 1, 1, 1, 1, false, 0, false, false, true)
|
|
|| test_convolution_int8(13, 11, 65, 97, 3, 1, 1, 1, 1, true, 0, false, false, true)
|
|
|| test_convolution_int8(7, 7, 15, 12, 3, 1, 1, 1, 0)
|
|
|| test_convolution_int8(5, 6, 31, 9, 5, 1, 1, 0, 1)
|
|
|| test_convolution_int8(5, 10, 5, 32, 3, 2, 1, 0, 1)
|
|
|| test_convolution_int8(3, 9, 16, 13, 2, 2, 1, 0, 0)
|
|
|| test_convolution_int8(23, 11, 33, 28, 5, 1, 1, 0, 1)
|
|
|| test_convolution_int8(7, 5, 4, 8, 1, 1, 1, 0, 1, false, 2)
|
|
|| test_convolution_int8(7, 5, 4, 8, 1, 1, 1, 0, 1, true, 102);
|
|
}
|
|
|
|
static int test_convolution_1_int8_input()
|
|
{
|
|
return 0
|
|
|| test_convolution_int8(7, 5, 1, 1, 3, 1, 1, 1, 1, false, 1, false, true)
|
|
|| test_convolution_int8(7, 5, 4, 4, 3, 1, 1, 1, 1, false, 1, false, true)
|
|
|| test_convolution_int8(8, 6, 4, 8, 1, 1, 1, 0, 1, false, 1, false, true)
|
|
|| test_convolution_int8(8, 6, 4, 8, 1, 1, 1, 0, 1, false, 2, false, true)
|
|
|| test_convolution_int8(8, 6, 4, 8, 1, 1, 1, 0, 1, true, 102, false, true)
|
|
|| test_convolution_int8(9, 7, 8, 8, 2, 1, 1, 1, 1, false, 1, true, true)
|
|
// large reduction depth with packed channels and scalar tails
|
|
|| test_convolution_int8(5, 7, 2048, 32, 3, 1, 1, 0, 1, false, 1, true, true)
|
|
|| test_convolution_int8(5, 7, 3072, 33, 3, 1, 1, 0, 1, false, 1, true, true)
|
|
// F(2,3) scalar transforms, split-K tails and threaded tile sizing
|
|
|| test_convolution_int8(7, 3, 384, 9, 3, 1, 1, 0, 1, false, 1, false, true, false, 23);
|
|
}
|
|
|
|
#endif // NCNN_INT8
|
|
|
|
int main()
|
|
{
|
|
SRAND(7767517);
|
|
|
|
#if NCNN_INT8
|
|
return 0
|
|
|| test_convolution_1()
|
|
|| test_convolution_1_int8_input();
|
|
#endif
|
|
|
|
return 0;
|
|
}
|