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
|
#include <catch2/catch_all.hpp>
#include <Eigen/Core>
#include "sopt/ort_session.h"
#include "sopt/utilities.h"
#include "tools_for_tests/directories.h"
using namespace sopt;
TEST_CASE("Load an example ORT model", "[ONNXrt]") {
const std::string path(sopt::tools::models_directory());
ORTsession model(path+"/example_grad_dynamic_CRR_sigma_5_t_5.onnx");
CHECK(true);
}
TEST_CASE("Check metadata of an example ORT model", "[ONNXrt]") {
const std::string path(sopt::tools::models_directory());
ORTsession model(path+"/example_grad_dynamic_CRR_sigma_5_t_5.onnx");
const double L = model.retrieve<double>("L_CRR");
const double L_ref = 0.769605;
CHECK(abs(L_ref - L) < 1e-6*L_ref);
}
TEST_CASE("Check forward folding of an example ORT model using std::vectors", "[ONNXrt]") {
const std::string path(sopt::tools::models_directory());
ORTsession model(path+"/example_grad_dynamic_CRR_sigma_5_t_5.onnx");
const size_t nROWS = 256, nCOLS = 256;
const size_t input_size = 1 * nROWS * nCOLS;
// This network metadata contains an example
// input vector and reference output vector
std::vector<float> input_values, expected_values;
input_values.reserve(input_size);
expected_values.reserve(input_size);
input_values = utilities::split<float>(model.retrieve<std::string>("x_init"), ",");
expected_values = utilities::split<float>(model.retrieve<std::string>("x_exp"), ",");
// forward fold
std::vector<float> output_values = model.compute(input_values, {1,nROWS,nCOLS});
// calculate mean squared error sum_i ( ( x_true(i) - x_est(i) ) **2 )
// check this is less than the number of pixels * 0.01
double mse = 0.0;
for (size_t i = 0; i < output_values.size(); ++i) {
double diff = output_values[i] - expected_values[i];
mse += diff*diff;
}
mse /= output_values.size();
CAPTURE(mse);
CHECK(mse < 0.01);
}
TEST_CASE("Check forward folding of an example ORT model using sopt::Vectors", "[ONNXrt]") {
const std::string path(sopt::tools::models_directory());
ORTsession model(path+"/example_grad_dynamic_CRR_sigma_5_t_5.onnx");
const size_t nROWS = 256, nCOLS = 256;
const size_t input_size = 1 * nROWS * nCOLS;
// This network metadata contains an example
// input vector and reference output vector
std::vector<float> input_values, expected_values;
input_values.reserve(input_size);
expected_values.reserve(input_size);
input_values = utilities::split<float>(model.retrieve<std::string>("x_init"), ",");
expected_values = utilities::split<float>(model.retrieve<std::string>("x_exp"), ",");
// convert flat vectors to sopt::Vector
Vector<double> inputT(input_values.size());
Vector<double> refT(expected_values.size());
for (size_t i = 0; i < input_values.size(); ++i) {
inputT[i] = input_values[i];
refT[i] = expected_values[i];
}
// forward fold using sopt::Vector
Vector<double> outputT = model.compute(inputT,{1,nROWS,nCOLS});
// compare output tensor to reference tensor
// calculate mean squared error sum_i ( ( x_true(i) - x_est(i) ) **2 )
// check this is less than the number of pixels * 0.01
auto mse = (outputT - refT).squaredNorm() / outputT.size();
CAPTURE(mse);
CHECK(mse < 0.01);
}
|
