GithubStar/ncnn
1
0
Fork 0
mirror of https://github.com/Tencent/ncnn.git synced 2026-10-18 21:54:39 +08:00
Code Issues Packages Projects Releases Wiki Activity
Files
master
ncnn/tests/test_lstm.cpp
T
nihui 5ea98309e3 layer param hardening (#7003) 
* harden layer parameter validation

Original patches and author identities:
https://github.com/Tencent/ncnn/pull/6912 (beilzx)
https://github.com/Tencent/ncnn/pull/6919 (FluxMind-O)
https://github.com/Tencent/ncnn/pull/6928 (lxcxjxhx)
https://github.com/Tencent/ncnn/pull/6948 (ktz03)
https://github.com/Tencent/ncnn/pull/6950 (ktz03)
https://github.com/Tencent/ncnn/pull/6971 (LI)
https://github.com/Tencent/ncnn/pull/6974 (LI)
https://github.com/Tencent/ncnn/pull/6977 (LI)

Co-authored-by: beilzx <2069360659@qq.com>
Co-authored-by: FluxMind-O <3283918317@qq.com>
Co-authored-by: lxcxjxhx <lxcxjxhx@users.noreply.github.com>
Co-authored-by: ktz03 <2484593937@qq.com>
Co-authored-by: LI <2484593937@qq.com>
2026-09-15 19:18:46 +08:00

690 lines
24 KiB
C++
Raw Permalink Blame History

// Copyright 2020 Tencent
// SPDX-License-Identifier: BSD-3-Clause
#include "testutil.h"
#include "layer_type.h"
#include <limits.h>
static int test_lstm(int size, int T, int outch, int direction, int hidden_size = 0)
{
ncnn::Mat a = RandomMat(size, T);
int num_directions = direction == 2 ? 2 : 1;
if (hidden_size == 0)
hidden_size = outch;
ncnn::ParamDict pd;
pd.set(0, outch);
pd.set(1, hidden_size * size * 4 * num_directions);
pd.set(2, direction);
pd.set(3, hidden_size);
std::vector<ncnn::Mat> weights(hidden_size == outch ? 3 : 4);
weights[0] = RandomMat(hidden_size * size * 4 * num_directions);
weights[1] = RandomMat(hidden_size * 4 * num_directions);
weights[2] = RandomMat(outch * hidden_size * 4 * num_directions);
if (hidden_size != outch)
{
weights[3] = RandomMat(hidden_size * outch * num_directions);
}
int ret = test_layer("LSTM", pd, weights, a);
if (ret != 0)
{
fprintf(stderr, "test_lstm failed size=%d T=%d outch=%d direction=%d hidden_size=%d\n", size, T, outch, direction, hidden_size);
}
return ret;
}
static int test_lstm_with_hidden(int size, int T, int outch, int direction, int hidden_size = 0)
{
ncnn::Mat a = RandomMat(size, T);
int num_directions = direction == 2 ? 2 : 1;
if (hidden_size == 0)
hidden_size = outch;
ncnn::ParamDict pd;
pd.set(0, outch);
pd.set(1, hidden_size * size * 4 * num_directions);
pd.set(2, direction);
pd.set(3, hidden_size);
std::vector<ncnn::Mat> weights(hidden_size == outch ? 3 : 4);
weights[0] = RandomMat(hidden_size * size * 4 * num_directions);
weights[1] = RandomMat(hidden_size * 4 * num_directions);
weights[2] = RandomMat(outch * hidden_size * 4 * num_directions);
if (hidden_size != outch)
{
weights[3] = RandomMat(hidden_size * outch * num_directions);
}
// initial hidden state
ncnn::Mat hidden = RandomMat(outch, num_directions);
// initial cell state
ncnn::Mat cell = RandomMat(hidden_size, num_directions);
std::vector<ncnn::Mat> as(3);
as[0] = a;
as[1] = hidden;
as[2] = cell;
int ret = test_layer("LSTM", pd, weights, as, 3);
if (ret != 0)
{
fprintf(stderr, "test_lstm_with_hidden failed size=%d T=%d outch=%d direction=%d hidden_size=%d\n", size, T, outch, direction, hidden_size);
}
return ret;
}
static int test_lstm_with_hidden_input(int size, int T, int outch, int direction, int hidden_size = 0)
{
ncnn::Mat a = RandomMat(size, T);
int num_directions = direction == 2 ? 2 : 1;
if (hidden_size == 0)
hidden_size = outch;
ncnn::ParamDict pd;
pd.set(0, outch);
pd.set(1, hidden_size * size * 4 * num_directions);
pd.set(2, direction);
pd.set(3, hidden_size);
std::vector<ncnn::Mat> weights(hidden_size == outch ? 3 : 4);
weights[0] = RandomMat(hidden_size * size * 4 * num_directions);
weights[1] = RandomMat(hidden_size * 4 * num_directions);
weights[2] = RandomMat(outch * hidden_size * 4 * num_directions);
if (hidden_size != outch)
{
weights[3] = RandomMat(hidden_size * outch * num_directions);
}
// initial hidden state
ncnn::Mat hidden = RandomMat(outch, num_directions);
// initial cell state
ncnn::Mat cell = RandomMat(hidden_size, num_directions);
std::vector<ncnn::Mat> as(3);
as[0] = a;
as[1] = hidden;
as[2] = cell;
int ret = test_layer("LSTM", pd, weights, as, 1);
if (ret != 0)
{
fprintf(stderr, "test_lstm_with_hidden_input failed size=%d T=%d outch=%d direction=%d hidden_size=%d\n", size, T, outch, direction, hidden_size);
}
return ret;
}
static int test_lstm_with_hidden_output(int size, int T, int outch, int direction, int hidden_size = 0)
{
ncnn::Mat a = RandomMat(size, T);
int num_directions = direction == 2 ? 2 : 1;
if (hidden_size == 0)
hidden_size = outch;
ncnn::ParamDict pd;
pd.set(0, outch);
pd.set(1, hidden_size * size * 4 * num_directions);
pd.set(2, direction);
pd.set(3, hidden_size);
std::vector<ncnn::Mat> weights(hidden_size == outch ? 3 : 4);
weights[0] = RandomMat(hidden_size * size * 4 * num_directions);
weights[1] = RandomMat(hidden_size * 4 * num_directions);
weights[2] = RandomMat(outch * hidden_size * 4 * num_directions);
if (hidden_size != outch)
{
weights[3] = RandomMat(hidden_size * outch * num_directions);
}
std::vector<ncnn::Mat> as(1);
as[0] = a;
int ret = test_layer("LSTM", pd, weights, as, 3);
if (ret != 0)
{
fprintf(stderr, "test_lstm_with_hidden_output failed size=%d T=%d outch=%d direction=%d hidden_size=%d\n", size, T, outch, direction, hidden_size);
}
return ret;
}
static int test_lstm_0()
{
return 0
|| test_lstm(4, 1, 2, 2)
|| test_lstm(8, 2, 2, 2)
|| test_lstm(16, 8, 7, 2)
|| test_lstm(17, 8, 8, 2)
|| test_lstm(19, 15, 8, 2)
|| test_lstm(5, 16, 16, 2)
|| test_lstm(3, 16, 8, 2)
|| test_lstm(8, 16, 16, 2)
|| test_lstm(31, 3, 31, 2)
|| test_lstm(2, 5, 17, 2, 15);
}
static int test_lstm_1()
{
return 0
|| test_lstm_with_hidden(4, 4, 1, 2)
|| test_lstm_with_hidden(8, 2, 2, 2)
|| test_lstm_with_hidden(16, 8, 7, 2)
|| test_lstm_with_hidden(17, 8, 8, 2)
|| test_lstm_with_hidden(19, 15, 8, 2)
|| test_lstm_with_hidden(5, 16, 16, 2)
|| test_lstm_with_hidden(3, 16, 8, 2)
|| test_lstm_with_hidden(2, 5, 79, 2, 33)
|| test_lstm_with_hidden(4, 4, 1, 1)
|| test_lstm_with_hidden(8, 2, 2, 1)
|| test_lstm_with_hidden(16, 8, 7, 1)
|| test_lstm_with_hidden(17, 8, 8, 1)
|| test_lstm_with_hidden(19, 15, 8, 1)
|| test_lstm_with_hidden(5, 16, 16, 1)
|| test_lstm_with_hidden(3, 16, 8, 1)
|| test_lstm_with_hidden(2, 5, 79, 1, 33)
|| test_lstm_with_hidden(4, 2, 1, 0)
|| test_lstm_with_hidden(8, 2, 2, 0)
|| test_lstm_with_hidden(16, 8, 7, 0)
|| test_lstm_with_hidden(17, 8, 8, 0)
|| test_lstm_with_hidden(19, 15, 8, 0)
|| test_lstm_with_hidden(5, 16, 16, 0)
|| test_lstm_with_hidden(3, 16, 8, 0)
|| test_lstm_with_hidden(2, 5, 17, 0, 15)
|| test_lstm_with_hidden_input(4, 4, 1, 2)
|| test_lstm_with_hidden_input(8, 2, 2, 2)
|| test_lstm_with_hidden_input(16, 8, 7, 2)
|| test_lstm_with_hidden_input(17, 8, 8, 2)
|| test_lstm_with_hidden_input(19, 15, 8, 2)
|| test_lstm_with_hidden_input(5, 16, 16, 2)
|| test_lstm_with_hidden_input(3, 16, 8, 2)
|| test_lstm_with_hidden_input(2, 5, 79, 2, 33)
|| test_lstm_with_hidden_input(4, 4, 1, 1)
|| test_lstm_with_hidden_input(8, 2, 2, 1)
|| test_lstm_with_hidden_input(16, 8, 7, 1)
|| test_lstm_with_hidden_input(17, 8, 8, 1)
|| test_lstm_with_hidden_input(19, 15, 8, 1)
|| test_lstm_with_hidden_input(5, 16, 16, 1)
|| test_lstm_with_hidden_input(3, 16, 8, 1)
|| test_lstm_with_hidden_input(2, 5, 79, 1, 33)
|| test_lstm_with_hidden_input(4, 2, 1, 0)
|| test_lstm_with_hidden_input(8, 2, 2, 0)
|| test_lstm_with_hidden_input(16, 8, 7, 0)
|| test_lstm_with_hidden_input(17, 8, 8, 0)
|| test_lstm_with_hidden_input(19, 15, 8, 0)
|| test_lstm_with_hidden_input(5, 16, 16, 0)
|| test_lstm_with_hidden_input(3, 16, 8, 0)
|| test_lstm_with_hidden_input(2, 5, 17, 0, 15)
|| test_lstm_with_hidden_output(4, 4, 1, 2)
|| test_lstm_with_hidden_output(8, 2, 2, 2)
|| test_lstm_with_hidden_output(16, 8, 7, 2)
|| test_lstm_with_hidden_output(17, 8, 8, 2)
|| test_lstm_with_hidden_output(19, 15, 8, 2)
|| test_lstm_with_hidden_output(5, 16, 16, 2)
|| test_lstm_with_hidden_output(3, 16, 8, 2)
|| test_lstm_with_hidden_output(2, 5, 79, 2, 33)
|| test_lstm_with_hidden_output(4, 4, 1, 1)
|| test_lstm_with_hidden_output(8, 2, 2, 1)
|| test_lstm_with_hidden_output(16, 8, 7, 1)
|| test_lstm_with_hidden_output(17, 8, 8, 1)
|| test_lstm_with_hidden_output(19, 15, 8, 1)
|| test_lstm_with_hidden_output(5, 16, 16, 1)
|| test_lstm_with_hidden_output(3, 16, 8, 1)
|| test_lstm_with_hidden_output(2, 5, 79, 1, 33)
|| test_lstm_with_hidden_output(4, 2, 1, 0)
|| test_lstm_with_hidden_output(8, 2, 2, 0)
|| test_lstm_with_hidden_output(16, 8, 7, 0)
|| test_lstm_with_hidden_output(17, 8, 8, 0)
|| test_lstm_with_hidden_output(19, 15, 8, 0)
|| test_lstm_with_hidden_output(5, 16, 16, 0)
|| test_lstm_with_hidden_output(3, 16, 8, 0)
|| test_lstm_with_hidden_output(2, 5, 17, 0, 15);
}
static int test_lstm_2()
{
return 0
|| test_lstm(4, 1, 1, 0)
|| test_lstm(8, 2, 2, 0)
|| test_lstm(16, 8, 7, 0)
|| test_lstm(17, 8, 8, 0)
|| test_lstm(19, 15, 8, 0)
|| test_lstm(5, 16, 16, 0)
|| test_lstm(3, 16, 8, 0)
|| test_lstm(8, 16, 16, 0)
|| test_lstm(2, 5, 17, 0, 15);
}
static int test_lstm_3()
{
return 0
|| test_lstm(4, 1, 1, 1)
|| test_lstm(8, 2, 2, 1)
|| test_lstm(16, 8, 7, 1)
|| test_lstm(17, 8, 8, 1)
|| test_lstm(19, 15, 8, 1)
|| test_lstm(5, 16, 16, 1)
|| test_lstm(3, 16, 8, 1)
|| test_lstm(8, 16, 16, 1)
|| test_lstm(2, 5, 17, 1, 15);
}
#if NCNN_INT8
static void RandomizeA(ncnn::Mat& m, float absmax)
{
absmax = ncnn::float16_to_float32(ncnn::float32_to_float16(absmax));
absmax = ncnn::bfloat16_to_float32(ncnn::float32_to_bfloat16(absmax));
const int h = m.h;
float* p = m;
for (int i = 0; i < h; i++)
{
float* p = m.row(i);
for (int j = 0; j < m.w; j++)
{
p[j] = RandomFloat(-absmax, absmax);
// drop 0.45 ~ 0.55
float v = p[j] * (127.f / absmax);
float vv = fabs(v - (int)v);
float hp = ncnn::float16_to_float32(ncnn::float32_to_float16(p[j]));
float hv = hp * (127.f / absmax);
float hvv = fabs(hv - (int)hv);
float bp = ncnn::bfloat16_to_float32(ncnn::float32_to_bfloat16(p[j]));
float bv = bp * (127.f / absmax);
float bvv = fabs(bv - (int)bv);
while ((vv > 0.45f && vv < 0.55f) || (hvv > 0.45f && hvv < 0.55f) || (bvv > 0.45f && bvv < 0.55f))
{
p[j] = RandomFloat(-absmax, absmax);
v = p[j] * (127.f / absmax);
vv = fabs(v - (int)v);
hp = ncnn::float16_to_float32(ncnn::float32_to_float16(p[j]));
hv = hp * (127.f / absmax);
hvv = fabs(hv - (int)hv);
bp = ncnn::bfloat16_to_float32(ncnn::float32_to_bfloat16(p[j]));
bv = bp * (127.f / absmax);
bvv = fabs(bv - (int)bv);
}
}
}
// set random a and b
m.row(RandomInt(0, h - 1))[RandomInt(0, m.w - 1)] = -absmax;
m.row(RandomInt(0, h - 1))[RandomInt(0, m.w - 1)] = absmax;
}
static int test_lstm_int8(int size, int T, int outch, int direction, int hidden_size = 0)
{
int num_directions = direction == 2 ? 2 : 1;
if (hidden_size == 0)
hidden_size = outch;
ncnn::ParamDict pd;
pd.set(0, outch);
pd.set(1, hidden_size * size * 4 * num_directions);
pd.set(2, direction);
pd.set(3, hidden_size);
pd.set(8, 2); // int8_scale_term
std::vector<ncnn::Mat> weights(hidden_size == outch ? 5 : 6);
weights[0] = RandomS8Mat(hidden_size * size * 4 * num_directions);
weights[1] = RandomMat(hidden_size * 4 * num_directions);
weights[2] = RandomS8Mat(outch * hidden_size * 4 * num_directions);
if (hidden_size != outch)
{
weights[3] = RandomMat(hidden_size * outch * num_directions);
weights[4] = RandomMat(hidden_size * 4 * num_directions, 100.f, 200.f);
weights[5] = RandomMat(hidden_size * 4 * num_directions, 100.f, 200.f);
}
else
{
weights[3] = RandomMat(hidden_size * 4 * num_directions, 100.f, 200.f);
weights[4] = RandomMat(hidden_size * 4 * num_directions, 100.f, 200.f);
}
ncnn::Mat a(size, T);
RandomizeA(a, 10.f);
int ret = test_layer("LSTM", pd, weights, a);
if (ret != 0)
{
fprintf(stderr, "test_lstm_int8 failed size=%d T=%d outch=%d direction=%d hidden_size=%d\n", size, T, outch, direction, hidden_size);
}
return ret;
}
static int test_lstm_int8_with_hidden(int size, int T, int outch, int direction, int hidden_size = 0)
{
int num_directions = direction == 2 ? 2 : 1;
if (hidden_size == 0)
hidden_size = outch;
ncnn::ParamDict pd;
pd.set(0, outch);
pd.set(1, hidden_size * size * 4 * num_directions);
pd.set(2, direction);
pd.set(3, hidden_size);
pd.set(8, 2); // int8_scale_term
std::vector<ncnn::Mat> weights(hidden_size == outch ? 5 : 6);
weights[0] = RandomS8Mat(hidden_size * size * 4 * num_directions);
weights[1] = RandomMat(hidden_size * 4 * num_directions);
weights[2] = RandomS8Mat(outch * hidden_size * 4 * num_directions);
if (hidden_size != outch)
{
weights[3] = RandomMat(hidden_size * outch * num_directions);
weights[4] = RandomMat(hidden_size * 4 * num_directions, 100.f, 200.f);
weights[5] = RandomMat(hidden_size * 4 * num_directions, 100.f, 200.f);
}
else
{
weights[3] = RandomMat(hidden_size * 4 * num_directions, 100.f, 200.f);
weights[4] = RandomMat(hidden_size * 4 * num_directions, 100.f, 200.f);
}
ncnn::Mat a(size, T);
RandomizeA(a, 10.f);
// initial hidden state
ncnn::Mat hidden(outch, num_directions);
RandomizeA(hidden, 10.f);
// initial cell state
ncnn::Mat cell(hidden_size, num_directions);
RandomizeA(cell, 10.f);
std::vector<ncnn::Mat> as(3);
as[0] = a;
as[1] = hidden;
as[2] = cell;
int ret = test_layer("LSTM", pd, weights, as, 3);
if (ret != 0)
{
fprintf(stderr, "test_lstm_int8_with_hidden failed size=%d T=%d outch=%d direction=%d hidden_size=%d\n", size, T, outch, direction, hidden_size);
}
return ret;
}
static int test_lstm_int8_with_hidden_input(int size, int T, int outch, int direction, int hidden_size = 0)
{
int num_directions = direction == 2 ? 2 : 1;
if (hidden_size == 0)
hidden_size = outch;
ncnn::ParamDict pd;
pd.set(0, outch);
pd.set(1, hidden_size * size * 4 * num_directions);
pd.set(2, direction);
pd.set(3, hidden_size);
pd.set(8, 2); // int8_scale_term
std::vector<ncnn::Mat> weights(hidden_size == outch ? 5 : 6);
weights[0] = RandomS8Mat(hidden_size * size * 4 * num_directions);
weights[1] = RandomMat(hidden_size * 4 * num_directions);
weights[2] = RandomS8Mat(outch * hidden_size * 4 * num_directions);
if (hidden_size != outch)
{
weights[3] = RandomMat(hidden_size * outch * num_directions);
weights[4] = RandomMat(hidden_size * 4 * num_directions, 100.f, 200.f);
weights[5] = RandomMat(hidden_size * 4 * num_directions, 100.f, 200.f);
}
else
{
weights[3] = RandomMat(hidden_size * 4 * num_directions, 100.f, 200.f);
weights[4] = RandomMat(hidden_size * 4 * num_directions, 100.f, 200.f);
}
ncnn::Mat a(size, T);
RandomizeA(a, 10.f);
// initial hidden state
ncnn::Mat hidden(outch, num_directions);
RandomizeA(hidden, 10.f);
// initial cell state
ncnn::Mat cell(hidden_size, num_directions);
RandomizeA(cell, 10.f);
std::vector<ncnn::Mat> as(3);
as[0] = a;
as[1] = hidden;
as[2] = cell;
int ret = test_layer("LSTM", pd, weights, as, 1);
if (ret != 0)
{
fprintf(stderr, "test_lstm_int8_with_hidden_input failed size=%d T=%d outch=%d direction=%d hidden_size=%d\n", size, T, outch, direction, hidden_size);
}
return ret;
}
static int test_lstm_int8_with_hidden_output(int size, int T, int outch, int direction, int hidden_size = 0)
{
int num_directions = direction == 2 ? 2 : 1;
if (hidden_size == 0)
hidden_size = outch;
ncnn::ParamDict pd;
pd.set(0, outch);
pd.set(1, hidden_size * size * 4 * num_directions);
pd.set(2, direction);
pd.set(3, hidden_size);
pd.set(8, 2); // int8_scale_term
std::vector<ncnn::Mat> weights(hidden_size == outch ? 5 : 6);
weights[0] = RandomS8Mat(hidden_size * size * 4 * num_directions);
weights[1] = RandomMat(hidden_size * 4 * num_directions);
weights[2] = RandomS8Mat(outch * hidden_size * 4 * num_directions);
if (hidden_size != outch)
{
weights[3] = RandomMat(hidden_size * outch * num_directions);
weights[4] = RandomMat(hidden_size * 4 * num_directions, 100.f, 200.f);
weights[5] = RandomMat(hidden_size * 4 * num_directions, 100.f, 200.f);
}
else
{
weights[3] = RandomMat(hidden_size * 4 * num_directions, 100.f, 200.f);
weights[4] = RandomMat(hidden_size * 4 * num_directions, 100.f, 200.f);
}
ncnn::Mat a(size, T);
RandomizeA(a, 10.f);
std::vector<ncnn::Mat> as(1);
as[0] = a;
int ret = test_layer("LSTM", pd, weights, as, 3);
if (ret != 0)
{
fprintf(stderr, "test_lstm_int8_with_hidden_output failed size=%d T=%d outch=%d direction=%d hidden_size=%d\n", size, T, outch, direction, hidden_size);
}
return ret;
}
static int test_lstm_4()
{
return 0
|| test_lstm_int8(4, 1, 2, 2)
|| test_lstm_int8(8, 2, 2, 2)
|| test_lstm_int8(16, 8, 7, 2)
|| test_lstm_int8(17, 8, 8, 2)
|| test_lstm_int8(19, 15, 8, 2)
|| test_lstm_int8(5, 16, 16, 2)
|| test_lstm_int8(3, 16, 8, 2)
|| test_lstm_int8(8, 16, 16, 2)
|| test_lstm_int8(31, 3, 31, 2)
|| test_lstm_int8(2, 5, 17, 2, 15);
}
static int test_lstm_5()
{
return 0
|| test_lstm_int8_with_hidden(4, 4, 1, 2)
|| test_lstm_int8_with_hidden(8, 2, 2, 2)
|| test_lstm_int8_with_hidden(16, 8, 7, 2)
|| test_lstm_int8_with_hidden(17, 8, 8, 2)
|| test_lstm_int8_with_hidden(19, 15, 8, 2)
|| test_lstm_int8_with_hidden(5, 16, 16, 2)
|| test_lstm_int8_with_hidden(3, 16, 8, 2)
|| test_lstm_int8_with_hidden(2, 5, 79, 2, 33)
|| test_lstm_int8_with_hidden(4, 4, 1, 1)
|| test_lstm_int8_with_hidden(8, 2, 2, 1)
|| test_lstm_int8_with_hidden(16, 8, 7, 1)
|| test_lstm_int8_with_hidden(17, 8, 8, 1)
|| test_lstm_int8_with_hidden(19, 15, 8, 1)
|| test_lstm_int8_with_hidden(5, 16, 16, 1)
|| test_lstm_int8_with_hidden(3, 16, 8, 1)
|| test_lstm_int8_with_hidden(2, 5, 79, 1, 33)
|| test_lstm_int8_with_hidden(4, 2, 1, 0)
|| test_lstm_int8_with_hidden(8, 2, 2, 0)
|| test_lstm_int8_with_hidden(16, 8, 7, 0)
|| test_lstm_int8_with_hidden(17, 8, 8, 0)
|| test_lstm_int8_with_hidden(19, 15, 8, 0)
|| test_lstm_int8_with_hidden(5, 16, 16, 0)
|| test_lstm_int8_with_hidden(3, 16, 8, 0)
|| test_lstm_int8_with_hidden(2, 5, 17, 0, 15)
|| test_lstm_int8_with_hidden_input(4, 4, 1, 2)
|| test_lstm_int8_with_hidden_input(8, 2, 2, 2)
|| test_lstm_int8_with_hidden_input(16, 8, 7, 2)
|| test_lstm_int8_with_hidden_input(17, 8, 8, 2)
|| test_lstm_int8_with_hidden_input(19, 15, 8, 2)
|| test_lstm_int8_with_hidden_input(5, 16, 16, 2)
|| test_lstm_int8_with_hidden_input(3, 16, 8, 2)
|| test_lstm_int8_with_hidden_input(2, 5, 79, 2, 33)
|| test_lstm_int8_with_hidden_input(4, 4, 1, 1)
|| test_lstm_int8_with_hidden_input(8, 2, 2, 1)
|| test_lstm_int8_with_hidden_input(16, 8, 7, 1)
|| test_lstm_int8_with_hidden_input(17, 8, 8, 1)
|| test_lstm_int8_with_hidden_input(19, 15, 8, 1)
|| test_lstm_int8_with_hidden_input(5, 16, 16, 1)
|| test_lstm_int8_with_hidden_input(3, 16, 8, 1)
|| test_lstm_int8_with_hidden_input(2, 5, 79, 1, 33)
|| test_lstm_int8_with_hidden_input(4, 2, 1, 0)
|| test_lstm_int8_with_hidden_input(8, 2, 2, 0)
|| test_lstm_int8_with_hidden_input(16, 8, 7, 0)
|| test_lstm_int8_with_hidden_input(17, 8, 8, 0)
|| test_lstm_int8_with_hidden_input(19, 15, 8, 0)
|| test_lstm_int8_with_hidden_input(5, 16, 16, 0)
|| test_lstm_int8_with_hidden_input(3, 16, 8, 0)
|| test_lstm_int8_with_hidden_input(2, 5, 17, 0, 15)
|| test_lstm_int8_with_hidden_output(4, 4, 1, 2)
|| test_lstm_int8_with_hidden_output(8, 2, 2, 2)
|| test_lstm_int8_with_hidden_output(16, 8, 7, 2)
|| test_lstm_int8_with_hidden_output(17, 8, 8, 2)
|| test_lstm_int8_with_hidden_output(19, 15, 8, 2)
|| test_lstm_int8_with_hidden_output(5, 16, 16, 2)
|| test_lstm_int8_with_hidden_output(3, 16, 8, 2)
|| test_lstm_int8_with_hidden_output(2, 5, 79, 2, 33)
|| test_lstm_int8_with_hidden_output(4, 4, 1, 1)
|| test_lstm_int8_with_hidden_output(8, 2, 2, 1)
|| test_lstm_int8_with_hidden_output(16, 8, 7, 1)
|| test_lstm_int8_with_hidden_output(17, 8, 8, 1)
|| test_lstm_int8_with_hidden_output(19, 15, 8, 1)
|| test_lstm_int8_with_hidden_output(5, 16, 16, 1)
|| test_lstm_int8_with_hidden_output(3, 16, 8, 1)
|| test_lstm_int8_with_hidden_output(2, 5, 79, 1, 33)
|| test_lstm_int8_with_hidden_output(4, 2, 1, 0)
|| test_lstm_int8_with_hidden_output(8, 2, 2, 0)
|| test_lstm_int8_with_hidden_output(16, 8, 7, 0)
|| test_lstm_int8_with_hidden_output(17, 8, 8, 0)
|| test_lstm_int8_with_hidden_output(19, 15, 8, 0)
|| test_lstm_int8_with_hidden_output(5, 16, 16, 0)
|| test_lstm_int8_with_hidden_output(3, 16, 8, 0)
|| test_lstm_int8_with_hidden_output(2, 5, 17, 0, 15);
}
static int test_lstm_6()
{
return 0
|| test_lstm_int8(4, 1, 1, 0)
|| test_lstm_int8(8, 2, 2, 0)
|| test_lstm_int8(16, 8, 7, 0)
|| test_lstm_int8(17, 8, 8, 0)
|| test_lstm_int8(19, 15, 8, 0)
|| test_lstm_int8(5, 16, 16, 0)
|| test_lstm_int8(3, 16, 8, 0)
|| test_lstm_int8(8, 16, 16, 0)
|| test_lstm_int8(2, 5, 17, 0, 15);
}
static int test_lstm_7()
{
return 0
|| test_lstm_int8(4, 1, 1, 1)
|| test_lstm_int8(8, 2, 2, 1)
|| test_lstm_int8(16, 8, 7, 1)
|| test_lstm_int8(17, 8, 8, 1)
|| test_lstm_int8(19, 15, 8, 1)
|| test_lstm_int8(5, 16, 16, 1)
|| test_lstm_int8(3, 16, 8, 1)
|| test_lstm_int8(8, 16, 16, 1)
|| test_lstm_int8(2, 5, 17, 1, 15);
}
#endif
#if NCNN_VALIDATION
static int test_lstm_load_param()
{
ncnn::ParamDict base;
base.set(0, 8);
base.set(1, 192);
if (test_layer_param(ncnn::LayerType::LSTM, base, 0) != 0)
return -1;
const int invalid[] = {0, -1, INT_MIN, INT_MAX};
for (int i = 0; i < 4; i++)
{
if (test_layer_param(ncnn::LayerType::LSTM, base, 0, invalid[i], -1) != 0)
return -1;
}
return 0
|| test_layer_param(ncnn::LayerType::LSTM, base, 2, 3, -1)
|| test_layer_param(ncnn::LayerType::LSTM, base, 1, 193, -1)
|| test_layer_param(ncnn::LayerType::LSTM, base, 3, 0, -1)
|| test_layer_param(ncnn::LayerType::LSTM, base, 3, INT_MAX, -1);
}
#endif // NCNN_VALIDATION
int main()
{
SRAND(7767517);
return 0
|| test_lstm_0()
|| test_lstm_1()
|| test_lstm_2()
|| test_lstm_3()
#if NCNN_INT8
|| test_lstm_4()
|| test_lstm_5()
|| test_lstm_6()
|| test_lstm_7()
#endif // NCNN_INT8
#if NCNN_VALIDATION
|| test_lstm_load_param()
#endif // NCNN_VALIDATION
;
}
Reference in New Issue View Git Blame Copy Permalink