sum.in.cpp 30.1 KB
Newer Older
1
//*****************************************************************************
2
// Copyright 2017-2020 Intel Corporation
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
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//*****************************************************************************

#include <algorithm>
#include <cinttypes>
#include <cmath>
#include <cstdlib>
#include <random>
#include <string>

#include "gtest/gtest.h"
#include "ngraph/ngraph.hpp"
#include "util/all_close.hpp"
#include "util/all_close_f.hpp"
#include "util/ndarray.hpp"
#include "util/random.hpp"
#include "util/test_control.hpp"
#include "util/test_tools.hpp"

static std::mt19937_64 random_generator;

using namespace std;
using namespace ngraph;

static string s_manifest = "${MANIFEST}";

// Trivial case with no summed axes.
NGRAPH_TEST(${BACKEND_NAME}, sum_trivial)
{
    Shape shape{2, 2};
    auto A = make_shared<op::Parameter>(element::f32, shape);
45
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{}), ParameterVector{A});
46 47 48 49 50 51 52 53

    auto backend = runtime::Backend::create("${BACKEND_NAME}");

    // Create some tensors for input/output
    auto a = backend->create_tensor(element::f32, shape);
    copy_data(a, vector<float>{1, 2, 3, 4});
    auto result = backend->create_tensor(element::f32, shape);

54
    auto handle = backend->compile(f);
55
    handle->call_with_validate({result}, {a});
56
    EXPECT_TRUE(test::all_close_f((vector<float>{1, 2, 3, 4}), read_vector<float>(result)));
57 58 59 60 61 62 63
}

// Failure has been reported at 5D for some reason
NGRAPH_TEST(${BACKEND_NAME}, sum_trivial_5d)
{
    Shape shape{2, 2, 2, 2, 2};
    auto A = make_shared<op::Parameter>(element::f32, shape);
64
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{}), ParameterVector{A});
65 66 67 68 69 70 71 72 73

    auto backend = runtime::Backend::create("${BACKEND_NAME}");

    // Create some tensors for input/output
    auto a = backend->create_tensor(element::f32, shape);
    copy_data(a, vector<float>{1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
                               1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1});
    auto result = backend->create_tensor(element::f32, shape);

74
    auto handle = backend->compile(f);
75
    handle->call_with_validate({result}, {a});
76 77 78
    EXPECT_TRUE(test::all_close_f((vector<float>{1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
                                                 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1}),
                                  read_vector<float>(result)));
79 80 81 82 83 84
}

NGRAPH_TEST(${BACKEND_NAME}, sum_to_scalar)
{
    Shape shape{2, 2};
    auto A = make_shared<op::Parameter>(element::f32, shape);
85
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{0, 1}), ParameterVector{A});
86 87 88 89 90 91 92 93

    auto backend = runtime::Backend::create("${BACKEND_NAME}");

    // Create some tensors for input/output
    auto a = backend->create_tensor(element::f32, shape);
    copy_data(a, vector<float>{1, 2, 3, 4});
    auto result = backend->create_tensor(element::f32, Shape{});

94
    auto handle = backend->compile(f);
95
    handle->call_with_validate({result}, {a});
96
    EXPECT_TRUE(test::all_close_f((vector<float>{10}), read_vector<float>(result)));
97 98 99

    // For some reason I'm feeling extra paranoid about making sure reduction doesn't clobber the
    // input tensors, so let's do this too.
100
    EXPECT_TRUE(test::all_close_f((vector<float>{1, 2, 3, 4}), read_vector<float>(a)));
101 102 103 104 105 106
}

NGRAPH_TEST(${BACKEND_NAME}, sum_large_1d_to_scalar)
{
    Shape shape{1000000};
    auto A = make_shared<op::Parameter>(element::f32, shape);
107
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{0}), ParameterVector{A});
108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123

    auto backend = runtime::Backend::create("${BACKEND_NAME}");

    // Create some tensors for input/output
    random_generator.seed(2);
    vector<float> v_a(1000000, 0);
    double r = 0;
    for (int i = 0; i < 1000000; i++)
    {
        v_a[i] = static_cast<float>(random_generator() % 255);
        r += static_cast<double>(v_a[i]);
    }
    auto a = backend->create_tensor(element::f32, shape);
    copy_data(a, v_a);
    auto result = backend->create_tensor(element::f32, Shape{});

124
    auto handle = backend->compile(f);
125
    handle->call_with_validate({result}, {a});
126 127 128 129 130 131 132 133 134 135

    EXPECT_TRUE(
        test::all_close_f(vector<float>{static_cast<float>(r)}, read_vector<float>(result)));
}

NGRAPH_TEST(${BACKEND_NAME}, sum_matrix_columns)
{
    Shape shape_a{3, 2};
    auto A = make_shared<op::Parameter>(element::f32, shape_a);
    Shape shape_rt{2};
136
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{0}), ParameterVector{A});
137 138 139 140 141 142 143 144

    auto backend = runtime::Backend::create("${BACKEND_NAME}");

    // Create some tensors for input/output
    auto a = backend->create_tensor(element::f32, shape_a);
    copy_data(a, vector<float>{1, 2, 3, 4, 5, 6});
    auto result = backend->create_tensor(element::f32, shape_rt);

145
    auto handle = backend->compile(f);
146
    handle->call_with_validate({result}, {a});
147
    EXPECT_TRUE(test::all_close_f((vector<float>{9, 12}), read_vector<float>(result)));
148 149 150

    // For some reason I'm feeling extra paranoid about making sure reduction doesn't clobber the
    // input tensors, so let's do this too.
151
    EXPECT_TRUE(test::all_close_f((vector<float>{1, 2, 3, 4, 5, 6}), read_vector<float>(a)));
152 153 154 155 156 157 158
}

NGRAPH_TEST(${BACKEND_NAME}, sum_matrix_6d)
{
    Shape shape_a{2, 6, 4, 5, 7, 3};
    auto A = make_shared<op::Parameter>(element::f32, shape_a);
    Shape shape_rt{2, 4, 5, 3};
159
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{1, 4}), ParameterVector{A});
160 161 162 163 164 165 166 167 168 169 170

    auto backend_wrk = runtime::Backend::create("${BACKEND_NAME}");
    auto backend_ref = runtime::Backend::create("INTERPRETER");

    // Create some tensors for input/output
    auto a_wrk = backend_wrk->create_tensor(element::f32, shape_a);
    auto a_ref = backend_ref->create_tensor(element::f32, shape_a);
    auto result_wrk = backend_wrk->create_tensor(element::f32, shape_rt);
    auto result_ref = backend_ref->create_tensor(element::f32, shape_rt);

    vector<float> inp_data(shape_size<const Shape>(shape_a));
171
    iota(inp_data.begin(), inp_data.end(), 1.f);
172 173 174
    copy_data(a_wrk, inp_data);
    copy_data(a_ref, inp_data);

175 176 177 178
    auto handle_wrk = backend_wrk->compile(f);
    auto handle_ref = backend_ref->compile(f);
    handle_wrk->call_with_validate({result_wrk}, {a_wrk});
    handle_ref->call_with_validate({result_ref}, {a_ref});
179

180
    EXPECT_TRUE(test::all_close_f(read_vector<float>(result_ref), read_vector<float>(result_wrk)));
181 182 183 184 185 186 187
}

NGRAPH_TEST(${BACKEND_NAME}, sum_matrix_rows)
{
    Shape shape_a{3, 2};
    auto A = make_shared<op::Parameter>(element::f32, shape_a);
    Shape shape_rt{3};
188
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{1}), ParameterVector{A});
189 190 191 192 193 194 195 196

    auto backend = runtime::Backend::create("${BACKEND_NAME}");

    // Create some tensors for input/output
    auto a = backend->create_tensor(element::f32, shape_a);
    copy_data(a, vector<float>{1, 2, 3, 4, 5, 6});
    auto result = backend->create_tensor(element::f32, shape_rt);

197
    auto handle = backend->compile(f);
198
    handle->call_with_validate({result}, {a});
199
    EXPECT_TRUE(test::all_close_f((vector<float>{3, 7, 11}), read_vector<float>(result)));
200 201 202

    // For some reason I'm feeling extra paranoid about making sure reduction doesn't clobber the
    // input tensors, so let's do this too.
203
    EXPECT_TRUE(test::all_close_f((vector<float>{1, 2, 3, 4, 5, 6}), read_vector<float>(a)));
204 205 206 207 208 209 210
}

NGRAPH_TEST(${BACKEND_NAME}, sum_matrix_rows_zero)
{
    Shape shape_a{3, 0};
    auto A = make_shared<op::Parameter>(element::f32, shape_a);
    Shape shape_rt{3};
211
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{1}), ParameterVector{A});
212 213 214 215 216 217 218 219 220

    auto backend = runtime::Backend::create("${BACKEND_NAME}");

    // Create some tensors for input/output
    auto a = backend->create_tensor(element::f32, shape_a);
    copy_data(a, vector<float>{});
    auto result = backend->create_tensor(element::f32, shape_rt);
    copy_data(result, vector<float>({3, 3, 3}));

221
    auto handle = backend->compile(f);
222
    handle->call_with_validate({result}, {a});
223
    EXPECT_TRUE(test::all_close_f((vector<float>{0, 0, 0}), read_vector<float>(result)));
224 225 226

    // For some reason I'm feeling extra paranoid about making sure reduction doesn't clobber the
    // input tensors, so let's do this too.
227
    EXPECT_TRUE(test::all_close_f((vector<float>{}), read_vector<float>(a)));
228 229 230 231 232 233 234 235
}

NGRAPH_TEST(${BACKEND_NAME}, sum_matrix_cols_zero)
{
    // Now the reduction (g(x:float32[2,2],y:float32[]) = reduce(x,y,f,axes={})).
    Shape shape_a{0, 2};
    auto A = make_shared<op::Parameter>(element::f32, shape_a);
    Shape shape_rt{2};
236
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{0}), ParameterVector{A});
237 238 239 240 241 242 243 244 245

    auto backend = runtime::Backend::create("${BACKEND_NAME}");

    // Create some tensors for input/output
    auto a = backend->create_tensor(element::f32, shape_a);
    copy_data(a, vector<float>{});
    auto result = backend->create_tensor(element::f32, shape_rt);
    copy_data(result, vector<float>({3, 3}));

246
    auto handle = backend->compile(f);
247
    handle->call_with_validate({result}, {a});
248
    EXPECT_TRUE(test::all_close_f((vector<float>{0, 0}), read_vector<float>(result)));
249 250 251

    // For some reason I'm feeling extra paranoid about making sure reduction doesn't clobber the
    // input tensors, so let's do this too.
252
    EXPECT_TRUE(test::all_close_f((vector<float>{}), read_vector<float>(a)));
253 254 255 256 257 258 259
}

NGRAPH_TEST(${BACKEND_NAME}, sum_vector_zero)
{
    Shape shape_a{0};
    auto A = make_shared<op::Parameter>(element::f32, shape_a);
    Shape shape_rt{};
260
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{0}), ParameterVector{A});
261 262 263 264 265 266 267 268 269

    auto backend = runtime::Backend::create("${BACKEND_NAME}");

    // Create some tensors for input/output
    auto a = backend->create_tensor(element::f32, shape_a);
    copy_data(a, vector<float>{});
    auto result = backend->create_tensor(element::f32, shape_rt);
    copy_data(result, vector<float>({3}));

270
    auto handle = backend->compile(f);
271
    handle->call_with_validate({result}, {a});
272
    EXPECT_TRUE(test::all_close_f((vector<float>{0}), read_vector<float>(result)));
273 274 275

    // For some reason I'm feeling extra paranoid about making sure reduction doesn't clobber the
    // input tensors, so let's do this too.
276
    EXPECT_TRUE(test::all_close_f((vector<float>{}), read_vector<float>(a)));
277 278 279 280 281 282 283
}

NGRAPH_TEST(${BACKEND_NAME}, sum_matrix_to_scalar_zero_by_zero)
{
    Shape shape_a{0, 0};
    auto A = make_shared<op::Parameter>(element::f32, shape_a);
    Shape shape_rt{};
284
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{0, 1}), ParameterVector{A});
285 286 287 288 289 290 291 292 293

    auto backend = runtime::Backend::create("${BACKEND_NAME}");

    // Create some tensors for input/output
    auto a = backend->create_tensor(element::f32, shape_a);
    copy_data(a, vector<float>{});
    auto result = backend->create_tensor(element::f32, shape_rt);
    copy_data(result, vector<float>({3}));

294
    auto handle = backend->compile(f);
295
    handle->call_with_validate({result}, {a});
296
    EXPECT_TRUE(test::all_close_f((vector<float>{0}), read_vector<float>(result)));
297 298 299

    // For some reason I'm feeling extra paranoid about making sure reduction doesn't clobber the
    // input tensors, so let's do this too.
300
    EXPECT_TRUE(test::all_close_f((vector<float>{}), read_vector<float>(a)));
301 302 303 304 305 306 307
}

NGRAPH_TEST(${BACKEND_NAME}, sum_3d_to_matrix_most_sig)
{
    Shape shape_a{3, 3, 3};
    auto A = make_shared<op::Parameter>(element::f32, shape_a);
    Shape shape_rt{3, 3};
308
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{0}), ParameterVector{A});
309 310 311 312 313 314 315 316 317

    auto backend = runtime::Backend::create("${BACKEND_NAME}");

    // Create some tensors for input/output
    auto a = backend->create_tensor(element::f32, shape_a);
    copy_data(a, vector<float>{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});
    auto result = backend->create_tensor(element::f32, shape_rt);

318
    auto handle = backend->compile(f);
319
    handle->call_with_validate({result}, {a});
320 321 322 323 324 325 326 327 328 329
    EXPECT_TRUE(test::all_close_f((vector<float>{1 + 10 + 19,
                                                 2 + 11 + 20,
                                                 3 + 12 + 21,
                                                 4 + 13 + 22,
                                                 5 + 14 + 23,
                                                 6 + 15 + 24,
                                                 7 + 16 + 25,
                                                 8 + 17 + 26,
                                                 9 + 18 + 27}),
                                  read_vector<float>(result)));
330 331 332 333 334 335 336
}

NGRAPH_TEST(${BACKEND_NAME}, sum_3d_to_matrix_least_sig)
{
    Shape shape_a{3, 3, 3};
    auto A = make_shared<op::Parameter>(element::f32, shape_a);
    Shape shape_rt{3, 3};
337
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{2}), ParameterVector{A});
338 339 340 341 342 343 344 345 346

    auto backend = runtime::Backend::create("${BACKEND_NAME}");

    // Create some tensors for input/output
    auto a = backend->create_tensor(element::f32, shape_a);
    copy_data(a, vector<float>{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});
    auto result = backend->create_tensor(element::f32, shape_rt);

347
    auto handle = backend->compile(f);
348
    handle->call_with_validate({result}, {a});
349 350 351 352 353 354 355 356 357 358
    EXPECT_TRUE(test::all_close_f((vector<float>{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}),
                                  read_vector<float>(result)));
359 360 361 362 363 364 365
}

NGRAPH_TEST(${BACKEND_NAME}, sum_3d_to_vector)
{
    Shape shape_a{3, 3, 3};
    auto A = make_shared<op::Parameter>(element::f32, shape_a);
    Shape shape_rt{3};
366
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{0, 1}), ParameterVector{A});
367 368 369 370 371 372 373 374 375

    auto backend = runtime::Backend::create("${BACKEND_NAME}");

    // Create some tensors for input/output
    auto a = backend->create_tensor(element::f32, shape_a);
    copy_data(a, vector<float>{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});
    auto result = backend->create_tensor(element::f32, shape_rt);

376
    auto handle = backend->compile(f);
377
    handle->call_with_validate({result}, {a});
378 379 380 381
    EXPECT_TRUE(test::all_close_f((vector<float>{1 + 10 + 19 + 4 + 13 + 22 + 7 + 16 + 25,
                                                 2 + 11 + 20 + 5 + 14 + 23 + 8 + 17 + 26,
                                                 3 + 12 + 21 + 6 + 15 + 24 + 9 + 18 + 27}),
                                  read_vector<float>(result)));
382 383 384 385 386 387 388
}

NGRAPH_TEST(${BACKEND_NAME}, sum_3d_to_scalar)
{
    Shape shape_a{3, 3, 3};
    auto A = make_shared<op::Parameter>(element::f32, shape_a);
    Shape shape_rt{};
389
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{0, 1, 2}), ParameterVector{A});
390 391 392 393 394 395 396 397 398

    auto backend = runtime::Backend::create("${BACKEND_NAME}");

    // Create some tensors for input/output
    auto a = backend->create_tensor(element::f32, shape_a);
    copy_data(a, vector<float>{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});
    auto result = backend->create_tensor(element::f32, shape_rt);

399
    auto handle = backend->compile(f);
400
    handle->call_with_validate({result}, {a});
401 402 403 404
    EXPECT_TRUE(test::all_close_f(
        (vector<float>{1 + 10 + 19 + 4 + 13 + 22 + 7 + 16 + 25 + 2 + 11 + 20 + 5 + 14 + 23 + 8 +
                       17 + 26 + 3 + 12 + 21 + 6 + 15 + 24 + 9 + 18 + 27}),
        read_vector<float>(result)));
405 406
}

407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428
NGRAPH_TEST(${BACKEND_NAME}, sum_3d_to_scalar_int32)
{
    Shape shape_a{3, 3, 3};
    auto A = make_shared<op::Parameter>(element::i32, shape_a);
    Shape shape_rt{};
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{0, 1, 2}), ParameterVector{A});

    auto backend = runtime::Backend::create("${BACKEND_NAME}");

    // Create some tensors for input/output
    auto a = backend->create_tensor(element::i32, shape_a);
    copy_data(a, vector<int32_t>{0x40000001, 10, 19, 4,  13, 22, 7,  16, 25, 2,  11, 20, 5, 14,
                                 23,         8,  17, 26, 3,  12, 21, 6,  15, 24, 9,  18, 27});
    auto result = backend->create_tensor(element::i32, shape_rt);

    auto handle = backend->compile(f);
    handle->call_with_validate({result}, {a});
    EXPECT_EQ((vector<int32_t>{0x40000001 + 10 + 19 + 4 + 13 + 22 + 7 + 16 + 25 + 2 + 11 + 20 + 5 +
                               14 + 23 + 8 + 17 + 26 + 3 + 12 + 21 + 6 + 15 + 24 + 9 + 18 + 27}),
              read_vector<int32_t>(result));
}

429 430 431 432 433
NGRAPH_TEST(${BACKEND_NAME}, sum_3d_eliminate_zero_dim)
{
    Shape shape_a{3, 0, 2};
    auto A = make_shared<op::Parameter>(element::f32, shape_a);
    Shape shape_rt{3, 2};
434
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{1}), ParameterVector{A});
435 436 437 438 439 440 441 442

    auto backend = runtime::Backend::create("${BACKEND_NAME}");

    // Create some tensors for input/output
    auto a = backend->create_tensor(element::f32, shape_a);
    copy_data(a, vector<float>{});
    auto result = backend->create_tensor(element::f32, shape_rt);

443 444
    // Overwrite the initial result vector to make sure we're not just coincidentally getting the
    // right value.
445 446
    copy_data(result, vector<float>{2112, 2112, 2112, 2112, 2112, 2112});

447
    auto handle = backend->compile(f);
448
    handle->call_with_validate({result}, {a});
449
    EXPECT_TRUE(test::all_close_f((vector<float>{0, 0, 0, 0, 0, 0}), read_vector<float>(result)));
450 451
}

452 453 454 455 456 457 458 459 460 461 462 463 464 465
NGRAPH_TEST(${BACKEND_NAME}, sum_3d_eliminate_zero_dim_int32)
{
    Shape shape_a{3, 0, 2};
    auto A = make_shared<op::Parameter>(element::i32, shape_a);
    Shape shape_rt{3, 2};
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{1}), ParameterVector{A});

    auto backend = runtime::Backend::create("${BACKEND_NAME}");

    // Create some tensors for input/output
    auto a = backend->create_tensor(element::i32, shape_a);
    copy_data(a, vector<int32_t>{});
    auto result = backend->create_tensor(element::i32, shape_rt);

466 467
    // Overwrite the initial result vector to make sure we're not just coincidentally getting the
    // right value.
468 469
    copy_data(result, vector<int32_t>{2112, 2112, 2112, 2112, 2112, 2112});

470
    auto handle = backend->compile(f);
471
    handle->call_with_validate({result}, {a});
472 473 474
    EXPECT_EQ((vector<int32_t>{0, 0, 0, 0, 0, 0}), read_vector<int32_t>(result));
}

475 476 477 478 479
NGRAPH_TEST(${BACKEND_NAME}, sum_5d_to_scalar)
{
    Shape shape_a{3, 3, 3, 3, 3};
    auto A = make_shared<op::Parameter>(element::f32, shape_a);
    Shape shape_rt{};
480 481
    auto f =
        make_shared<Function>(make_shared<op::Sum>(A, AxisSet{0, 1, 2, 3, 4}), ParameterVector{A});
482 483 484 485 486 487 488 489

    auto backend = runtime::Backend::create("${BACKEND_NAME}");

    // Create some tensors for input/output
    auto a = backend->create_tensor(element::f32, shape_a);
    copy_data(a, std::vector<float>(std::pow(3, 5), 1));
    auto result = backend->create_tensor(element::f32, shape_rt);

490
    auto handle = backend->compile(f);
491
    handle->call_with_validate({result}, {a});
492
    EXPECT_TRUE(test::all_close_f(std::vector<float>{243.}, read_vector<float>(result)));
493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509
}

NGRAPH_TEST(${BACKEND_NAME}, sum_5d_to_scalar_int32)
{
    Shape shape_a{3, 3, 3, 3, 3};
    auto A = make_shared<op::Parameter>(element::i32, shape_a);
    Shape shape_rt{};
    auto f =
        make_shared<Function>(make_shared<op::Sum>(A, AxisSet{0, 1, 2, 3, 4}), ParameterVector{A});

    auto backend = runtime::Backend::create("${BACKEND_NAME}");

    // Create some tensors for input/output
    auto a = backend->create_tensor(element::i32, shape_a);
    copy_data(a, std::vector<int32_t>(std::pow(3, 5), 1));
    auto result = backend->create_tensor(element::i32, shape_rt);

510
    auto handle = backend->compile(f);
511
    handle->call_with_validate({result}, {a});
512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528
    EXPECT_EQ(std::vector<int32_t>{243}, read_vector<int32_t>(result));
}

NGRAPH_TEST(${BACKEND_NAME}, sum_2d_to_scalar_int8)
{
    Shape shape_a{3, 3};
    auto A = make_shared<op::Parameter>(element::i8, shape_a);
    Shape shape_rt{};
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{0, 1}), ParameterVector{A});

    auto backend = runtime::Backend::create("${BACKEND_NAME}");

    // Create some tensors for input/output
    auto a = backend->create_tensor(element::i8, shape_a);
    copy_data(a, std::vector<int8_t>{1, 2, 3, 4, 5, 6, 7, 8, 9});
    auto result = backend->create_tensor(element::i8, shape_rt);

529
    auto handle = backend->compile(f);
530
    handle->call_with_validate({result}, {a});
531
    EXPECT_EQ(std::vector<int8_t>{45}, read_vector<int8_t>(result));
532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547
}

NGRAPH_TEST(${BACKEND_NAME}, sum_trivial_in_double)
{
    Shape shape{4, 3};
    Shape rshape{3};
    auto A = make_shared<op::Parameter>(element::f64, shape);
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{0}), ParameterVector{A});

    auto backend = runtime::Backend::create("${BACKEND_NAME}");

    // Create some tensors for input/output
    auto a = backend->create_tensor(element::f64, shape);
    copy_data(a, vector<double>{12, 2, 10, 9, 8, 4, 6, 1, 5, 3, 11, 7});
    auto result = backend->create_tensor(element::f64, rshape);

548
    auto handle = backend->compile(f);
549
    handle->call_with_validate({result}, {a});
550
    EXPECT_TRUE(test::all_close_f((vector<double>{30, 22, 26}), read_vector<double>(result)));
551
}
552 553 554

#if NGRAPH_INTERPRETER_ENABLE

555
#ifndef _WIN32
556 557 558 559 560
NGRAPH_TEST(${BACKEND_NAME}, sum_stable_acc)
{
    std::string backend_name = "${BACKEND_NAME}";
    if (backend_name == "INTERPRETER")
    {
561
        return;
562 563 564 565 566
    }
    Shape shape_a{10, 10, 10, 30};
    auto A = make_shared<op::Parameter>(element::f32, shape_a);

    Shape shape_rt{10};
567
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{1, 2, 3}), ParameterVector{A});
568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583

    test::Uniform<float> rng(1000.0f, 1000.1f, 2112);
    vector<vector<float>> args;
    for (shared_ptr<op::Parameter> param : f->get_parameters())
    {
        vector<float> tensor_val(shape_size(param->get_shape()));
        rng.initialize(tensor_val);
        args.push_back(tensor_val);
    }

    auto ref_func = clone_function(*f);
    auto bk_func = clone_function(*f);

    auto ref_results = execute(ref_func, args, "INTERPRETER");
    auto bk_results = execute(bk_func, args, "${BACKEND_NAME}");

584 585
    EXPECT_TRUE(
        test::all_close_f(ref_results.at(0), bk_results.at(0), DEFAULT_FLOAT_TOLERANCE_BITS + 1));
586
}
587
#endif
588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643

NGRAPH_TEST(${BACKEND_NAME}, sum_stable_acc_double)
{
    std::string backend_name = "${BACKEND_NAME}";
    if (backend_name == "INTERPRETER")
    {
        return;
    }
    Shape shape_a{10, 10, 20, 300};
    auto A = make_shared<op::Parameter>(element::f64, shape_a);

    Shape shape_rt{10};
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{1, 2, 3}), ParameterVector{A});

    test::Uniform<double> rng(1000000000.0L, 1000000000.001L, 2112);
    vector<vector<double>> args;
    for (shared_ptr<op::Parameter> param : f->get_parameters())
    {
        vector<double> tensor_val(shape_size(param->get_shape()));
        rng.initialize(tensor_val);
        args.push_back(tensor_val);
    }

    auto ref_func = clone_function(*f);
    auto bk_func = clone_function(*f);

    auto ref_results = execute(ref_func, args, "INTERPRETER");
    auto bk_results = execute(bk_func, args, "${BACKEND_NAME}");

    EXPECT_TRUE(test::all_close(ref_results.at(0), bk_results.at(0), 0.0, 1e-5));
}

NGRAPH_TEST(${BACKEND_NAME}, sum_stable_simple_float)
{
    std::string backend_name = "${BACKEND_NAME}";
    if (backend_name == "INTERPRETER")
    {
        return;
    }
    Shape shape_a{20};
    auto A = make_shared<op::Parameter>(element::f32, shape_a);

    Shape shape_rt{};
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{0}), ParameterVector{A});

    vector<vector<float>> args;
    args.push_back(vector<float>{10000000.0f, 0.9f, 0.3f, 0.4f, 0.5f, 0.6f, 0.7f,
                                 0.8f,        0.1f, 0.9f, 0.5f, 0.2f, 0.3f, 0.4f,
                                 0.5f,        0.6f, 0.7f, 0.8f, 0.9f, 0.1f});

    auto ref_func = clone_function(*f);
    auto bk_func = clone_function(*f);

    auto ref_results = execute(ref_func, args, "INTERPRETER");
    auto bk_results = execute(bk_func, args, "${BACKEND_NAME}");

644 645
    EXPECT_TRUE(
        test::all_close_f(ref_results.at(0), bk_results.at(0), DEFAULT_FLOAT_TOLERANCE_BITS - 1));
646 647
}

648
#ifndef _WIN32
649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691
NGRAPH_TEST(${BACKEND_NAME}, sum_stable_simple_double)
{
    std::string backend_name = "${BACKEND_NAME}";
    if (backend_name == "INTERPRETER")
    {
        return;
    }
    Shape shape_a{20};
    auto A = make_shared<op::Parameter>(element::f64, shape_a);

    Shape shape_rt{};
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{0}), ParameterVector{A});

    vector<vector<double>> args;
    args.push_back(vector<double>{10000000000000000.0L,
                                  0.2L,
                                  0.3L,
                                  0.4L,
                                  0.5L,
                                  0.6L,
                                  0.7L,
                                  0.8L,
                                  0.9L,
                                  0.7L,
                                  0.9L,
                                  0.7L,
                                  0.3L,
                                  0.6L,
                                  0.8L,
                                  0.4L,
                                  0.6L,
                                  0.5L,
                                  0.8L,
                                  0.7L});

    auto ref_func = clone_function(*f);
    auto bk_func = clone_function(*f);

    auto ref_results = execute(ref_func, args, "INTERPRETER");
    auto bk_results = execute(bk_func, args, "${BACKEND_NAME}");

    EXPECT_TRUE(test::all_close(ref_results.at(0), bk_results.at(0), 0.0, 2.0));
}
692
#endif
693 694

#endif
695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744

NGRAPH_TEST(${BACKEND_NAME}, sum_dynamic)
{
    // Create a graph for f(x,axes:int32) = Sum(x,Convert<int64>(axes)).
    auto x = make_shared<op::Parameter>(element::f32, PartialShape::dynamic());
    auto axes = make_shared<op::Parameter>(element::i32, PartialShape{Dimension::dynamic()});
    auto axes_i64 = make_shared<op::Convert>(axes, element::i64);

    auto sum = make_shared<op::Sum>(x, axes_i64);
    ASSERT_TRUE(sum->get_output_partial_shape(0).rank().is_dynamic());

    auto f = make_shared<Function>(NodeVector{sum}, ParameterVector{x, axes});

    auto backend = runtime::Backend::create("${BACKEND_NAME}", true);

    auto ex = backend->compile(f);

    auto t_r = backend->create_dynamic_tensor(element::f32, PartialShape::dynamic());

    std::vector<Shape> x_shapes{
        Shape{2, 3}, Shape{2, 3}, Shape{2, 3}, Shape{2, 3}, Shape{5}, Shape{5}};
    std::vector<std::vector<int32_t>> axeses{{}, {0}, {1}, {0, 1}, {}, {0}};
    std::vector<std::vector<float>> inputs{{1, 2, 3, 4, 5, 6},
                                           {1, 2, 3, 4, 5, 6},
                                           {1, 2, 3, 4, 5, 6},
                                           {1, 2, 3, 4, 5, 6},
                                           {1, 2, 3, 4, 5},
                                           {1, 2, 3, 4, 5}};
    std::vector<Shape> expected_result_shapes{
        Shape{2, 3}, Shape{3}, Shape{2}, Shape{}, Shape{5}, Shape{}};
    std::vector<std::vector<float>> expected_results{
        {1, 2, 3, 4, 5, 6}, {5, 7, 9}, {6, 15}, {21}, {1, 2, 3, 4, 5}, {15}};

    for (size_t i = 0; i < x_shapes.size(); i++)
    {
        auto t_x = backend->create_tensor(element::f32, x_shapes[i]);
        auto t_axes = backend->create_tensor(element::i32, Shape{axeses[i].size()});

        copy_data(t_x, inputs[i]);
        copy_data(t_axes, axeses[i]);

        ex->call_with_validate({t_r}, {t_x, t_axes});

        ASSERT_EQ(t_r->get_shape(), expected_result_shapes[i]);

        auto results = read_vector<float>(t_r);

        ASSERT_TRUE(test::all_close_f(results, expected_results[i], MIN_FLOAT_TOLERANCE_BITS));
    }
}
745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780

NGRAPH_TEST(${BACKEND_NAME}, sum_inf)
{
    Shape shape{7, 4};
    auto A = make_shared<op::Parameter>(element::f32, shape);
    auto f = make_shared<Function>(make_shared<op::Sum>(A, AxisSet{1}), ParameterVector{A});

    auto infi = std::numeric_limits<float>::infinity();

    auto backend = runtime::Backend::create("${BACKEND_NAME}");

    // Create some tensors for input/output
    auto a = backend->create_tensor(element::f32, shape);
    copy_data(a,
              test::NDArray<float, 2>({{-infi, 0, 0, infi},
                                       {infi, 100, -100, -infi},
                                       {infi, 0, 100, infi},
                                       {-infi, -100, 0, -infi},
                                       {infi, infi, infi, infi},
                                       {infi, infi, infi, -infi},
                                       {infi, std::nanf(""), 42, infi}})
                  .get_vector());
    auto result = backend->create_tensor(element::f32, Shape{7});

    auto handle = backend->compile(f);
    handle->call_with_validate({result}, {a});
    auto r = read_vector<float>(result);
    ASSERT_EQ(r.size(), 7);
    EXPECT_TRUE(isnan(r[0]));
    EXPECT_TRUE(isnan(r[1]));
    EXPECT_TRUE(r[2] > 0 && isinf(r[2]));
    EXPECT_TRUE(r[3] < 0 && isinf(r[3]));
    EXPECT_TRUE(r[4] > 0 && isinf(r[4]));
    EXPECT_TRUE(isnan(r[5]));
    EXPECT_TRUE(isnan(r[6]));
}